Figure 5. Schematic view of flap and support structure [32]

Design Optimization of Ocean Renewable Energy Converter Using a Combined Bi-level Metaheuristic Approach

결합된 Bi-level 메타휴리스틱 접근법을 사용한 해양 재생 에너지 변환기의 설계 최적화

Erfan Amini a1, Mahdieh Nasiri b1, Navid Salami Pargoo a, Zahra Mozhgani c, Danial Golbaz d, Mehrdad Baniesmaeil e, Meysam Majidi Nezhad f, Mehdi Neshat gj, Davide Astiaso Garcia h, Georgios Sylaios i


In recent years, there has been an increasing interest in renewable energies in view of the fact that fossil fuels are the leading cause of catastrophic environmental consequences. Ocean wave energy is a renewable energy source that is particularly prevalent in coastal areas. Since many countries have tremendous potential to extract this type of energy, a number of researchers have sought to determine certain effective factors on wave converters’ performance, with a primary emphasis on ambient factors. In this study, we used metaheuristic optimization methods to investigate the effects of geometric factors on the performance of an Oscillating Surge Wave Energy Converter (OSWEC), in addition to the effects of hydrodynamic parameters. To do so, we used CATIA software to model different geometries which were then inserted into a numerical model developed in Flow3D software. A Ribed-surface design of the converter’s flap is also introduced in this study to maximize wave-converter interaction. Besides, a Bi-level Hill Climbing Multi-Verse Optimization (HCMVO) method was also developed for this application. The results showed that the converter performs better with greater wave heights, flap freeboard heights, and shorter wave periods. Additionally, the added ribs led to more wave-converter interaction and better performance, while the distance between the flap and flume bed negatively impacted the performance. Finally, tracking the changes in the five-dimensional objective function revealed the optimum value for each parameter in all scenarios. This is achieved by the newly developed optimization algorithm, which is much faster than other existing cutting-edge metaheuristic approaches.


Wave Energy Converter


Hydrodynamic Effects

Geometric Design

Metaheuristic Optimization

Multi-Verse Optimizer


The increase in energy demand, the limitations of fossil fuels, as well as environmental crises, such as air pollution and global warming, are the leading causes of calling more attention to harvesting renewable energy recently [1][2][3]. While still in its infancy, ocean wave energy has neither reached commercial maturity nor technological convergence. In recent decades, remarkable progress has been made in the marine energy domain, which is still in the early stage of development, to improve the technology performance level (TPL) [4][5]and technology readiness level (TRL) of wave energy converters (WECs). This has been achieved using novel modeling techniques [6][7][8][9][10][11][12][13][14] to gain the following advantages [15]: (i) As a source of sustainable energy, it contributes to the mix of energy resources that leads to greater diversity and attractiveness for coastal cities and suppliers. [16] (ii) Since wave energy can be exploited offshore and does not require any land, in-land site selection would be less expensive and undesirable visual effects would be reduced. [17] (iii) When the best layout and location of offshore site are taken into account, permanent generation of energy will be feasible (as opposed to using solar energy, for example, which is time-dependent) [18].

In general, the energy conversion process can be divided into three stages in a WEC device, including primary, secondary, and tertiary stages [19][20]. In the first stage of energy conversion, which is the subject of this study, the wave power is converted to mechanical power by wave-structure interaction (WSI) between ocean waves and structures. Moreover, the mechanical power is transferred into electricity in the second stage, in which mechanical structures are coupled with power take-off systems (PTO). At this stage, optimal control strategies are useful to tune the system dynamics to maximize power output [10][13][12]. Furthermore, the tertiary energy conversion stage revolves around transferring the non-standard AC power into direct current (DC) power for energy storage or standard AC power for grid integration [21][22]. We discuss only the first stage regardless of the secondary and tertiary stages. While Page 1 of 16 WECs include several categories and technologies such as terminators, point absorbers, and attenuators [15][23], we focus on oscillating surge wave energy converters (OSWECs) in this paper due to its high capacity for industrialization [24].

Over the past two decades, a number of studies have been conducted to understand how OSWECs’ structures and interactions between ocean waves and flaps affect converters performance. Henry et al.’s experiment on oscillating surge wave energy converters is considered as one of the most influential pieces of research [25], which demonstrated how the performance of oscillating surge wave energy converters (OSWECs) is affected by seven different factors, including wave period, wave power, flap’s relative density, water depth, free-board of the flap, the gap between the tubes, gap underneath the flap, and flap width. These parameters were assessed in their two models in order to estimate the absorbed energy from incoming waves [26][27]. In addition, Folly et al. investigated the impact of water depth on the OSWECs performance analytically, numerically, and experimentally. According to this and further similar studies, the average annual incident wave power is significantly reduced by water depth. Based on the experimental results, both the surge wave force and the power capture of OSWECs increase in shallow water [28][29]. Following this, Sarkar et al. found that under such circumstances, the device that is located near the coast performs much better than those in the open ocean [30]. On the other hand, other studies are showing that the size of the converter, including height and width, is relatively independent of the location (within similar depth) [31]. Subsequently, Schmitt et al. studied OSWECs numerically and experimentally. In fact, for the simulation of OSWEC, OpenFOAM was used to test the applicability of Reynolds-averaged Navier-Stokes (RANS) solvers. Then, the experimental model reproduced the numerical results with satisfying accuracy [32]. In another influential study, Wang et al. numerically assessed the effect of OSWEC’s width on their performance. According to their findings, as converter width increases, its efficiency decreases in short wave periods while increases in long wave periods [33]. One of the main challenges in the analysis of the OSWEC is the coupled effect of hydrodynamic and geometric variables. As a result, numerous cutting-edge geometry studies have been performed in recent years in order to find the optimal structure that maximizes power output and minimizes costs. Garcia et al. reviewed hull geometry optimization studies in the literature in [19]. In addition, Guo and Ringwood surveyed geometric optimization methods to improve the hydrodynamic performance of OSWECs at the primary stage [14]. Besides, they classified the hull geometry of OSWECs based on Figure 1. Subsequently, Whittaker et al. proposed a different design of OSWEC called Oyster2. There have been three examples of different geometries of oysters with different water depths. Based on its water depth, they determined the width and height of the converter. They also found that in the constant wave period the less the converter’s width, the less power captures the converter has [34]. Afterward, O’Boyle et al. investigated a type of OSWEC called Oyster 800. They compared the experimental and numerical models with the prototype model. In order to precisely reproduce the shape, mass distribution, and buoyancy properties of the prototype, a 40th-scale experimental model has been designed. Overall, all the models were fairly accurate according to the results [35].

Inclusive analysis of recent research avenues in the area of flap geometry has revealed that the interaction-based designs of such converters are emerging as a novel approach. An initiative workflow is designed in the current study to maximizing the wave energy extrication by such systems. To begin with, a sensitivity analysis plays its role of determining the best hydrodynamic values for installing the converter’s flap. Then, all flap dimensions and characteristics come into play to finalize the primary model. Following, interactive designs is proposed to increase the influence of incident waves on the body by adding ribs on both sides of the flap as a novel design. Finally, a new bi-level metaheuristic method is proposed to consider the effects of simultaneous changes in ribs properties and other design parameters. We hope this novel approach will be utilized to make big-scale projects less costly and justifiable. The efficiency of the method is also compared with four well known metaheuristic algorithms and out weight them for this application.

This paper is organized as follows. First, the research methodology is introduced by providing details about the numerical model implementation. To that end, we first introduced the primary model’s geometry and software details. That primary model is later verified with a benchmark study with regard to the flap angle of rotation and water surface elevation. Then, governing equations and performance criteria are presented. In the third part of the paper, we discuss the model’s sensitivity to lower and upper parts width (we proposed a two cross-sectional design for the flap), bottom elevation, and freeboard. Finally, the novel optimization approach is introduced in the final part and compared with four recent metaheuristic algorithms.

2. Numerical Methods

In this section, after a brief introduction of the numerical software, Flow3D, boundary conditions are defined. Afterwards, the numerical model implementation, along with primary model properties are described. Finally, governing equations, as part of numerical process, are discussed.

2.1Model Setup

FLOW-3D is a powerful and comprehensive CFD simulation platform for studying fluid dynamics. This software has several modules to solve many complex engineering problems. In addition, modeling complex flows is simple and effective using FLOW-3D’s robust meshing capabilities [36]. Interaction between fluid and moving objects might alter the computational range. Dynamic meshes are used in our modeling to take these changes into account. At each time step, the computational node positions change in order to adapt the meshing area to the moving object. In addition, to choose mesh dimensions, some factors are taken into account such as computational accuracy, computational time, and stability. The final grid size is selected based on the detailed procedure provided in [37]. To that end, we performed grid-independence testing on a CFD model using three different mesh grid sizes of 0.01, 0.015, and 0.02 meters. The problem geometry and boundary conditions were defined the same, and simulations were run on all three grids under the same conditions. The predicted values of the relevant variable, such as velocity, was compared between the grids. The convergence behavior of the numerical solution was analyzed by calculating the relative L2 norm error between two consecutive grids. Based on the results obtained, it was found that the grid size of 0.02 meters showed the least error, indicating that it provided the most accurate and reliable solution among the three grids. Therefore, the grid size of 0.02 meters was selected as the optimal spatial resolution for the mesh grid.

In this work, the flume dimensions are 10 meters long, 0.1 meters wide, and 2.2 meters high, which are shown in figure2. In addition, input waves with linear characteristics have a height of 0.1 meters and a period of 1.4 seconds. Among the linear wave methods included in this software, RNGk-ε and k- ε are appropriate for turbulence model. The research of Lopez et al. shows that RNGk- ε provides the most accurate simulation of turbulence in OSWECs [21]. We use CATIA software to create the flap primary model and other innovative designs for this project. The flap measures 0.1 m x 0.65 m x 0.360 m in x, y and z directions, respectively. In Figure 3, the primary model of flap and its dimensions are shown. In this simulation, five boundaries have been defined, including 1. Inlet, 2. Outlet, 3. Converter flap, 4. Bed flume, and 5. Water surface, which are shown in figure 2. Besides, to avoid wave reflection in inlet and outlet zones, Flow3D is capable of defining some areas as damping zones, the length of which has to be one to one and a half times the wavelength. Therefore, in the model, this length is considered equal to 2 meters. Furthermore, there is no slip in all the boundaries. In other words, at every single time step, the fluid velocity is zero on the bed flume, while it is equal to the flap velocity on the converter flap. According to the wave theory defined in the software, at the inlet boundary, the water velocity is called from the wave speed to be fed into the model.


In the current study, we utilize the Schmitt experimental model as a benchmark for verification, which was developed at the Queen’s University of Belfast. The experiments were conducted on the flap of the converter, its rotation, and its interaction with the water surface. Thus, the details of the experiments are presented below based up on the experimental setup’s description [38]. In the experiment, the laboratory flume has a length of 20m and a width of 4.58m. Besides, in order to avoid incident wave reflection, a wave absorption source is devised at the end of the left flume. The flume bed, also, includes two parts with different slops. The flap position and dimensions of the flume can be seen in Figure4. In addition, a wave-maker with 6 paddles is installed at one end. At the opposite end, there is a beach with wire meshes. Additionally, there are 6 indicators to extract the water level elevation. In the flap model, there are three components: the fixed support structure, the hinge, and the flap. The flap measures 0.1m x 0.65m x 0.341m in x, y and z directions, respectively. In Figure5, the details are given [32]. The support structure consists of a 15 mm thick stainless steel base plate measuring 1m by 1.4m, which is screwed onto the bottom of the tank. The hinge is supported by three bearing blocks. There is a foam centerpiece on the front and back of the flap which is sandwiched between two PVC plates. Enabling changes of the flap, three metal fittings link the flap to the hinge. Moreover, in this experiment, the selected wave is generated based on sea wave data at scale 1:40. The wave height and the wave period are equal to 0.038 (m) and 2.0625 (s), respectively, which are tantamount to a wave with a period of 13 (s) and a height of 1.5 (m).

Two distinct graphs illustrate the numerical and experi-mental study results. Figure6 and Figure7 are denoting the angle of rotation of flap and surface elevation in computational and experimental models, respectively. The two figures roughly represent that the numerical and experimental models are a good match. However, for the purpose of verifying the match, we calculated the correlation coefficient (C) and root mean square error (RMSE). According to Figure6, correlation coefficient and RMSE are 0.998 and 0.003, respectively, and in Figure7 correlation coefficient and RMSE are respectively 0.999 and 0.001. Accordingly, there is a good match between the numerical and empirical models. It is worth mentioning that the small differences between the numerical and experimental outputs may be due to the error of the measuring devices and the calibration of the data collection devices.

Including continuity equation and momentum conserva- tion for incompressible fluid are given as [32][39]:(1)

where P represents the pressure, g denotes gravitational acceleration, u represents fluid velocity, and Di is damping coefficient. Likewise, the model uses the same equation. to calculate the fluid velocity in other directions as well. Considering the turbulence, we use the two-equation model of RNGK- ε. These equations are:

(3)��t(��)+����(����)=����[�eff�������]+��-��and(4)���(��)+����(����)=����[�eff�������]+�1�∗����-��2��2�Where �2� and �1� are constants. In addition, �� and �� represent the turbulent Prandtl number of � and k, respectively.

�� also denote the production of turbulent kinetic energy of k under the effect of velocity gradient, which is calculated as follows:(5)��=�eff[�����+�����]�����(6)�eff=�+��(7)�eff=�+��where � is molecular viscosity,�� represents turbulence viscosity, k denotes kinetic energy, and ∊∊ is energy dissipation rate. The values of constant coefficients in the two-equation RNGK ∊-∊ model is as shown in the Table 1 [40].Table 2.

Table 1. Constant coefficients in RNGK- model


Table 2. Flap properties

Joint height (m)0.476
Height of the center of mass (m)0.53
Weight (Kg)10.77

It is worth mentioning that the volume of fluid method is used to separate water and air phases in this software [41]. Below is the equation of this method [40].(8)����+����(���)=0where α and 1 − α are portion of water phase and air phase, respectively. As a weighting factor, each fluid phase portion is used to determine the mixture properties. Finally, using the following equations, we calculate the efficiency of converters [42][34][43]:(9)�=14|�|2�+�2+(�+�a)2(�n2-�2)2where �� represents natural frequency, I denotes the inertia of OSWEC, Ia is the added inertia, F is the complex wave force, and B denotes the hydrodynamic damping coefficient. Afterward, the capture factor of the converter is calculated by [44]:(10)��=�1/2��2����gw where �� represents the capture factor, which is the total efficiency of device per unit length of the wave crest at each time step [15], �� represent the dimensional amplitude of the incident wave, w is the flap’s width, and Cg is the group velocity of the incident wave, as below:(11)��=��0·121+2�0ℎsinh2�0ℎwhere �0 denotes the wave number, h is water depth, and H is the height of incident waves.

According to previous sections ∊,����-∊ modeling is used for all models simulated in this section. For this purpose, the empty boundary condition is used for flume walls. In order to preventing wave reflection at the inlet and outlet of the flume, the length of wave absorption is set to be at least one incident wavelength. In addition, the structured mesh is chosen, and the mesh dimensions are selected in two distinct directions. In each model, all grids have a length of 2 (cm) and a height of 1 (cm). Afterwards, as an input of the software for all of the models, we define the time step as 0.001 (s). Moreover, the run time of every simulation is 30 (s). As mentioned before, our primary model is Schmitt model, and the flap properties is given in table2. For all simulations, the flume measures 15 meters in length and 0.65 meters in width, and water depth is equal to 0.335 (m). The flap is also located 7 meters from the flume’s inlet.

Finally, in order to compare the results, the capture factor is calculated for each simulation and compared to the primary model. It is worth mentioning that capture factor refers to the ratio of absorbed wave energy to the input wave energy.

According to primary model simulation and due to the decreasing horizontal velocity with depth, the wave crest has the highest velocity. Considering the fact that the wave’s orbital velocity causes the flap to move, the contact between the upper edge of the flap and the incident wave can enhance its performance. Additionally, the numerical model shows that the dynamic pressure decreases as depth increases, and the hydrostatic pressure increases as depth increases.

To determine the OSWEC design, it is imperative to understand the correlation between the capture factor, wave period, and wave height. Therefore, as it is shown in Figure8, we plot the change in capture factor over the variations in wave period and wave height in 3D and 2D. In this diagram, the first axis features changes in wave period, the second axis displays changes in wave height, and the third axis depicts changes in capture factor. According to our wave properties in the numerical model, the wave period and wave height range from 2 to 14 seconds and 2 to 8 meters, respectively. This is due to the fact that the flap does not oscillate if the wave height is less than 2 (m), and it does not reverse if the wave height is more than 8 (m). In addition, with wave periods more than 14 (s), the wavelength would be so long that it would violate the deep-water conditions, and with wave periods less than 2 (s), the flap would not oscillate properly due to the shortness of wavelength. The results of simulation are shown in Figure 8. As it can be perceived from Figure 8, in a constant wave period, the capture factor is in direct proportion to the wave height. It is because of the fact that waves with more height have more energy to rotate the flap. Besides, in a constant wave height, the capture factor increases when the wave period increases, until a given wave period value. However, the capture factor falls after this point. These results are expected since the flap’s angular displacement is not high in lower wave periods, while the oscillating motion of that is not fast enough to activate the power take-off system in very high wave periods.

As is shown in Figure 9, we plot the change in capture factor over the variations in wave period (s) and water depth (m) in 3D. As it can be seen in this diagram, the first axis features changes in water depth (m), the second axis depicts the wave period (s), and the third axis displays OSWEC’s capture factor. The wave period ranges from 0 to 10 seconds based on our wave properties, which have been adopted from Schmitt’s model, while water depth ranges from 0 to 0.5 meters according to the flume and flap dimensions and laboratory limitations. According to Figure9, for any specific water depth, the capture factor increases in a varying rate when the wave period increases, until a given wave period value. However, the capture factor falls steadily after this point. In fact, the maximum capture factor occurs when the wave period is around 6 seconds. This trend is expected since, in a specific water depth, the flap cannot oscillate properly when the wavelength is too short. As the wave period increases, the flap can oscillate more easily, and consequently its capture factor increases. However, the capture factor drops in higher wave periods because the wavelength is too large to move the flap. Furthermore, in a constant wave period, by changing the water depth, the capture factor does not alter. In other words, the capture factor does not depend on the water depth when it is around its maximum value.

3Sensitivity Analysis

Based on previous studies, in addition to the flap design, the location of the flap relative to the water surface (freeboard) and its elevation relative to the flume bed (flap bottom elevation) play a significant role in extracting energy from the wave energy converter. This study measures the sensitivity of the model to various parameters related to the flap design including upper part width of the flap, lower part width of the flap, the freeboard, and the flap bottom elevation. Moreover, as a novel idea, we propose that the flap widths differ in the lower and upper parts. In Figure10, as an example, a flap with an upper thickness of 100 (mm) and a lower thickness of 50 (mm) and a flap with an upper thickness of 50 (mm) and a lower thickness of 100 (mm) are shown. The influence of such discrepancy between the widths of the upper and lower parts on the interaction between the wave and the flap, or in other words on the capture factor, is evaluated. To do so, other parameters are remained constant, such as the freeboard, the distance between the flap and the flume bed, and the wave properties.

In Figure11, models are simulated with distinct upper and lower widths. As it is clear in this figure, the first axis depicts the lower part width of the flap, the second axis indicates the upper part width of the flap, and the colors represent the capture factor values. Additionally, in order to consider a sufficient range of change, the flap thickness varies from half to double the value of the primary model for each part.

According to this study, the greater the discrepancy in these two parts, the lower the capture factor. It is on account of the fact that when the lower part of the flap is thicker than the upper part, and this thickness difference in these two parts is extremely conspicuous, the inertia against the motion is significant at zero degrees of rotation. Consequently, it is difficult to move the flap, which results in a low capture factor. Similarly, when the upper part of the flap is thicker than the lower part, and this thickness difference in these two parts is exceedingly noticeable, the inertia is so great that the flap can not reverse at the maximum degree of rotation. As the results indicate, the discrepancy can enhance the performance of the converter if the difference between these two parts is around 20%. As it is depicted in the Figure11, the capture factor reaches its own maximum amount, when the lower part thickness is from 5 to 6 (cm), and the upper part thickness is between 6 and 7 (cm). Consequently, as a result of this discrepancy, less material will be used, and therefore there will be less cost.

As illustrated in Figure12, this study examines the effects of freeboard (level difference between the flap top and water surface) and the flap bottom elevation (the distance between the flume bed and flap bottom) on the converter performance. In this diagram, the first axis demonstrates the freeboard and the second axis on the left side displays the flap bottom elevation, while the colors indicate the capture factor. In addition, the feasible range of freeboard is between -15 to 15 (cm) due to the limitation of the numerical model, so that we can take the wave slamming and the overtopping into consideration. Additionally, based on the Schmitt model and its scaled model of 1:40 of the base height, the flap bottom should be at least 9 (cm) high. Since the effect of surface waves is distributed over the depth of the flume, it is imperative to maintain a reasonable flap height exposed to incoming waves. Thus, the maximum flap bottom elevation is limited to 19 (cm). As the Figure12 pictures, at constant negative values of the freeboard, the capture factor is in inverse proportion with the flap bottom elevation, although slightly.

Furthermore, at constant positive values of the freeboard, the capture factor fluctuates as the flap bottom elevation decreases while it maintains an overall increasing trend. This is on account of the fact that increasing the flap bottom elevation creates turbulence flow behind the flap, which encumbers its rotation, as well as the fact that the flap surface has less interaction with the incoming waves. Furthermore, while keeping the flap bottom elevation constant, the capture factor increases by raising the freeboard. This is due to the fact that there is overtopping with adverse impacts on the converter performance when the freeboard is negative and the flap is under the water surface. Besides, increasing the freeboard makes the wave slam more vigorously, which improves the converter performance.

Adding ribs to the flap surface, as shown in Figure13, is a novel idea that is investigated in the next section. To achieve an optimized design for the proposed geometry of the flap, we determine the optimal number and dimensions of ribs based on the flap properties as our decision variables in the optimization process. As an example, Figure13 illustrates a flap with 3 ribs on each side with specific dimensions.

Figure14 shows the flow velocity field around the flap jointed to the flume bed. During the oscillation of the flap, the pressure on the upper and lower surfaces of the flap changes dynamically due to the changing angle of attack and the resulting change in the direction of fluid flow. As the flap moves upwards, the pressure on the upper surface decreases, and the pressure on the lower surface increases. Conversely, as the flap moves downwards, the pressure on the upper surface increases, and the pressure on the lower surface decreases. This results in a cyclic pressure variation around the flap. Under certain conditions, the pressure field around the flap can exhibit significant variations in magnitude and direction, forming vortices and other flow structures. These flow structures can affect the performance of the OSWEC by altering the lift and drag forces acting on the flap.

4Design Optimization

We consider optimizing the design parameters of the flap of converter using a nature-based swarm optimization method, that fall in the category of metaheuristic algorithms [45]. Accordingly, we choose four state-of-the-art algorithms to perform an optimization study. Then, based on their performances to achieve the highest capture factor, one of them will be chosen to be combined with the Hill Climb algorithm to carry out a local search. Therefore, in the remainder of this section, we discuss the search process of each algorithm and visualize their performance and convergence curve as they try to find the best values for decision variables.

4.1. Metaheuristic Approaches

As the first considered algorithm, the Gray Wolf Optimizer (GWO) algorithm simulates the natural leadership and hunting performance of gray wolves which tend to live in colonies. Hunters must obey the alpha wolf, the leader, who is responsible for hunting. Then, the beta wolf is at the second level of the gray wolf hierarchy. A subordinate of alpha wolf, beta stands under the command of the alpha. At the next level in this hierarchy, there are the delta wolves. They are subordinate to the alpha and beta wolves. This category of wolves includes scouts, sentinels, elders, hunters, and caretakers. In this ranking, omega wolves are at the bottom, having the lowest level and obeying all other wolves. They are also allowed to eat the prey just after others have eaten. Despite the fact that they seem less important than others, they are really central to the pack survival. Since, it has been shown that without omega wolves, the entire pack would experience some problems like fighting, violence, and frustration. In this simulation, there are three primary steps of hunting including searching, surrounding, and finally attacking the prey. Mathematically model of gray wolves’ hunting technique and their social hierarchy are applied in determined by optimization. this study. As mentioned before, gray wolves can locate their prey and surround them. The alpha wolf also leads the hunt. Assuming that the alpha, beta, and delta have more knowledge about prey locations, we can mathematically simulate gray wolf hunting behavior. Hence, in addition to saving the top three best solutions obtained so far, we compel the rest of the search agents (also the omegas) to adjust their positions based on the best search agent. Encircling behavior can be mathematically modeled by the following equations: [46].(12)�→=|�→·��→(�)-�→(�)|(13)�→(�+1)=��→(�)-�→·�→(14)�→=2.�2→(15)�→=2�→·�1→-�→Where �→indicates the position vector of gray wolf, ��→ defines the vector of prey, t indicates the current iteration, and �→and �→are coefficient vectors. To force the search agent to diverge from the prey, we use �→ with random values greater than 1 or less than -1. In addition, C→ contains random values in the range [0,2], and �→ 1 and �2→ are random vectors in [0,1]. The second considered technique is the Moth Flame Optimizer (MFO) algorithm. This method revolves around the moths’ navigation mechanism, which is realized by positioning themselves and maintaining a fixed angle relative to the moon while flying. This effective mechanism helps moths to fly in a straight path. However, when the source of light is artificial, maintaining an angle with the light leads to a spiral flying path towards the source that causes the moth’s death [47]. In MFO algorithm, moths and flames are both solutions. The moths are actual search agents that fly in hyper-dimensional space by changing their position vectors, and the flames are considered pins that moths drop when searching the search space [48]. The problem’s variables are the position of moths in the space. Each moth searches around a flame and updates it in case of finding a better solution. The fitness value is the return value of each moth’s fitness (objective) function. The position vector of each moth is passed to the fitness function, and the output of the fitness function is assigned to the corresponding moth. With this mechanism, a moth never loses its best solution [49]. Some attributes of this algorithm are as follows:

  • •It takes different values to converge moth in any point around the flame.
  • •Distance to the flame is lowered to be eventually minimized.
  • •When the position gets closer to the flame, the updated positions around the flame become more frequent.

As another method, the Multi-Verse Optimizer is based on a multiverse theory which proposes there are other universes besides the one in which we all live. According to this theory, there are more than one big bang in the universe, and each big bang leads to the birth of a new universe [50]. Multi-Verse Optimizer (MVO) is mainly inspired by three phenomena in cosmology: white holes, black holes, and wormholes. A white hole has never been observed in our universe, but physicists believe the big bang could be considered a white hole [51]. Black holes, which behave completely in contrast to white holes, attract everything including light beams with their extremely high gravitational force [52]. In the multiverse theory, wormholes are time and space tunnels that allow objects to move instantly between any two corners of a universe (or even simultaneously from one universe to another) [53]. Based on these three concepts, mathematical models are designed to perform exploration, exploitation, and local search, respectively. The concept of white and black holes is implied as an exploration phase, while the concept of wormholes is considered as an exploitation phase by MVO. Additionally, each solution is analogous to a universe, and each variable in the solution represents an object in that universe. Furthermore, each solution is assigned an inflation rate, and the time is used instead of iterations. Following are the universe rules in MVO:

  • •The possibility of having white hole increases with the inflation rate.
  • •The possibility of having black hole decreases with the inflation rate.
  • •Objects tend to pass through black holes more frequently in universes with lower inflation rates.
  • •Regardless of inflation rate, wormholes may cause objects in universes to move randomly towards the best universe. [54]

Modeling the white/black hole tunnels and exchanging objects of universes mathematically was accomplished by using the roulette wheel mechanism. With every iteration, the universes are sorted according to their inflation rates, then, based on the roulette wheel, the one with the white hole is selected as the local extremum solution. This is accomplished through the following steps:

Assume that


Where ��� represents the jth parameter of the ith universe, Ui indicates the ith universe, NI(Ui) is normalized inflation rate of the ith universe, r1 is a random number in [0,1], and j xk shows the jth parameter of the kth universe selected by a roulette wheel selection mechanism [54]. It is assumed that wormhole tunnels always exist between a universe and the best universe formed so far. This mechanism is as follows:(17)���=if�2<���:��+���×((���-���)×�4+���)�3<0.5��-���×((���-���)×�4+���)�3≥0.5����:���where Xj indicates the jth parameter of the best universe formed so far, TDR and WEP are coefficients, where Xj indicates the jth parameter of the best universelbjshows the lower bound of the jth variable, ubj is the upper bound of the jth variable, and r2, r3, and r4 are random numbers in [1][54].

Finally, one of the newest optimization algorithms is WOA. The WOA algorithm simulates the movement of prey and the whale’s discipline when looking for their prey. Among several species, Humpback whales have a specific method of hunting [55]. Humpback whales can recognize the location of prey and encircle it before hunting. The optimal design position in the search space is not known a priori, and the WOA algorithm assumes that the best candidate solution is either the target prey or close to the optimum. This foraging behavior is called the bubble-net feeding method. Two maneuvers are associated with bubbles: upward spirals and double loops. A unique behavior exhibited only by humpback whales is bubble-net feeding. In fact, The WOA algorithm starts with a set of random solutions. At each iteration, search agents update their positions for either a randomly chosen search agent or the best solution obtained so far [56][55]. When the best search agent is determined, the other search agents will attempt to update their positions toward that agent. It is important to note that humpback whales swim around their prey simultaneously in a circular, shrinking circle and along a spiral-shaped path. By using a mathematical model, the spiral bubble-net feeding maneuver is optimized. The following equation represents this behavior:(18)�→(�+1)=�′→·�bl·cos(2��)+�∗→(�)


X→(t+ 1) indicates the distance of the it h whale to the prey (best solution obtained so far),� is a constant for defining the shape of the logarithmic spiral, l is a random number in [−1, 1], and dot (.) is an element-by-element multiplication [55].

Comparing the four above-mentioned methods, simulations are run with 10 search agents for 400 iterations. In Figure 15, there are 20 plots the optimal values of different parameters in optimization algorithms. The five parameters of this study are freeboard, bottom elevations, number of ribs on the converter, rib thickness, and rib Height. The optimal value for each was found by optimization algorithms, naming WOA, MVO, MFO, and GWO. By looking through the first row, the freeboard parameter converges to its maximum possible value in the optimization process of GWO after 300 iterations. Similarly, MFO finds the same result as GWO. In contrast, the freeboard converges to its minimum possible value in MVO optimizing process, which indicates positioning the converter under the water. Furthermore, WOA found the optimal value of freeboard as around 0.02 after almost 200 iterations. In the second row, the bottom elevation is found at almost 0.11 (m) in all algorithms; however, the curves follow different trends in each algorithm. The third row shows the number of ribs, where results immediately reveal that it should be over 4. All algorithms coincide at 5 ribs as the optimal number in this process. The fourth row displays the trends of algorithms to find optimal rib thickness. MFO finds the optimal value early and sets it to around 0.022, while others find the same value in higher iterations. Finally, regarding the rib height, MVO, MFO, and GWO state that the optimal value is 0.06 meters, but WOA did not find a higher value than 0.039.

4.2. HCMVO Bi-level Approach

Despite several strong search characteristics of MVO and its high performance in various optimization problems, it suffers from a few deficiencies in local and global search mechanisms. For instance, it is trapped in the local optimum when wormholes stochastically generate many solutions near the best universe achieved throughout iterations, especially in solving complex multimodal problems with high dimensions [57]. Furthermore, MVO needs to be modified by an escaping strategy from the local optima to enhance the global search abilities. To address these shortages, we propose a fast and effective meta-algorithm (HCMVO) to combine MVO with a Random-restart hill-climbing local search. This meta-algorithm uses MVO on the upper level to develop global tracking and provide a range of feasible and proper solutions. The hill-climbing algorithm is designed to develop a comprehensive neighborhood search around the best-found solution proposed by the upper-level (MVO) when MVO is faced with a stagnation issue or falling into a local optimum. The performance threshold is formulated as follows.(20)Δ����THD=∑�=1�����TH��-����TH��-1�where BestTHDis the best-found solution per generation, andM is related to the domain of iterations to compute the average performance of MVO. If the proposed best solution by the local search is better than the initial one, the global best of MVO will be updated. HCMVO iteratively runs hill climbing when the performance of MVO goes down, each time with an initial condition to prepare for escaping such undesirable situations. In order to get a better balance between exploration and exploitation, the search step size linearly decreases as follows:(21)��=��-����Ma�iter��+1where iter and Maxiter are the current iteration and maximum number of evaluation, respectively. �� stands for the step size of the neighborhood search. Meanwhile, this strategy can improve the convergence rate of MVO compared with other algorithms.

Algorithm 1 shows the technical details of the proposed optimization method (HCMVO). The initial solution includes freeboard (�), bottom elevation (�), number of ribs (Nr), rib thickness (�), and rib height(�).

5. Conclusion

The high trend of diminishing worldwide energy resources has entailed a great crisis upon vulnerable societies. To withstand this effect, developing renewable energy technologies can open doors to a more reliable means, among which the wave energy converters will help the coastal residents and infrastructure. This paper set out to determine the optimized design for such devices that leads to the highest possible power output. The main goal of this research was to demonstrate the best design for an oscillating surge wave energy converter using a novel metaheuristic optimization algorithm. In this regard, the methodology was devised such that it argued the effects of influential parameters, including wave characteristics, WEC design, and interaction criteria.

To begin with, a numerical model was developed in Flow 3D software to simulate the response of the flap of a wave energy converter to incoming waves, followed by a validation study based upon a well-reputed experimental study to verify the accuracy of the model. Secondly, the hydrodynamics of the flap was investigated by incorporating the turbulence. The effect of depth, wave height, and wave period are also investigated in this part. The influence of two novel ideas on increasing the wave-converter interaction was then assessed: i) designing a flap with different widths in the upper and lower part, and ii) adding ribs on the surface of the flap. Finally, four trending single-objective metaheuristic optimization methods

Empty CellAlgorithm 1: Hill Climb Multiverse Optimization
01:procedure HCMVO
04:Initialize parameters�ER,�DR,�EP,Best�,���ite��▹Wormhole existence probability (WEP)
06:��=Normalize the inflation rate��
07:for iter in[1,⋯,���iter]do
13:White HoleIndex=Roulette�heelSelection(-��)
14:�(Black HoleIndex,�)=��(White HoleIndex,�)
15:end if
24:end if
25:end if
26:end for
27:end for
31:ifΔBestTHD<��then▹Perform hill climbing local search
33:end if
34:end for
35:return�,BestTHD▹Final configuration
36:end procedure

The implementation details of the hill-climbing algorithm applied in HCMPA can be seen in Algorithm 2. One of the critical parameters isg, which denotes the resolution of the neighborhood search around the proposed global best by MVO. If we set a small step size for hill-climbing, the convergence speed will be decreased. On the other hand, a large step size reinforces the exploration ability. Still, it may reduce the exploitation ability and in return increase the act of jumping from a global optimum or surfaces with high-potential solutions. Per each decision variable, the neighborhood search evaluates two different direct searches, incremental or decremental. After assessing the generated solutions, the best candidate will be selected to iterate the search algorithm. It is noted that the hill-climbing algorithm should not be applied in the initial iteration of the optimization process due to the immense tendency for converging to local optima. Meanwhile, for optimizing largescale problems, hill-climbing is not an appropriate selection. In order to improve understanding of the proposed hybrid optimization algorithm’s steps, the flowchart of HCMVO is designed and can be seen in Figure 16.

Figure 17 shows the observed capture factor (which is the absorbed energy with respect to the available energy) by each optimization algorithm from iterations 1 to 400. The algorithms use ten search agents in their modified codes to find the optimal solutions. While GWO and MFO remain roughly constant after iterations 54 and 40, the other three algorithms keep improving the capture factor. In this case, HCMVO and MVO worked very well in the optimizing process with a capture factor obtained by the former as 0.594 and by the latter as 0.593. MFO almost found its highest value before the iteration 50, which means the exploration part of the algorithm works out well. Similarly, HCMVO does the same. However, it keeps finding the better solution during the optimization process until the last iteration, indicating the strong exploitation part of the algorithm. GWO reveals a weakness in exploration and exploitation because not only does it evoke the least capture factor value, but also the curve remains almost unchanged throughout 350 iterations.

Figure 18 illustrates complex interactions between the five optimization parameters and the capture factor for HCMVO (a), MPA (b), and MFO (c) algorithms. The first interesting observation is that there is a high level of nonlinear relationships among the setting parameters that can make a multi-modal search space. The dark blue lines represent the best-found configuration throughout the optimisation process. Based on both HCMVO (a) and MVO (b), we can infer that the dark blue lines concentrate in a specific range, showing the high convergence ability of both HCMVO and MVO. However, MFO (c) could not find the exact optimal range of the decision variables, and the best-found solutions per generation distribute mostly all around the search space.

Empty CellAlgorithm 1: Hill Climb Multiverse Optimization
01:procedure HCMVO
03:Initialize the constraints��1�,��1�
04:�1�=Mi�1�+���1�/�▹Compute the step size,�is search resolution
07:Main loop
08:for iter≤���ita=do
13:t = t+1
14:end while
18:end for
20:end procedure

were utilized to illuminate the optimum values of the design parameters, and the best method was chosen to develop a new algorithm that performs both local and global search methods.

The correlation between hydrodynamic parameters and the capture factor of the converter was supported by the results. For any given water depth, the capture factor increases as the wave period increases, until a certain wave period value (6 seconds) is reached, after which the capture factor gradually decreases. It is expected since the flap cannot oscillate effectively when the wavelength is too short for a certain water depth. Conversely, when the wavelength is too long, the capture factor decreases. Furthermore, under a constant wave period, increasing the water depth does not affect the capture factor. Regarding the sensitivity analysis, the study found that increasing the flap bottom elevation causes turbulence flow behind the flap and limitation of rotation, which leads to less interaction with the incoming waves. Furthermore, while keeping the flap bottom elevation constant, increasing the freeboard improves the capture factor. Overtopping happens when the freeboard is negative and the flap is below the water surface, which has a detrimental influence on converter performance. Furthermore, raising the freeboard causes the wave impact to become more violent, which increases converter performance.

In the last part, we discussed the search process of each algorithm and visualized their performance and convergence curves as they try to find the best values for decision variables. Among the four selected metaheuristic algorithms, the Multi-verse Optimizer proved to be the most effective in achieving the best answer in terms of the WEC capture factor. However, the MVO needed modifications regarding its escape approach from the local optima in order to improve its global search capabilities. To overcome these constraints, we presented a fast and efficient meta-algorithm (HCMVO) that combines MVO with a Random-restart hill-climbing local search. On a higher level, this meta-algorithm employed MVO to generate global tracking and present a range of possible and appropriate solutions. Taken together, the results demonstrated that there is a significant degree of nonlinearity among the setup parameters that might result in a multimodal search space. Since MVO was faced with a stagnation issue or fell into a local optimum, we constructed a complete neighborhood search around the best-found solution offered by the upper level. In sum, the newly-developed algorithm proved to be highly effective for the problem compared to other similar optimization methods. The strength of the current findings may encourage future investigation on design optimization of wave energy converters using developed geometry as well as the novel approach.

CRediT authorship contribution statement

Erfan Amini: Conceptualization, Methodology, Validation, Data curation, Writing – original draft, Writing – review & editing, Visualization. Mahdieh Nasiri: Conceptualization, Methodology, Validation, Data curation, Writing – original draft, Writing – review & editing, Visualization. Navid Salami Pargoo: Writing – original draft, Writing – review & editing. Zahra Mozhgani: Conceptualization, Methodology. Danial Golbaz: Writing – original draft. Mehrdad Baniesmaeil: Writing – original draft. Meysam Majidi Nezhad: . Mehdi Neshat: Supervision, Conceptualization, Writing – original draft, Writing – review & editing, Visualization. Davide Astiaso Garcia: Supervision. Georgios Sylaios: Supervision.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.


This research has been carried out within ILIAD (Inte-grated Digital Framework for Comprehensive Maritime Data and Information Services) project that received funding from the European Union’s H2020 programme.

Data availability

Data will be made available on request.


Figure 1: Mold drawings

3D Flow and Temperature Analysis of Filling a Plutonium Mold

플루토늄 주형 충전의 3D 유동 및 온도 분석

Authors: Orenstein, Nicholas P. [1]

Publication Date:2013-07-24
Research Org.: Los Alamos National Lab. (LANL), Los Alamos, NM (United States)
Sponsoring Org.: DOE/LANL
OSTI Identifier: 1088904
Report Number(s): LA-UR-13-25537
DOE Contract Number: AC52-06NA25396
Resource Type: Technical Report
Country of Publication: United States
Language: English
Subject: Engineering(42); Materials Science(36); Radiation Chemistry, Radiochemistry, & Nuclear Chemistry(38)


The plutonium foundry at Los Alamos National Laboratory casts products for various special nuclear applications. However, plutonium’s radioactivity, material properties, and security constraints complicate the ability to perform experimental analysis of mold behavior. The Manufacturing Engineering and Technologies (MET-2) group previously developed a graphite mold to vacuum cast small plutonium disks to be used by the Department of Homeland Security as point sources for radiation sensor testing.

A two-stage pouring basin consisting of a funnel and an angled cavity directs the liquid into a vertical runner. A stack of ten disk castings connect to the runner by horizontal gates. Volumetric flow rates were implemented to limit overflow into the funnel and minimize foundry returns. Models using Flow-3D computational fluid dynamics software are employed here to determine liquid Pu flow paths, optimal pour regimes, temperature changes, and pressure variations.


Hardcopy drawings provided necessary information to create 3D .stl models for import into Flow-3D (Figs. 1 and 2). The mesh was refined over several iterations to isolate the disk cavities, runner, angled cavity, funnel, and input pour. The final flow and mold-filling simulation utilizes a fine mesh with ~5.5 million total cells. For the temperature study, the mesh contained 1/8 as many cells to reduce computational time and set temperatures to 850 °C for the molten plutonium and 500 °C for the solid graphite mold components (Fig. 3).

Flow-3D solves mass continuity and Navier-Stokes momentum equations over the structured rectangular grid model using finite difference and finite volume numerical algorithms. The solver includes terms in the momentum equation for body and viscous accelerations and uses convective heat transfer.

Simulation settings enabled Flow-3D physics calculations for gravity at 980.665 cm/s 2 in the negative Z direction (top of mold to bottom); viscous, turbulent, incompressible flow using dynamically-computed Renormalized Group Model turbulence calculations and no-slip/partial slip wall shear, and; first order, full energy equation heat transfer.

Mesh boundaries were all set to symmetric boundary conditions except for the Zmin boundary set to outflow and the Zmax boundary set to a volume flow. Vacuum casting conditions and the high reactivity of remaining air molecules with Pu validate the assumption of an initially fluidless void.


The flow follows a unique three-dimensional path. The mold fills upwards with two to three disks receiving fluid in a staggered sequence. Figures 5-9 show how the fluid fills the cavity, and Figure 7 includes the color scale for pressure levels in these four figures. The narrow gate causes a high pressure region which forces the fluid to flow down the cavity centerline.

It proceeds to splash against the far wall and then wrap around the circumference back to the gate (Figs. 5 and 6). Flow in the angled region of the pouring basin cascades over the bottom ledge and attaches to the far wall of the runner, as seen in Figure 7.

This channeling becomes less pronounced as fluid volume levels increase. Finally, two similar but non-uniform depressed regions form about the centerline. These regions fill from their perimeter and bottom until completion (Fig. 8). Such a pattern is counter, for example, to a steady scenario in which a circle of molten Pu encompassing the entire bottom surface rises as a growing cylinder.

Cavity pressure becomes uniform when the cavity is full. Pressure levels build in the rising well section of the runner, where impurities were found to settle in actual casting. Early test simulations optimized the flow as three pours so that the fluid would never overflow to the funnel, the cavities would all fill completely, and small amounts of fluid would remain as foundry returns in the angled cavity.

These rates and durations were translated to the single 2.7s pour at 100 cm 3 per second used here. Figure 9 shows anomalous pressure fluctuations which occurred as the cavities became completely filled. Multiple simulations exhibited a rapid change in pressure from positive to negative and back within the newly-full disk and surrounding, already-full disks.

The time required to completely fill each cavity is plotted in Figure 10. Results show negligible temperature change within the molten Pu during mold filling and, as seen in Figure 11, at fill completion.

Figure 1: Mold drawings
Figure 1: Mold drawings
Figure 2: Mold Assembly
Figure 2: Mold Assembly
Figure 4: Actual mold and cast Pu
Figure 4: Actual mold and cast Pu
Figure 5: Bottom cavity filling
from runner
Figure 5: Bottom cavity filling from runner
Figure 6: Pouring and filling
Figure 6: Pouring and filling
Figure 8: Edge detection of cavity fill geometry. Two similar depressed areas form
about the centerline. Top cavity shown; same pressure scale as other figures
Figure 8: Edge detection of cavity fill geometry. Two similar depressed areas form about the centerline. Top cavity shown; same pressure scale as other figures
Figure 10: Cavity fill times,from first fluid contact with pouring basin, Figure 11:Fluid temperature remains essentially constant
Figure 10: Cavity fill times,from first fluid contact with pouring basin, Figure 11:Fluid temperature remains essentially constant


Non-uniform cavity filling could cause crystal microstructure irregularities during solidification. However, the small temperature changes seen – due to large differences in specific heat between Pu and graphite – over a relatively short time make such problems unlikely in this case.

In the actual casting, cooling required approximately ten minutes. This large difference in time scales further reduces the chance for temperature effects in such a superheated scenario. Pouring basin emptying decreases pressure at the gate which extends fill time of the top two cavities.

The bottom cavity takes longer to fill because fluid must first enter the runner and fill the well. Fill times continue linearly until the top two cavities. The anomalous pressure fluctuations may be due to physical attempts by the system to reach equilibrium, but they are more likely due to numerical errors in the Flow3D solver.

Unsuccessful tests were performed to remove them by halving fluid viscosity. The fine mesh reduced, but did not eliminate, the extent of the fluctuations. Future work is planned to study induction and heat transfer in the full Pu furnace system, including quantifying temporal lag of the cavity void temperature to the mold wall temperature during pre-heat and comparing heat flux levels between furnace components during cool-down.

Thanks to Doug Kautz for the opportunity to work with MET-2 and for assigning an interesting unclassified project. Additional thanks to Mike Bange for CFD guidance, insight of the project’s history, and draft review.

Sketch of approach channel and spillway of the Kamal-Saleh dam

CFD modeling of flow pattern in spillway’s approach channel

Sustainable Water Resources Management volume 1, pages245–251 (2015)Cite this article


Analysis of behavior and hydraulic characteristics of flow over the dam spillway is a complicated task that takes lots of money and time in water engineering projects planning. To model those hydraulic characteristics, several methods such as physical and numerical methods can be used. Nowadays, by utilizing new methods in computational fluid dynamics (CFD) and by the development of fast computers, the numerical methods have become accessible for use in the analysis of such sophisticated flows. The CFD softwares have the capability to analyze two- and three-dimensional flow fields. In this paper, the flow pattern at the guide wall of the Kamal-Saleh dam was modeled by Flow 3D. The results show that the current geometry of the left wall causes instability in the flow pattern and making secondary and vortex flow at beginning approach channel. This shape of guide wall reduced the performance of weir to remove the peak flood discharge.

댐 여수로 흐름의 거동 및 수리학적 특성 분석은 물 공학 프로젝트 계획에 많은 비용과 시간이 소요되는 복잡한 작업입니다. 이러한 수력학적 특성을 모델링하기 위해 물리적, 수치적 방법과 같은 여러 가지 방법을 사용할 수 있습니다. 요즘에는 전산유체역학(CFD)의 새로운 방법을 활용하고 빠른 컴퓨터의 개발로 이러한 정교한 흐름의 해석에 수치 방법을 사용할 수 있게 되었습니다. CFD 소프트웨어에는 2차원 및 3차원 유동장을 분석하는 기능이 있습니다. 본 논문에서는 Kamal-Saleh 댐 유도벽의 흐름 패턴을 Flow 3D로 모델링하였다. 결과는 왼쪽 벽의 현재 형상이 흐름 패턴의 불안정성을 유발하고 시작 접근 채널에서 2차 및 와류 흐름을 만드는 것을 보여줍니다. 이러한 형태의 안내벽은 첨두방류량을 제거하기 위해 둑의 성능을 저하시켰다.


Spillways are one of the main structures used in the dam projects. Design of the spillway in all types of dams, specifically earthen dams is important because the inability of the spillway to remove probable maximum flood (PMF) discharge may cause overflow of water which ultimately leads to destruction of the dam (Das and Saikia et al. 2009; E 2013 and Novak et al. 2007). So study on the hydraulic characteristics of this structure is important. Hydraulic properties of spillway including flow pattern at the entrance of the guide walls and along the chute. Moreover, estimating the values of velocity and pressure parameters of flow along the chute is very important (Chanson 2004; Chatila and Tabbara 2004). The purpose of the study on the flow pattern is the effect of wall geometry on the creation transverse waves, flow instability, rotating and reciprocating flow through the inlet of spillway and its chute (Parsaie and Haghiabi 2015ab; Parsaie et al. 2015; Wang and Jiang 2010). The purpose of study on the values of velocity and pressure is to calculate the potential of the structure to occurrence of phenomena such as cavitation (Fattor and Bacchiega 2009; Ma et al. 2010). Sometimes, it can be seen that the spillway design parameters of pressure and velocity are very suitable, but geometry is considered not suitable for conducting walls causing unstable flow pattern over the spillway, rotating flows at the beginning of the spillway and its design reduced the flood discharge capacity (Fattor and Bacchiega 2009). Study on spillway is usually conducted using physical models (Su et al. 2009; Suprapto 2013; Wang and Chen 2009; Wang and Jiang 2010). But recently, with advances in the field of computational fluid dynamics (CFD), study on hydraulic characteristics of this structure has been done with these techniques (Chatila and Tabbara 2004; Zhenwei et al. 2012). Using the CFD as a powerful technique for modeling the hydraulic structures can reduce the time and cost of experiments (Tabbara et al. 2005). In CFD field, the Navier–Stokes equation is solved by powerful numerical methods such as finite element method and finite volumes (Kim and Park 2005; Zhenwei et al. 2012). In order to obtain closed-form Navier–Stokes equations turbulence models, such k − ε and Re-Normalisation Group (RNG) models have been presented. To use the technique of computational fluid dynamics, software packages such as Fluent and Flow 3D, etc., are provided. Recently, these two software packages have been widely used in hydraulic engineering because the performance and their accuracy are very suitable (Gessler 2005; Kim 2007; Kim et al. 2012; Milési and Causse 2014; Montagna et al. 2011). In this paper, to assess the flow pattern at Kamal-Saleh guide wall, numerical method has been used. All the stages of numerical modeling were conducted in the Flow 3D software.

Materials and methods

Firstly, a three-dimensional model was constructed according to two-dimensional map that was prepared for designing the spillway. Then a small model was prepared with scale of 1:80 and entered into the Flow 3D software; all stages of the model construction was conducted in AutoCAD 3D. Flow 3D software numerically solved the Navier–Stokes equation by finite volume method. Below is a brief reference on the equations that used in the software. Figure 1 shows the 3D sketch of Kamal-Saleh spillway and Fig. 2 shows the uploading file of the Kamal-Saleh spillway in Flow 3D software.

figure 1
Fig. 1
figure 2
Fig. 2

Review of the governing equations in software Flow 3D

Continuity equation at three-dimensional Cartesian coordinates is given as Eq (1).



where uvz are velocity component in the x, y, z direction; A xA yA z cross-sectional area of the flow; ρ fluid density; PSOR the source term; v f is the volume fraction of the fluid and three-dimensional momentum equations given in Eq (2).



where P is the fluid pressure; G xG yG z the acceleration created by body fluids; f xf yf z viscosity acceleration in three dimensions and v f is related to the volume of fluid, defined by Eq. (3). For modeling of free surface profile the VOF technique based on the volume fraction of the computational cells has been used. Since the volume fraction F represents the amount of fluid in each cell, it takes value between 0 and 1.



Turbulence models

Flow 3D offers five types of turbulence models: Prantl mixing length, k − ε equation, RNG models, Large eddy simulation model. Turbulence models that have been proposed recently are based on Reynolds-averaged Navier–Stokes equations. This approach involves statistical methods to extract an averaged equation related to the turbulence quantities.

Steps of solving a problem in Flow 3D software

(1) Preparing the 3D model of spillway by AutoCAD software. (2) Uploading the file of 3D model in Flow 3D software and defining the problem in the software and checking the final mesh. (3) Choosing the basic equations that should be solved. (4) Defining the characteristics of fluid. (5) Defining the boundary conditions; it is notable that this software has a wide range of boundary conditions. (6) Initializing the flow field. (7) Adjusting the output. (8) Adjusting the control parameters, choice of the calculation method and solution formula. (9) Start of calculation. Figure 1 shows the 3D model of the Kamal-Saleh spillway; in this figure, geometry of the left and right guide wall is shown.

Figure 2 shows the uploading of the 3D spillway dam in Flow 3D software. Moreover, in this figure the considered boundary condition in software is shown. At the entrance and end of spillway, the flow rate or fluid elevation and outflow was considered as BC. The bottom of spillway was considered as wall and left and right as symmetry.

Model calibration

Calibration of the Flow 3D for modeling the effect of geometry of guide wall on the flow pattern is included for comparing the results of Flow 3D with measured water surface profile. Calibration the Flow 3D software could be conducted in two ways: first, changing the value of upstream boundary conditions is continued until the results of water surface profile of the Flow 3D along the spillway successfully covered the measurement water surface profile; second is the assessment the mesh sensitivity. Analyzing the size of mesh is a trial-and-error process where the size of mesh is evaluated form the largest to the smallest. With fining the size of mesh the accuracy of model is increased; whereas, the cost of computation is increased. In this research, the value of upstream boundary condition was adjusted with measured data during the experimental studies on the scaled model and the mesh size was equal to 1 × 1 × 1 cm3.

Results and discussion

The behavior of water in spillway is strongly affected by the flow pattern at the entrance of the spillway, the flow pattern formation at the entrance is affected by the guide wall, and choice of an optimized form for the guide wall has a great effect on rising the ability of spillway for easy passing the PMF, so any nonuniformity in flow in the approach channel can cause reduction of spillway capacity, reduction in discharge coefficient of spillway, and even probability of cavitation. Optimizing the flow guiding walls (in terms of length, angle and radius) can cause the loss of turbulence and flow disturbances on spillway. For this purpose, initially geometry proposed for model for the discharge of spillway dam, Kamal-Saleh, 80, 100, and 120 (L/s) were surveyed. These discharges of flow were considered with regard to the flood return period, 5, 100 and 1000 years. Geometric properties of the conducting guidance wall are given in Table 1.Table 1 Characteristics and dimensions of the guidance walls tested

Full size table

Results of the CFD simulation for passing the flow rate 80 (L/s) are shown in Fig. 3. Figure 3 shows the secondary flow and vortex at the left guide wall.

figure 3
Fig. 3

For giving more information about flow pattern at the left and right guide wall, Fig. 4 shows the flow pattern at the right side guide wall and Fig. 5 shows the flow pattern at the left side guide wall.

figure 4
Fig. 4
figure 5
Fig. 5

With regard to Figs. 4 and 5 and observing the streamlines, at discharge equal to 80 (L/s), the right wall has suitable performance but the left wall has no suitable performance and the left wall of the geometric design creates a secondary and circular flow, and vortex motion in the beginning of the entrance of spillway that creates cross waves at the beginning of spillway. By increasing the flow rate (Q = 100 L/s), at the inlet spillway secondary flows and vortex were removed, but the streamline is severely distorted. Results of the guide wall performances at the Q = 100 (L/s) are shown in Fig. 6.

figure 6
Fig. 6

Also more information about the performance of each guide wall can be derived from Figs. 7 and 8. These figures uphold that the secondary and vortex flows were removed, but the streamlines were fully diverted specifically near the left side guide wall.

figure 7
Fig. 7
figure 8
Fig. 8

As mentioned in the past, these secondary and vortex flows and diversion in streamline cause nonuniformity and create cross wave through the spillway. Figure 9 shows the cross waves at the crest of the spillway.

figure 9
Fig. 9

The performance of guide walls at the Q = 120 (L/s) also was assessed. The result of simulation is shown in Fig. 10. Figures 11 and 12 show a more clear view of the streamlines near to right and left side guide wall, respectively. As seen in Fig. 12, the left side wall still causes vortex flow and creation of and diversion in streamline.

figure 10
Fig. 10
figure 11
Fig. 11
figure 12
Fig. 12

The results of the affected left side guide wall shape on the cross wave creation are shown in Fig. 13. As seen from Fig. 3, the left side guide wall also causes cross wave at the spillway crest.

figure 13
Fig. 13

As can be seen clearly in Figs. 9 and 13, by moving from the left side to the right side of the spillway, the cross waves and the nonuniformity in flow is removed. By reviewing Figs. 9 and 13, it is found that the right side guide wall removes the cross waves and nonuniformity. With this point as aim, a geometry similar to the right side guide wall was considered instead of the left side guide wall. The result of simulation for Q = 120 (L/s) is shown in Fig. 14. As seen from this figure, the proposed geometry for the left side wall has suitable performance smoothly passing the flow through the approach channel and spillway.

figure 14
Fig. 14

More information about the proposed shape for the left guide wall is shown in Fig. 15. As seen from this figure, this shape has suitable performance for removing the cross waves and vortex flows.

figure 15
Fig. 15

Figure 16 shows the cross section of flow at the crest of spillway. As seen in this figure, the proposed shape for the left side guide wall is suitable for removing the cross waves and secondary flows.

figure 16
Fig. 16


Analysis of behavior and hydraulic properties of flow over the spillway dam is a complicated task which is cost and time intensive. Several techniques suitable to the purposes of study have been undertaken in this research. Physical modeling, usage of expert experience, usage of mathematical models on simulation flow in one-dimensional, two-dimensional and three-dimensional techniques, are some of the techniques utilized to study this phenomenon. The results of the modeling show that the CFD technique is a suitable tool for simulating the flow pattern in the guide wall. Using this tools helps the designer for developing the optimal shape for hydraulic structure which the flow pattern through them are important.


  • Chanson H (2004) 19—Design of weirs and spillways. In: Chanson H (ed) Hydraulics of open channel flow, 2nd edn. Butterworth-Heinemann, Oxford, pp 391–430Chapter Google Scholar 
  • Chatila J, Tabbara M (2004) Computational modeling of flow over an ogee spillway. Comput Struct 82:1805–1812Article Google Scholar 
  • Das MM, Saikia MD (2009) Irrigation and water power engineering. PHI Learning, New DelhiGoogle Scholar 
  • E, Department Of Army: U.S. Army Corps (2013) Hydraulic Design of Spillways. BiblioBazaar, CharlestonGoogle Scholar 
  • Fattor C, Bacchiega J (2009) Design conditions for morning-glory spillways: application to potrerillos dam spillway. Adv Water Res Hydraul Eng Springer, Berlin, pp 2123–2128Google Scholar 
  • Gessler D (2005) CFD modeling of spillway performance. Impacts Glob Clim Change. doi:10.1061/40792(173)398
  • Kim D-G (2007) Numerical analysis of free flow past a sluice gate. KSCE J Civ Eng 11:127–132Article Google Scholar 
  • Kim D, Park J (2005) Analysis of flow structure over ogee-spillway in consideration of scale and roughness effects by using CFD model. KSCE J Civ Eng 9:161–169Article Google Scholar 
  • Kim S, Yu K, Yoon B, Lim Y (2012) A numerical study on hydraulic characteristics in the ice Harbor-type fishway. KSCE J Civ Eng 16:265–272Article Google Scholar 
  • Ma X-D, Dai G-Q, Yang Q, Li G-J, Zhao L (2010) Analysis of influence factors of cavity length in the spillway tunnel downstream of middle gate chamber outlet with sudden lateral enlargement and vertical drop aerator. J Hydrodyn Ser B 22:680–686Article Google Scholar 
  • Milési G, Causse S (2014) 3D numerical modeling of a side-channel spillway. In: Gourbesville P, Cunge J, Caignaert G (eds) Advances in hydroinformatics. Springer, Singapore, pp 487–498Chapter Google Scholar 
  • Montagna F, Bellotti G, Di Risio M (2011) 3D numerical modeling of landslide-generated tsunamis around a conical island. Nat Hazards 58:591–608Article Google Scholar 
  • Novak P, Moffat AIB, Nalluri C, Narayanan R (2007) Hydraulic structures. Taylor & Francis, LondonGoogle Scholar 
  • Parsaie A, Haghiabi A (2015a) Computational modeling of pollution transmission in rivers. Appl Water Sci. doi:10.1007/s13201-015-0319-6
  • Parsaie A, Haghiabi A (2015b) The effect of predicting discharge coefficient by neural network on increasing the numerical modeling accuracy of flow over side weir. Water Res Manag 29:973–985Article Google Scholar 
  • Parsaie A, Yonesi H, Najafian S (2015) Predictive modeling of discharge in compound open channel by support vector machine technique. Model Earth Syst Environ 1:1–6Article Google Scholar 
  • Su P-L, Liao H-S, Qiu Y, Li CJ (2009) Experimental study on a new type of aerator in spillway with low Froude number and mild slope flow. J Hydrodyn Ser B 21:415–422Article Google Scholar 
  • Suprapto M (2013) Increase spillway capacity using Labyrinth Weir. Procedia Eng 54:440–446Article Google Scholar 
  • Tabbara M, Chatila J, Awwad R (2005) Computational simulation of flow over stepped spillways. Comput Struct 83:2215–2224Article Google Scholar 
  • Wang J, Chen H (2009) Experimental study of elimination of vortices along guide wall of bank spillway. Adv Water Res Hydraul Eng Springer, Berlin, pp 2059–2063Google Scholar 
  • Wang Y, Jiang C (2010) Investigation of the surface vortex in a spillway tunnel intake. Tsinghua Sci Technol 15:561–565Article Google Scholar 
  • Zhenwei MU, Zhiyan Z, Tao Z (2012) Numerical simulation of 3-D flow field of spillway based on VOF method. Procedia Eng 28:808–812Article Google Scholar 

Download references

Author information

Authors and Affiliations

  1. Department of Water Engineering, Lorestan University, Khorram Abad, IranAbbas Parsaie, Amir Hamzeh Haghiabi & Amir Moradinejad

Corresponding author

Correspondence to Abbas Parsaie.

Reprints and Permissions

About this article

Cite this article

Parsaie, A., Haghiabi, A.H. & Moradinejad, A. CFD modeling of flow pattern in spillway’s approach channel. Sustain. Water Resour. Manag. 1, 245–251 (2015).

Download citation

  • Received28 April 2015
  • Accepted28 August 2015
  • Published15 September 2015
  • Issue DateSeptember 2015
  • DOI

Share this article

Anyone you share the following link with will be able to read this content:Get shareable link

Provided by the Springer Nature SharedIt content-sharing initiative


  • Approach channel
  • Kamal-Saleh dam
  • Guide wall
  • Flow pattern
  • Numerical modeling
  • Flow 3D software
    Effect of tailwater depth on non-cohesive earth dam failure due to overtopping

    Effect of tailwater depth on non-cohesive earth dam failure due to overtopping

    범람으로 인한 비점착성 흙댐 붕괴에 대한 테일워터 깊이의 영향

    ShaimaaAmanaMohamedAbdelrazek RezkbRabieaNasrc


    본 연구에서는 범람으로 인한 토사댐 붕괴에 대한 테일워터 깊이의 영향을 실험적으로 조사하였다. 테일워터 깊이의 네 가지 다른 값을 검사합니다. 각 실험에 대해 댐 수심 측량 프로파일의 진화, 고장 기간, 침식 체적 및 유출 수위곡선을 관찰하고 기록합니다.

    결과는 tailwater 깊이를 늘리면 고장 시간이 최대 57% 감소하고 상대적으로 침식된 마루 높이가 최대 77.6% 감소한다는 것을 보여줍니다. 또한 상대 배수 깊이가 3, 4, 5인 경우 누적 침식 체적의 감소는 각각 23, 36.5 및 75%인 반면 최대 유출량의 감소는 각각 7, 14 및 17.35%입니다.

    실험 결과는 침식 과정을 복제할 때 Flow 3D 소프트웨어의 성능을 평가하는 데 활용됩니다. 수치 모델은 비응집성 흙댐의 침식 과정을 성공적으로 시뮬레이션합니다.

    The influence of tailwater depth on earth dam failure due to overtopping is investigated experimentally in this work. Four different values of tailwater depths are examined. For each experiment, the evolution of the dam bathymetry profile, the duration of failure, the eroded volume, and the outflow hydrograph are observed and recorded. The results reveal that increasing the tailwater depth reduces the time of failure by up to 57% and decreases the relative eroded crest height by up to 77.6%. In addition, for relative tailwater depths equal to 3, 4, and 5, the reduction in the cumulative eroded volume is 23, 36.5, and 75%, while the reduction in peak discharge is 7, 14, and 17.35%, respectively. The experimental results are utilized to evaluate the performance of the Flow 3D software in replicating the erosion process. The numerical model successfully simulates the erosion process of non-cohesive earth dams.


    Earth dam, Eroded volume, Flow 3D model, Non-cohesive soil, Overtopping failure, Tailwater depth



    Mean partical diameterWc

    Optimum water contentZo

    Dam height (cm)do

    Tailwater depth (cm)Zeroded

    Eroded height of the dam measured at distance of 0.7 m from the dam heel (cm)t

    Total time of failure (sec)t1

    Time of crest width erosion (sec)Zcrest

    The crest height (cm)Vtotal

    Total volume of the dam (m3)Veroded

    Cumulative eroded volume (m3)RMSE

    The statistical variable root- mean- square errord

    Degree of agreement indexyu.s.

    The upstream water depth (cm)yd.s

    The downstream water depth (cm)H

    Water surface elevation over sharp crested weir (cm)Q

    Outflow discharge (liter/sec)Qpeak

    Peak discharge (liter/sec)

    1. Introduction

    Earth dams are compacted structures composed of natural materials that are usually mined or quarried from local locations. The failures of the earth dams have proven to be deadly, destructive, and costly. According to People’s Daily, two earthen dams, Yong’an Dam and Xinfa Dam located in Hulun Buir City in North China’s Inner Mongolia failed on 2021, due to a surge in the water level of the Nuomin River caused by heavy rain. The dam breach affected 16,660 people, flooded 325,622 mu of farmland (21708.1 ha), and destroyed 22 bridges, 124 culverts, and 15.6 km of roadways. Also, the failure of south fork dam (earth and rock fill dam) near Johnstown on 1889 is considered the worst U.S dam disaster in terms of loss of life. The dam was overtopped and washed away due to unexpected heavy rains, releasing 20 million tons of water which destroyed Johnstown and resulted in 2209 deaths, [1][2]. Piping or shear sliding, failure due to natural factors, and failure due to overtopping are all possible causes of earth dam failure. However, overtopping failure is the most frequent cause of dam failure. According to The International Committee on Large Dams (ICOLD, 1995), and [3], more than one-third of the total known dam failures were caused by dam overtopping.

    Overtopping occurs as the result of insufficient flood design or freeboard in some cases. Extreme rainstorms can cause floods which can overtop the dam and cause it to fail. The size and geometry of the reservoir or the dam (side slopes, top width, height, etc.), the homogeneity of the material used in the construction of the dam, overtopping depth, and the presence or absence of tailwater are all elements that influence this type of failure which will be illustrated in the following literature. Overtopping failures of earth dams may be divided into several failure mechanisms based on the material composition and the inner structure of the dam. For cohesive earth dams because of low permeability, no seepage exists on the slopes. Erosion often begins at the earth dam toe during turbulent erosion and moves upstream, undercutting the slope, causing the removal of large chunks of materials. While for non-cohesive earth dams the downstream face of the dam flattens progressively and is often said to rotate around a point near the downstream toe [4][5][6] In the last few decades, the study of failures due to overtopping has gained popularity among researchers. The overtopping failure, in fact, has been widely investigated in coastal and river hydraulics and morpho dynamic. In addition, several laboratory experimental studies have been conducted in this field in order to better understand different involved factors. Also, many numerical types of research have been conducted to investigate the process of overtopping failure as well as the elements that influence this type of failure.

    Tabrizi et al. [5] conducted a series of embankment overtopping tests to find the effect of compaction on the failure of a homogenous sand embankment. A plane breach process occurred across the flume width due to the narrow flume width. They measured the downstream hydrographs and embankment surface profile for every case. They concluded that the peak discharge decreased with a high compaction level, while the time to peak increased. Kansoh et al. [6] studied experimentally the failure of compacted homogeneous non-cohesive earthen embankment due to overtopping. They investigated the influence of different shape parameters including the downstream slope, the crest width, and the height of the embankment on the erosion process. The erosion process was initiated by carving a pilot channel into the embankment crest. They evaluated the time of embankment failure for different shape parameters. They concluded that the failure time increases with increasing the downstream slope and the crest width. Zhu et al. [7] investigated experimentally the breaching of five embankments, one constructed with pure sand, and four with different sand-silt–clay mixtures. The erosion pattern was similar across the flume width. They stated that for cohesive soil mixtures the head cut erosion was the most important factor that affected the breach growth, while for non-cohesive soil the breach erosion was affected by shear erosion.

    Amaral et al. [8] studied experimentally the failure by overtopping for two embankments built from silt sand material. They studied the effect of the degree of compaction of the embankment and the geometry of the pilot channel carved at the centre of the dam crest. They studied two shapes of pilot channel a rectangular shape and triangular shape. They stated that the breach development is influenced by a higher degree of compaction, however, the pilot channel geometry did not influence the breach’s final form. Bereta et al. [9] studied experimentally the breach formation of five dam models, three of them were homogenous clay soil while two were sandy-clay mixtures. The erosion process was initiated by cutting a pilot channel at the centre of the dam crest. They observed the initiation of erosion, flow shear erosion, sidewall bottom erosion, and distinguished the soil mechanical slope mass failure from the head cut vertically and laterally during these tests. Verma et al. [10] investigated experimentally a two-dimensional erosion phenomenon due to overtopping by using a wooden fuse plug model and five different soils. They concluded that the erosion process was affected mostly by cohesiveness and degree of compaction. For cohesive soils, a head cut erosion was observed, while for non-cohesive soils surface erosion occurred gradually. Also, the dimensions of fuse plug, type of fill material, reservoir capacity, and inflow were found to affect the behaviour of the overall breaching process.

    Wu and Qin [11] studied the effect of adding coarse grains to the downstream face of a non-cohesive dam as a result of tailings deposition. The process of overtopping during tailings dam failures is analyzed and its effect on delaying the dam-break process and disaster mitigation are investigated. They found that the tested protective measures decreased the breach area, the maximum breaching flow discharge and flow velocity, and the downstream inundated area. Khankandi et al. [12] studied experimentally the effect of reservoir geometry on dam break flow in case of dry and wet bed conditions. They considered four different reservoir shapes, a long reservoir, a wide, a trapezoidal shaped and one with a 90◦ bend all with identical water volume and horizontal bed. The dam break is simulated by the sudden gate removal using a pneumatic jack. They measured the variation of water level over time with ultrasonic sensors and flow velocity component with an acoustic Doppler velocimeter. Also, the experimental results of water level variation are compared with Ritters solution (1892) [13]. They stated that for dry bed condition the long and 90 bend reservoirs results are close to the analytical solution by ritter also in these two shapes a 1D flow is noticed. However, for wide and trapezoidal reservoirs a 2D effect is significant due to flow contraction at channel entrance.

    Rifai et al. [14] conducted a series of experiments to investigate the effect of tailwater depth on the outflow discharge and breach geometry during non-cohesive homogenous fluvial dikes overtopping failure. They cut an initial notch in the crest at 0.8 m from the upstream end of the dike to initiate overtopping. They compared their results to previous experiments under different main channel inflow discharges combined with a free floodplain. They divided the dike breaching process into three stages: gradual start of overtopping flow resulting in slow initiation of dike erosion, deepening and widening breach due to large flow depth and velocity, finally the flow depth starts stabilizing at its minimal level with or without sustained breach expansion. They stated that breach discharge has lower values than in free floodplain tests. Jiang [15] studied the effect of bed slope on breach parameters and peak discharge in non-cohesive embankment failure. An initial triangular breach with a depth and width of 4 cm was pre-set on one side of the dam. He stated that peak discharge increases with the increase of bed slope and then decreases.

    Ozmen-cagatay et al. [16] studied experimentally flood wave propagation resulted from a sudden dam break event. For dam-break modelling, they used a mechanism that permitted the rapid removal of a vertical plate with a thickness of 4 mm and made of rigid plastic. They conducted three tests, one with dry bed condition and two tests with tailwater depths equal 0.025 m and 0.1 m respectively. They recorded the free surface profile during initial stages of dam break by using digital image processing. Finally, they compared the experimental results with the with a commercially available VOF-based CFD program solving the Reynolds-averaged Navier –Stokes equations (RANS) with the k– Ɛ turbulence model and the shallow water equations (SWEs). They concluded that Wave breaking was delayed with increasing the tailwater depth to initial reservoir depth ratio. They also stated that the SWE approach is sufficient more to represent dam break flows for wet bed condition. Evangelista [17] investigated experimentally and numerically using a depth-integrated two-phase model, the erosion of sand dike caused by the impact of a dam break wave. The dam break is simulated by a sudden opening of an upstream reservoir gate resulting in the overtopping of a downstream trapezoidal sand dike. The evolution of the water wave caused from the gate opening and dike erosion process are recorded by using a computer-controlled camera. The experimental results demonstrated that the progression of the wave front and dike erosion have a considerable influence on each other during the process. In addition, the dike constructed from fine sands was more resistant to erosion than the one built with coarse sand. They also stated that the numerical model can is capable of accurately predicting wave front position and dike erosion. Also, Di Cristo et al. [18] studied the effect of dam break wave propagation on a sand embankment both experimentally and numerically using a two-phase shallow-water model. The evolution of free surface and of the embankment bottom are recorded and used in numerical model assessment. They stated that the model allows reasonable simulation of the experimental trends of the free surface elevation regardeless of the geofailure operator.

    Lots of numerical models have been developed over the past few years to simulate the dam break flooding problem. A one-dimensional model, such as Hec-Ras, DAMBRK and MIKE 11, ect. A two-dimensional model such as iRIC Nay2DH is used in earth embankment breach simulation. Other researchers studied the failure process numerically using (3D) computational fluid dynamics (CFD) models, such as FLOW-3D, and FLUENT. Goharnejad et al. [19] determined the outflow hydrograph which results from the embankment dam break due to overtopping. Hu et al. [20] performed a comparison between Flow-3D and MIKE3 FM numerical models in simulating a dam break event under dry and wet bed conditions with different tailwater depths. Kaurav et al. [21] simulated a planar dam breach process due to overtopping. They conducted a sensitivity analysis to find the effect of dam material, dam height, downstream slope, crest width, and inlet discharge on the erosion process and peak discharge through breach. They concluded that downstream slope has a significant influence on breaching process. Yusof et al. [22] studied the effect of embankment sediment sizes and inflow rates on breaching geometric and hydrodynamic parameters. They stated that the peak outflow hydrograph increases with increasing sediment size and inflow rates while time of failure decreases.

    In the present work, the effect of tailwater depth on earth dam failure during overtopping is studied experimentally. The relation between the eroded volume of the dam and the tailwater depth is presented. Also, the percentage of reduction in peak discharge due to tailwater existence is calculated. An assessment of Flow 3D software performance in simulating the erosion process during earth dam failure is introduced. The statistical variable root- mean- square error, RMSE, and the agreement degree index, d, are used in model assessment.

    2. Material and methods

    The tests are conducted in a straight rectangular flume in the laboratory of Irrigation Engineering and Hydraulics Department, Faculty of Engineering, Alexandria University, Egypt. The flume dimensions are 10 m long, 0.86 m wide, and 0.5 m deep. The front part of the flume is connected to a storage basin 1 m long by 0.86 m wide. The storage basin is connected to a collecting tank for water recirculation during the experiments as shown in Fig. 1Fig. 2. A sharp-crested weir is placed at a distance of 4 m downstream the constructed dam to keep a constant tailwater depth in each experiment and to measure the outflow discharge.

    To measure the eroded volume with time a rods technique is used. This technique consists of two parallel wooden plates with 10 cm distance in between and five rows of stainless-steel rods passing vertically through the wooden plates at a spacing of 20 cm distributed across flume width. Each row consists of four rods with 15 cm spacing between them. Also, a graph board is provided to measure the drop in each rod with time as shown in Fig. 3Fig. 4. After dam construction the rods are carefully rested on the dam, with the first line of rods resting in the middle of the dam crest and then a constant distance of 15 cm between rods lines is maintained.

    A soil sample is taken and tested in the laboratory of the soil mechanics to find the soil geotechnical parameters. The soil particle size distribution is also determined by sieve analysis as shown in Fig. 5. The soil mean diameter d50,equals 0.38 mm and internal friction angle equals 32.6°.

    2.1. Experimental procedures

    To investigate the effect of the tailwater depth (do), the tailwater depth is changed four times 5, 15, 20, and 25 cm on the sand dam model. The dam profile is 35 cm height, with crest width = 15 cm, the dam base width is 155 cm, and the upstream and downstream slopes are 2:1 as shown in Fig. 6. The dam dimensions are set as the flume permitted to allow observation of the dam erosion process under the available flume dimensions and conditions. All of the conducted experiments have the same dimensions and configurations.

    The optimum water content, Wc, from the standard proctor test is found to be 8 % and the maximum dry unit weight is 19.42 kN/m3. The soil and water are mixed thoroughly to ensure consistency and then placed on three horizontal layers. Each layer is compacted according to ASTM standard with 25 blows by using a rammer (27 cm × 20.5 cm) weighing 4 kg. Special attention is paid to the compaction of the soil to guarantee the repeatability of the tests.

    After placing and compacting the three layers, the dam slopes are trimmed carefully to form the trapezoidal shape of the dam. A small triangular pilot channel with 1 cm height and 1:1 side slopes is cut into the dam crest to initiate the erosion process. The position of triangular pilot channel is presented in Fig. 1. Three digital video cameras with a resolution of 1920 × 1080 pixels and a frame rate of 60 fps are placed in three different locations. One camera on one side of the flume to record the progress of the dam profile during erosion. Another to track the water level over the sharp-crested rectangular weir placed at the downstream end of the flume. And the third camera is placed above the flume at the downstream side of the dam and in front of the rods to record the drop of the tip of the rods with time as shown previously in Fig. 1.

    Before starting the experiment, the water is pumped into the storage basin by using pump with capacity 360 m3/hr, and then into the upstream section of the flume. The upstream boundary is an inflow condition. The flow discharge provided to the storage basin is kept at a constant rate of 6 L/sec for all experiments, while the downstream boundary is an outflow boundary condition.

    Also, the required tailwater depth for each experiment is filled to the desired depth. A dye container valve is opened to color the water upstream of the dam to make it easy to distinguish the dam profile from the water profile. A wooden board is placed just upstream of the dam to prevent water from overtopping the dam until the water level rises to a certain level above the dam crest and then the wooden board is removed slowly to start the experiment.

    2.2. Repeatability

    To verify the accuracy of the results, each experiment is repeated two times under the same conditions. Fig. 7 shows the relative eroded crest height, Zeroded / Zo, with time for 5 cm tailwater depth. From the Figure, it can be noticed that results for all runs are consistent, and accuracy is achieved.

    3. Numerical model

    The commercially available numerical model, Flow 3D is used to simulate the dam failure due to overtopping for the cases of 15 cm, 20 cm and 25 cm tailwater depths. For numerical model calibration, experimental results for dam surface evolution are used. The numerical model is calibrated for selection of the optimal turbulence model (RNG, K-e, and k-w) and sediment scour equations (Van Rin, Meyer- peter and Muller, and Nielsen) that produce the best results. In this, the flow field is solved by the RNG turbulence model, and the van Rijn equation is used for the sediment scour model. A geometry file is imported before applying the mesh.

    A Mesh sensitivity is analyzed and checked for various cell sizes, and it is found that decreasing the cell size significantly increases the simulation time with insignificant differences in the result. It is noticed that the most important factor influencing cell size selection is the value of the dam’s upstream and downstream slopes. For example, the slopes in the dam model are 2:1, thus the cell size ratio in X and Z directions should be 2:1 as well. The cell size in a mesh block is set to be 0.02 m, 0.025 m, and 0.01 m in X, Y and Z directions respectively.

    In the numerical computations, the boundary conditions employed are the walls for sidewalls and the channel bottom. The pressure boundary condition is applied at the top, at the air–water interface, to account for atmospheric pressure on the free surface. The upstream boundary is volume flow rate while the downstream boundary is outflow discharge.

    The initial condition is a fluid region, which is used to define fluid areas both upstream and downstream of the dam. To assess the model accuracy, the statistical variable root- mean- square error, RMSE, and the agreement degree index, d, are calculated as(1)RMSE=1N∑i=1N(Pi-Mi)2(2)d=1-∑Mi-Pi2∑Mi-M¯+Pi-P¯2

    where N is the number of samples, Pi and Mi are the models and experimental values, P and M are the means of the model and experimental values. The best fit between the experimental and model results would have an RMSE = 0 and degree of agreement, d = 1.

    4. Results of experimental work

    The results of the total time of failure, t (defined as the time from when the water begins to overtop the dam crest until the erosion reaches a steady state, when no erosion occurs), time of crest width erosion t1, cumulative eroded volume Veroded, and peak discharge Qpeak for each experiment are listed in Table 1. The case of 5 cm tailwater depth is considered as a reference case in this work.

    Table 1. Results of experimental work.

    Tailwater depth, do (cm)Total time of failure, t (sec)Time of crest width erosion, t1 (sec)cumulative eroded volume, Veroded (m3)Peak discharge, Qpeak (liter/sec)

    5. Discussion

    5.1. Side erosion

    The evolution of the bathymetry of the erosion line recorded by the video camera1. The videos are split into frames (60 frames/sec) by the Free Video to JPG Converter v.5.063 build and then converted into an excel spreadsheet using MATLAB code as shown in Fig. 8.

    Fig. 9 shows a sample of numerical model output. Fig. 10Fig. 11Fig. 12 show a dam profile development for different time steps from both experimental and numerical model, for tailwater depths equal 15 cm, 20 cm and 25 cm. Also, the values of RMSE and d for each figure are presented. The comparison shows that the Flow 3D software can simulate the erosion process of non-cohesive earth dam during overtopping with an RMSE value equals 0.023, 0.0218, and 0.0167 and degree of agreement, d, equals 0.95, 0.968, and 0.988 for relative tailwater depths, do/(do)ref, = 3, 4 and 5, respectively. The low values of RMSE and high values of d show that the Flow 3D can effectively simulate the erosion process. From Fig. 10Fig. 11Fig. 12, it can be noticed that the model is not capable of reproducing the head cut, while it can simulate well the degradation of the crest height with a minor difference from experimental work. The reason of this could be due to inability of simulation of all physical conditions which exists in the experimental work, such as channel friction and the grain size distribution of the dam soil which is surely has a great effect on the erosion process and breach development. In the experimental work the grain size distribution is shown in Fig. 5, while the numerical model considers that the soil is uniform and exactly 50 % of the dam particles diameter are equal to the d50 value. Another reason is that the model is not considering the increased resistance of the dam due to the apparent cohesion which happens due to dam saturation [23].

    It is clear from both the experimental and numerical results that for a 5 cm tailwater depth, do/(do)ref = 1.0, erosion begins near the dam toe and continues upward on the downstream slope until it reaches the crest. After eroding the crest width, the crest is lowered, resulting in increased flow rates and the speeding up of the erosion process. While for relative tailwater depths, do/(do)ref = 3, 4, and 5 erosion starts at the point of intersection between the downstream slope and tailwater. The existence of tailwater works as an energy dissipater for the falling water which reduces the erosion process and prevents the dam from failure as shown in Fig. 13. It is found that the time of the failure decreases with increasing the tailwater depth because most of the dam height is being submerged with water which decreases the erosion process. The reduction in time of failure from the referenced case is found to be 35.3, 45, and 57 % for relative tailwater depth, do /(do)ref equals 3, 4, and 5, respectively.

    The relation between the relative eroded crest height, Zeroded /Zo, with time is drawn as shown in Fig. 14. It is found that the relative eroded crest height decreases with increasing tailwater depth by 10, 41, and 77.6 % for relative tailwater depth, do /(do)ref equals 3, 4, and 5, respectively. The time required for the erosion of the crest width, t1, is calculated for each experiment. The relation between relative tailwater depth and relative time of crest width erosion is shown in Fig. 15. It is found that the time of crest width erosion increases linearly with increasing, do /Zo. The percent of increase is 36.4, 54.5 and 77.3 % for relative tailwater depth, do /(do)ref = 3, 4 and 5, respectively.

    Crest height, Zcrest is calculated from the experimental results and the Flow 3D results for relative tailwater depths, do/(do)ref, = 3, 4, and 5. A relation between relative crest height, Zcrest/Zo with time from experimental and numerical results is presented in Fig. 16. From Fig. 16, it is seen that there is a good consistency between the results of numerical model and the experimental results in the case of tracking the erosion of the crest height with time.

    5.2. Upstream and downstream water depths

    It is noticed that at the beginning of the erosion process, both upstream and downstream water depths increase linearly with time as long as erosion of the crest height did not take place. However, when the crest height starts to lower the upstream water depth decreases with time while the downstream water depth increases. At the end of the experiment, the two depths are nearly equal. A relation between relative downstream and upstream water depths with time is drawn for each experiment as shown in Fig. 17.

    5.3. Eroded volume

    A MATLAB code is used to calculate the cumulative eroded volume every time interval for each experiment. The total volume of the dam, Vtotal is 0.256 m3. The cumulative eroded volume, Veroded is 0.21, 0.16, 0.13, and 0.05 m3 for tailwater depths, do = 5, 15, 20, and 25 cm, respectively. Fig. 18 presents the relation between cumulative eroded volume, Veroded and time. From Fig. 18, it is observed that the cumulative eroded volume decreases with increasing the tailwater depth. The reduction in cumulative eroded volume is 23, 36.5, and 75 % for relative tailwater depth, do /(do)ref = 3, 4, and 5, respectively. The relative remained volume of the dam equals 0.18, 0.375, 0.492, and 0.8 for tailwater depths = 5, 15, 20, and 25 cm, respectively. Fig. 19 shows a relation between relative tailwater depth and relative cumulative eroded volume from experimental results. From that figure, it is noticed that the eroded volume decreases exponentially with increasing relative tailwater depth.

    5.4. The outflow discharge

    The inflow discharge provided to the storage tank is maintained constant for all experiments. The water surface elevation, H, over the sharp-crested weir placed at the downstream side is recorded by the video camera 2. For each experiment, the outflow discharge is then calculated by using the sharp-crested rectangular weir equation every 10 sec.

    The outflow discharge is found to increase rapidly until it reaches its peak then it decreases until it is constant. For high values of tailwater depths, the peak discharge becomes less than that in the case of small tailwater depth as shown in Fig. 20 which agrees well with the results of Rifai et al. [14] The reduction in peak discharge is 7, 14, and 17.35 % for relative tailwater depth, do /(do)ref = 3, 4, and 5, respectively.

    The scenario presented in this article in which the tailwater depth rises due to unexpected heavy rainfall, is investigated to find the effect of rising tailwater depth on earth dam failure. The results revealed that rising tailwater depth positively affects the process of dam failure in terms of preventing the dam from complete failure and reducing the outflow discharge.

    6. Conclusions

    The effect of tailwater depth on earth dam failure due to overtopping is investigated experimentally in this work. The study focuses on the effect of tailwater depth on side erosion, upstream and downstream water depths, eroded volume, outflow hydrograph, and duration of the failure process. The Flow 3D numerical software is used to simulate the dam failure, and a comparison is made between the experimental and numerical results to find the ability of this software to simulate the erosion process. The following are the results of the investigation:

    The existence of tailwater with high depths prevents the dam from completely collapsing thereby turning it into a broad crested weir. The failure time decreases with increasing the tailwater depth and the reduction from the reference case is found to be 35.3, 45, and 57 % for relative tailwater depth, do /(do)ref = 3, 4, and 5, respectively. The difference between the upstream and downstream water depths decreases with time till it became almost negligible at the end of the experiment. The reduction in cumulative eroded volume is 23, 36.5, and 75 % for relative tailwater depth, do /(do)ref = 3, 4, and 5, respectively. The peak discharge decreases by 7, 14, and 17.35 % for relative tailwater depth, do /(do)ref = 3, 4, and 5, respectively. The relative eroded crest height decreases linearly with increasing the tailwater depth by 10, 41, and 77.6 % for relative tailwater depth, do /(do)ref = 3, 4, and 5, respectively. The numerical model can reproduce the erosion process with a minor deviation from the experimental results, particularly in terms of tracking the degradation of the crest height with time.

    Declaration of Competing Interest

    The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.



    D. McCullough

    The Johnstown Flood

    Simon and Schuster, NY (1968)

    Google Scholar[2]Rose AT. The influence of dam failures on dam safety laws in Pennsylvania. Association of State Dam Safety Officials Annual Conference 2013, Dam Safety 2013. 2013;1:738–56.

    Google Scholar[3]

    M. Foster, R. Fell, M. Spannagle

    The statistics of embankment dam failures and accidents

    Can Geotech J, 37 (5) (2000), pp. 1000-1024, 10.1139/t00-030 View PDF

    View Record in ScopusGoogle Scholar[4]Pickert, G., Jirka, G., Bieberstein, A., Brauns, J. Soil/water interaction during the breaching process of overtopped embankments. In: Greco, M., Carravetta, A., Morte, R.D. (Eds.), Proceedings of the Conference River-Flow 2004, Balkema.

    Google Scholar[5]

    A. Asghari Tabrizi, E. Elalfy, M. Elkholy, M.H. Chaudhry, J. Imran

    Effects of compaction on embankment breach due to overtopping

    J Hydraul Res, 55 (2) (2017), pp. 236-247, 10.1080/00221686.2016.1238014 View PDF

    View Record in ScopusGoogle Scholar[6]

    R.M. Kansoh, M. Elkholy, G. Abo-Zaid

    Effect of Shape Parameters on Failure of Earthen Embankment due to Overtopping

    KSCE J Civ Eng, 24 (5) (2020), pp. 1476-1485, 10.1007/s12205-020-1107-x View PDF

    View Record in ScopusGoogle Scholar[7]

    YongHui Zhu, P.J. Visser, J.K. Vrijling, GuangQian Wang

    Experimental investigation on breaching of embankments

    Experimental investigation on breaching of embankments, 54 (1) (2011), pp. 148-155 View PDF

    CrossRefView Record in ScopusGoogle Scholar[8]Amaral S, Jónatas R, Bento AM, Palma J, Viseu T, Cardoso R, et al. Failure by overtopping of earth dams. Quantification of the discharge hydrograph. Proceedings of the 3rd IAHR Europe Congress: 14-15 April 2014, Portugal. 2014;(1):182–93.

    Google Scholar[9]

    G. Bereta, P. Hui, H. Kai, L. Guang, P. Kefan, Y.Z. Zhao

    Experimental study of cohesive embankment dam breach formation due to overtopping

    Periodica Polytechnica Civil Engineering, 64 (1) (2020), pp. 198-211, 10.3311/PPci.14565 View PDF

    View Record in ScopusGoogle Scholar[10]

    D.K. Verma, B. Setia, V.K. Arora

    Experimental study of breaching of an earthen dam using a fuse plug model

    Int J Eng Trans A, 30 (4) (2017), pp. 479-485, 10.5829/idosi.ije.2017.30.04a.04 View PDF

    View Record in ScopusGoogle Scholar[11]Wu T, Qin J. Experimental Study of a Tailings Impoundment Dam Failure Due to Overtopping. Mine Water and the Environment [Internet]. 2018;37(2):272–80. Available from: doi: 10.1007/s10230-018-0529-x.

    Google Scholar[12]

    A. Feizi Khankandi, A. Tahershamsi, S. Soares-Frazo

    Experimental investigation of reservoir geometry effect on dam-break flow

    J Hydraul Res, 50 (4) (2012), pp. 376-387 View PDF

    CrossRefView Record in ScopusGoogle Scholar[13]

    A. Ritter

    Die Fortpflanzung der Wasserwellen (The propagation of water waves)

    Zeitschrift Verein Deutscher Ingenieure, 36 (33) (1892), pp. 947-954

    [in German]

    View Record in ScopusGoogle Scholar[14]

    I. Rifai, K. El Kadi Abderrezzak, S. Erpicum, P. Archambeau, D. Violeau, M. Pirotton, et al.

    Floodplain Backwater Effect on Overtopping Induced Fluvial Dike Failure

    Water Resour Res, 54 (11) (2018), pp. 9060-9073 View PDF

    This article is free to access.

    CrossRefView Record in ScopusGoogle Scholar[15]

    X. Jiang

    Laboratory Experiments on Breaching Characteristics of Natural Dams on Sloping Beds

    Advances in Civil Engineering, 2019 (2019), pp. 1-14

    View Record in ScopusGoogle Scholar[16]

    H. Ozmen-Cagatay, S. Kocaman

    Dam-break flows during initial stage using SWE and RANS approaches

    J Hydraul Res, 48 (5) (2010), pp. 603-611 View PDF

    CrossRefView Record in ScopusGoogle Scholar[17]

    S. Evangelista

    Experiments and numerical simulations of dike erosion due to a wave impact

    Water (Switzerland), 7 (10) (2015), pp. 5831-5848 View PDF

    CrossRefView Record in ScopusGoogle Scholar[18]

    C. Di Cristo, S. Evangelista, M. Greco, M. Iervolino, A. Leopardi, A. Vacca

    Dam-break waves over an erodible embankment: experiments and simulations

    J Hydraul Res, 56 (2) (2018), pp. 196-210 View PDF

    CrossRefView Record in ScopusGoogle Scholar[19]Goharnejad H, Sm M, Zn M, Sadeghi L, Abadi K. Numerical Modeling and Evaluation of Embankment Dam Break Phenomenon (Case Study : Taleghan Dam) ISSN : 2319-9873. 2016;5(3):104–11.

    Google Scholar[20]Hu H, Zhang J, Li T. Dam-Break Flows : Comparison between Flow-3D , MIKE 3 FM , and Analytical Solutions with Experimental Data. 2018;1–24. doi: 10.3390/app8122456.

    Google Scholar[21]

    R. Kaurav, P.K. Mohapatra, D. Ph

    Studying the Peak Discharge through a Planar Dam Breach, 145 (6) (2019), pp. 1-8 View PDF


    Z.M. Yusof, Z.A.L. Shirling, A.K.A. Wahab, Z. Ismail, S. Amerudin

    A hydrodynamic model of an embankment breaching due to overtopping flow using FLOW-3D

    IOP Conference Series: Earth and Environmental Science, 920 (1) (2021)

    Google Scholar[23]

    G. Pickert, V. Weitbrecht, A. Bieberstein

    Breaching of overtopped river embankments controlled by apparent cohesion

    J Hydraul Res, 49 (2) (Apr. 2011), pp. 143-156, 10.1080/00221686.2011.552468 View PDF

    View Record in ScopusGoogle Scholar

    Cited by (0)

    My name is Shaimaa Ibrahim Mohamed Aman and I am a teaching assistant in Irrigation and Hydraulics department, Faculty of Engineering, Alexandria University. I graduated from the Faculty of Engineering, Alexandria University in 2013. I had my MSc in Irrigation and Hydraulic Engineering in 2017. My research interests lie in the area of earth dam Failures.

    Peer review under responsibility of Ain Shams University.

    © 2022 THE AUTHORS. Published by Elsevier BV on behalf of Faculty of Engineering, Ain Shams University.

    Extratropical cyclone damage to the seawall in Dawlish, UK: eyewitness accounts, sea level analysis and numerical modelling

    영국 Dawlish의 방파제에 대한 온대 저기압 피해: 목격자 설명, 해수면 분석 및 수치 모델링

    Extratropical cyclone damage to the seawall in Dawlish, UK: eyewitness accounts, sea level analysis and numerical modelling

    Natural Hazards (2022)Cite this article


    2014년 2월 영국 해협(영국)과 특히 Dawlish에 영향을 미친 온대 저기압 폭풍 사슬은 남서부 지역과 영국의 나머지 지역을 연결하는 주요 철도에 심각한 피해를 입혔습니다.

    이 사건으로 라인이 두 달 동안 폐쇄되어 5천만 파운드의 피해와 12억 파운드의 경제적 손실이 발생했습니다. 이 연구에서는 폭풍의 파괴력을 해독하기 위해 목격자 계정을 수집하고 해수면 데이터를 분석하며 수치 모델링을 수행합니다.

    우리의 분석에 따르면 이벤트의 재난 관리는 성공적이고 효율적이었으며 폭풍 전과 도중에 인명과 재산을 구하기 위해 즉각적인 조치를 취했습니다. 파도 부이 분석에 따르면 주기가 4–8, 8–12 및 20–25초인 복잡한 삼중 봉우리 바다 상태가 존재하는 반면, 조위계 기록에 따르면 최대 0.8m의 상당한 파도와 최대 1.5m의 파도 성분이 나타났습니다.

    이벤트에서 가능한 기여 요인으로 결합된 진폭. 최대 286 KN의 상당한 임펄스 파동이 손상의 시작 원인일 가능성이 가장 높았습니다. 수직 벽의 반사는 파동 진폭의 보강 간섭을 일으켜 파고가 증가하고 최대 16.1m3/s/m(벽의 미터 너비당)의 상당한 오버탑핑을 초래했습니다.

    이 정보와 우리의 공학적 판단을 통해 우리는 이 사고 동안 다중 위험 계단식 실패의 가장 가능성 있는 순서는 다음과 같다고 결론을 내립니다. 조적 파괴로 이어지는 파도 충격력, 충전물 손실 및 연속적인 조수에 따른 구조물 파괴.

    The February 2014 extratropical cyclonic storm chain, which impacted the English Channel (UK) and Dawlish in particular, caused significant damage to the main railway connecting the south-west region to the rest of the UK. The incident caused the line to be closed for two months, £50 million of damage and an estimated £1.2bn of economic loss. In this study, we collate eyewitness accounts, analyse sea level data and conduct numerical modelling in order to decipher the destructive forces of the storm. Our analysis reveals that the disaster management of the event was successful and efficient with immediate actions taken to save lives and property before and during the storm. Wave buoy analysis showed that a complex triple peak sea state with periods at 4–8, 8–12 and 20–25 s was present, while tide gauge records indicated that significant surge of up to 0.8 m and wave components of up to 1.5 m amplitude combined as likely contributing factors in the event. Significant impulsive wave force of up to 286 KN was the most likely initiating cause of the damage. Reflections off the vertical wall caused constructive interference of the wave amplitudes that led to increased wave height and significant overtopping of up to 16.1 m3/s/m (per metre width of wall). With this information and our engineering judgement, we conclude that the most probable sequence of multi-hazard cascading failure during this incident was: wave impact force leading to masonry failure, loss of infill and failure of the structure following successive tides.


    The progress of climate change and increasing sea levels has started to have wide ranging effects on critical engineering infrastructure (Shakou et al. 2019). The meteorological effects of increased atmospheric instability linked to warming seas mean we may be experiencing more frequent extreme storm events and more frequent series or chains of events, as well as an increase in the force of these events, a phenomenon called storminess (Mölter et al. 2016; Feser et al. 2014). Features of more extreme weather events in extratropical latitudes (30°–60°, north and south of the equator) include increased gusting winds, more frequent storm squalls, increased prolonged precipitation and rapid changes in atmospheric pressure and more frequent and significant storm surges (Dacre and Pinto 2020). A recent example of these events impacting the UK with simultaneous significant damage to coastal infrastructure was the extratropical cyclonic storm chain of winter 2013/2014 (Masselink et al. 2016; Adams and Heidarzadeh 2021). The cluster of storms had a profound effect on both coastal and inland infrastructure, bringing widespread flooding events and large insurance claims (RMS 2014).

    The extreme storms of February 2014, which had a catastrophic effect on the seawall of the south Devon stretch of the UK’s south-west mainline, caused a two-month closure of the line and significant disruption to the local and regional economy (Fig. 1b) (Network Rail 2014; Dawson et al. 2016; Adams and Heidarzadeh 2021). Restoration costs were £35 m, and economic effects to the south-west region of England were estimated up to £1.2bn (Peninsula Rail Taskforce 2016). Adams and Heidarzadeh (2021) investigated the disparate cascading failure mechanisms which played a part in the failure of the railway through Dawlish and attempted to put these in the context of the historical records of infrastructure damage on the line. Subsequent severe storms in 2016 in the region have continued to cause damage and disruption to the line in the years since 2014 (Met Office 2016). Following the events of 2014, Network Rail Footnote1 who owns the network has undertaken a resilience study. As a result, it has proposed a £400 m refurbishment of the civil engineering assets that support the railway (Fig. 1) (Network Rail 2014). The new seawall structure (Fig. 1a,c), which is constructed of pre-cast concrete sections, encases the existing Brunel seawall (named after the project lead engineer, Isambard Kingdom Brunel) and has been improved with piled reinforced concrete foundations. It is now over 2 m taller to increase the available crest freeboard and incorporates wave return features to minimise wave overtopping. The project aims to increase both the resilience of the assets to extreme weather events as well as maintain or improve amenity value of the coastline for residents and visitors.

    figure 1
    Fig. 1

    In this work, we return to the Brunel seawall and the damage it sustained during the 2014 storms which affected the assets on the evening of the 4th and daytime of the 5th of February and eventually resulted in a prolonged closure of the line. The motivation for this research is to analyse and model the damage made to the seawall and explain the damage mechanisms in order to improve the resilience of many similar coastal structures in the UK and worldwide. The innovation of this work is the multidisciplinary approach that we take comprising a combination of analysis of eyewitness accounts (social science), sea level and wave data analysis (physical science) as well as numerical modelling and engineering judgement (engineering sciences). We investigate the contemporary wave climate and sea levels by interrogating the real-time tide gauge and wave buoys installed along the south-west coast of the English Channel. We then model a typical masonry seawall (Fig. 2), applying the computational fluid dynamics package FLOW3D-Hydro,Footnote2 to quantify the magnitude of impact forces that the seawall would have experienced leading to its failure. We triangulate this information to determine the probable sequence of failures that led to the disaster in 2014.

    figure 2
    Fig. 2

    Data and methods

    Our data comprise eyewitness accounts, sea level records from coastal tide gauges and offshore wave buoys as well as structural details of the seawall. As for methodology, we analyse eyewitness data, process and investigate sea level records through Fourier transform and conduct numerical simulations using the Flow3D-Hydro package (Flow Science 2022). Details of the data and methodology are provided in the following.

    Eyewitness data

    The scale of damage to the seawall and its effects led the local community to document the first-hand accounts of those most closely affected by the storms including residents, local businesses, emergency responders, politicians and engineering contractors involved in the post-storm restoration work. These records now form a permanent exhibition in the local museum in DawlishFootnote3, and some of these accounts have been transcribed into a DVD account of the disaster (Dawlish Museum 2015). We have gathered data from the Dawlish Museum, national and international news reports, social media tweets and videos. Table 1 provides a summary of the eyewitness accounts. Overall, 26 entries have been collected around the time of the incident. Our analysis of the eyewitness data is provided in the third column of Table 1 and is expanded in Sect. 3.Table 1 Eyewitness accounts of damage to the Dawlish railway due to the February 2014 storm and our interpretations

    Full size table

    Sea level data and wave environment

    Our sea level data are a collection of three tide gauge stations (Newlyn, Devonport and Swanage Pier—Fig. 5a) owned and operated by the UK National Tide and Sea Level FacilityFootnote4 for the Environment Agency and four offshore wave buoys (Dawlish, West Bay, Torbay and Chesil Beach—Fig. 6a). The tide gauge sites are all fitted with POL-EKO ( data loggers. Newlyn has a Munro float gauge with one full tide and one mid-tide pneumatic bubbler system. Devonport has a three-channel data pneumatic bubbler system, and Swanage Pier consists of a pneumatic gauge. Each has a sampling interval of 15 min, except for Swanage Pier which has a sampling interval of 10 min. The tide gauges are located within the port areas, whereas the offshore wave buoys are situated approximately 2—3.3 km from the coast at water depths of 10–15 m. The wave buoys are all Datawell Wavemaker Mk III unitsFootnote5 and come with sampling interval of 0.78 s. The buoys have a maximum saturation amplitude of 20.5 m for recording the incident waves which implies that every wave larger than this threshold will be recorded at 20.5 m. The data are provided by the British Oceanographic Data CentreFootnote6 for tide gauges and the Channel Coastal ObservatoryFootnote7 for wave buoys.

    Sea level analysis

    The sea level data underwent quality control to remove outliers and spikes as well as gaps in data (e.g. Heidarzadeh et al. 2022; Heidarzadeh and Satake 2015). We processed the time series of the sea level data using the Matlab signal processing tool (MathWorks 2018). For calculations of the tidal signals, we applied the tidal package TIDALFIT (Grinsted 2008), which is based on fitting tidal harmonics to the observed sea level data. To calculate the surge signals, we applied a 30-min moving average filter to the de-tided data in order to remove all wind, swell and infra-gravity waves from the time series. Based on the surge analysis and the variations of the surge component before the time period of the incident, an error margin of approximately ± 10 cm is identified for our surge analysis. Spectral analysis of the wave buoy data is performed using the fast Fourier transform (FFT) of Matlab package (Mathworks 2018).

    Numerical modelling

    Numerical modelling of wave-structure interaction is conducted using the computational fluid dynamics package Flow3D-Hydro version 1.1 (Flow Science 2022). Flow3D-Hydro solves the transient Navier–Stokes equations of conservation of mass and momentum using a finite difference method and on Eulerian and Lagrangian frameworks (Flow Science 2022). The aforementioned governing equations are:





    where uu is the velocity vector, PP is the pressure, ρρ is the water density, υυ is the kinematic viscosity and gg is the gravitational acceleration. A Fractional Area/Volume Obstacle Representation (FAVOR) is adapted in Flow3D-Hydro, which applies solid boundaries within the Eulerian grid and calculates the fraction of areas and volume in partially blocked volume in order to compute flows on corresponding boundaries (Hirt and Nichols 1981). We validated the numerical modelling through comparing the results with Sainflou’s analytical equation for the design of vertical seawalls (Sainflou 1928; Ackhurst 2020), which is as follows:



    where pdpd is the hydrodynamic pressure, ρρ is the water density, gg is the gravitational acceleration, HH is the wave height, dd is the water depth, kk is the wavenumber, zz is the difference in still water level and mean water level, σσ is the angular frequency and tt is the time. The Sainflou’s equation (Eq. 3) is used to calculate the dynamic pressure from wave action, which is combined with static pressure on the seawall.

    Using Flow3D-Hydro, a model of the Dawlish seawall was made with a computational domain which is 250.0 m in length, 15.0 m in height and 0.375 m in width (Fig. 3a). The computational domain was discretised using a single uniform grid with a mesh size of 0.125 m. The model has a wave boundary at the left side of the domain (x-min), an outflow boundary on the right side (x-max), a symmetry boundary at the bottom (z-min) and a wall boundary at the top (z-max). A wall boundary implies that water or waves are unable to pass through the boundary, whereas a symmetry boundary means that the two edges of the boundary are identical and therefore there is no flow through it. The water is considered incompressible in our model. For volume of fluid advection for the wave boundary (i.e. the left-side boundary) in our simulations, we utilised the “Split Lagrangian Method”, which guarantees the best accuracy (Flow Science, 2022).

    figure 3
    Fig. 3

    The stability of the numerical scheme is controlled and maintained through checking the Courant number (CC) as given in the following:



    where VV is the velocity of the flow, ΔtΔt is the time step and ΔxΔx is the spatial step (i.e. grid size). For stability and convergence of the numerical simulations, the Courant number must be sufficiently below one (Courant et al. 1928). This is maintained by a careful adjustment of the ΔxΔx and ΔtΔt selections. Flow3D-Hydro applies a dynamic Courant number, meaning the program adjusts the value of time step (ΔtΔt) during the simulations to achieve a balance between accuracy of results and speed of simulation. In our simulation, the time step was in the range ΔtΔt = 0.0051—0.051 s.

    In order to achieve the most efficient mesh resolution, we varied cell size for five values of ΔxΔx = 0.1 m, 0.125 m, 0.15 m, 0.175 m and 0.20 m. Simulations were performed for all mesh sizes, and the results were compared in terms of convergence, stability and speed of simulation (Fig. 3). A linear wave with an amplitude of 1.5 m and a period of 6 s was used for these optimisation simulations. We considered wave time histories at two gauges A and B and recorded the waves from simulations using different mesh sizes (Fig. 3). Although the results are close (Fig. 3), some limited deviations are observed for larger mesh sizes of 0.20 m and 0.175 m. We therefore selected mesh size of 0.125 m as the optimum, giving an extra safety margin as a conservative solution.

    The pressure from the incident waves on the vertical wall is validated in our model by comparing them with the analytical equation of Sainflou (1928), Eq. (3), which is one of the most common set of equations for design of coastal structures (Fig. 4). The model was tested by running a linear wave of period 6 s and wave amplitude of 1.5 m against the wall, with a still water level of 4.5 m. It can be seen that the model results are very close to those from analytical equations of Sainflou (1928), indicating that our numerical model is accurately modelling the wave-structure interaction (Fig. 4).

    figure 4
    Fig. 4

    Eyewitness account analysis

    Contemporary reporting of the 4th and 5th February 2014 storms by the main national news outlets in the UK highlights the extreme nature of the events and the significant damage and disruption they were likely to have on the communities of the south-west of England. In interviews, this was reinforced by Network Rail engineers who, even at this early stage, were forecasting remedial engineering works to last for at least 6 weeks. One week later, following subsequent storms the cascading nature of the events was obvious. Multiple breaches of the seawall had taken place with up to 35 separate landslide events and significant damage to parapet walls along the coastal route also were reported. Residents of the area reported extreme effects of the storm, one likening it to an earthquake and reporting water ingress through doors windows and even through vertical chimneys (Table 1). This suggests extreme wave overtopping volumes and large wave impact forces. One resident described the structural effects as: “the house was jumping up and down on its footings”.

    Disaster management plans were quickly and effectively put into action by the local council, police service and National Rail. A major incident was declared, and decisions regarding evacuation of the residents under threat were taken around 2100 h on the night of 4th February when reports of initial damage to the seawall were received (Table 1). Local hotels were asked to provide short-term refuge to residents while local leisure facilities were prepared to accept residents later that evening. Initial repair work to the railway line was hampered by successive high spring tides and storms in the following days although significant progress was still made when weather conditions permitted (Table 1).

    Sea level observations and spectral analysis

    The results of surge and wave analyses are presented in Figs. 5 and 6. A surge height of up to 0.8 m was recorded in the examined tide gauge stations (Fig. 5b-d). Two main episodes of high surge heights are identified: the first surge started on 3rd February 2014 at 03:00 (UTC) and lasted until 4th of February 2014 at 00:00; the second event occurred in the period 4th February 2014 15:00 to 5th February 2014 at 17:00 (Fig. 5b-d). These data imply surge durations of 21 h and 26 h for the first and the second events, respectively. Based on the surge data in Fig. 5, we note that the storm event of early February 2014 and the associated surges was a relatively powerful one, which impacted at least 230 km of the south coast of England, from Land’s End to Weymouth, with large surge heights.

    figure 5
    Fig. 5
    figure 6
    Fig. 6

    Based on wave buoy records, the maximum recorded amplitudes are at least 20.5 m in Dawlish and West Bay, 1.9 m in Tor Bay and 4.9 m in Chesil (Fig. 6a-b). The buoys at Tor Bay and Chesil recorded dual peak period bands of 4–8 and 8–12 s, whereas at Dawlish and West Bay registered triple peak period bands at 4–8, 8–12 and 20–25 s (Fig. 6c, d). It is important to note that the long-period waves at 20–25 s occur with short durations (approximately 2 min) while the waves at the other two bands of 4–8 and 8–12 s appear to be present at all times during the storm event.

    The wave component at the period band of 4–8 s can be most likely attributed to normal coastal waves while the one at 8–12 s, which is longer, is most likely the swell component of the storm. Regarding the third component of the waves with long period of 20 -25 s, which occurs with short durations of 2 min, there are two hypotheses; it is either the result of a local (port and harbour) and regional (the Lyme Bay) oscillations (eg. Rabinovich 1997; Heidarzadeh and Satake 2014; Wang et al. 1992), or due to an abnormally long swell. To test the first hypothesis, we consider various water bodies such as Lyme Bay (approximate dimensions of 70 km × 20 km with an average water depth of 30 m; Fig. 6), several local bays (approximate dimensions of 3.6 km × 0.6 km with an average water depth of 6 m) and harbours (approximate dimensions of 0.5 km × 0.5 km with an average water depth of 4 m). Their water depths are based on the online Marine navigation website.Footnote8 According to Rabinovich (2010), the oscillation modes of a semi-enclosed rectangle basin are given by the following equation:



    where TmnTmn is the oscillation period, gg is the gravitational acceleration, dd is the water depth, LL is the length of the basin, WW is the width of the basin, m=1,2,3,…m=1,2,3,… and n=0,1,2,3,…n=0,1,2,3,…; mm and nn are the counters of the different modes. Applying Eq. (5) to the aforementioned water bodies results in oscillation modes of at least 5 min, which is far longer than the observed period of 20–25 s. Therefore, we rule out the first hypothesis and infer that the long period of 20–25 s is most likely a long swell wave coming from distant sources. As discussed by Rabinovich (1997) and Wang et al. (2022), comparison between sea level spectra before and after the incident is a useful method to distinguish the spectrum of the weather event. A visual inspection of Fig. 6 reveals that the forcing at the period band of 20–25 s is non-existent before the incident.

    Numerical simulations of wave loading and overtopping

    Based on the results of sea level data analyses in the previous section (Fig. 6), we use a dual peak wave spectrum with peak periods of 10.0 s and 25.0 s for numerical simulations because such a wave would be comprised of the most energetic signals of the storm. For variations of water depth (2.0–4.0 m), coastal wave amplitude (0.5–1.5 m) (Fig. 7) and storm surge height (0.5–0.8 m) (Fig. 5), we developed 20 scenarios (Scn) which we used in numerical simulations (Table 2). Data during the incident indicated that water depth was up to the crest level of the seawall (approximately 4 m water depth); therefore, we varied water depth from 2 to 4 m in our simulation scenarios. Regarding wave amplitudes, we referred to the variations at a nearby tide gauge station (West Bay) which showed wave amplitude up to 1.2 m (Fig. 7). Therefore, wave amplitude was varied from 0.5 m to 1.5 m by considering a factor a safety of 25% for the maximum wave amplitude. As for the storm surge component, time series of storm surges calculated at three coastal stations adjacent to Dawlish showed that it was in the range of 0.5 m to 0.8 m (Fig. 5). These 20 scenarios would help to study uncertainties associated with wave amplitudes and pressures. Figure 8 shows snapshots of wave propagation and impacts on the seawall at different times.

    figure 7
    Fig. 7

    Table 2 The 20 scenarios considered for numerical simulations in this study

    Full size table

    figure 8
    Fig. 8

    Results of wave amplitude simulations

    Large wave amplitudes can induce significant wave forcing on the structure and cause overtopping of the seawall, which could eventually cascade to other hazards such as erosion of the backfill and scour (Adams and Heidarzadeh, 2021). The first 10 scenarios of our modelling efforts are for the same incident wave amplitudes of 0.5 m, which occur at different water depths (2.0–4.0 m) and storm surge heights (0.5–0.8 m) (Table 2 and Fig. 9). This is because we aim at studying the impacts of effective water depth (deff—the sum of mean sea level and surge height) on the time histories of wave amplitudes as the storm evolves. As seen in Fig. 9a, by decreasing effective water depth, wave amplitude increases. For example, for Scn-1 with effective depth of 4.5 m, the maximum amplitude of the first wave is 1.6 m, whereas it is 2.9 m for Scn-2 with effective depth of 3.5 m. However, due to intensive reflections and interferences of the waves in front of the vertical seawall, such a relationship is barely seen for the second and the third wave peaks. It is important to note that the later peaks (second or third) produce the largest waves rather than the first wave. Extraordinary wave amplifications are seen for the Scn-2 (deff = 3.5 m) and Scn-7 (deff = 3.3 m), where the corresponding wave amplitudes are 4.5 m and 3.7 m, respectively. This may indicate that the effective water depth of deff = 3.3–3.5 m is possibly a critical water depth for this structure resulting in maximum wave amplitudes under similar storms. In the second wave impact, the combined wave height (i.e. the wave amplitude plus the effective water depth), which is ultimately an indicator of wave overtopping, shows that the largest wave heights are generated by Scn-2, 7 and 8 (Fig. 9a) with effective water depths of 3.5 m, 3.3 m and 3.8 m and combined heights of 8.0 m, 7.0 m and 6.9 m (Fig. 9b). Since the height of seawall is 5.4 m, the combined wave heights for Scn-2, 7 and 8 are greater than the crest height of the seawall by 2.6 m, 1.6 m and 1.5 m, respectively, which indicates wave overtopping.

    figure 9
    Fig. 9

    For scenarios 11–20 (Fig. 10), with incident wave amplitudes of 1.5 m (Table 2), the largest wave amplitudes are produced by Scn-17 (deff = 3.3 m), Scn-13 (deff = 2.5 m) and Scn-12 (deff = 3.5 m), which are 5.6 m, 5.1 m and 4.5 m. The maximum combined wave heights belong to Scn-11 (deff = 4.5 m) and Scn-17 (deff = 3.3 m), with combined wave heights of 9.0 m and 8.9 m (Fig. 10b), which are greater than the crest height of the seawall by 4.6 m and 3.5 m, respectively.

    figure 10
    Fig. 10

    Our simulations for all 20 scenarios reveal that the first wave is not always the largest and wave interactions, reflections and interferences play major roles in amplifying the waves in front of the seawall. This is primarily because the wall is fully vertical and therefore has a reflection coefficient of close to one (i.e. full reflection). Simulations show that the combined wave height is up to 4.6 m higher than the crest height of the wall, implying that severe overtopping would be expected.

    Results of wave loading calculations

    The pressure calculations for scenarios 1–10 are given in Fig. 11 and those of scenarios 11–20 in Fig. 12. The total pressure distribution in Figs. 1112 mostly follows a triangular shape with maximum pressure at the seafloor as expected from the Sainflou (1928) design equations. These pressure plots comprise both static (due to mean sea level in front of the wall) and dynamic (combined effects of surge and wave) pressures. For incident wave amplitudes of 0.5 m (Fig. 11), the maximum wave pressure varies in the range of 35–63 kPa. At the sea surface, it is in the range of 4–20 kPa (Fig. 11). For some scenarios (Scn-2 and 7), the pressure distribution deviates from a triangular shape and shows larger pressures at the top, which is attributed to the wave impacts and partial breaking at the sea surface. This adds an additional triangle-shaped pressure distribution at the sea surface elevation consistent with the design procedure developed by Goda (2000) for braking waves. The maximum force on the seawall due to scenarios 1–10, which is calculated by integrating the maximum pressure distribution over the wave-facing surface of the seawall, is in the range of 92–190 KN (Table 2).

    figure 11
    Fig. 11
    figure 12
    Fig. 12

    For scenarios 11–20, with incident wave amplitude of 1.5 m, wave pressures of 45–78 kPa and 7–120 kPa, for  the bottom and top of the wall, respectively, were observed (Fig. 12). Most of the plots show a triangular pressure distribution, except for Scn-11 and 15. A significant increase in wave impact pressure is seen for Scn-15 at the top of the structure, where a maximum pressure of approximately 120 kPa is produced while other scenarios give a pressure of 7–32 kPa for the sea surface. In other words, the pressure from Scn-15 is approximately four times larger than the other scenarios. Such a significant increase of the pressure at the top is most likely attributed to the breaking wave impact loads as detailed by Goda (2000) and Cuomo et al. (2010). The wave simulation snapshots in Fig. 8 show that the wave breaks before reaching the wall. The maximum force due to scenarios 11–20 is 120–286 KN.

    The breaking wave impacts peaking at 286 KN in our simulations suggest destabilisation of the upper masonry blocks, probably by grout malfunction. This significant impact force initiated the failure of the seawall which in turn caused extensive ballast erosion. Wave impact damage was proposed by Adams and Heidarzadeh (2021) as one of the primary mechanisms in the 2014 Dawlish disaster. In the multi-hazard risk model proposed by these authors, damage mechanism III (failure pathway 5 in Adams and Heidarzadeh, 2021) was characterised by wave impact force causing damage to the masonry elements, leading to failure of the upper sections of the seawall and loss of infill material. As blocks were removed, access to the track bed was increased for inbound waves allowing infill material from behind the seawall to be fluidised and subsequently removed by backwash. The loss of infill material critically compromised the stability of the seawall and directly led to structural failure. In parallel, significant wave overtopping (discussed in the next section) led to ballast washout and cascaded, in combination with masonry damage, to catastrophic failure of the wall and suspension of the rails in mid-air (Fig. 1b), leaving the railway inoperable for two months.

    Wave Overtopping

    The two most important factors contributing to the 2014 Dawlish railway catastrophe were wave impact forces and overtopping. Figure 13 gives the instantaneous overtopping rates for different scenarios, which experienced overtopping. It can be seen that the overtopping rates range from 0.5 m3/s/m to 16.1 m3/s/m (Fig. 13). Time histories of the wave overtopping rates show that the phenomenon occurs intermittently, and each time lasts 1.0–7.0 s. It is clear that the longer the overtopping time, the larger the volume of the water poured on the structure. The largest wave overtopping rates of 16.1 m3/s/m and 14.4 m3/s/m belong to Scn-20 and 11, respectively. These are the two scenarios that also give the largest combined wave heights (Fig. 10b).

    figure 13
    Fig. 13

    The cumulative overtopping curves (Figs. 1415) show the total water volume overtopped the structure during the entire simulation time. This is an important hazard factor as it determines the level of soil saturation, water pore pressure in the soil and soil erosion (Van der Meer et al. 2018). The maximum volume belongs to Scn-20, which is 65.0 m3/m (m-cubed of water per metre length of the wall). The overtopping volumes are 42.7 m3/m for Scn-11 and 28.8 m3/m for Scn-19. The overtopping volume is in the range of 0.7–65.0 m3/m for all scenarios.

    figure 14
    Fig. 14
    figure 15
    Fig. 15

    For comparison, we compare our modelling results with those estimated using empirical equations. For the case of the Dawlish seawall, we apply the equation proposed by Van Der Meer et al. (2018) to estimate wave overtopping rates, based on a set of decision criteria which are the influence of foreshore, vertical wall, possible breaking waves and low freeboard:



    where qq is the mean overtopping rate per metre length of the seawall (m3/s/m), gg is the acceleration due to gravity, HmHm is the incident wave height at the toe of the structure, RcRc is the wall crest height above mean sea level, hshs is the deep-water significant wave height and e(x)e(x) is the exponential function. It is noted that Eq. (6) is valid for 0.1<RcHm<1.350.1<RcHm<1.35. For the case of the Dawlish seawall and considering the scenarios with larger incident wave amplitude of 1.5 m (hshs= 1.5 m), the incident wave height at the toe of the structure is HmHm = 2.2—5.6 m, and the wall crest height above mean sea level is RcRc = 0.6–2.9 m. As a result, Eq. (6) gives mean overtopping rates up to approximately 2.9 m3/s/m. A visual inspection of simulated overtopping rates in Fig. 13 for Scn 11–20 shows that the mean value of the simulated overtopping rates (Fig. 13) is close to estimates using Eq. (6).

    Discussion and conclusions

    We applied a combination of eyewitness account analysis, sea level data analysis and numerical modelling in combination with our engineering judgement to explain the damage to the Dawlish railway seawall in February 2014. Main findings are:

    • Eyewitness data analysis showed that the extreme nature of the event was well forecasted in the hours prior to the storm impact; however, the magnitude of the risks to the structures was not well understood. Multiple hazards were activated simultaneously, and the effects cascaded to amplify the damage. Disaster management was effective, exemplified by the establishment of an emergency rendezvous point and temporary evacuation centre during the storm, indicating a high level of hazard awareness and preparedness.
    • Based on sea level data analysis, we identified triple peak period bands at 4–8, 8–12 and 20–25 s in the sea level data. Storm surge heights and wave oscillations were up to 0.8 m and 1.5 m, respectively.
    • Based on the numerical simulations of 20 scenarios with different water depths, incident wave amplitudes, surge heights and peak periods, we found that the wave oscillations at the foot of the seawall result in multiple wave interactions and interferences. Consequently, large wave amplitudes, up to 4.6 m higher than the height of the seawall, were generated and overtopped the wall. Extreme impulsive wave impact forces of up to 286 KN were generated by the waves interacting with the seawall.
    • We measured maximum wave overtopping rates of 0.5–16.1 m3/s/m for our scenarios. The cumulative overtopping water volumes per metre length of the wall were 0.7–65.0 m3/m.
    • Analysis of all the evidence combined with our engineering judgement suggests that the most likely initiating cause of the failure was impulsive wave impact forces destabilising one or more grouted joints between adjacent masonry blocks in the wall. Maximum observed pressures of 286 KN in our simulations are four times greater in magnitude than background pressures leading to block removal and initiating failure. Therefore, the sequence of cascading events was :1) impulsive wave impact force causing damage to masonry, 2) failure of the upper sections of the seawall, 3) loss of infill resulting in a reduction of structural strength in the landward direction, 4) ballast washout as wave overtopping and inbound wave activity increased and 5) progressive structural failure following successive tides.

    From a risk mitigation point of view, the stability of the seawall in the face of future energetic cyclonic storm events and sea level rise will become a critical factor in protecting the rail network. Mitigation efforts will involve significant infrastructure investment to strengthen the civil engineering assets combined with improved hazard warning systems consisting of meteorological forecasting and real-time wave observations and instrumentation. These efforts must take into account the amenity value of coastal railway infrastructure to local communities and the significant number of tourists who visit every year. In this regard, public awareness and active engagement in the planning and execution of the project will be crucial in order to secure local stakeholder support for the significant infrastructure project that will be required for future resilience.




    Download references


    We are grateful to Brunel University London for administering the scholarship awarded to KA. The Flow3D-Hydro used in this research for numerical modelling is licenced to Brunel University London through an academic programme contract. We sincerely thank Prof Harsh Gupta (Editor-in-Chief) and two anonymous reviewers for their constructive review comments.


    This project was funded by the UK Engineering and Physical Sciences Research Council (EPSRC) through a PhD scholarship to Keith Adams.

    Author information

    Authors and Affiliations

    1. Department of Civil and Environmental Engineering, Brunel University London, Uxbridge, UB8 3PH, UKKeith Adams
    2. Department of Architecture and Civil Engineering, University of Bath, Bath, BA2 7AY, UKMohammad Heidarzadeh

    Corresponding author

    Correspondence to Keith Adams.

    Ethics declarations

    Conflict of interest

    The authors have no relevant financial or non-financial interests to disclose.

    Availability of data

    All data used in this study are provided in the body of the article.

    Additional information

    Publisher’s Note

    Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

    Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit

    Reprints and Permissions

    About this article

    Verify currency and authenticity via CrossMark

    Cite this article

    Adams, K., Heidarzadeh, M. Extratropical cyclone damage to the seawall in Dawlish, UK: eyewitness accounts, sea level analysis and numerical modelling. Nat Hazards (2022).

    Download citation

    • Received17 May 2022
    • Accepted17 October 2022
    • Published14 November 2022
    • DOI

    Share this article

    Anyone you share the following link with will be able to read this content:Get shareable link

    Provided by the Springer Nature SharedIt content-sharing initiative


    • Storm surge
    • Cyclone
    • Railway
    • Climate change
    • Infrastructure
    • Resilience
    Fig. 1. (a) Dimensions of the casting with runners (unit: mm), (b) a melt flow simulation using Flow-3D software together with Reilly's model[44], predicted that a large amount of bifilms (denoted by the black particles) would be contained in the final casting. (c) A solidification simulation using Pro-cast software showed that no shrinkage defect was contained in the final casting.

    AZ91 합금 주물 내 연행 결함에 대한 캐리어 가스의 영향

    aUniversity of Birmingham, Birmingham B15 2TT, United Kingdom
    bGrainger and Worrall Ltd, Bridgnorth WV15 5HP, United Kingdom
    cBrunel Centre for Advanced Solidification Technology, Brunel University London, Kingston Ln, London, Uxbridge UB8 3PH, United Kingdom


    An entrainment defect (also known as a double oxide film defect or bifilm) acts a void containing an entrapped gas when submerged into a light-alloy melt, thus reducing the quality and reproducibility of the final castings. Previous publications, carried out with Al-alloy castings, reported that this trapped gas could be subsequently consumed by the reaction with the surrounding melt, thus reducing the void volume and negative effect of entrainment defects. Compared with Al-alloys, the entrapped gas within Mg-alloy might be more efficiently consumed due to the relatively high reactivity of magnesium. However, research into the entrainment defects within Mg alloys has been significantly limited. In the present work, AZ91 alloy castings were produced under different carrier gas atmospheres (i.e., SF6/CO2, SF6/air). The evolution processes of the entrainment defects contained in AZ91 alloy were suggested according to the microstructure inspections and thermodynamic calculations. The defects formed in the different atmospheres have a similar sandwich-like structure, but their oxide films contained different combinations of compounds. The use of carrier gases, which were associated with different entrained-gas consumption rates, affected the reproducibility of AZ91 castings.

    연행 결함(이중 산화막 결함 또는 이중막이라고도 함)은 경합금 용융물에 잠길 때 갇힌 가스를 포함하는 공극으로 작용하여 최종 주물의 품질과 재현성을 저하시킵니다. Al-합금 주물을 사용하여 수행된 이전 간행물에서는 이 갇힌 가스가 주변 용융물과의 반응에 의해 후속적으로 소모되어 공극 부피와 연행 결함의 부정적인 영향을 줄일 수 있다고 보고했습니다. Al-합금에 비해 마그네슘의 상대적으로 높은 반응성으로 인해 Mg-합금 내에 포집된 가스가 더 효율적으로 소모될 수 있습니다. 그러나 Mg 합금 내 연행 결함에 대한 연구는 상당히 제한적이었습니다. 현재 작업에서 AZ91 합금 주물은 다양한 캐리어 가스 분위기(즉, SF6/CO2, SF6/공기)에서 생산되었습니다. AZ91 합금에 포함된 연행 결함의 진화 과정은 미세 조직 검사 및 열역학 계산에 따라 제안되었습니다. 서로 다른 분위기에서 형성된 결함은 유사한 샌드위치 구조를 갖지만 산화막에는 서로 다른 화합물 조합이 포함되어 있습니다. 다른 동반 가스 소비율과 관련된 운반 가스의 사용은 AZ91 주물의 재현성에 영향을 미쳤습니다.


    Magnesium alloy, Casting, Oxide film, Bifilm, Entrainment defect, Reproducibility

    1. Introduction

    As the lightest structural metal available on Earth, magnesium became one of the most attractive light metals over the last few decades. The magnesium industry has consequently experienced a rapid development in the last 20 years [1,2], indicating a large growth in demand for Mg alloys all over the world. Nowadays, the use of Mg alloys can be found in the fields of automobiles, aerospace, electronics and etc.[3,4]. It has been predicted that the global consumption of Mg metals will further increase in the future, especially in the automotive industry, as the energy efficiency requirement of both traditional and electric vehicles further push manufactures lightweight their design [3,5,6].

    The sustained growth in demand for Mg alloys motivated a wide interest in the improvement of the quality and mechanical properties of Mg-alloy castings. During a Mg-alloy casting process, surface turbulence of the melt can lead to the entrapment of a doubled-over surface film containing a small quantity of the surrounding atmosphere, thus forming an entrainment defect (also known as a double oxide film defect or bifilm) [7][8][9][10]. The random size, quantity, orientation, and placement of entrainment defects are widely accepted to be significant factors linked to the variation of casting properties [7]. In addition, Peng et al. [11] found that entrained oxides films in AZ91 alloy melt acted as filters to Al8Mn5 particles, trapping them as they settle. Mackie et al. [12] further suggested that entrained oxide films can act to trawl the intermetallic particles, causing them to cluster and form extremely large defects. The clustering of intermetallic compounds made the entrainment defects more detrimental for the casting properties.

    Most of the previous studies regarding entrainment defects were carried out on Al-alloys [7,[13][14][15][16][17][18], and a few potential methods have been suggested for diminishing their negative effect on the quality of Al-alloy castings. Nyahumwa et al.,[16] shows that the void volume within entrainment defects could be reduced by a hot isostatic pressing (HIP) process. Campbell [7] suggested the entrained gas within the defects could be consumed due to reaction with the surrounding melt, which was further verified by Raiszedeh and Griffiths [19].The effect of the entrained gas consumption on the mechanical properties of Al-alloy castings has been investigated by [8,9], suggesting that the consumption of the entrained gas promoted the improvement of the casting reproducibility.

    Compared with the investigation concerning the defects within Al-alloys, research into the entrainment defects within Mg-alloys has been significantly limited. The existence of entrainment defects has been demonstrated in Mg-alloy castings [20,21], but their behaviour, evolution, as well as entrained gas consumption are still not clear.

    In a Mg-alloy casting process, the melt is usually protected by a cover gas to avoid magnesium ignition. The cavities of sand or investment moulds are accordingly required to be flushed with the cover gas prior to the melt pouring [22]. Therefore, the entrained gas within Mg-alloy castings should contain the cover gas used in the casting process, rather than air only, which may complicate the structure and evolution of the corresponding entrainment defects.

    SF6 is a typical cover gas widely used for Mg-alloy casting processes [23][24][25]. Although this cover gas has been restricted to use in European Mg-alloy foundries, a commercial report has pointed out that this cover is still popular in global Mg-alloy industry, especially in the countries which dominated the global Mg-alloy production, such as China, Brazil, India, etc. [26]. In addition, a survey in academic publications also showed that this cover gas was widely used in recent Mg-alloy studies [27]. The protective mechanism of SF6 cover gas (i.e., the reaction between liquid Mg-alloy and SF6 cover gas) has been investigated by several previous researchers, but the formation process of the surface oxide film is still not clearly understood, and even some published results are conflicting with each other. In early 1970s, Fruehling [28] found that the surface film formed under SF6 was MgO mainly with traces of fluorides, and suggested that SF6 was absorbed in the Mg-alloy surface film. Couling [29] further noticed that the absorbed SF6 reacted with the Mg-alloy melt to form MgF2. In last 20 years, different structures of the Mg-alloy surface films have been reported, as detailed below.(1)

    Single-layered film. Cashion [30,31] used X-ray Photoelectron Spectroscopy (XPS) and Auger Spectroscopy (AES) to identify the surface film as MgO and MgF2. He also found that composition of the film was constant throughout the thickness and the whole experimental holding time. The film observed by Cashion had a single-layered structure created from a holding time from 10 min to 100 min.(2)

    Double-layered film. Aarstad et. al [32] reported a doubled-layered surface oxide film in 2003. They observed several well-distributed MgF2 particles attached to the preliminary MgO film and grew until they covered 25–50% of the total surface area. The inward diffusion of F through the outer MgO film was the driving force for the evolution process. This double-layered structure was also supported by Xiong’s group [25,33] and Shih et al. [34].(3)

    Triple-layered film. The triple-layered film and its evolution process were reported in 2002 by Pettersen [35]. Pettersen found that the initial surface film was a MgO phase and then gradually evolved to the stable MgF2 phase by the inward diffusion of F. In the final stage, the film has a triple-layered structure with a thin O-rich interlayer between the thick top and bottom MgF2 layers.(4)

    Oxide film consisted of discrete particles. Wang et al [36] stirred the Mg-alloy surface film into the melt under a SF6 cover gas, and then inspect the entrained surface film after the solidification. They found that the entrained surface films were not continues as the protective surface films reported by other researchers but composed of discrete particles. The young oxide film was composed of MgO nano-sized oxide particles, while the old oxide films consist of coarse particles (about 1  µm in average size) on one side that contained fluorides and nitrides.

    The oxide films of a Mg-alloy melt surface or an entrained gas are both formed due to the reaction between liquid Mg-alloy and the cover gas, thus the above-mentioned research regarding the Mg-alloy surface film gives valuable insights into the evolution of entrainment defects. The protective mechanism of SF6 cover gas (i.e., formation of a Mg-alloy surface film) therefore indicated a potential complicated evolution process of the corresponding entrainment defects.

    However, it should be noted that the formation of a surface film on a Mg-alloy melt is in a different situation to the consumption of an entrained gas that is submerged into the melt. For example, a sufficient amount of cover gas was supported during the surface film formation in the studies previously mentioned, which suppressed the depletion of the cover gas. In contrast, the amount of entrained gas within a Mg-alloy melt is finite, and the entrained gas may become fully depleted. Mirak [37] introduced 3.5%SF6/air bubbles into a pure Mg-alloy melt solidifying in a specially designed permanent mould. It was found that the gas bubbles were entirely consumed, and the corresponding oxide film was a mixture of MgO and MgF2. However, the nucleation sites (such as the MgF2 spots observed by Aarstad [32] and Xiong [25,33]) were not observed. Mirak also speculated that the MgF2 formed prior to MgO in the oxide film based on the composition analysis, which was opposite to the surface film formation process reported in previous literatures (i.e., MgO formed prior to MgF2). Mirak’s work indicated that the oxide-film formation of an entrained gas may be quite different from that of surface films, but he did not reveal the structure and evolution of the oxide films.

    In addition, the use of carrier gas in the cover gases also influenced the reaction between the cover gas and the liquid Mg-alloy. SF6/air required a higher content of SF6 than did a SF6/CO2 carrier gas [38], to avoid the ignition of molten magnesium, revealing different gas-consumption rates. Liang [39] suggested that carbon was formed in the surface film when CO2 was used as a carrier gas, which was different from the films formed in SF6/air. An investigation into Mg combustion [40] reported a detection of Mg2C3 in the Mg-alloy sample after burning in CO2, which not only supported Liang’s results, but also indicated a potential formation of Mg carbides in double oxide film defects.

    The work reported here is an investigation into the behaviour and evolution of entrainment defects formed in AZ91 Mg-alloy castings, protected by different cover gases (i.e., SF6/air and SF6/CO2). These carrier gases have different protectability for liquid Mg alloy, which may be therefore associated with different consumption rates and evolution processes of the corresponding entrained gases. The effect of the entrained-gas consumption on the reproducibility of AZ91 castings was also studied.

    2. Experiment

    2.1. Melting and casting

    Three kilograms AZ91 alloy was melted in a mild steel crucible at 700 ± 5 °C. The composition of the AZ91 alloy has been shown in Table 1. Prior to heating, all oxide scale on the ingot surface was removed by machining. The cover gases used were 0.5%SF6/air or 0.5%SF6/CO2 (vol.%) at a flow rate of 6 L/min for different castings. The melt was degassed by argon with a flow rate of 0.3 L/min for 15 min [41,42], and then poured into sand moulds. Prior to pouring, the sand mould cavity was flushed with the cover gas for 20 min [22]. The residual melt (around 1 kg) was solidified in the crucible.

    Table 1. Composition (wt.%) of the AZ91 alloy used in this study.


    Fig. 1(a) shows the dimensions of the casting with runners. A top-filling system was deliberately used to generate entrainment defects in the final castings. Green and Campbell [7,43] suggested that a top-filling system caused more entrainment events (i.e., bifilms) during a casting process, compared with a bottom-filling system. A melt flow simulation (Flow-3D software) of this mould, using Reilly’s model [44] regarding the entrainment events, also predicted that a large amount of bifilms would be contained in the final casting (denoted by the black particles in Fig. 1b).

    Fig. 1. (a) Dimensions of the casting with runners (unit: mm), (b) a melt flow simulation using Flow-3D software together with Reilly's model[44], predicted that a large amount of bifilms (denoted by the black particles) would be contained in the final casting. (c) A solidification simulation using Pro-cast software showed that no shrinkage defect was contained in the final casting.

    Shrinkage defects also affect the mechanical properties and reproducibility of castings. Since this study focused on the effect of bifilms on the casting quality, the mould has been deliberately designed to avoid generating shrinkage defects. A solidification simulation using ProCAST software showed that no shrinkage defect would be contained in the final casting, as shown in Fig. 1c. The casting soundness has also been confirmed using a real time X-ray prior to the test bar machining.

    The sand moulds were made from resin-bonded silica sand, containing 1wt. % PEPSET 5230 resin and 1wt. % PEPSET 5112 catalyst. The sand also contained 2 wt.% Na2SiF6 to act as an inhibitor [45]. The pouring temperature was 700 ± 5 °C. After the solidification, a section of the runner bars was sent to the Sci-Lab Analytical Ltd for a H-content analysis (LECO analysis), and all the H-content measurements were carried out on the 5th day after the casting process. Each of the castings was machined into 40 test bars for a tensile strength test, using a Zwick 1484 tensile test machine with a clip extensometer. The fracture surfaces of the broken test bars were examined using Scanning Electron Microscope (SEM, Philips JEOL7000) with an accelerating voltage of 5–15 kV. The fractured test bars, residual Mg-alloy solidified in the crucible, and the casting runners were then sectioned, polished and also inspected using the same SEM. The cross-section of the oxide film found on the test-bar fracture surface was exposed by the Focused Ion Beam milling technique (FIB), using a CFEI Quanta 3D FEG FIB-SEM. The oxide film required to be analysed was coated with a platinum layer. Then, a gallium ion beam, accelerated to 30 kV, milled the material substrate surrounding the platinum coated area to expose the cross section of the oxide film. EDS analysis of the oxide film’s cross section was carried out using the FIB equipment at accelerating voltage of 30 kV.

    2.2. Oxidation cell

    As previously mentioned, several past researchers investigated the protective film formed on a Mg-alloy melt surface [38,39,[46][47][48][49][50][51][52]. During these experiments, the amount of cover gas used was sufficient, thus suppressing the depletion of fluorides in the cover gas. The experiment described in this section used a sealed oxidation cell, which limited the supply of cover gas, to study the evolution of the oxide films of entrainment defects. The cover gas contained in the oxidation cell was regarded as large-size “entrained bubble”.

    As shown in Fig. 2, the main body of the oxidation cell was a closed-end mild steel tube which had an inner length of 400 mm, and an inner diameter of 32 mm. A water-cooled copper tube was wrapped around the upper section of the cell. When the tube was heated, the cooling system created a temperature difference between the upper and lower sections, causing the interior gas to convect within the tube. The temperature was monitored by a type-K thermocouple located at the top of the crucible. Nie et al. [53] suggested that the SF6 cover gas would react with the steel wall of the holding furnace when they investigated the surface film of a Mg-alloy melt. To avoid this reaction, the interior surface of the steel oxidation cell (shown in Fig. 2) and the upper half section of the thermocouple were coated with boron nitride (the Mg-alloy was not in contact with boron nitride).

    Fig. 2. Schematic of the oxidation cell used to study the evolution of the oxide films of the entrainment defects (unit mm).

    During the experiment, a block of solid AZ91 alloy was placed in a magnesia crucible located at the bottom of the oxidation cell. The cell was heated to 100 °C in an electric resistance furnace under a gas flow rate of 1 L/min. The cell was held at this temperature for 20 min, to replace the original trapped atmosphere (i.e. air). Then, the oxidation cell was further heated to 700 °C, melting the AZ91 sample. The gas inlet and exit valves were then closed, creating a sealed environment for oxidation under a limited supply of cover gas. The oxidation cell was then held at 700 ± 10 °C for periods of time from 5 min to 30 min in 5-min intervals. At the end of each holding time, the cell was quenched in water. After cooling to room temperature, the oxidised sample was sectioned, polished, and subsequently examined by SEM.

    3. Results

    3.1. Structure and composition of the entrainment defects formed in SF6/air

    The structure and composition of the entrainment defect formed in the AZ91 castings under a cover gas of 0.5%SF6/air was observed by SEM and EDS. The results indicate that there exist two types of entrainment defects which are sketched in Fig. 3: (1) Type A defect whose oxide film has a traditional single-layered structure and (2) Type B defect, whose oxide film has two layers. The details of these defects were introduced in the following. Here it should be noticed that, as the entrainment defects are also known as biofilms or double oxide film, the oxide films of Type B defect were referred to as “multi-layered oxide film” or “multi-layered structure” in the present work to avoid a confusing description such as “the double-layered oxide film of a double oxide film defect”.

    Fig. 3. Schematic of the different types of entrainment defects found in AZ91 castings. (a) Type A defect with a single-layered oxide film and (b) Type B defect with two-layered oxide film.

    Fig. 4(a-b) shows a Type A defect having a compact single-layered oxide film with about 0.4 µm thickness. Oxygen, fluorine, magnesium and aluminium were detected in this film (Fig. 4c). It is speculated that oxide film is the mixture of fluoride and oxide of magnesium and aluminium. The detection of fluorine revealed that an entrained cover gas was contained in the formation of this defect. That is to say that the pores shown in Fig. 4(a) were not shrinkage defects or hydrogen porosity, but entrainment defects. The detection of aluminium was different with Xiong and Wang’s previous study [47,48], which showed that no aluminium was contained in their surface film of an AZ91 melt protected by a SF6 cover gas. Sulphur could not be clearly recognized in the element map, but there was a S-peak in the corresponding ESD spectrum.

    Fig. 4. (a) A Type A entrainment defect formed in SF6/air and having a single-layered oxide film, (b) the oxide film of this defect, (c) SEM-EDS element maps (using Philips JEOL7000) corresponding to the area highlighted in (b).

    Fig. 5(a-b) shows a Type B entrainment defect having a multi-layered oxide film. The compact outer layers of the oxide films were enriched with fluorine and oxygen (Fig. 5c), while their relatively porous inner layers were only enriched with oxygen (i.e., poor in fluorine) and partly grew together, thus forming a sandwich-like structure. Therefore, it is speculated that the outer layer is the mixture of fluoride and oxide, while the inner layer is mainly oxide. Sulphur could only be recognized in the EDX spectrum and could not be clearly identified in the element map, which might be due to the small S-content in the cover gas (i.e., 0.5% volume content of SF6 in the cover gas). In this oxide film, aluminium was contained in the outer layer of this oxide film but could not be clearly detected in the inner layer. Moreover, the distribution of Al seems to be uneven. It can be found that, in the right side of the defect, aluminium exists in the film but its concentration can not be identified to be higher than the matrix. However, there is a small area with much higher aluminium concentration in the left side of the defect. Such an uneven distribution of aluminium was also observed in other defects (shown in the following), and it is the result of the formation of some oxide particles in or under the film.

    Fig. 5. (a) A Type B entrainment defect formed in SF6/air and having a multi-layered oxide film, (b) the oxide films of this defect have grown together, (c) SEM-EDS element maps (using Philips JEOL7000) corresponding to the area shown in (b).

    Figs. 4 and 5 show cross sectional observations of the entrainment defects formed in the AZ91 alloy sample cast under a cover gas of SF6/air. It is not sufficient to characterize the entrainment defects only by the figures observed from the two-dimensional section. To have a further understanding, the surface of the entrainment defects (i.e. the oxide film) was further studied by observing the fracture surface of the test bars.

    Fig. 6(a) shows fracture surfaces of an AZ91 alloy tensile test bar produced in SF6/air. Symmetrical dark regions can be seen on both sides of the fracture surfaces. Fig. 6(b) shows boundaries between the dark and bright regions. The bright region consisted of jagged and broken features, while the surface of the dark region was relatively smooth and flat. In addition, the EDS results (Fig. 6c-d and Table 2) show that fluorine, oxygen, sulphur, and nitrogen were only detected in the dark regions, indicating that the dark regions were surface protective films entrained into the melt. Therefore, it could be suggested that the dark regions were an entrainment defect with consideration of their symmetrical nature. Similar defects on fracture surfaces of Al-alloy castings have been previously reported [7]Nitrides were only found in the oxide films on the test-bar fracture surfaces but never detected in the cross-sectional samples shown in Figs. 4 and 5. An underlying reason is that the nitrides contained in these samples may have hydrolysed during the sample polishing process [54].

    Fig. 6. (a) A pair of the fracture surfaces of a AZ91 alloy tensile test bar produced under a cover gas of SF6/air. The dimension of the fracture surface is 5 mm × 6 mm, (b) a section of the boundary between the dark and bright regions shown in (a), (c-d) EDS spectrum of the (c) bright regions and (d) dark regions, (e) schematic of an entrainment defect contained in a test bar.

    Table 2. EDS results (wt.%) corresponding to the regions shown in Fig. 6 (cover gas: SF6/air).

    Empty CellCOMgFAlZnSN
    Dark region in Fig. 6(b)3.481.3279.130.4713.630.570.080.73
    Bright region in Fig. 6(b)3.5884.4811.250.68

    In conjunction with the cross-sectional observation of the defects shown in Figs. 4 and 5, the structure of an entrainment defect contained in a tensile test bar was sketched as shown in Fig. 6(e). The defect contained an entrained gas enclosed by its oxide film, creating a void section inside the test bar. When the tensile force applied on the defect during the fracture process, the crack was initiated at the void section and propagated along the entrainment defect, since cracks would be propagated along the weakest path [55]. Therefore, when the test bar was finally fractured, the oxide films of entrainment defect appeared on both fracture surfaces of the test bar, as shown in Fig. 6(a).

    3.2. Structure and composition of the entrainment defects formed in SF6/CO2

    Similar to the entrainment defect formed in SF6/air, the defects formed under a cover gas of 0.5%SF6/CO2 also had two types of oxide films (i.e., single-layered and multi-layered types). Fig. 7(a) shows an example of the entrainment defects containing a multi-layered oxide film. A magnified observation to the defect (Fig. 7b) shows that the inner layers of the oxide films had grown together, presenting a sandwich-like structure, which was similar to the defects formed in an atmosphere of SF6/air (Fig. 5b). An EDS spectrum (Fig. 7c) revealed that the joint area (inner layer) of this sandwich-like structure mainly contained magnesium oxides. Peaks of fluorine, sulphur, and aluminium were recognized in this EDS spectrum, but their amount was relatively small. In contrast, the outer layers of the oxide films were compact and composed of a mixture of fluorides and oxides (Fig. 7d-e).

    Fig. 7. (a) An example of entrainment defects formed in SF6/CO2 and having a multi-layered oxide film, (b) magnified observation of the defect, showing the inner layer of the oxide films has grown together, (c) EDS spectrum of the point denoted in (b), (d) outer layer of the oxide film, (e) SEM-EDS element maps (using Philips JEOL7000) corresponding to the area shown in (d).

    Fig. 8(a) shows an entrainment defect on the fracture surfaces of an AZ91 alloy tensile test bar, which was produced in an atmosphere of 0.5%SF6/CO2. The corresponding EDS results (Table 3) showed that oxide film contained fluorides and oxides. Sulphur and nitrogen were not detected. Besides, a magnified observation (Fig. 8b) indicated spots on the oxide film surface. The diameter of the spots ranged from hundreds of nanometres to a few micron meters.

    Fig. 8. (a) A pair of the fracture surfaces of a AZ91 alloy tensile test bar, produced in an atmosphere of SF6/CO2. The dimension of the fracture surface is 5 mm × 6 mm, (b) surface appearance of the oxide films on the fracture surfaces, showing spots on the film surface.

    To further reveal the structure and composition of the oxide film clearly, the cross-section of the oxide film on a test-bar fracture surface was onsite exposed using the FIB technique (Fig. 9). As shown in Fig. 9a, a continuous oxide film was found between the platinum coating layer and the Mg-Al alloy substrate. Fig. 9 (b-c) shows a magnified observation to oxide films, indicating a multi-layered structure (denoted by the red box in Fig. 9c). The bottom layer was enriched with fluorine and oxygen and should be the mixture of fluoride and oxide, which was similar to the “outer layer” shown in Figs. 5 and 7, while the only-oxygen-enriched top layer was similar to the “inner layer” shown in Figs. 5 and 7.

    Fig. 9. (a) A cross-sectional observation of the oxide film on the fracture surface of the AZ91 casting produced in SF6/CO2, exposed by FIB, (b) a magnified observation of area highlighted in (a), and (c) SEM-EDS elements map of the area shown in (b), obtained by CFEI Quanta 3D FEG FIB-SEM.

    Except the continuous film, some individual particles were also observed in or below the continuous film, as shown in Fig. 9. An Al-enriched particle was detected in the left side of the oxide film shown in Fig. 9b and might be speculated to be spinel Mg2AlO4 because it also contains abundant magnesium and oxygen elements. The existing of such Mg2AlO4 particles is responsible for the high concentration of aluminium in small areas of the observed film and the uneven distribution of aluminium, as shown in Fig. 5(c). Here it should be emphasized that, although the other part of the bottom layer of the continuous oxide film contains less aluminium than this Al-enriched particle, the Fig. 9c indicated that the amount of aluminium in this bottom layer was still non-negligible, especially when comparing with the outer layer of the film. Below the right side of the oxide film shown in Fig. 9b, a particle was detected and speculated to be MgO because it is rich in Mg and O. According to Wang’s result [56], lots of discrete MgO particles can be formed on the surface of the Mg melt by the oxidation of Mg melt and Mg vapor. The MgO particles observed in our present work may be formed due to the same reasons. While, due to the differences in experimental conditions, less Mg melt can be vapored or react with O2, thus only a few of MgO particles formed in our work. An enrichment of carbon was also found in the film, revealing that CO2 was able to react with the melt, thus forming carbon or carbides. This carbon concentration was consistent with the relatively high carbon content of the oxide film shown in Table 3 (i.e., the dark region). In the area next to the oxide film.

    Table 3. EDS results (wt.%) corresponding to the regions shown in Fig. 8 (cover gas: SF6/ CO2).

    Empty CellCOMgFAlZnSN
    Dark region in Fig. 8(a)7.253.6469.823.827.030.86
    Bright region in Fig. 8(a)2.100.4482.8313.261.36

    This cross-sectional observation of the oxide film on a test bar fracture surface (Fig. 9) further verified the schematic of the entrainment defect shown in Fig. 6(e). The entrainment defects formed in different atmospheres of SF6/CO2 and SF6/air had similar structures, but their compositions were different.

    3.3. Evolution of the oxide films in the oxidation cell

    The results in Section 3.1 and 3.2 have shown the structures and compositions of entrainment defects formed in AZ91 castings under cover gases of SF6/air and SF6/CO2. Different stages of the oxidation reaction may lead to the different structures and compositions of entrainment defects. Although Campbell has conjectured that an entrained gas may react with the surrounding melt, it is rarely reported that the reaction occurring between the Mg-alloy melt and entrapped cover gas. Previous researchers normally focus on the reaction between a Mg-alloy melt and the cover gas in an open environment [38,39,[46][47][48][49][50][51][52], which was different from the situation of a cover gas trapped into the melt. To further understand the formation of the entrainment defect in an AZ91 alloy, the evolution process of oxide films of the entrainment defect was further studied using an oxidation cell.

    Fig. 10 (a and d) shows a surface film held for 5 min in the oxidation cell, protected by 0.5%SF6/air. There was only one single layer consisting of fluoride and oxide (MgF2 and MgO). In this surface film. Sulphur was detected in the EDS spectrum, but its amount was too small to be recognized in the element map. The structure and composition of this oxide film was similar to the single-layered films of entrainment defects shown in Fig. 4.

    Fig. 10. Oxide films formed in the oxidation cell under a cover gas of 0.5%SF6/air and held at 700 °C for (a) 5 min; (b) 10 min; (c) 30 min, and (d-f) the SEM-EDS element maps (using Philips JEOL7000) corresponding to the oxide film shown in (a-c) respectively, (d) 5 min; (e) 10 min; (f) 30 min. The red points in (c and f) are the location references, denoting the boundary of the F-enriched layer in different element maps.

    After a holding time of 10 min, a thin (O, S)-enriched top layer (around 700 nm) appeared upon the preliminary F-enriched film, forming a multi-layered structure, as shown in Fig. 10(b and e). The thickness of the (O, S)-enriched top layer increased with increased holding time. As shown in Fig. 10(c and f), the oxide film held for 30 min also had a multi-layered structure, but the thickness of its (O, S)-enriched top layer (around 2.5 µm) was higher than the that of the 10-min oxide film. The multi-layered oxide films shown in Fig. 10(b-c) presented a similar appearance to the films of the sandwich-like defect shown in Fig. 5.

    The different structures of the oxide films shown in Fig. 10 indicated that fluorides in the cover gas would be preferentially consumed due to the reaction with the AZ91 alloy melt. After the depletion of fluorides, the residual cover gas reacted further with the liquid AZ91 alloy, forming the top (O, S)-enriched layer in the oxide film. Therefore, the different structures and compositions of entrainment defects shown in Figs. 4 and 5 may be due to an ongoing oxidation reaction between melt and entrapped cover gas.

    This multi-layered structure has not been reported in previous publications concerning the protective surface film formed on a Mg-alloy melt [38,[46][47][48][49][50][51]. This may be due to the fact that previous researchers carried out their experiments with an un-limited amount of cover gas, creating a situation where the fluorides in the cover gas were not able to become depleted. Therefore, the oxide film of an entrainment defect had behaviour traits similar to the oxide films shown in Fig. 10, but different from the oxide films formed on the Mg-alloy melt surface reported in [38,[46][47][48][49][50][51].

    Similar with the oxide films held in SF6/air, the oxide films formed in SF6/CO2 also had different structures with different holding times in the oxidation cell. Fig. 11(a) shows an oxide film, held on an AZ91 melt surface under a cover gas of 0.5%SF6/CO2 for 5 min. This film had a single-layered structure consisting of MgF2. The existence of MgO could not be confirmed in this film. After the holding time of 30 min, the film had a multi-layered structure; the inner layer was of a compact and uniform appearance and composed of MgF2, while the outer layer is the mixture of MgF2 and MgO. Sulphur was not detected in this film, which was different from the surface film formed in 0.5%SF6/air. Therefore, fluorides in the cover gas of 0.5%SF6/CO2 were also preferentially consumed at an early stage of the film growth process. Compared with the film formed in SF6/air, the MgO in film formed in SF6/CO2 appeared later and sulphide did not appear within 30 min. It may mean that the formation and evolution of film in SF6/air is faster than SF6/CO2. CO2 may have subsequently reacted with the melt to form MgO, while sulphur-containing compounds accumulated in the cover gas and reacted to form sulphide in very late stage (may after 30 min in oxidation cell).

    Fig. 11. Oxide films formed in the oxidation cell under a cover gas of 0.5%SF6/CO2, and their SEM-EDS element maps (using Philips JEOL7000). They were held at 700 °C for (a) 5 min; (b) 30 min. The red points in (b) are the location references, denoting the boundary between the top and bottom layers in the oxide film.

    4. Discussion

    4.1. Evolution of entrainment defects formed in SF6/air

    HSC software from Outokumpu HSC Chemistry for Windows ( was used to carry out thermodynamic calculations needed to explore the reactions which might occur between the trapped gases and liquid AZ91 alloy. The solutions to the calculations suggest which products are most likely to form in the reaction process between a small amount of cover gas (i.e., the amount within a trapped bubble) and the AZ91-alloy melt.

    In the trials, the pressure was set to 1 atm, and the temperature set to 700 °C. The amount of the cover gas was assumed to be 7 × 10−7 kg, with a volume of approximately 0.57 cm3 (3.14 × 10−8 kmol) for 0.5%SF6/air, and 0.35 cm3 (3.12 × 10−8 kmol) for 0.5%SF6/CO2. The amount of the AZ91 alloy melt in contact with the trapped gas was assumed to be sufficient to complete all reactions. The decomposition products of SF6 were SF5, SF4, SF3, SF2, F2, S(g), S2(g) and F(g) [57][58][59][60].

    Fig. 12 shows the equilibrium diagram of the thermodynamic calculation of the reaction between the AZ91 alloy and 0.5%SF6/air. In the diagram, the reactants and products with less than 10−15 kmol have not been shown, as this was 5 orders of magnitude less than the amount of SF6 present (≈ 1.57 × 10−10 kmol) and therefore would not affect the observed process in a practical way.

    Fig. 12. An equilibrium diagram for the reaction between 7e-7 kg 0.5%SF6/air and a sufficient amount of AZ91 alloy. The X axis is the amount of AZ91 alloy melt having reacted with the entrained gas, and the vertical Y-axis is the amount of the reactants and products.

    This reaction process could be divided into 3 stages.

    Stage 1: The formation of fluorides. the AZ91 melt preferentially reacted with SF6 and its decomposition products, producing MgF2, AlF3, and ZnF2. However, the amount of ZnF2 may have been too small to be detected practically (1.25 × 10−12 kmol of ZnF2 compared with 3 × 10−10 kmol of MgF2), which may be the reason why Zn was not detected in any the oxide films shown in Sections 3.13.3. Meanwhile, sulphur accumulated in the residual gas as SO2.

    Stage 2: The formation of oxides. After the liquid AZ91 alloy had depleted all the available fluorides in the entrapped gas, the amount of AlF3 and ZnF2 quickly reduced due to a reaction with Mg. O2(g) and SO2 reacted with the AZ91 melt, forming MgO, Al2O3, MgAl2O4, ZnO, ZnSO4 and MgSO4. However, the amount of ZnO and ZnSO4 would have been too small to be found practically by EDS (e.g. 9.5 × 10−12 kmol of ZnO,1.38 × 10−14 kmol of ZnSO4, in contrast to 4.68 × 10−10 kmol of MgF2, when the amount of AZ91 on the X-axis is 2.5 × 10−9 kmol). In the experimental cases, the concentration of F in the cover gas is very low, whole the concentration f O is much higher. Therefore, the stage 1 and 2, i.e, the formation of fluoride and oxide may happen simultaneously at the beginning of the reaction, resulting in the formation of a singer-layered mixture of fluoride and oxide, as shown in Figs. 4 and 10(a). While an inner layer consisted of oxides but fluorides could form after the complete depletion of F element in the cover gas.

    Stages 1- 2 theoretically verified the formation process of the multi-layered structure shown in Fig. 10.

    The amount of MgAl2O4 and Al2O3 in the oxide film was of a sufficient amount to be detected, which was consistent with the oxide films shown in Fig. 4. However, the existence of aluminium could not be recognized in the oxide films grown in the oxidation cell, as shown in Fig. 10. This absence of Al may be due to the following reactions between the surface film and AZ91 alloy melt:(1)

    Al2O3 + 3Mg + = 3MgO + 2Al, △G(700 °C) = -119.82 kJ/mol(2)

    Mg + MgAl2O4 = MgO + Al, △G(700 °C) =-106.34 kJ/molwhich could not be simulated by the HSC software since the thermodynamic calculation was carried out under an assumption that the reactants were in full contact with each other. However, in a practical process, the AZ91 melt and the cover gas would not be able to be in contact with each other completely, due to the existence of the protective surface film.

    Stage 3: The formation of Sulphide and nitride. After a holding time of 30 min, the gas-phase fluorides and oxides in the oxidation cell had become depleted, allowing the melt reaction with the residual gas, forming an additional sulphur-enriched layer upon the initial F-enriched or (F, O)-enriched surface film, thus resulting in the observed multi-layered structure shown in Fig. 10 (b and c). Besides, nitrogen reacted with the AZ91 melt until all reactions were completed. The oxide film shown in Fig. 6 may correspond to this reaction stage due to its nitride content. However, the results shows that the nitrides were not detected in the polished samples shown in Figs. 4 and 5, but only found on the test bar fracture surfaces. The nitrides may have hydrolysed during the sample preparation process, as follows [54]:(3)

    Mg3N2 + 6H2O =3Mg(OH)2 + 2NH3↑(4)

    AlN+ 3H2O =Al(OH)3 + NH3

    In addition, Schmidt et al. [61] found that Mg3N2 and AlN could react to form ternary nitrides (Mg3AlnNn+2, n= 1, 2, 3…). HSC software did not contain the database of ternary nitrides, and it could not be added into the calculation. The oxide films in this stage may also contain ternary nitrides.

    4.2. Evolution of entrainment defects formed in SF6/CO2

    Fig. 13 shows the results of the thermodynamic calculation between AZ91 alloy and 0.5%SF6/CO2. This reaction processes can also be divided into three stages.

    Fig. 13. An equilibrium diagram for the reaction between 7e-7 kg 0.5%SF6/CO2 and a sufficient amount of AZ91 alloy. The X axis denotes the amount of Mg alloy melt having reacted with the entrained gas, and the vertical Y-axis denotes the amounts of the reactants and products.

    Stage 1: The formation of fluorides. SF6 and its decomposition products were consumed by the AZ91 melt, forming MgF2, AlF3, and ZnF2. As in the reaction of AZ91 in 0.5%SF6/air, the amount of ZnF2 was too small to be detected practically (1.51 × 10−13 kmol of ZnF2 compared with 2.67 × 10−10 kmol of MgF2). Sulphur accumulated in the residual trapped gas as S2(g) and a portion of the S2(g) reacted with CO2, to form SO2 and CO. The products in this reaction stage were consistent with the film shown in Fig. 11(a), which had a single layer structure that contained fluorides only.

    Stage 2: The formation of oxides. AlF3 and ZnF2 reacted with the Mg in the AZ91 melt, forming MgF2, Al and Zn. The SO2 began to be consumed, producing oxides in the surface film and S2(g) in the cover gas. Meanwhile, the CO2 directly reacted with the AZ91 melt, forming CO, MgO, ZnO, and Al2O3. The oxide films shown in Figs. 9 and 11(b) may correspond to this reaction stage due to their oxygen-enriched layer and multi-layered structure.

    The CO in the cover gas could further react with the AZ91 melt, producing C. This carbon may further react with Mg to form Mg carbides, when the temperature reduced (during solidification period) [62]. This may be the reason for the high carbon content in the oxide film shown in Figs. 89. Liang et al. [39] also reported carbon-detection in an AZ91 alloy surface film protected by SO2/CO2. The produced Al2O3 may be further combined with MgO, forming MgAl2O4 [63]. As discussed in Section 4.1, the alumina and spinel can react with Mg, causing an absence of aluminium in the surface films, as shown in Fig. 11.

    Stage 3: The formation of Sulphide. the AZ91 melt began to consume S2(g) in the residual entrapped gas, forming ZnS and MgS. These reactions did not occur until the last stage of the reaction process, which could be the reason why the S-content in the defect shown Fig. 7(c) was small.

    In summary, thermodynamic calculations indicate that the AZ91 melt will react with the cover gas to form fluorides firstly, then oxides and sulphides in the last. The oxide film in the different reaction stages would have different structures and compositions.

    4.3. Effect of the carrier gases on consumption of the entrained gas and the reproducibility of AZ91 castings

    The evolution processes of entrainment defects, formed in SF6/air and SF6/CO2, have been suggested in Sections 4.1 and 4.2. The theoretical calculations were verified with respect to the corresponding oxide films found in practical samples. The atmosphere within an entrainment defect could be efficiently consumed due to the reaction with liquid Mg-alloy, in a scenario dissimilar to the Al-alloy system (i.e., nitrogen in an entrained air bubble would not efficiently react with Al-alloy melt [64,65], however, nitrogen would be more readily consumed in liquid Mg alloys, commonly referred to as “nitrogen burning” [66]).

    The reaction between the entrained gas and the surrounding liquid Mg-alloy converted the entrained gas into solid compounds (e.g. MgO) within the oxide film, thus reducing the void volume of the entrainment defect and hence probably causing a collapse of the defect (e.g., if an entrained gas of air was depleted by the surrounding liquid Mg-alloy, under an assumption that the melt temperature is 700 °C and the depth of liquid Mg-alloy is 10 cm, the total volume of the final solid products would be 0.044% of the initial volume taken by the entrapped air).

    The relationship between the void volume reduction of entrainment defects and the corresponding casting properties has been widely studied in Al-alloy castings. Nyahumwa and Campbell [16] reported that the Hot Isostatic Pressing (HIP) process caused the entrainment defects in Al-alloy castings to collapse and their oxide surfaces forced into contact. The fatigue lives of their castings were improved after HIP. Nyahumwa and Campbell [16] also suggested a potential bonding of the double oxide films that were in contact with each other, but there was no direct evidence to support this. This binding phenomenon was further investigated by Aryafar[8], who re-melted two Al-alloy bars with oxide skins in a steel tube and then carried out a tensile strength test on the solidified sample. They found that the oxide skins of the Al-alloy bars strongly bonded with each other and became even stronger with an extension of the melt holding time, indicating a potential “healing” phenomenon due to the consumption of the entrained gas within the double oxide film structure. In addition, Raidszadeh and Griffiths [9,19] successfully reduced the negative effect of entrainment defects on the reproducibility of Al-alloy castings, by extending the melt holding time before solidification, which allowed the entrained gas to have a longer time to react with the surrounding melt.

    With consideration of the previous work mentioned, the consumption of the entrained gas in Mg-alloy castings may diminish the negative effect of entrainment defects in the following two ways.

    (1) Bonding phenomenon of the double oxide films. The sandwich-like structure shown in Fig. 5 and 7 indicated a potential bonding of the double oxide film structure. However, more evidence is required to quantify the increase in strength due to the bonding of the oxide films.

    (2) Void volume reduction of entrainment defects. The positive effect of void-volume reduction on the quality of castings has been widely demonstrated by the HIP process [67]. As the evolution processes discussed in Section 4.14.2, the oxide films of entrainment defects can grow together due to an ongoing reaction between the entrained gas and surrounding AZ91 alloy melt. The volume of the final solid products was significant small compared with the entrained gas (i.e., 0.044% as previously mentioned).

    Therefore, the consumption rate of the entrained gas (i.e., the growth rate of oxide films) may be a critical parameter for improving the quality of AZ91 alloy castings. The oxide film growth rate in the oxidization cell was accordingly further investigated.

    Fig. 14 shows a comparison of the surface film growth rates in different cover gases (i.e., 0.5%SF6/air and 0.5%SF6/CO2). 15 random points on each sample were selected for film thickness measurements. The 95% confidence interval (95%CI) was computed under an assumption that the variation of the film thickness followed a Gaussian distribution. It can be seen that all the surface films formed in 0.5%SF6/air grew faster than those formed in 0.5%SF6/CO2. The different growth rates suggested that the entrained-gas consumption rate of 0.5%SF6/air was higher than that of 0.5%SF6/CO2, which was more beneficial for the consumption of the entrained gas.

    Fig. 14. A comparison of the AZ91 alloy oxide film growth rates in 0.5%SF6/air and 0.5%SF6/CO2

    It should be noted that, in the oxidation cell, the contact area of liquid AZ91 alloy and cover gas (i.e. the size of the crucible) was relatively small with consideration of the large volume of melt and gas. Consequently, the holding time for the oxide film growth within the oxidation cell was comparatively long (i.e., 5–30 min). However, the entrainment defects contained in a real casting are comparatively very small (i.e., a few microns size as shown in Figs. 36, and [7]), and the entrained gas is fully enclosed by the surrounding melt, creating a relatively large contact area. Hence the reaction time for cover gas and the AZ91 alloy melt may be comparatively short. In addition, the solidification time of real Mg-alloy sand castings can be a few minutes (e.g. Guo [68] reported that a Mg-alloy sand casting with 60 mm diameter required 4 min to be solidified). Therefore, it can be expected that an entrained gas trapped during an Mg-alloy melt pouring process will be readily consumed by the surrounding melt, especially for sand castings and large-size castings, where solidification times are long.

    Therefore, the different cover gases (0.5%SF6/air and 0.5%SF6/CO2) associated with different consumption rates of the entrained gases may affect the reproducibility of the final castings. To verify this assumption, the AZ91 castings produced in 0.5%SF6/air and 0.5%SF6/CO2 were machined into test bars for mechanical evaluation. A Weibull analysis was carried out using both linear least square (LLS) method and non-linear least square (non-LLS) method [69].

    Fig. 15(a-b) shows a traditional 2-p linearized Weibull plot of the UTS and elongation of the AZ91 alloy castings, obtained by the LLS method. The estimator used is P= (i-0.5)/N, which was suggested to cause the lowest bias among all the popular estimators [69,70]. The casting produced in SF6/air has an UTS Weibull moduli of 16.9, and an elongation Weibull moduli of 5.0. In contrast, the UTS and elongation Weibull modulus of the casting produced in SF6/CO2 are 7.7 and 2.7 respectively, suggesting that the reproducibility of the casting protected by SF6/CO2 were much lower than that produced in SF6/air.

    Fig. 15. The Weibull modulus of AZ91 castings produced in different atmospheres, estimated by (a-b) the linear least square method, (c-d) the non-linear least square method, where SSR is the sum of residual squares.

    In addition, the author’s previous publication [69] demonstrated a shortcoming of the linearized Weibull plots, which may cause a higher bias and incorrect R2 interruption of the Weibull estimation. A Non-LLS Weibull estimation was therefore carried out, as shown in Fig. 15 (c-d). The UTS Weibull modulus of the SF6/air casting was 20.8, while the casting produced under SF6/CO2 had a lower UTS Weibull modulus of 11.4, showing a clear difference in their reproducibility. In addition, the SF6/air elongation (El%) dataset also had a Weibull modulus (shape = 5.8) higher than the elongation dataset of SF6/CO2 (shape = 3.1). Therefore, both the LLS and Non-LLS estimations suggested that the SF6/air casting has a higher reproducibility than the SF6/CO2 casting. It supports the method that the use of air instead of CO2 contributes to a quicker consumption of the entrained gas, which may reduce the void volume within the defects. Therefore, the use of 0.5%SF6/air instead of 0.5%SF6/CO2 (which increased the consumption rate of the entrained gas) improved the reproducibility of the AZ91 castings.

    However, it should be noted that not all the Mg-alloy foundries followed the casting process used in present work. The Mg-alloy melt in present work was degassed, thus reducing the effect of hydrogen on the consumption of the entrained gas (i.e., hydrogen could diffuse into the entrained gas, potentially suppressing the depletion of the entrained gas [7,71,72]). In contrast, in Mg-alloy foundries, the Mg-alloy melt is not normally degassed, since it was widely believed that there is not a ‘gas problem’ when casting magnesium and hence no significant change in tensile properties [73]. Although studies have shown the negative effect of hydrogen on the mechanical properties of Mg-alloy castings [41,42,73], a degassing process is still not very popular in Mg-alloy foundries.

    Moreover, in present work, the sand mould cavity was flushed with the SF6 cover gas prior to pouring [22]. However, not all the Mg-alloy foundries flushed the mould cavity in this way. For example, the Stone Foundry Ltd (UK) used sulphur powder instead of the cover-gas flushing. The entrained gas within their castings may be SO2/air, rather than the protective gas.

    Therefore, although the results in present work have shown that using air instead of CO2 improved the reproducibility of the final casting, it still requires further investigations to confirm the effect of carrier gases with respect to different industrial Mg-alloy casting processes.

    7. Conclusion

    Entrainment defects formed in an AZ91 alloy were observed. Their oxide films had two types of structure: single-layered and multi-layered. The multi-layered oxide film can grow together forming a sandwich-like structure in the final casting.2.

    Both the experimental results and the theoretical thermodynamic calculations demonstrated that fluorides in the trapped gas were depleted prior to the consumption of sulphur. A three-stage evolution process of the double oxide film defects has been suggested. The oxide films contained different combinations of compounds, depending on the evolution stage. The defects formed in SF6/air had a similar structure to those formed in SF6/CO2, but the compositions of their oxide films were different. The oxide-film formation and evolution process of the entrainment defects were different from that of the Mg-alloy surface films previous reported (i.e., MgO formed prior to MgF2).3.

    The growth rate of the oxide film was demonstrated to be greater under SF6/air than SF6/CO2, contributing to a quicker consumption of the damaging entrapped gas. The reproducibility of an AZ91 alloy casting improved when using SF6/air instead of SF6/CO2.


    The authors acknowledge funding from the EPSRC LiME grant EP/H026177/1, and the help from Dr W.D. Griffiths and Mr. Adrian Carden (University of Birmingham). The casting work was carried out in University of Birmingham.



    M.K. McNutt, SALAZAR K.

    Magnesium, Compounds & Metal, U.S. Geological Survey and U.S. Department of the Interior

    Reston, Virginia (2013)

    Google Scholar[2]


    Compounds & Metal, U.S. Geological Survey and U.S. Department of the Interior


    Google Scholar[3]

    I. Ostrovsky, Y. Henn

    ASTEC’07 International Conference-New Challenges in Aeronautics, Moscow (2007), pp. 1-5

    Aug 19-22

    View Record in ScopusGoogle Scholar[4]

    Y. Wan, B. Tang, Y. Gao, L. Tang, G. Sha, B. Zhang, N. Liang, C. Liu, S. Jiang, Z. Chen, X. Guo, Y. Zhao

    Acta Mater., 200 (2020), pp. 274-286

    ArticleDownload PDFView Record in Scopus[5]

    J.T.J. Burd, E.A. Moore, H. Ezzat, R. Kirchain, R. Roth

    Appl. Energy, 283 (2021), Article 116269

    ArticleDownload PDFView Record in Scopus[6]

    A.M. Lewis, J.C. Kelly, G.A. Keoleian

    Appl. Energy, 126 (2014), pp. 13-20

    ArticleDownload PDFView Record in Scopus[7]

    J. Campbell


    Butterworth-Heinemann, Oxford (2004)

    Google Scholar[8]

    M. Aryafar, R. Raiszadeh, A. Shalbafzadeh

    J. Mater. Sci., 45 (2010), pp. 3041-3051 View PDF

    CrossRefView Record in Scopus[9]

    R. Raiszadeh, W.D. Griffiths

    Metall. Mater. Trans. B-Process Metall. Mater. Process. Sci., 42 (2011), pp. 133-143 View PDF

    CrossRefView Record in Scopus[10]

    R. Raiszadeh, W.D. Griffiths

    J. Alloy. Compd., 491 (2010), pp. 575-580

    ArticleDownload PDFView Record in Scopus[11]

    L. Peng, G. Zeng, T.C. Su, H. Yasuda, K. Nogita, C.M. Gourlay

    JOM, 71 (2019), pp. 2235-2244 View PDF

    CrossRefView Record in Scopus[12]

    S. Ganguly, A.K. Mondal, S. Sarkar, A. Basu, S. Kumar, C. Blawert

    Corros. Sci., 166 (2020)[13]

    G.E. Bozchaloei, N. Varahram, P. Davami, S.K. Kim

    Mater. Sci. Eng. A-Struct. Mater. Prop. Microstruct. Process., 548 (2012), pp. 99-105

    View Record in Scopus[14]

    S. Fox, J. Campbell

    Scr. Mater., 43 (2000), pp. 881-886

    ArticleDownload PDFView Record in Scopus[15]

    M. Cox, R.A. Harding, J. Campbell

    Mater. Sci. Technol., 19 (2003), pp. 613-625

    View Record in Scopus[16]

    C. Nyahumwa, N.R. Green, J. Campbell

    Metall. Mater. Trans. A-Phys. Metall. Mater. Sci., 32 (2001), pp. 349-358

    View Record in Scopus[17]

    A. Ardekhani, R. Raiszadeh

    J. Mater. Eng. Perform., 21 (2012), pp. 1352-1362 View PDF

    CrossRefView Record in Scopus[18]

    X. Dai, X. Yang, J. Campbell, J. Wood

    Mater. Sci. Technol., 20 (2004), pp. 505-513

    View Record in Scopus[19]

    E.M. Elgallad, M.F. Ibrahim, H.W. Doty, F.H. Samuel

    Philos. Mag., 98 (2018), pp. 1337-1359 View PDF

    CrossRefView Record in Scopus[20]

    W.D. Griffiths, N.W. Lai

    Metall. Mater. Trans. A-Phys. Metall. Mater. Sci., 38A (2007), pp. 190-196 View PDF

    CrossRefView Record in Scopus[21]

    A.R. Mirak, M. Divandari, S.M.A. Boutorabi, J. Campbell

    Int. J. Cast Met. Res., 20 (2007), pp. 215-220 View PDF

    CrossRefView Record in Scopus[22]

    C. Cingi

    Laboratory of Foundry Engineering

    Helsinki University of Technology, Espoo, Finland (2006)

    Google Scholar[23]

    Y. Jia, J. Hou, H. Wang, Q. Le, Q. Lan, X. Chen, L. Bao

    J. Mater. Process. Technol., 278 (2020), Article 116542

    ArticleDownload PDFView Record in Scopus[24]

    S. Ouyang, G. Yang, H. Qin, S. Luo, L. Xiao, W. Jie

    Mater. Sci. Eng. A, 780 (2020), Article 139138

    ArticleDownload PDFView Record in Scopus[25]

    S.-m. Xiong, X.-F. Wang

    Trans. Nonferrous Met. Soc. China, 20 (2010), pp. 1228-1234

    ArticleDownload PDFView Record in Scopus[26]

    G.V. Research

    Grand View Research



    Google Scholar[27]

    T. Li, J. Davies

    Metall. Mater. Trans. A, 51 (2020), pp. 5389-5400 View PDF

    CrossRefView Record in Scopus[28]J.F. Fruehling, The University of Michigan, 1970.

    Google Scholar[29]

    S. Couling

    36th Annual World Conference on Magnesium, Norway (1979), pp. 54-57

    View Record in ScopusGoogle Scholar[30]

    S. Cashion, N. Ricketts, P. Hayes

    J. Light Met., 2 (2002), pp. 43-47

    ArticleDownload PDFView Record in Scopus[31]

    S. Cashion, N. Ricketts, P. Hayes

    J. Light Met., 2 (2002), pp. 37-42

    ArticleDownload PDFView Record in Scopus[32]

    K. Aarstad, G. Tranell, G. Pettersen, T.A. Engh

    Various Techniques to Study the Surface of Magnesium Protected by SF6

    TMS (2003)

    Google Scholar[33]

    S.-M. Xiong, X.-L. Liu

    Metall. Mater. Trans. A, 38 (2007), pp. 428-434 View PDF

    CrossRefView Record in Scopus[34]

    T.-S. Shih, J.-B. Liu, P.-S. Wei

    Mater. Chem. Phys., 104 (2007), pp. 497-504

    ArticleDownload PDFView Record in Scopus[35]

    G. Pettersen, E. Øvrelid, G. Tranell, J. Fenstad, H. Gjestland

    Mater. Sci. Eng. A, 332 (2002), pp. 285-294

    ArticleDownload PDFView Record in Scopus[36]

    H. Bo, L.B. Liu, Z.P. Jin

    J. Alloy. Compd., 490 (2010), pp. 318-325

    ArticleDownload PDFView Record in Scopus[37]

    A. Mirak, C. Davidson, J. Taylor

    Corros. Sci., 52 (2010), pp. 1992-2000

    ArticleDownload PDFView Record in Scopus[38]

    B.D. Lee, U.H. Beak, K.W. Lee, G.S. Han, J.W. Han

    Mater. Trans., 54 (2013), pp. 66-73 View PDF

    View Record in Scopus[39]

    W.Z. Liang, Q. Gao, F. Chen, H.H. Liu, Z.H. Zhao

    China Foundry, 9 (2012), pp. 226-230 View PDF


    U.I. Gol’dshleger, E.Y. Shafirovich

    Combust. Explos. Shock Waves, 35 (1999), pp. 637-644[41]

    A. Elsayed, S.L. Sin, E. Vandersluis, J. Hill, S. Ahmad, C. Ravindran, S. Amer Foundry

    Trans. Am. Foundry Soc., 120 (2012), pp. 423-429[42]

    E. Zhang, G.J. Wang, Z.C. Hu

    Mater. Sci. Technol., 26 (2010), pp. 1253-1258

    View Record in Scopus[43]

    N.R. Green, J. Campbell

    Mater. Sci. Eng. A-Struct. Mater. Prop. Microstruct. Process., 173 (1993), pp. 261-266

    ArticleDownload PDFView Record in Scopus[44]

    C Reilly, MR Jolly, NR Green

    Proceedings of MCWASP XII – 12th Modelling of Casting, Welding and Advanced Solidifcation Processes, Vancouver, Canada (2009)

    Google Scholar[45]H.E. Friedrich, B.L. Mordike, Springer, Germany, 2006.

    Google Scholar[46]

    C. Zheng, B.R. Qin, X.B. Lou

    Proceedings of the 2010 International Conference on Mechanical, Industrial, and Manufacturing Technologies, ASME (2010), pp. 383-388

    Mimt 2010 View PDF

    CrossRefView Record in ScopusGoogle Scholar[47]

    S.M. Xiong, X.F. Wang

    Trans. Nonferrous Met. Soc. China, 20 (2010), pp. 1228-1234

    ArticleDownload PDFView Record in Scopus[48]

    S.M. Xiong, X.L. Liu

    Metall. Mater. Trans. A-Phys. Metall. Mater. Sci., 38A (2007), pp. 428-434 View PDF

    CrossRefView Record in Scopus[49]

    T.S. Shih, J.B. Liu, P.S. Wei

    Mater. Chem. Phys., 104 (2007), pp. 497-504

    ArticleDownload PDFView Record in Scopus[50]

    K. Aarstad, G. Tranell, G. Pettersen, T.A. Engh

    Magn. Technol. (2003), pp. 5-10[51]

    G. Pettersen, E. Ovrelid, G. Tranell, J. Fenstad, H. Gjestland

    Mater. Sci. Eng. A-Struct. Mater. Prop. Microstruct. Process., 332 (2002), pp. 285-294

    ArticleDownload PDFView Record in Scopus[52]

    X.F. Wang, S.M. Xiong

    Corros. Sci., 66 (2013), pp. 300-307

    ArticleDownload PDFView Record in Scopus[53]

    S.H. Nie, S.M. Xiong, B.C. Liu

    Mater. Sci. Eng. A-Struct. Mater. Prop. Microstruct. Process., 422 (2006), pp. 346-351

    ArticleDownload PDFView Record in Scopus[54]

    C. Bauer, A. Mogessie, U. Galovsky

    Zeitschrift Fur Metallkunde, 97 (2006), pp. 164-168 View PDF


    Q.G. Wang, D. Apelian, D.A. Lados

    J. Light Met., 1 (2001), pp. 73-84

    ArticleDownload PDFView Record in Scopus[56]

    S. Wang, Y. Wang, Q. Ramasse, Z. Fan

    Metall. Mater. Trans. A, 51 (2020), pp. 2957-2974[57]

    S. Hayashi, W. Minami, T. Oguchi, H.J. Kim

    Kag. Kog. Ronbunshu, 35 (2009), pp. 411-415 View PDF

    CrossRefView Record in Scopus[58]

    K. Aarstad

    Norwegian University of Science and Technology


    Google Scholar[59]

    R.L. Wilkins

    J. Chem. Phys., 51 (1969), p. 853


    View Record in Scopus[60]

    O. Kubaschewski, K. Hesselemam

    Thermo-Chemical Properties of Inorganic Substances

    Springer-Verlag, Belin (1991)

    Google Scholar[61]

    R. Schmidt, M. Strobele, K. Eichele, H.J. Meyer

    Eur. J. Inorg. Chem. (2017), pp. 2727-2735 View PDF

    CrossRefView Record in Scopus[62]

    B. Hu, Y. Du, H. Xu, W. Sun, W.W. Zhang, D. Zhao

    J. Min. Metall. Sect. B-Metall., 46 (2010), pp. 97-103

    View Record in Scopus[63]

    O. Salas, H. Ni, V. Jayaram, K.C. Vlach, C.G. Levi, R. Mehrabian

    J. Mater. Res., 6 (1991), pp. 1964-1981

    View Record in Scopus[64]

    S.S.S. Kumari, U.T.S. Pillai, B.C. Pai

    J. Alloy. Compd., 509 (2011), pp. 2503-2509

    ArticleDownload PDFView Record in Scopus[65]

    H. Scholz, P. Greil

    J. Mater. Sci., 26 (1991), pp. 669-677

    View Record in Scopus[66]

    P. Biedenkopf, A. Karger, M. Laukotter, W. Schneider

    Magn. Technol., 2005 (2005), pp. 39-42

    View Record in Scopus[67]

    H.V. Atkinson, S. Davies

    Metall. Mater. Trans. A, 31 (2000), pp. 2981-3000 View PDF

    CrossRefView Record in Scopus[68]

    E.J. Guo, L. Wang, Y.C. Feng, L.P. Wang, Y.H. Chen

    J. Therm. Anal. Calorim., 135 (2019), pp. 2001-2008 View PDF

    CrossRefView Record in Scopus[69]

    T. Li, W.D. Griffiths, J. Chen

    Metall. Mater. Trans. A-Phys. Metall. Mater. Sci., 48A (2017), pp. 5516-5528 View PDF

    CrossRefView Record in Scopus[70]

    M. Tiryakioglu, D. Hudak

    J. Mater. Sci., 42 (2007), pp. 10173-10179 View PDF

    CrossRefView Record in Scopus[71]

    Y. Yue, W.D. Griffiths, J.L. Fife, N.R. Green

    Proceedings of the 1st International Conference on 3d Materials Science (2012), pp. 131-136 View PDF

    CrossRefView Record in ScopusGoogle Scholar[72]

    R. Raiszadeh, W.D. Griffiths

    Metall. Mater. Trans. B-Process Metall. Mater. Process. Sci., 37 (2006), pp. 865-871

    View Record in Scopus[73]

    Z.C. Hu, E.L. Zhang, S.Y. Zeng

    Mater. Sci. Technol., 24 (2008), pp. 1304-1308 View PDF

    CrossRefView Record in Scopus

    Fig. 1. Schematic figure showing the PREP with additional gas flowing on the end face of electrode.

    플라즈마 회전 전극 공정 중 분말 형성에 대한 공정 매개변수 및 냉각 가스의 영향

    Effects of process parameters and cooling gas on powder formation during the plasma rotating electrode process

    Yujie Cuia Yufan Zhaoa1 Haruko Numatab Kenta Yamanakaa Huakang Biana Kenta Aoyagia AkihikoChibaa
    aInstitute for Materials Research, Tohoku University, Sendai 980-8577, JapanbDepartment of Materials Processing, Graduate School of Engineering, Tohoku University, Sendai 980-8577, Japan


    •The limitation of increasing the rotational speed in decreasing powder size was clarified.

    •Cooling and disturbance effects varied with the gas flowing rate.

    •Inclined angle of the residual electrode end face affected powder formation.

    •Additional cooling gas flowing could be applied to control powder size.


    The plasma rotating electrode process (PREP) is rapidly becoming an important powder fabrication method in additive manufacturing. However, the low production rate of fine PREP powder limits the development of PREP. Herein, we investigated different factors affecting powder formation during PREP by combining experimental methods and numerical simulations. The limitation of increasing the rotation electrode speed in decreasing powder size is attributed to the increased probability of adjacent droplets recombining and the decreased tendency of granulation. The effects of additional Ar/He gas flowing on the rotational electrode on powder formation is determined through the cooling effect, the disturbance effect, and the inclined effect of the residual electrode end face simultaneously. A smaller-sized powder was obtained in the He atmosphere owing to the larger inclined angle of the residual electrode end face compared to the Ar atmosphere. Our research highlights the route for the fabrication of smaller-sized powders using PREP.

    플라즈마 회전 전극 공정(PREP)은 적층 제조 에서 중요한 분말 제조 방법으로 빠르게 자리잡고 있습니다. 그러나 미세한 PREP 분말의 낮은 생산율은 PREP의 개발을 제한합니다. 여기에서 우리는 실험 방법과 수치 시뮬레이션을 결합하여 PREP 동안 분말 형성에 영향을 미치는 다양한 요인을 조사했습니다. 분말 크기 감소에서 회전 전극 속도 증가의 한계는 인접한 액적 재결합 확률 증가 및 과립화 경향 감소에 기인합니다.. 회전 전극에 흐르는 추가 Ar/He 가스가 분말 형성에 미치는 영향은 냉각 효과, 외란 효과 및 잔류 전극 단면의 경사 효과를 통해 동시에 결정됩니다. He 분위기에서는 Ar 분위기에 비해 잔류 전극 단면의 경사각이 크기 때문에 더 작은 크기의 분말이 얻어졌다. 우리의 연구는 PREP를 사용하여 더 작은 크기의 분말을 제조하는 경로를 강조합니다.


    Plasma rotating electrode process

    Ti-6Al-4 V alloy, Rotating speed, Numerical simulation, Gas flowing, Powder size


    With the development of additive manufacturing, there has been a significant increase in high-quality powder production demand [1,2]. The initial powder characteristics are closely related to the uniform powder spreading [3,4], packing density [5], and layer thickness observed during additive manufacturing [6], thus determining the mechanical properties of the additive manufactured parts [7,8]. Gas atomization (GA) [9–11], centrifugal atomization (CA) [12–15], and the plasma rotating electrode process (PREP) are three important powder fabrication methods.

    Currently, GA is the dominant powder fabrication method used in additive manufacturing [16] for the fabrication of a wide range of alloys [11]. GA produces powders by impinging a liquid metal stream to droplets through a high-speed gas flow of nitrogen, argon, or helium. With relatively low energy consumption and a high fraction of fine powders, GA has become the most popular powder manufacturing technology for AM.

    The entrapped gas pores are generally formed in the powder after solidification during GA, in which the molten metal is impacted by a high-speed atomization gas jet. In addition, satellites are formed in GA powder when fine particles adhere to partially molten particles.

    The gas pores of GA powder result in porosity generation in the additive manufactured parts, which in turn deteriorates its mechanical properties because pores can become crack initiation sites [17]. In CA, a molten metal stream is poured directly onto an atomizer disc spinning at a high rotational speed. A thin film is formed on the surface of the disc, which breaks into small droplets due to the centrifugal force. Metal powder is obtained when these droplets solidify.

    Compared with GA powder, CA powder exhibits higher sphericity, lower impurity content, fewer satellites, and narrower particle size distribution [12]. However, very high speed is required to obtain fine powder by CA. In PREP, the molten metal, melted using the plasma arc, is ejected from the rotating rod through centrifugal force. Compared with GA powder, PREP-produced powders also have higher sphericity and fewer pores and satellites [18].

    For instance, PREP-fabricated Ti6Al-4 V alloy powder with a powder size below 150 μm exhibits lower porosity than gas-atomized powder [19], which decreases the porosity of additive manufactured parts. Furthermore, the process window during electron beam melting was broadened using PREP powder compared to GA powder in Inconel 718 alloy [20] owing to the higher sphericity of the PREP powder.

    In summary, PREP powder exhibits many advantages and is highly recommended for powder-based additive manufacturing and direct energy deposition-type additive manufacturing. However, the low production rate of fine PREP powder limits the widespread application of PREP powder in additive manufacturing.

    Although increasing the rotating speed is an effective method to decrease the powder size [21,22], the reduction in powder size becomes smaller with the increased rotating speed [23]. The occurrence of limiting effects has not been fully clarified yet.

    Moreover, the powder size can be decreased by increasing the rotating electrode diameter [24]. However, these methods are quite demanding for the PREP equipment. For instance, it is costly to revise the PREP equipment to meet the demand of further increasing the rotating speed or electrode diameter.

    Accordingly, more feasible methods should be developed to further decrease the PREP powder size. Another factor that influences powder formation is the melting rate [25]. It has been reported that increasing the melting rate decreases the powder size of Inconel 718 alloy [26].

    In contrast, the powder size of SUS316 alloy was decreased by decreasing the plasma current within certain ranges. This was ascribed to the formation of larger-sized droplets from fluid strips with increased thickness and spatial density at higher plasma currents [27]. The powder size of NiTi alloy also decreases at lower melting rates [28]. Consequently, altering the melting rate, varied with the plasma current, is expected to regulate the PREP powder size.

    Furthermore, gas flowing has a significant influence on powder formation [27,29–31]. On one hand, the disturbance effect of gas flowing promotes fluid granulation, which in turn contributes to the formation of smaller-sized powder [27]. On the other hand, the cooling effect of gas flowing facilitates the formation of large-sized powder due to increased viscosity and surface tension. However, there is a lack of systematic research on the effect of different gas flowing on powder formation during PREP.

    Herein, the authors systematically studied the effects of rotating speed, electrode diameter, plasma current, and gas flowing on the formation of Ti-6Al-4 V alloy powder during PREP as additive manufactured Ti-6Al-4 V alloy exhibits great application potential [32]. Numerical simulations were conducted to explain why increasing the rotating speed is not effective in decreasing powder size when the rotation speed reaches a certain level. In addition, the different factors incited by the Ar/He gas flowing on powder formation were clarified.

    Fig. 1. Schematic figure showing the PREP with additional gas flowing on the end face of electrode.
    Fig. 1. Schematic figure showing the PREP with additional gas flowing on the end face of electrode.


    [1] W. Ding, G. Chen, M. Qin, Y. He, X. Qu, Low-cost Ti powders for additive manufacturing treated by fluidized bed, Powder Technol. 350 (2019) 117–122,
    [2] W.S.W. Harun, M.S.I.N. Kamariah, N. Muhamad, S.A.C. Ghani, F. Ahmad, Z. Mohamed,
    A review of powder additive manufacturing processes for metallic biomaterials,
    Powder Technol. 327 (2018) 128–151,
    [3] M. Ahmed, M. Pasha, W. Nan, M. Ghadiri, A simple method for assessing powder
    spreadability for additive manufacturing, Powder Technol. 367 (2020) 671–679,
    [4] W. Nan, M. Pasha, M. Ghadiri, Numerical simulation of particle flow and segregation
    during roller spreading process in additive manufacturing, Powder Technol. 364
    (2020) 811–821,
    [5] A. Averardi, C. Cola, S.E. Zeltmann, N. Gupta, Effect of particle size distribution on the
    packing of powder beds : a critical discussion relevant to additive manufacturing,
    Mater. Today Commun. 24 (2020) 100964,
    [6] K. Riener, N. Albrecht, S. Ziegelmeier, R. Ramakrishnan, L. Haferkamp, A.B. Spierings,
    G.J. Leichtfried, Influence of particle size distribution and morphology on the properties of the powder feedstock as well as of AlSi10Mg parts produced by laser powder bed fusion (LPBF), Addit. Manuf. 34 (2020) 101286,
    [7] W.S.W. Harun, N.S. Manam, M.S.I.N. Kamariah, S. Sharif, A.H. Zulkifly, I. Ahmad, H.
    Miura, A review of powdered additive manufacturing techniques for Ti-6Al-4V biomedical applications, Powder Technol. 331 (2018) 74–97,
    [8] A.T. Sutton, C.S. Kriewall, M.C. Leu, J.W. Newkirk, Powder characterisation techniques and effects of powder characteristics on part properties in powder-bed fusion processes, Virtual Phys. Prototyp. 12 (2017) (2017) 3–29,
    [9] G. Chen, Q. Zhou, S.Y. Zhao, J.O. Yin, P. Tan, Z.F. Li, Y. Ge, J. Wang, H.P. Tang, A pore
    morphological study of gas-atomized Ti-6Al-4V powders by scanning electron microscopy and synchrotron X-ray computed tomography, Powder Technol. 330
    (2018) 425–430,
    [10] Y. Feng, T. Qiu, Preparation, characterization and microwave absorbing properties of
    FeNi alloy prepared by gas atomization method, J. Alloys Compd. 513 (2012)

    [11] I.E. Anderson, R.L. Terpstra, Progress toward gas atomization processing with increased uniformity and control, Mater. Sci. Eng. A 326 (2002) 101–109, https://
    [12] P. Phairote, T. Plookphol, S. Wisutmethangoon, Design and development of a centrifugal atomizer for producing zinc metal powder, Int. J. Appl. Phys. Math. 2 (2012)
    [13] L. Tian, I. Anderson, T. Riedemann, A. Russell, Production of fine calcium powders by
    centrifugal atomization with rotating quench bath, Powder Technol. 308 (2017)
    [14] M. Eslamian, J. Rak, N. Ashgriz, Preparation of aluminum/silicon carbide metal matrix composites using centrifugal atomization, Powder Technol. 184 (2008) 11–20,
    [15] T. Plookphol, S. Wisutmethangoon, S. Gonsrang, Influence of process parameters on
    SAC305 lead-free solder powder produced by centrifugal atomization, Powder
    Technol. 214 (2011) 506–512,
    [16] M.Z. Gao, B. Ludwig, T.A. Palmer, Impact of atomization gas on characteristics of austenitic stainless steel powder feedstocks for additive manufacturing, Powder
    Technol. 383 (2021) 30–42,
    [17] X. Shui, K. Yamanaka, M. Mori, Y. Nagata, A. Chiba, Effects of post-processing on cyclic fatigue response of a titanium alloy additively manufactured by electron beam
    melting, Mater. Sci. Eng. A 680 (2017) 239–248,
    [18] C. Wang, X.H. Zhao, Y.C. Ma, Q.X. Wang, Y.J. Lai, S.J. Liang, Study of the spherical
    HoCu powders prepared by supreme-speed plasma rotating electrode process,
    Powder Metallurgy Technology 38 (3) (2020) 227–233,
    j.cnki.cn11-1974/tf.2020.03.011 (in Chinese).
    [19] G. Chen, S.Y. Zhao, P. Tan, J. Wang, C.S. Xiang, H.P. Tang, A comparative study of Ti6Al-4V powders for additive manufacturing by gas atomization, plasma rotating
    electrode process and plasma atomization, Powder Technol. 333 (2018) 38–46,
    [20] Y. Zhao, K. Aoyagi, Y. Daino, K. Yamanaka, A. Chiba, Significance of powder feedstock
    characteristics in defect suppression of additively manufactured Inconel 718, Addit.
    Manuf. 34 (2020) 101277,
    [21] Y. Nie, J. Tang, B. Yang, Q. Lei, S. Yu, Y. Li, Comparison in characteristic and atomization behavior of metallic powders produced by plasma rotating electrode process,
    Adv. Powder Technol. 31 (2020) 2152–2160,
    [22] Y. Cui, Y. Zhao, H. Numata, H. Bian, K. Wako, K. Yamanaka, K. Aoyagi, C. Zhang, A.
    Chiba, Effects of plasma rotating electrode process parameters on the particle size
    distribution and microstructure of Ti-6Al-4 V alloy powder, Powder Technol 376
    (2020) 363–372,
    [23] J. Tang, Y. Nie, Q. Lei, Y. Li, Characteristics and atomization behavior of Ti-6Al-4V
    powder produced by plasma rotating electrode process Adv, Powder Technol. 10
    (2019) 2330–2337,
    [24] M. Zdujić, D. Uskoković, Production of atomized metal and alloy powders by the rotating electrode process, Sov. Powder Metall. Met. Ceram. 29 (1990) 673–683,
    [25] L. Zhang, Y. Zhao, Particle size distribution of tin powder produced by centrifugal
    atomisation using rotating cups, Powder Technol. 318 (2017) 62–67, https://doi.
    [26] Y. Liu, S. Liang, Z. Han, J. Song, Q. Wang, A novel model of calculating particle sizes in
    plasma rotating electrode process for superalloys, Powder Technol. 336 (2018)
    [27] Y. Zhao, Y. Cui, H. Numata, H. Bian, K. Wako, K. Yamanaka, Centrifugal granulation
    behavior in metallic powder fabrication by plasma rotating electrode process, Sci.
    Rep. (2020) 1–15,
    [28] T. Hsu, C. Wei, L. Wu, Y. Li, A. Chiba, M. Tsai, Nitinol powders generate from plasma
    rotation electrode process provide clean powder for biomedical devices used with
    suitable size, spheroid surface and pure composition, Sci. Rep. 8 (2018) 1–8,
    [29] M. Wei, S. Chen, M. Sun, J. Liang, C. Liu, M. Wang, Atomization simulation and preparation of 24CrNiMoY alloy steel powder using VIGA technology at high gas pressure, Powder Technol. 367 (2020) 724–739,
    [30] Y. Tan, X. Zhu, X.Y. He, B. Ding, H. Wang, Q. Liao, H. Li, Granulation characteristics of
    molten blast furnace slag by hybrid centrifugal-air blast technique, Powder Technol.
    323 (2018) 176–185,
    [31] P. Xu, D.H. Liu, J. Hu, G.Y. Lin, Synthesis of Ni-Ti composite powder by radio frequency plasma spheroidization process, Nonferrous Metals Science and Engineering
    39 (1) (2020) 67–71 , (in Chinese) 10.13264/j.cnki.ysjskx.2020.01.011.
    [32] H. Mehboob, F. Tarlochan, A. Mehboob, S.H. Chang, S. Ramesh, W.S.W. Harun, K.
    Kadirgama, A novel design, analysis and 3D printing of Ti-6Al-4V alloy bioinspired porous femoral stem, J. Mater. Sci. Mater. Med. 31 (2020) 78, https://doi.
    [33] FLOW-3D® Version 11.2 [Computer software]. , Flow Science, Inc., Santa Fe, NM,
    [34] M. Boivineau, C. Cagran, D. Doytier, V. Eyraud, M.H. Nadal, B. Wilthan, G. Pottlacher,
    Thermophysical properties of solid and liquid Ti-6Al-4V (TA6V) alloy, Int. J.
    Thermophys. 27 (2006) 507–529,
    [35] J. Liu, Q. Qin, Q. Yu, The effect of size distribution of slag particles obtained in dry
    granulation on blast furnace slag cement strength, Powder Technol. 362 (2020)
    [36] M. Tanaka, S. Tashiro, A study of thermal pinch effect of welding arcs, J. Japan Weld.
    Soc. 25 (2007) 336–342, (in Japanese).
    [37] T. Kamiya, A. Kayano, Disintegration of viscous fluid in the ligament state purged
    from a rotating disk, J. Chem. Eng. JAPAN. 4 (1971) 364–369,
    [38] T. Kamiya, An analysis of the ligament-type disintegration of thin liquid film at the
    edge of a rotating disk, J. Chem. Eng. Japan. 5 (1972) 391–396,
    [39] J. Burns, C. Ramshaw, R. Jachuck, Measurement of liquid film thickness and the determination of spin-up radius on a rotating disc using an electrical resistance technique, Chem. Eng. Sci. 58 (2003) 2245–2253,
    [40] J. Rauscher, R. Kelly, J. Cole, An asymptotic solution for the laminar flow of a thin film
    on a rotating disk, J. Appl. Mech. Trans. ASME 40 (1973) 43–47,

    Figure 2. Schematic diagram for pilot-scale cooling-water circulation system (a) along with a real picture of the system (b).

    Application of Computational Fluid Dynamics in Chlorine-Dynamics Modeling of In-Situ Chlorination Systems for Cooling Systems

    Jongchan Yi 1, Jonghun Lee 1, Mohd Amiruddin Fikri 2,3, Byoung-In Sang 4 and Hyunook Kim 1,*


    염소화는 상대적인 효율성과 저렴한 비용으로 인해 발전소 냉각 시스템에서 생물학적 오염을 제어하는​​데 선호되는 방법입니다. 해안 지역에 발전소가 있는 경우 바닷물을 사용하여 현장에서 염소를 전기화학적으로 생성할 수 있습니다. 이를 현장 전기염소화라고 합니다. 이 접근 방식은 유해한 염소화 부산물이 적고 염소를 저장할 필요가 없다는 점을 포함하여 몇 가지 장점이 있습니다. 그럼에도 불구하고, 이 전기화학적 공정은 실제로는 아직 초기 단계에 있습니다. 이 연구에서는 파일럿 규모 냉각 시스템에서 염소 붕괴를 시뮬레이션하기 위해 병렬 1차 동역학을 적용했습니다. 붕괴가 취수관을 따라 발생하기 때문에 동역학은 전산유체역학(CFD) 코드에 통합되었으며, 이후에 파이프의 염소 거동을 시뮬레이션하는데 적용되었습니다. 실험과 시뮬레이션 데이터는 강한 난류가 형성되는 조건하에서도 파이프 벽을 따라 염소 농도가 점진적인 것으로 나타났습니다. 염소가 중간보다 파이프 표면을 따라 훨씬 더 집중적으로 남아 있다는 사실은 전기 염소화를 기반으로 하는 시스템의 전체 염소 요구량을 감소시킬 수 있었습니다. 현장 전기 염소화 방식의 냉각 시스템은 직접 주입 방식에 필요한 염소 사용량의 1/3만 소비했습니다. 따라서 현장 전기염소화는 해안 지역의 발전소에서 바이오파울링 제어를 위한 비용 효율적이고 환경 친화적인 접근 방식으로 사용될 수 있다고 결론지었습니다.

    Chlorination is the preferred method to control biofouling in a power plant cooling system due to its comparative effectiveness and low cost. If a power plant is located in a coastal area, chlorine can be electrochemically generated in-situ using seawater, which is called in-situ electrochlorination; this approach has several advantages including fewer harmful chlorination byproducts and no need for chlorine storage. Nonetheless, this electrochemical process is still in its infancy in practice. In this study, a parallel first-order kinetics was applied to simulate chlorine decay in a pilot-scale cooling system. Since the decay occurs along the water-intake pipe, the kinetics was incorporated into computational fluid dynamics (CFD) codes, which were subsequently applied to simulate chlorine behavior in the pipe. The experiment and the simulation data indicated that chlorine concentrations along the pipe wall were incremental, even under the condition where a strong turbulent flow was formed. The fact that chlorine remained much more concentrated along the pipe surface than in the middle allowed for the reduction of the overall chlorine demand of the system based on the electro-chlorination. The cooling system, with an in-situ electro-chlorination, consumed only 1/3 of the chlorine dose demanded by the direct injection method. Therefore, it was concluded that in-situ electro-chlorination could serve as a cost-effective and environmentally friendly approach for biofouling control at power plants on coastal areas.


    computational fluid dynamics; power plant; cooling system; electro-chlorination; insitu chlorination

    Figure 1. Electrodes and batch experiment set-up. (a) Two cylindrical electrodes used in this study. (b) Batch experiment set-up for kinetic tests.
    Figure 1. Electrodes and batch experiment set-up. (a) Two cylindrical electrodes used in this study. (b) Batch experiment set-up for kinetic tests.
    Figure 2. Schematic diagram for pilot-scale cooling-water circulation system (a) along with a real picture of the system (b).
    Figure 2. Schematic diagram for pilot-scale cooling-water circulation system (a) along with a real picture of the system (b).
    Figure 3. Free chlorine decay curves in seawater with different TOC and initial chlorine concentration. Each line represents the predicted concentration of chlorine under a given condition. (a) Artificial seawater solution with 1 mg L−1 of TOC; (b) artificial seawater solution with 2 mg L−1 of TOC; (c) artificial seawater solution with 3 mg L−1 of TOC; (d) West Sea water (1.3 mg L−1 of TOC).
    Figure 3. Free chlorine decay curves in seawater with different TOC and initial chlorine concentration. Each line represents the predicted concentration of chlorine under a given condition. (a) Artificial seawater solution with 1 mg L−1 of TOC; (b) artificial seawater solution with 2 mg L−1 of TOC; (c) artificial seawater solution with 3 mg L−1 of TOC; (d) West Sea water (1.3 mg L−1 of TOC).
    Figure 4. Correlation between model and experimental data in the chlorine kinetics using seawater.
    Figure 4. Correlation between model and experimental data in the chlorine kinetics using seawater.
    Figure 5. Free chlorine concentrations in West Sea water under different current conditions in an insitu electro-chlorination system.
    Figure 5. Free chlorine concentrations in West Sea water under different current conditions in an insitu electro-chlorination system.
    Figure 6. Free chlorine distribution along the sampling ports under different flow rates. Each dot represents experimental data, and each point on the black line is the expected chlorine concentration obtained from computational fluid dynamics (CFD) simulation with a parallel first-order decay model. The red-dotted line is the desirable concentration at the given flow rate: (a) 600 L min−1 of flow rate, (b) 700 L min−1 of flow rate, (c) 800 L min−1 of flow rate, (d) 900 L min−1 of flow rate.
    Figure 6. Free chlorine distribution along the sampling ports under different flow rates. Each dot represents experimental data, and each point on the black line is the expected chlorine concentration obtained from computational fluid dynamics (CFD) simulation with a parallel first-order decay model. The red-dotted line is the desirable concentration at the given flow rate: (a) 600 L min−1 of flow rate, (b) 700 L min−1 of flow rate, (c) 800 L min−1 of flow rate, (d) 900 L min−1 of flow rate.
    Figure 7. Fluid contour images from CFD simulation of the electro-chlorination experiment. Inlet flow rate is 800 L min−1. Outlet pressure was set to 10.8 kPa. (a) Chlorine concentration; (b) expanded view of electrode side in image (a); (c) velocity magnitude; (d) pressure.
    Figure 7. Fluid contour images from CFD simulation of the electro-chlorination experiment. Inlet flow rate is 800 L min−1. Outlet pressure was set to 10.8 kPa. (a) Chlorine concentration; (b) expanded view of electrode side in image (a); (c) velocity magnitude; (d) pressure.
    Figure 8. Chlorine concentration contour in the simulation of full-scale in-situ electro-chlorination with different cathode positions. The pipe diameter is 2 m and the flow rate is 14 m3 s−1. The figure shows 10 m of the pipeline. (a) The simulation result when the cathode is placed on the surface of the pipe wall. (b) The simulation result when the cathode is placed on the inside of the pipe with 100 mm of distance from the pipe wall.
    Figure 8. Chlorine concentration contour in the simulation of full-scale in-situ electro-chlorination with different cathode positions. The pipe diameter is 2 m and the flow rate is 14 m3 s−1. The figure shows 10 m of the pipeline. (a) The simulation result when the cathode is placed on the surface of the pipe wall. (b) The simulation result when the cathode is placed on the inside of the pipe with 100 mm of distance from the pipe wall.
    Figure 9. Comparison of in-situ electro-chlorination and direct chlorine injection in full-scale applications. (a) Estimated chlorine concentrations along the pipe surface. (b) Relative chlorine demands.
    Figure 9. Comparison of in-situ electro-chlorination and direct chlorine injection in full-scale applications. (a) Estimated chlorine concentrations along the pipe surface. (b) Relative chlorine demands.


    1. Macknick, J.; Newmark, R.; Heath, G.; Hallett, K.C. Operational water consumption and withdrawal factors for electricity generating technologies: A review of existing literature. Environ. Res. Lett. 2012, 7, 045802.
    2. Pan, S.-Y.; Snyder, S.W.; Packman, A.I.; Lin, Y.J.; Chiang, P.-C. Cooling water use in thermoelectric power generation and its associated challenges for addressing water-energy nexus. Water-Energy Nexus 2018, 1, 26–41.
    3. Feeley, T.J., III; Skone, T.J.; Stiegel, G.J., Jr.; McNemar, A.; Nemeth, M.; Schimmoller, B.; Murphy, J.T.;
      Manfredo, L. Water: A critical resource in the thermoelectric power industry. Energy 2008, 33, 1–11.
    4. World Nuclear Association. World Nuclear Performance Report 2016; World Nuclear Association: London, UK, 2016.
    5. Pugh, S.; Hewitt, G.; Müller-Steinhagen, H. Fouling during the use of seawater as coolant—The development of a user guide. Heat Transf. Eng. 2005, 26, 35–43.
    6. Satpathy, K.K.; Mohanty, A.K.; Sahu, G.; Biswas, S.; Prasad, M.; Slvanayagam, M. Biofouling and its control in seawater cooled power plant cooling water system—A review. Nucl. Power 2010, 17, 191–242.
    7. Cristiani, P.; Perboni, G. Antifouling strategies and corrosion control in cooling circuits. Bioelectrochemistry 2014, 97, 120–126.
    8. Walker, M.E.; Safari, I.; Theregowda, R.B.; Hsieh, M.-K.; Abbasian, J.; Arastoopour, H.; Dzombak, D.A.; Miller, D.C. Economic impact of condenser fouling in existing thermoelectric power plants. Energy 2012,44, 429–437.
    9. Yi, J.; Ahn, Y.; Hong, M.; Kim, G.-H.; Shabnam, N.; Jeon, B.; Sang, B.-I.; Kim, H. Comparison between OCl−-Injection and In Situ Electrochlorination in the Formation of Chlorate and Perchlorate in Seawater. Appl.Sci. 2019, 9, 229.
    10. Xue, Y.; Zhao, J.; Qiu, R.; Zheng, J.; Lin, C.; Ma, B.; Wang, P. In Situ glass antifouling using Pt nanoparticle coating for periodic electrolysis of seawater. Appl. Surf. Sci. 2015, 357, 60–68.
    11. Mahfouz, A.B.; Atilhan, S.; Batchelor, B.; Linke, P.; Abdel-Wahab, A.; El-Halwagi, M.M. Optimal scheduling of biocide dosing for seawater-cooled power and desalination plants. Clean Technol. Environ. Policy 2011, 13, 783–796.
    12. Rubio, D.; López-Galindo, C.; Casanueva, J.F.; Nebot, E. Monitoring and assessment of an industrial antifouling treatment. Seasonal effects and influence of water velocity in an open once-through seawater cooling system. Appl. Therm. Eng. 2014, 67, 378–387.
    13. European Integrated Pollution Prevention and Control (IPPC) Bureau, European Commission. Reference Document on the Application of Best Available Techniques to Industrial Cooling Systems December 2001; European Commission, Tech. Rep: Brussels, Belgium, 2001.
    14. Venkatesan R.; Murthy P. S. Macrofouling Control in Power Plants. In Springer Series on Biofilms; Springer: Berlin/Heidelberg, Germany, 2008.
    15. Kastl, G.; Fisher, I.; Jegatheesan, V. Evaluation of chlorine decay kinetics expressions for drinking water distribution systems modelling. J. Water Supply Res. Technol. AQUA 1999, 48, 219–226.
    16. Fisher, I.; Kastl, G.; Sathasivan, A.; Cook, D.; Seneverathne, L. General model of chlorine decay in blends of surface waters, desalinated water, and groundwaters. J. Environ. Eng. 2015, 141, 04015039.
    17. Fisher, I.; Kastl, G.; Sathasivan, A.; Jegatheesan, V. Suitability of chlorine bulk decay models for planning and management of water distribution systems. Crit. Rev. Environ. Sci. Technol. 2011, 41, 1843–1882.
    18. Fisher, I.; Kastl, G.; Sathasivan, A. Evaluation of suitable chlorine bulk-decay models for water distribution systems. Water Res. 2011, 45, 4896–4908.
    19. Haas, C.N.; Karra, S. Kinetics of wastewater chlorine demand exertion. J. (Water Pollut. Control Fed.) 1984, 56, 170–173.
    20. Zeng, J.; Jiang, Z.; Chen, Q.; Zheng, P.; Huang, Y. The decay kinetics of residual chlorine in cooling seawater simulation experiments. Acta Oceanol. Sin. 2009, 28, 54–59.
    21. Saeed, S.; Prakash, S.; Deb, N.; Campbell, R.; Kolluru, V.; Febbo, E.; Dupont, J. Development of a sitespecific kinetic model for chlorine decay and the formation of chlorination by-products in seawater. J. Mar. Sci. Eng. 2015, 3, 772–792.
    22. Al Heboos, S.; Licskó, I. Application and comparison of two chlorine decay models for predicting bulk chlorine residuals. Period. Polytech. Civ. Eng. 2017, 61, 7–13.
    23. Shadloo, M.S.; Oger, G.; Le Touzé, D. Smoothed particle hydrodynamics method for fluid flows, towards industrial applications: Motivations, current state, and challenges. Comput. Fluids 2016, 136, 11–34.
    24. Wols, B.; Hofman, J.; Uijttewaal, W.; Rietveld, L.; Van Dijk, J. Evaluation of different disinfection calculation methods using CFD. Environ. Model. Softw. 2010, 25, 573–582.
    25. Angeloudis, A.; Stoesser, T.; Falconer, R.A. Predicting the disinfection efficiency range in chlorine contact tanks through a CFD-based approach. Water Res. 2014, 60, 118–129.
    26. Zhang, J.; Tejada-Martínez, A.E.; Zhang, Q. Developments in computational fluid dynamics-based modeling for disinfection technologies over the last two decades: A review. Environ. Model. Softw. 2014, 58,71–85.
    27. Lim, Y.H.; Deering, D.D. In Modeling Chlorine Residual in a Ground Water Supply Tank for a Small Community in Cold Conditions, World Environmental and Water Resources Congress 2017; American Society of Civil Engineers: Reston, Virginia, USA, 2017; pp. 124–138.
    28. Hernández-Cervantes, D.; Delgado-Galván, X.; Nava, J.L.; López-Jiménez, P.A.; Rosales, M.; Mora Rodríguez, J. Validation of a computational fluid dynamics model for a novel residence time distribution analysis in mixing at cross-junctions. Water 2018, 10, 733.
    29. Hua, F.; West, J.; Barker, R.; Forster, C. Modelling of chlorine decay in municipal water supplies. Water Res. 1999, 33, 2735–2746.
    30. Jonkergouw, P.M.; Khu, S.-T.; Savic, D.A.; Zhong, D.; Hou, X.Q.; Zhao, H.-B. A variable rate coefficient chlorine decay model. Environ. Sci. Technol. 2009, 43, 408–414.
    31. Nejjari, F.; Puig, V.; Pérez, R.; Quevedo, J.; Cugueró, M.; Sanz, G.; Mirats, J. Chlorine decay model calibration and comparison: Application to a real water network. Procedia Eng. 2014, 70, 1221–1230.
    32. Kohpaei, A.J.; Sathasivan, A.; Aboutalebi, H. Effectiveness of parallel second order model over second and first order models. Desalin. Water Treat. 2011, 32, 107–114.
    33. Powell, J.C.; Hallam, N.B.; West, J.R.; Forster, C.F.; Simms, J. Factors which control bulk chlorine decay rates. Water Res. 2000, 34, 117–126.
    34. Clark, R.M.; Sivaganesan, M. Predicting chlorine residuals in drinking water: Second order model. J. Water Resour. Plan. Manag. 2002, 128, 152–161.
    35. Li, X.; Li, C.; Bayier, M.; Zhao, T.; Zhang, T.; Chen, X.; Mao, X. Desalinated seawater into pilot-scale drinking water distribution system: Chlorine decay and trihalomethanes formation. Desalin. Water Treat. 2016, 57,19149–19159.
    36. United States Environmental Protection Agency (EPA). Chlorine, Total Residual (Spectrophotometric, DPD); EPA-NERL: 330.5; EPA: Cincinnati, OH, USA, 1978.
    37. Polman, H.; Verhaart, F.; Bruijs, M. Impact of biofouling in intake pipes on the hydraulics and efficiency of pumping capacity. Desalin. Water Treat. 2013, 51, 997–1003.
    38. Rajagopal, S.; Van der Velde, G.; Van der Gaag, M.; Jenner, H.A. How effective is intermittent chlorination to control adult mussel fouling in cooling water systems? Water Res. 2003, 37, 329–338.
    39. Bruijs, M.C.; Venhuis, L.P.; Daal, L. Global Experiences in Optimizing Biofouling Control through PulseChlorination®. 2017. Available online: ntrol_through_Pulse-ChlorinationR (accessed on 1 May 2020).
    40. Kim, H.; Hao, O.J.; McAvoy, T.J. Comparison between model-and pH/ORP-based process control for an AAA system. Tamkang J. Sci. Eng. 2000, 3, 165–172.
    41. Brdys, M.; Chang, T.; Duzinkiewicz, K. Intelligent Model Predictive Control of Chlorine Residuals in Water Distribution Systems, Bridging the Gap: Meeting the World’s Water and Environmental Resources Challenges. In Proceedings of the ASCE Water Resource Engineering and Water Resources Planning and Management, July 30–August 2, 2000; pp. 1–11
    Fig. 5. The predicted shapes of initial breach (a) Rectangular (b) V-notch. Fig. 6. Dam breaching stages.

    Investigating the peak outflow through a spatial embankment dam breach

    공간적 제방댐 붕괴를 통한 최대 유출량 조사

    Mahmoud T.GhonimMagdy H.MowafyMohamed N.SalemAshrafJatwaryFaculty of Engineering, Zagazig University, Zagazig 44519, Egypt


    Investigating the breach outflow hydrograph is an essential task to conduct mitigation plans and flood warnings. In the present study, the spatial dam breach is simulated by using a three-dimensional computational fluid dynamics model, FLOW-3D. The model parameters were adjusted by making a comparison with a previous experimental model. The different parameters (initial breach shape, dimensions, location, and dam slopes) are studied to investigate their effects on dam breaching. The results indicate that these parameters have a significant impact. The maximum erosion rate and peak outflow for the rectangular shape are higher than those for the V-notch by 8.85% and 5%, respectively. Increasing breach width or decreasing depth by 5% leads to increasing maximum erosion rate by 11% and 15%, respectively. Increasing the downstream slope angle by 4° leads to an increase in both peak outflow and maximum erosion rate by 2.0% and 6.0%, respectively.

    유출 유출 수문곡선을 조사하는 것은 완화 계획 및 홍수 경보를 수행하는 데 필수적인 작업입니다. 본 연구에서는 3차원 전산유체역학 모델인 FLOW-3D를 사용하여 공간 댐 붕괴를 시뮬레이션합니다. 이전 실험 모델과 비교하여 모델 매개변수를 조정했습니다.

    다양한 매개변수(초기 붕괴 형태, 치수, 위치 및 댐 경사)가 댐 붕괴에 미치는 영향을 조사하기 위해 연구됩니다. 결과는 이러한 매개변수가 상당한 영향을 미친다는 것을 나타냅니다. 직사각형 형태의 최대 침식율과 최대 유출량은 V-notch보다 각각 8.85%, 5% 높게 나타났습니다.

    위반 폭을 늘리거나 깊이를 5% 줄이면 최대 침식률이 각각 11% 및 15% 증가합니다. 하류 경사각을 4° 증가시키면 최대 유출량과 최대 침식률이 각각 2.0% 및 6.0% 증가합니다.


    Spatial dam breach; FLOW-3D; Overtopping erosion; Computational fluid dynamics (CFD)

    1. Introduction

    There are many purposes for dam construction, such as protection from flood disasters, water storage, and power generationEmbankment failures may have a catastrophic impact on lives and infrastructure in the downstream regions. One of the most common causes of embankment dam failure is overtopping. Once the overtopping of the dam begins, the breach formation will start in the dam body then end with the dam failure. This failure occurs within a very short time, which threatens to be very dangerous. Therefore, understanding and modeling the embankment breaching processes is essential for conducting mitigation plans, flood warnings, and forecasting flood damage.

    The analysis of the dam breaching process is implemented by different techniques: comparative methods, empirical models with dimensional and dimensionless solutions, physical-based models, and parametric models. These models were described in detail [1]Parametric modeling is commonly used to simulate breach growth as a time-dependent linear process and calculate outflow discharge from the breach using hydraulics principles [2]. Alhasan et al. [3] presented a simple one-dimensional mathematical model and a computer code to simulate the dam breaching process. These models were validated by small dams breaching during the floods in 2002 in the Czech Republic. Fread [4] developed an erosion model (BREACH) based on hydraulics principles, sediment transport, and soil mechanics to estimate breach size, time of formation, and outflow discharge. Říha et al. [5] investigated the dam break process for a cascade of small dams using a simple parametric model for piping and overtopping erosion, as well as a 2D shallow-water flow model for the flood in downstream areas. Goodarzi et al. [6] implemented mathematical and statistical methods to assess the effect of inflows and wind speeds on the dam’s overtopping failure.

    Dam breaching studies can be divided into two main modes of erosion. The first mode is called “planar dam breach” where the flow overtops the whole dam width. While the second mode is called “spatial dam breach” where the flow overtops through the initial pilot channel (i.e., a channel created in the dam body). Therefore, the erosion will be in both vertical and horizontal directions [7].

    The erosion process through the embankment dams occurs due to the shear stress applied by water flows. The dam breaching evolution can be divided into three stages [8][9], but Y. Yang et al. [10] divided the breach development into five stages: Stage I, the seepage erosion; Stage II, the initial breach formation; Stage III, the head erosion; Stage IV, the breach expansion; and Stage V, the re-equilibrium of the river channel through the breach. Many experimental tests have been carried out on non-cohesive embankment dams with an initial breach to examine the effect of upstream inflow discharges on the longitudinal profile evolution and the time to inflection point [11].

    Zhang et al. [12] studied the effect of changing downstream slope angle, sediment grain size, and dam crest length on erosion rates. They noticed that increasing dam crest length and decreasing downstream slope angle lead to decreasing sediment transport rate. While the increase in sediment grain size leads to an increased sediment transport rate at the initial stages. Höeg et al. [13] presented a series of field tests to investigate the stability of embankment dams made of various materials. Overtopping and piping were among the failure tests carried out for the dams composed of homogeneous rock-fill, clay, or gravel with a height of up to 6.0 m. Hakimzadeh et al. [14] constructed 40 homogeneous cohesive and non-cohesive embankment dams to study the effect of changing sediment diameter and dam height on the breaching process. They also used genetic programming (GP) to estimate the breach outflow. Refaiy et al. [15] studied different scenarios for the downstream drain geometry, such as length, height, and angle, to minimize the effect of piping phenomena and therefore increase dam safety.

    Zhu et al. [16] examined the effect of headcut erosion on dam breach growth, especially in the case of cohesive dams. They found that the breach growth in non-cohesive embankments is slower than cohesive embankments due to the little effect of headcut. Schmocker and Hager [7] proposed a relationship for estimating peak outflow from the dam breach process.(1)QpQin-1=1.7exp-20hc23d5013H0

    where: Qp = peak outflow discharge.

    Qin = inflow discharge.

    hc = critical flow depth.

    d50 = mean sediment diameter.

    Ho = initial dam height.

    Yu et al. [17] carried out an experimental study for homogeneous non-cohesive embankment dams in a 180° bending rectangular flume to determine the effect of overtopping flows on breaching formation. They found that the main factors influencing breach formation are water level, river discharge, and embankment material diameter.

    Wu et al. [18] carried out a series of experiments to investigate the effect of breaching geometry on both non-cohesive and cohesive embankment dams in a U-bend flume due to overtopping flows. In the case of non-cohesive embankments, the non-symmetrical lateral expansion was noticed during the breach formation. This expansion was described by a coefficient ranging from 2.7 to 3.3.

    The numerical models of the dam breach can be categorized according to different parameters, such as flow dimensions (1D, 2D, or 3D), flow governing equations, and solution methods. The 1D models are mainly used to predict the outflow hydrograph from the dam breach. Saberi et al. [19] applied the 1D Saint-Venant equation, which is solved by the finite difference method to investigate the outflow hydrograph during dam overtopping failure. Because of the ability to study dam profile evolution and breach formation, 2D models are more applicable than 1D models. Guan et al. [20] and Wu et al. [21] employed both 2D shallow water equations (SWEs) and sediment erosion equations, which are solved by the finite volume method to study the effect of the dam’s geometry parameters on outflow hydrograph and dam profile evolution. Wang et al. [22] also proposed a second-order hybrid-type of total variation diminishing (TVD) finite-difference to estimate the breach outflow by solving the 2D (SWEs). The accuracy of (SWEs) for both vertical flow contraction and surface roughness has been assessed [23]. They noted that the accuracy of (SWEs) is acceptable for milder slopes, but in the case of steeper slopes, modelers should be more careful. Generally, the accuracy of 2D models is still low, especially with velocity distribution over the flow depth, lateral momentum exchange, density-driven flows, and bottom friction [24]. Therefore, 3D models are preferred. Larocque et al. [25] and Yang et al. [26] started to use three-dimensional (3D) models that depend on the Reynolds-averaged Navier-Stokes (RANS) equations.

    Previous experimental studies concluded that there is no clear relationship between the peak outflow from the dam breach and the initial breach characteristics. Some of these studies depend on the sharp-crested weir fixed at the end of the flume to determine the peak outflow from the breach, which leads to a decrease in the accuracy of outflow calculations at the microscale. The main goals of this study are to carry out a numerical simulation for a spatial dam breach due to overtopping flows by using (FLOW-3D) software to find an empirical equation for the peak outflow discharge from the breach and determine the worst-case that leads to accelerating the dam breaching process.

    2. Numerical simulation

    The current study for spatial dam breach is simulated by using (FLOW-3D) software [27], which is a powerful computational fluid dynamics (CFD) program.

    2.1. Geometric presentations

    A stereolithographic (STL) file is prepared for each change in the initial breach geometry and dimensions. The CAD program is useful for creating solid objects and converting them to STL format, as shown in Fig. 1.

    2.2. Governing equations

    The governing equations for water flow are three-dimensional Reynolds Averaged Navier-Stokes equations (RANS).

    The continuity equation:(2)∂ui∂xi=0

    The momentum equation:(3)∂ui∂t+1VFuj∂ui∂xj=1ρ∂∂xj-pδij+ν∂ui∂xj+∂uj∂xi-ρu`iu`j¯

    where u is time-averaged velocity,ν is kinematic viscosity, VF is fractional volume open to flow, p is averaged pressure and -u`iu`j¯ are components of Reynold’s stress. The Volume of Fluid (VOF) technique is used to simulate the free surface profile. Hirt et al. [28] presented the VOF algorithm, which employs the function (F) to express the occupancy of each grid cell with fluid. The value of (F) varies from zero to unity. Zero value refers to no fluid in the grid cell, while the unity value refers to the grid cell being fully occupied with fluid. The free surface is formed in the grid cells having (F) values between zero and unity.(4)∂F∂t+1VF∂∂xFAxu+∂∂yFAyv+∂∂zFAzw=0

    where (u, v, w) are the velocity components in (x, y, z) coordinates, respectively, and (AxAyAz) are the area fractions.

    2.3. Boundary and initial conditions

    To improve the accuracy of the results, the boundary conditions should be carefully determined. In this study, two mesh blocks are used to minimize the time consumed in the simulation. The boundary conditions for mesh block 1 are as follows: The inlet and sides boundaries are defined as a wall boundary condition (wall boundary condition is usually used for bound fluid by solid regions. In the case of viscous flows, no-slip means that the tangential velocity is equal to the wall velocity and the normal velocity is zero), the outlet is defined as a symmetry boundary condition (symmetry boundary condition is usually used to reduce computational effort during CFD simulation. This condition allows the flow to be transferred from one mesh block to another. No inputs are required for this boundary condition except that its location should be defined accurately), the bottom boundary is defined as a uniform flow rate boundary condition, and the top boundary is defined as a specific pressure boundary condition with assigned atmospheric pressure. The boundary conditions for mesh block 2 are as follows: The inlet is defined as a symmetry boundary condition, the outlet is defined as a free flow boundary condition, the bottom and sides boundaries are defined as a wall boundary condition, and the top boundary is defined as a specific pressure boundary condition with assigned atmospheric pressure as shown in Fig. 2. The initial conditions required to be set for the fluid (i.e., water) inside of the domain include configuration, temperature, velocities, and pressure distribution. The configuration of water depends on the dimensions and shape of the dam reservoir. While the other conditions have been assigned as follows: temperature is normal water temperature (25 °c) and pressure distribution is hydrostatic with no initial velocity.

    2.4. Numerical method

    FLOW-3D uses the finite volume method (FVM) to solve the governing equation (Reynolds-averaged Navier-Stokes) over the computational domain. A finite-volume method is an Eulerian approach for representing and evaluating partial differential equations in algebraic equations form [29]. At discrete points on the mesh geometry, values are determined. Finite volume expresses a small volume surrounding each node point on a mesh. In this method, the divergence theorem is used to convert volume integrals with a divergence term to surface integrals. After that, these terms are evaluated as fluxes at each finite volume’s surfaces.

    2.5. Turbulent models

    Turbulence is the chaotic, unstable motion of fluids that occurs when there are insufficient stabilizing viscous forces. In FLOW-3D, there are six turbulence models available: the Prandtl mixing length model, the one-equation turbulent energy model, the two-equation (k – ε) model, the Renormalization-Group (RNG) model, the two-equation (k – ω) models, and a large eddy simulation (LES) model. For simulating flow motion, the RNG model is adopted to simulate the motion behavior better than the k – ε and k – ω.

    models [30]. The RNG model consists of two main equations for the turbulent kinetic energy KT and its dissipation.εT(5)∂kT∂t+1VFuAx∂kT∂x+vAy∂kT∂y+wAz∂kT∂z=PT+GT+DiffKT-εT(6)∂εT∂t+1VFuAx∂εT∂x+vAy∂εT∂y+wAz∂εT∂z=C1.εTKTPT+c3.GT+Diffε-c2εT2kT

    where KT is the turbulent kinetic energy, PT is the turbulent kinetic energy production, GT is the buoyancy turbulence energy, εT is the turbulent energy dissipation rate, DiffKT and Diffε are terms of diffusion, c1, c2 and c3 are dimensionless parameters, in which c1 and c3 have a constant value of 1.42 and 0.2, respectively, c2 is computed from the turbulent kinetic energy (KT) and turbulent production (PT) terms.

    2.6. Sediment scour model

    The sediment scour model available in FLOW-3D can calculate all the sediment transport processes including Entrainment transport, Bedload transport, Suspended transport, and Deposition. The erosion process starts once the water flows remove the grains from the packed bed and carry them into suspension. It happens when the applied shear stress by water flows exceeds critical shear stress. This process is represented by entrainment transport in the numerical model. After entrained, the grains carried by water flow are represented by suspended load transport. After that, some suspended grains resort to settling because of the combined effect of gravity, buoyancy, and friction. This process is described through a deposition. Finally, the grains sliding motions are represented by bedload transport in the model. For the entrainment process, the shear stress applied by the fluid motion on the packed bed surface is calculated using the standard wall function as shown in Eq.7.(7)ks,i=Cs,i∗d50

    where ks,i is the Nikuradse roughness and Cs,i is a user-defined coefficient. The critical bed shear stress is defined by a dimensionless parameter called the critical shields number as expressed in Eq.8.(8)θcr,i=τcr,i‖g‖diρi-ρf

    where θcr,i is the critical shields number, τcr,i is the critical bed shear stress, g is the absolute value of gravity acceleration, di is the diameter of the sediment grain, ρi is the density of the sediment species (i) and ρf is the density of the fluid. The value of the critical shields number is determined according to the Soulsby-Whitehouse equation.(9)θcr,i=0.31+1.2d∗,i+0.0551-exp-0.02d∗,i

    where d∗,i is the dimensionless diameter of the sediment, given by Eq.10.(10)d∗,i=diρfρi-ρf‖g‖μf213

    where μf is the fluid dynamic viscosity. For the sloping bed interface, the value of the critical shields number is modified according to Eq.11.(11)θ`cr,i=θcr,icosψsinβ+cos2βtan2φi-sin2ψsin2βtanφi

    where θ`cr,i is the modified critical shields number, φi is the angle of repose for the sediment, β is the angle of bed slope and ψ is the angle between the flow and the upslope direction. The effects of the rolling, hopping, and sliding motions of grains along the packed bed surface are taken by the bedload transport process. The volumetric bedload transport rate (qb,i) per width of the bed is expressed in Eq.12.(12)qb,i=Φi‖g‖ρi-ρfρfdi312

    where Φi is the dimensionless bedload transport rate is calculated by using Meyer Peter and Müller equation.(13)Φi=βMPM,iθi-θ`cr,i1.5cb,i

    where βMPM,i is the Meyer Peter and Müller user-defined coefficient and cb,i is the volume fraction of species i in the bed material. The suspended load transport is calculated as shown in Eq.14.(14)∂Cs,i∂t+∇∙Cs,ius,i=∇∙∇DCs,i

    where Cs,i is the suspended sediment mass concentration, D is the diffusivity, and us,i is the grain velocity of species i. Entrainment and deposition are two opposing processes that take place at the same time. The lifting and settling velocities for both entrainment and deposition processes are calculated according to Eq.15 and Eq.16, respectively.(15)ulifting,i=αid∗,i0.3θi-θ`cr,igdiρiρf-1(16)usettling,i=υfdi10.362+1.049d∗,i3-10.36

    where αi is the entrainment coefficient of species i and υf is the kinematic viscosity of the fluid.

    2.7. Grid type

    Using simple rectangular orthogonal elements in planes and hexahedral in volumes in the (FLOW-3D) program makes the mesh generation process easier, decreases the required memory, and improves numerical accuracy. Two mesh blocks were used in a joined form with a size ratio of 2:1. The first mesh block is coarser, which contains the reservoir water, and the second mesh block is finer, which contains the dam. For achieving accuracy and efficiency in results, the mesh size is determined by using a grid convergence test. The optimum uniform cell size for the first mesh block is 0.012 m and for the second mesh block is 0.006 m.

    2.8. Time step

    The maximum time step size is determined by using a Courant number, which controls the distance that the flow will travel during the simulation time step. In this study, the Courant number was taken equal to 0.25 to prevent the flow from traveling through more than one cell in the time step. Based on the Courant number, a maximum time step value of 0.00075 s was determined.

    2.9. Numerical model validation

    The numerical model accuracy was achieved by comparing the numerical model results with previous experimental results. The experimental study of Schmocker and Hager [7] was based on 31 tests with changes in six parameters (d50, Ho, Bo, Lk, XD, and Qin). All experimental tests were conducted in a straight open glass-sided flume. The horizontal flume has a rectangular cross-section with a width of 0.4 m and a height of 0.7 m. The flume was provided with a flow straightener and an intake with a length of 0.66 m. All tested dams were inserted at various distances (XD) from the intake. Test No.1 from this experimental program was chosen to validate the numerical model. The different parameters used in test No.1 are as follows:

    (1) uniform sediment with a mean diameter (d50 = 0.31 mm), (2) Ho = 0.2 m, (3) Bo = 0.2 m, (4) Lk = 0.1 m,

    (5) XD = 1.0 m, (6) Qin = 6.0 lit/s, (7) Su and Sd = 2:1, (8) mass density (ρs = 2650 kg/m3(9) Homogenous and non-cohesive embankment dam. As shown in Fig. 2, the simulation is contained within a rectangular grid with dimensions: 3.56 m in the x-direction (where 0.66 m is used as inlet, 0.9 m as dam base width, and 1.0 m as outlet), in y-direction 0.2 m (dam length), and in the z-direction 0.3 m, which represents the dam height (0.2 m) with a free distance (0.1 m) above the dam. There are two main reasons that this experimental program is preferred for the validation process. The first reason is that this program deals with homogenous, non-cohesive soil, which is available in FLOW-3D. The second reason is that this program deals with small-scale models which saves time for numerical simulation. Finally, some important assumptions were considered during the validation process. The flow is assumed to be incompressible, viscous, turbulent, and three-dimensional.

    By comparing dam profiles at different time instants for the experimental test with the current numerical model, it appears that the numerical model gives good agreement as shown in Fig. 3 and Fig. 4, with an average error percentage of 9% between the experimental results and the numerical model.

    3. Analysis and discussions

    The current model is used to study the effects of different parameters such as (initial breach shapes, dimensions, locations, upstream and downstream dam slopes) on the peak outflow discharge, QP, time of peak outflow, tP, and rate of erosion, E.

    This study consists of a group of scenarios. The first scenario is changing the shapes of the initial breach according to Singh [1], the most predicted shapes are rectangular and V-notch as shown in Fig. 5. The second scenario is changing the initial breach dimensions (i.e., width and depth). While the third scenario is changing the location of the initial breach. Eventually, the last scenario is changing the upstream and downstream dam slopes.

    All scenarios of this study were carried out under the same conditions such as inflow discharge value (Qin=1.0lit/s), dimensions of the tested dam, where dam height (Ho=0.20m), crest width.

    (Lk=0.1m), dam length (Bo=0.20m), and homogenous & non-cohesive soil with a mean diameter (d50=0.31mm).

    3.1. Dam breaching process evolution

    The dam breaching process is a very complex process due to the quick changes in hydrodynamic conditions during dam failure. The dam breaching process starts once water flows reach the downstream face of the dam. During the initial stage of dam breaching, the erosion process is relatively quiet due to low velocities of flow. As water flows continuously, erosion rates increase, especially in two main zones: the crest and the downstream face. As soon as the dam crest is totally eroded, the water levels in the dam reservoir decrease rapidly, accompanied by excessive erosion in the dam body. The erosion process continues until the water levels in the dam reservoir equal the remaining height of the dam.

    According to Zhou et al. [11], the breaching process consists of three main stages. The first stage starts with beginning overtopping flow, then ends when the erosion point directed upstream and reached the inflection point at the inflection time (ti). The second stage starts from the end of the stage1 until the occurrence of peak outflow discharge at the peak outflow time (tP). The third stage starts from the end of the stage2 until the value of outflow discharge becomes the same as the value of inflow discharge at the final time (tf). The outflow discharge from the dam breach increases rapidly during stage1 and stage2 because of the large dam storage capacity (i.e., the dam reservoir is totally full of water) and excessive erosion. While at stage3, the outflow values start to decrease slowly because most of the dam’s storage capacity was run out. The end of stage3 indicates that the dam storage capacity was totally run out, so the outflow equalized with the inflow discharge as shown in Fig. 6 and Fig. 7.

    3.2. The effect of initial breach shape

    To identify the effect of the initial breach shape on the evolution of the dam breaching process. Three tests were carried out with different cross-section areas for each shape. The initial breach is created at the center of the dam crest. Each test had an ID to make the process of arranging data easier. The rectangular shape had an ID (Rec5h & 5b), which means that its depth and width are equal to 5% of the dam height, and the V-notch shape had an ID (V-noch5h & 1:1) which means that its depth is equal to 5% of the dam height and its side slope is equal to 1:1. The comparison between rectangular and V-notch shapes is done by calculating the ratio between maximum dam height at different times (ZMax) to the initial dam height (Ho), rate of erosion, and hydrograph of outflow discharge for each test. The rectangular shape achieves maximum erosion rate and minimum inflection time, in addition to a rapid decrease in the dam reservoir levels. Therefore, the dam breaching is faster in the case of a rectangular shape than in a V-notch shape, which has the same cross-section area as shown in Fig. 8.

    Also, by comparing the hydrograph for each test, the peak outflow discharge value in the case of a rectangular shape is higher than the V-notch shape by 5% and the time of peak outflow for the rectangular shape is shorter than the V-notch shape by 9% as shown in Fig. 9.

    3.3. The effect of initial breach dimensions

    The results of the comparison between the different initial breach shapes indicate that the worst initial breach shape is rectangular, so the second scenario from this study concentrated on studying the effect of a change in the initial rectangular breach dimensions. Groups of tests were carried out with different depths and widths for the rectangular initial breach. The first group had a depth of 5% from the dam height and with three different widths of 5,10, and 15% from the dam height, the second group had a depth of 10% with three different widths of 5,10, and 15%, the third group had a depth of 15% with three different widths of 5,10, and 15% and the final group had a width of 15% with three different heights of 5, 10, and 15% for a rectangular breach shape. The comparison was made as in the previous section to determine the worst case that leads to the quick dam failure as shown in Fig. 10.

    The results show that the (Rec 5 h&15b) test achieves a maximum erosion rate for a shorter period of time and a minimum ratio for (Zmax / Ho) as shown in Fig. 10, which leads to accelerating the dam failure process. The dam breaching process is faster with the minimum initial breach depth and maximum initial breach width. In the case of a minimum initial breach depth, the retained head of water in the dam reservoir is high and the crest width at the bottom of the initial breach (L`K) is small, so the erosion point reaches the inflection point rapidly. While in the case of the maximum initial breach width, the erosion perimeter is large.

    3.4. The effect of initial breach location

    The results of the comparison between the different initial rectangular breach dimensions indicate that the worst initial breach dimension is (Rec 5 h&15b), so the third scenario from this study concentrated on studying the effect of a change in the initial breach location. Three locations were checked to determine the worst case for the dam failure process. The first location is at the center of the dam crest, which was named “Center”, the second location is at mid-distance between the dam center and dam edge, which was named “Mid”, and the third location is at the dam edge, which was named “Edge” as shown in Fig. 11. According to this scenario, the results indicate that the time of peak outflow discharge (tP) is the same in the three cases, but the maximum value of the peak outflow discharge occurs at the center location. The difference in the peak outflow values between the three cases is relatively small as shown in Fig. 12.

    The rates of erosion were also studied for the three cases. The results show that the maximum erosion rate occurs at the center location as shown in Fig. 13. By making a comparison between the three cases for the dam storage volume. The results show that the center location had the minimum values for the dam storage volume, which means that a large amount of water has passed to the downstream area as shown in Fig. 14. According to these results, the center location leads to increased erosion rate and accelerated dam failure process compared with the two other cases. Because the erosion occurs on both sides, but in the case of edge location, the erosion occurs on one side.

    3.5. The effect of upstream and downstream dam slopes

    The results of the comparison between the different initial rectangular breach locations indicate that the worst initial breach location is the center location, so the fourth scenario from this study concentrated on studying the effect of a change in the upstream (Su) and downstream (Sd) dam slopes. Three slopes were checked individually for both upstream and downstream slopes to determine the worst case for the dam failure process. The first slope value is (2H:1V), the second slope value is (2.5H:1V), and the third slope value is (3H:1V). According to this scenario, the results show that the decreasing downstream slope angle leads to increasing time of peak outflow discharge (tP) and decreasing value of peak outflow discharge. The difference in the peak outflow values between the three cases for the downstream slope is 2%, as shown in Fig. 15, but changing the upstream slope has a negligible impact on the peak outflow discharge and its time as shown in Fig. 16.

    The rates of erosion were also studied in the three cases for both upstream and downstream slopes. The results show that the maximum erosion rate increases by 6.0% with an increasing downstream slope angle by 4°, as shown in Fig. 17. The results also indicate that the erosion rates aren’t affected by increasing or decreasing the upstream slope angle, as shown in Fig. 18. According to these results, increasing the downstream slope angle leads to increased erosion rate and accelerated dam failure process compared with the upstream slope angle. Because of increasing shear stress applied by water flows in case of increasing downstream slope.

    According to all previous scenarios, the dimensionless peak outflow discharge QPQin is presented for a fixed dam height (Ho) and inflow discharge (Qin). Fig. 19 illustrates the relationship between QP∗=QPQin and.

    Lr=ho2/3∗bo2/3Ho. The deduced relationship achieves R2=0.96.(17)QP∗=2.2807exp-2.804∗Lr

    4. Conclusions

    A spatial dam breaching process was simulated by using FLOW-3D Software. The validation process was performed by making a comparison between the simulated results of dam profiles and the dam profiles obtained by Schmocker and Hager [7] in their experimental study. And also, the peak outflow value recorded an error percentage of 12% between the numerical model and the experimental study. This model was used to study the effect of initial breach shape, dimensions, location, and dam slopes on peak outflow discharge, time of peak outflow, and the erosion process. By using the parameters obtained from the validation process, the results of this study can be summarized in eight points as follows.1.

    The rectangular initial breach shape leads to an accelerating dam failure process compared with the V-notch.2.

    The value of peak outflow discharge in the case of a rectangular initial breach is higher than the V-notch shape by 5%.3.

    The time of peak outflow discharge for a rectangular initial breach is shorter than the V-notch shape by 9%.4.

    The minimum depth and maximum width for the initial breach achieve maximum erosion rates (increasing breach width, b0, or decreasing breach depth, h0, by 5% from the dam height leads to an increase in the maximum rate of erosion by 11% and 15%, respectively), so the dam failure is rapid.5.

    The center location of the initial breach leads to an accelerating dam failure compared with the edge location.6.

    The initial breach location has a negligible effect on the peak outflow discharge value and its time.7.

    Increasing the downstream slope angle by 4° leads to an increase in both peak outflow discharge and maximum rate of erosion by 2.0% and 6.0%, respectively.8.

    The upstream slope has a negligible effect on the dam breaching process.


    [1]V. SinghDam breach modeling technologySpringer Science & Business Media (1996)Google Scholar[2]Wahl TL. Prediction of embankment dam breach parameters: a literature review and needs assessment. 1998.Google Scholar[3]Z. Alhasan, J. Jandora, J. ŘíhaStudy of dam-break due to overtopping of four small dams in the Czech RepublicActa Universitatis Agriculturae et Silviculturae Mendelianae Brunensis, 63 (3) (2015), pp. 717-729 View PDFCrossRefView Record in ScopusGoogle Scholar[4]D. FreadBREACH, an erosion model for earthen dam failures: Hydrologic Research LaboratoryNOAA, National Weather Service (1988)Google Scholar[5]J. Říha, S. Kotaška, L. PetrulaDam Break Modeling in a Cascade of Small Earthen Dams: Case Study of the Čižina River in the Czech RepublicWater, 12 (8) (2020), p. 2309, 10.3390/w12082309 View PDFView Record in ScopusGoogle Scholar[6]E. Goodarzi, L. Teang Shui, M. ZiaeiDam overtopping risk using probabilistic concepts–Case study: The Meijaran DamIran Ain Shams Eng J, 4 (2) (2013), pp. 185-197ArticleDownload PDFView Record in ScopusGoogle Scholar[7]L. Schmocker, W.H. HagerPlane dike-breach due to overtopping: effects of sediment, dike height and dischargeJ Hydraul Res, 50 (6) (2012), pp. 576-586 View PDFCrossRefView Record in ScopusGoogle Scholar[8]J.S. Walder, R.M. Iverson, J.W. Godt, M. Logan, S.A. SolovitzControls on the breach geometry and flood hydrograph during overtopping of noncohesive earthen damsWater Resour Res, 51 (8) (2015), pp. 6701-6724View Record in ScopusGoogle Scholar[9]H. Wei, M. Yu, D. Wang, Y. LiOvertopping breaching of river levees constructed with cohesive sedimentsNat Hazards Earth Syst Sci, 16 (7) (2016), pp. 1541-1551 View PDFCrossRefView Record in ScopusGoogle Scholar[10]Y. Yang, S.-Y. Cao, K.-J. Yang, W.-P. LiYang K-j, Li W-p. Experimental study of breach process of landslide dams by overtopping and its initiation mechanismsJ Hydrodynamics, 27 (6) (2015), pp. 872-883ArticleDownload PDFCrossRefView Record in ScopusGoogle Scholar[11]G.G.D. Zhou, M. Zhou, M.S. Shrestha, D. Song, C.E. Choi, K.F.E. Cui, et al.Experimental investigation on the longitudinal evolution of landslide dam breaching and outburst floodsGeomorphology, 334 (2019), pp. 29-43ArticleDownload PDFView Record in ScopusGoogle Scholar[12]J. Zhang, Z.-x. Guo, S.-y. CaoYang F-g. Experimental study on scour and erosion of blocked damWater Sci Eng, 5 (2012), pp. 219-229ArticleDownload PDFView Record in ScopusGoogle Scholar[13]K. Höeg, A. Løvoll, K. VaskinnStability and breaching of embankment dams: Field tests on 6 m high damsInt J Hydropower Dams, 11 (2004), pp. 88-92View Record in ScopusGoogle Scholar[14]H. Hakimzadeh, V. Nourani, A.B. AminiGenetic programming simulation of dam breach hydrograph and peak outflow dischargeJ Hydrol Eng, 19 (4) (2014), pp. 757-768View Record in ScopusGoogle Scholar[15]A.R. Refaiy, N.M. AboulAtta, N.Y. Saad, D.A. El-MollaModeling the effect of downstream drain geometry on seepage through earth damsAin Shams Eng J, 12 (3) (2021), pp. 2511-2531ArticleDownload PDFView Record in ScopusGoogle Scholar[16]Y. Zhu, P.J. Visser, J.K. Vrijling, G. WangExperimental investigation on breaching of embankmentsScience China Technological Sci, 54 (1) (2011), pp. 148-155 View PDFCrossRefView Record in ScopusGoogle Scholar[17]M.-H. Yu, H.-Y. Wei, Y.-J. Liang, Y. ZhaoInvestigation of non-cohesive levee breach by overtopping flowJ Hydrodyn, 25 (4) (2013), pp. 572-579ArticleDownload PDFCrossRefView Record in ScopusGoogle Scholar[18]S. Wu, M. Yu, H. Wei, Y. Liang, J. ZengNon-symmetrical levee breaching processes in a channel bend due to overtoppingInt J Sedim Res, 33 (2) (2018), pp. 208-215ArticleDownload PDFView Record in ScopusGoogle Scholar[19]O. Saberi, G. ZenzNumerical investigation on 1D and 2D embankment dams failure due to overtopping flowInt J Hydraulic Engineering, 5 (2016), pp. 9-18View Record in ScopusGoogle Scholar[20]M. Guan, N.G. Wright, P.A. Sleigh2D Process-Based Morphodynamic Model for Flooding by Noncohesive Dyke BreachJ Hydraul Eng, 140 (7) (2014), p. 04014022, 10.1061/(ASCE)HY.1943-7900.0000861 View PDFView Record in ScopusGoogle Scholar[21]W. Wu, R. Marsooli, Z. HeDepth-Averaged Two-Dimensional Model of Unsteady Flow and Sediment Transport due to Noncohesive Embankment Break/BreachingJ Hydraul Eng, 138 (6) (2012), pp. 503-516View Record in ScopusGoogle Scholar[22]Z. Wang, D.S. BowlesThree-dimensional non-cohesive earthen dam breach model. Part 1: Theory and methodologyAdv Water Resour, 29 (10) (2006), pp. 1528-1545ArticleDownload PDFView Record in ScopusGoogle Scholar[23]Říha J, Duchan D, Zachoval Z, Erpicum S, Archambeau P, Pirotton M, et al. Performance of a shallow-water model for simulating flow over trapezoidal broad-crested weirs. J Hydrology Hydromechanics. 2019;67:322-8.Google Scholar[24]C.B. VreugdenhilNumerical methods for shallow-water flowSpringer Science & Business Media (1994)Google Scholar[25]L.A. Larocque, J. Imran, M.H. Chaudhry3D numerical simulation of partial breach dam-break flow using the LES and k–∊ turbulence modelsJ Hydraul Res, 51 (2) (2013), pp. 145-157 View PDFCrossRefView Record in ScopusGoogle Scholar[26]C. Yang, B. Lin, C. Jiang, Y. LiuPredicting near-field dam-break flow and impact force using a 3D modelJ Hydraul Res, 48 (6) (2010), pp. 784-792 View PDFCrossRefView Record in ScopusGoogle Scholar[27]FLOW-3D. Version 11.1.1 Flow Science, Inc., Santa Fe, NM. https://wwwflow3dcom.Google Scholar[28]C.W. Hirt, B.D. NicholsVolume of fluid (VOF) method for the dynamics of free boundariesJ Comput Phys, 39 (1) (1981), pp. 201-225ArticleDownload PDFGoogle Scholar[29]S.V. PatankarNumerical heat transfer and fluid flow, Hemisphere PublCorp, New York, 58 (1980), p. 288View Record in ScopusGoogle Scholar[30]M. Alemi, R. MaiaNumerical simulation of the flow and local scour process around single and complex bridge piersInt J Civil Eng, 16 (5) (2018), pp. 475-487 View PDFCrossRefView Record in ScopusGoogle Scholar

    Figure 3 Simulation PTC pipes enhanced with copper foam and nanoparticles in FLOW-3D software.

    다공성 미디어 및 나노유체에 의해 강화된 수집기로 태양광 CCHP 시스템의 최적화

    Optimization of Solar CCHP Systems with Collector Enhanced by Porous Media and Nanofluid

    Navid Tonekaboni,1Mahdi Feizbahr,2 Nima Tonekaboni,1Guang-Jun Jiang,3,4 and Hong-Xia Chen3,4


    태양열 집열기의 낮은 효율은 CCHP(Solar Combined Cooling, Heating, and Power) 사이클의 문제점 중 하나로 언급될 수 있습니다. 태양계를 개선하기 위해 나노유체와 다공성 매체가 태양열 집열기에 사용됩니다.

    다공성 매질과 나노입자를 사용하는 장점 중 하나는 동일한 조건에서 더 많은 에너지를 흡수할 수 있다는 것입니다. 이 연구에서는 평균 일사량이 1b인 따뜻하고 건조한 지역의 600 m2 건물의 전기, 냉방 및 난방을 생성하기 위해 다공성 매질과 나노유체를 사용하여 태양열 냉난방 복합 발전(SCCHP) 시스템을 최적화했습니다.

    본 논문에서는 침전물이 형성되지 않는 lb = 820 w/m2(이란) 정도까지 다공성 물질에서 나노유체의 최적량을 계산하였다. 이 연구에서 태양열 집열기는 구리 다공성 매체(95% 다공성)와 CuO 및 Al2O3 나노 유체로 향상되었습니다.

    나노유체의 0.1%-0.6%가 작동 유체로 물에 추가되었습니다. 나노유체의 0.5%가 태양열 집열기 및 SCCHP 시스템에서 가장 높은 에너지 및 엑서지 효율 향상으로 이어지는 것으로 밝혀졌습니다.

    본 연구에서 포물선형 집열기(PTC)의 최대 에너지 및 엑서지 효율은 각각 74.19% 및 32.6%입니다. 그림 1은 태양 CCHP의 주기를 정확하게 설명하기 위한 그래픽 초록으로 언급될 수 있습니다.

    The low efficiency of solar collectors can be mentioned as one of the problems in solar combined cooling, heating, and power (CCHP) cycles. For improving solar systems, nanofluid and porous media are used in solar collectors. One of the advantages of using porous media and nanoparticles is to absorb more energy under the same conditions. In this research, a solar combined cooling, heating, and power (SCCHP) system has been optimized by porous media and nanofluid for generating electricity, cooling, and heating of a 600 m2 building in a warm and dry region with average solar radiation of Ib = 820 w/m2 in Iran. In this paper, the optimal amount of nanofluid in porous materials has been calculated to the extent that no sediment is formed. In this study, solar collectors were enhanced with copper porous media (95% porosity) and CuO and Al2O3 nanofluids. 0.1%–0.6% of the nanofluids were added to water as working fluids; it is found that 0.5% of the nanofluids lead to the highest energy and exergy efficiency enhancement in solar collectors and SCCHP systems. Maximum energy and exergy efficiency of parabolic thermal collector (PTC) riches in this study are 74.19% and 32.6%, respectively. Figure 1 can be mentioned as a graphical abstract for accurately describing the cycle of solar CCHP.

    1. Introduction

    Due to the increase in energy consumption, the use of clean energy is one of the important goals of human societies. In the last four decades, the use of cogeneration cycles has increased significantly due to high efficiency. Among clean energy, the use of solar energy has become more popular due to its greater availability [1]. Low efficiency of energy production, transmission, and distribution system makes a new system to generate simultaneously electricity, heating, and cooling as an essential solution to be widely used. The low efficiency of the electricity generation, transmission, and distribution system makes the CCHP system a basic solution to eliminate waste of energy. CCHP system consists of a prime mover (PM), a power generator, a heat recovery system (produce extra heating/cooling/power), and thermal energy storage (TES) [2]. Solar combined cooling, heating, and power (SCCHP) has been started three decades ago. SCCHP is a system that receives its propulsive force from solar energy; in this cycle, solar collectors play the role of propulsive for generating power in this system [3].

    Increasing the rate of energy consumption in the whole world because of the low efficiency of energy production, transmission, and distribution system causes a new cogeneration system to generate electricity, heating, and cooling energy as an essential solution to be widely used. Building energy utilization fundamentally includes power required for lighting, home electrical appliances, warming and cooling of building inside, and boiling water. Domestic usage contributes to an average of 35% of the world’s total energy consumption [4].

    Due to the availability of solar energy in all areas, solar collectors can be used to obtain the propulsive power required for the CCHP cycle. Solar energy is the main source of energy in renewable applications. For selecting a suitable area to use solar collectors, annual sunshine hours, the number of sunny days, minus temperature and frosty days, and the windy status of the region are essentially considered [5]. Iran, with an average of more than 300 sunny days, is one of the suitable countries to use solar energy. Due to the fact that most of the solar radiation is in the southern regions of Iran, also the concentration of cities is low in these areas, and transmission lines are far apart, one of the best options is to use CCHP cycles based on solar collectors [6]. One of the major problems of solar collectors is their low efficiency [7]. Low efficiency increases the area of collectors, which increases the initial cost of solar systems and of course increases the initial payback period. To increase the efficiency of solar collectors and improve their performance, porous materials and nanofluids are used to increase their workability.

    There are two ways to increase the efficiency of solar collectors and mechanical and fluid improvement. In the first method, using porous materials or helical filaments inside the collector pipes causes turbulence of the flow and increases heat transfer. In the second method, using nanofluids or salt and other materials increases the heat transfer of water. The use of porous materials has grown up immensely over the past twenty years. Porous materials, especially copper porous foam, are widely used in solar collectors. Due to the high contact surface area, porous media are appropriate candidates for solar collectors [8]. A number of researchers investigated Solar System performance in accordance with energy and exergy analyses. Zhai et al. [9] reviewed the performance of a small solar-powered system in which the energy efficiency was 44.7% and the electrical efficiency was 16.9%.

    Abbasi et al. [10] proposed an innovative multiobjective optimization to optimize the design of a cogeneration system. Results showed the CCHP system based on an internal diesel combustion engine was the applicable alternative at all regions with different climates. The diesel engine can supply the electrical requirement of 31.0% and heating demand of 3.8% for building.

    Jiang et al. [11] combined the experiment and simulation together to analyze the performance of a cogeneration system. Moreover, some research focused on CCHP systems using solar energy. It integrated sustainable and renewable technologies in the CCHP, like PV, Stirling engine, and parabolic trough collector (PTC) [21215].

    Wang et al. [16] optimized a cogeneration solar cooling system with a Rankine cycle and ejector to reach the maximum total system efficiency of 55.9%. Jing et al. analyzed a big-scale building with the SCCHP system and auxiliary heaters to produced electrical, cooling, and heating power. The maximum energy efficiency reported in their work is 46.6% [17]. Various optimization methods have been used to improve the cogeneration system, minimum system size, and performance, such as genetic algorithm [1819].

    Hirasawa et al. [20] investigated the effect of using porous media to reduce thermal waste in solar systems. They used the high-porosity metal foam on top of the flat plate solar collector and observed that thermal waste decreased by 7% due to natural heat transfer. Many researchers study the efficiency improvement of the solar collector by changing the collector’s shapes or working fluids. However, the most effective method is the use of nanofluids in the solar collector as working fluid [21]. In the experimental study done by Jouybari et al. [22], the efficiency enhancement up to 8.1% was achieved by adding nanofluid in a flat plate collector. In this research, by adding porous materials to the solar collector, collector efficiency increased up to 92% in a low flow regime. Subramani et al. [23] analyzed the thermal performance of the parabolic solar collector with Al2O3 nanofluid. They conducted their experiments with Reynolds number range 2401 to 7202 and mass flow rate 0.0083 to 0.05 kg/s. The maximum efficiency improvement in this experiment was 56% at 0.05 kg/s mass flow rate.

    Shojaeizadeh et al. [24] investigated the analysis of the second law of thermodynamic on the flat plate solar collector using Al2O3/water nanofluid. Their research showed that energy efficiency rose up to 1.9% and the exergy efficiency increased by a maximum of 0.72% compared to pure water. Tiwari et al. [25] researched on the thermal performance of solar flat plate collectors for working fluid water with different nanofluids. The result showed that using 1.5% (optimum) particle volume fraction of Al2O3 nanofluid as an absorbing medium causes the thermal efficiency to enhance up to 31.64%.

    The effect of porous media and nanofluids on solar collectors has already been investigated in the literature but the SCCHP system with a collector embedded by both porous media and nanofluid for enhancing the ratio of nanoparticle in nanofluid for preventing sedimentation was not discussed. In this research, the amount of energy and exergy of the solar CCHP cycles with parabolic solar collectors in both base and improved modes with a porous material (copper foam with 95% porosity) and nanofluid with different ratios of nanoparticles was calculated. In the first step, it is planned to design a CCHP system based on the required load, and, in the next step, it will analyze the energy and exergy of the system in a basic and optimize mode. In the optimize mode, enhanced solar collectors with porous material and nanofluid in different ratios (0.1%–0.7%) were used to optimize the ratio of nanofluids to prevent sedimentation.

    2. Cycle Description

    CCHP is one of the methods to enhance energy efficiency and reduce energy loss and costs. The SCCHP system used a solar collector as a prime mover of the cogeneration system and assisted the boiler to generate vapor for the turbine. Hot water flows from the expander to the absorption chiller in summer or to the radiator or fan coil in winter. Finally, before the hot water wants to flow back to the storage tank, it flows inside a heat exchanger for generating domestic hot water [26].

    For designing of solar cogeneration system and its analysis, it is necessary to calculate the electrical, heating (heating load is the load required for the production of warm water and space heating), and cooling load required for the case study considered in a residential building with an area of 600 m2 in the warm region of Iran (Zahedan). In Table 1, the average of the required loads is shown for the different months of a year (average of electrical, heating, and cooling load calculated with CARRIER software).Table 1 The average amount of electric charges, heating load, and cooling load used in the different months of the year in the city of Zahedan for a residential building with 600 m2.

    According to Table 1, the maximum magnitude of heating, cooling, and electrical loads is used to calculate the cogeneration system. The maximum electric load is 96 kW, the maximum amount of heating load is 62 kW, and the maximum cooling load is 118 kW. Since the calculated loads are average, all loads increased up to 10% for the confidence coefficient. With the obtained values, the solar collector area and other cogeneration system components are calculated. The cogeneration cycle is capable of producing 105 kW electric power, 140 kW cooling capacity, and 100 kW heating power.

    2.1. System Analysis Equations

    An analysis is done by considering the following assumptions:(1)The system operates under steady-state conditions(2)The system is designed for the warm region of Iran (Zahedan) with average solar radiation Ib = 820 w/m2(3)The pressure drops in heat exchangers, separators, storage tanks, and pipes are ignored(4)The pressure drop is negligible in all processes and no expectable chemical reactions occurred in the processes(5)Potential, kinetic, and chemical exergy are not considered due to their insignificance(6)Pumps have been discontinued due to insignificance throughout the process(7)All components are assumed adiabatic

    Schematic shape of the cogeneration cycle is shown in Figure 1 and all data are given in Table 2.

    Figure 1 Schematic shape of the cogeneration cycle.Table 2 Temperature and humidity of different points of system.

    Based on the first law of thermodynamic, energy analysis is based on the following steps.

    First of all, the estimated solar radiation energy on collector has been calculated:where α is the heat transfer enhancement coefficient based on porous materials added to the collector’s pipes. The coefficient α is increased by the porosity percentage, the type of porous material (in this case, copper with a porosity percentage of 95), and the flow of fluid to the collector equation.

    Collector efficiency is going to be calculated by the following equation [9]:

    Total energy received by the collector is given by [9]

    Also, the auxiliary boiler heat load is [2]

    Energy consumed from vapor to expander is calculated by [2]

    The power output form by the screw expander [9]:

    The efficiency of the expander is 80% in this case [11].

    In this step, cooling and heating loads were calculated and then, the required heating load to reach sanitary hot water will be calculated as follows:

    First step: calculating the cooling load with the following equation [9]:

    Second step: calculating heating loads [9]:

    Then, calculating the required loud for sanitary hot water will be [9]

    According to the above-mentioned equations, efficiency is [9]

    In the third step, calculated exergy analysis as follows.

    First, the received exergy collector from the sun is calculated [9]:

    In the previous equation, f is the constant of air dilution.

    The received exergy from the collector is [9]

    In the case of using natural gas in an auxiliary heater, the gas exergy is calculated from the following equation [12]:

    Delivering exergy from vapor to expander is calculated with the following equation [9]:

    In the fourth step, the exergy in cooling and heating is calculated by the following equation:

    Cooling exergy in summer is calculated [9]:

    Heating exergy in winter is calculated [9]:

    In the last step based on thermodynamic second law, exergy efficiency has been calculated from the following equation and the above-mentioned calculated loads [9]:

    3. Porous Media

    The porous medium that filled the test section is copper foam with a porosity of 95%. The foams are determined in Figure 2 and also detailed thermophysical parameters and dimensions are shown in Table 3.

    Figure 2 Copper foam with a porosity of 95%.Table 3 Thermophysical parameters and dimensions of copper foam.

    In solar collectors, copper porous materials are suitable for use at low temperatures and have an easier and faster manufacturing process than ceramic porous materials. Due to the high coefficient conductivity of copper, the use of copper metallic foam to increase heat transfer is certainly more efficient in solar collectors.

    Porous media and nanofluid in solar collector’s pipes were simulated in FLOW-3D software using the finite-difference method [27]. Nanoparticles Al2O3 and CUO are mostly used in solar collector enhancement. In this research, different concentrations of nanofluid are added to the parabolic solar collectors with porous materials (copper foam with porosity of 95%) to achieve maximum heat transfer in the porous materials before sedimentation. After analyzing PTC pipes with the nanofluid flow in FLOW-3D software, for energy and exergy efficiency analysis, Carrier software results were used as EES software input. Simulation PTC with porous media inside collector pipe and nanofluids sedimentation is shown in Figure 3.

    Figure 3 Simulation PTC pipes enhanced with copper foam and nanoparticles in FLOW-3D software.

    3.1. Nano Fluid

    In this research, copper and silver nanofluids (Al2O3, CuO) have been added with percentages of 0.1%–0.7% as the working fluids. The nanoparticle properties are given in Table 4. Also, system constant parameters are presented in Table 4, which are available as default input in the EES software.Table 4 Properties of the nanoparticles [9].

    System constant parameters for input in the software are shown in Table 5.Table 5 System constant parameters.

    The thermal properties of the nanofluid can be obtained from equations (18)–(21). The basic fluid properties are indicated by the index (bf) and the properties of the nanoparticle silver with the index (np).

    The density of the mixture is shown in the following equation [28]:where ρ is density and ϕ is the nanoparticles volume fraction.

    The specific heat capacity is calculated from the following equation [29]:

    The thermal conductivity of the nanofluid is calculated from the following equation [29]:

    The parameter β is the ratio of the nanolayer thickness to the original particle radius and, usually, this parameter is taken equal to 0.1 for the calculated thermal conductivity of the nanofluids.

    The mixture viscosity is calculated as follows [30]:

    In all equations, instead of water properties, working fluids with nanofluid are used. All of the above equations and parameters are entered in the EES software for calculating the energy and exergy of solar collectors and the SCCHP cycle. All calculation repeats for both nanofluids with different concentrations of nanofluid in the solar collector’s pipe.

    4. Results and Discussion

    In the present study, relations were written according to Wang et al. [16] and the system analysis was performed to ensure the correctness of the code. The energy and exergy charts are plotted based on the main values of the paper and are shown in Figures 4 and 5. The error rate in this simulation is 1.07%.

    Figure 4 Verification charts of energy analysis results.

    Figure 5 Verification charts of exergy analysis results.

    We may also investigate the application of machine learning paradigms [3141] and various hybrid, advanced optimization approaches that are enhanced in terms of exploration and intensification [4255], and intelligent model studies [5661] as well, for example, methods such as particle swarm optimizer (PSO) [6062], differential search (DS) [63], ant colony optimizer (ACO) [616465], Harris hawks optimizer (HHO) [66], grey wolf optimizer (GWO) [5367], differential evolution (DE) [6869], and other fusion and boosted systems [4146485054557071].

    At the first step, the collector is modified with porous copper foam material. 14 cases have been considered for the analysis of the SCCHP system (Table 6). It should be noted that the adding of porous media causes an additional pressure drop inside the collector [922263072]. All fourteen cases use copper foam with a porosity of 95 percent. To simulate the effect of porous materials and nanofluids, the first solar PTC pipes have been simulated in the FLOW-3D software and then porous media (copper foam with porosity of 95%) and fluid flow with nanoparticles (AL2O3 and CUO) are generated in the software. After analyzing PTC pipes in FLOW-3D software, for analyzing energy and exergy efficiency, software outputs were used as EES software input for optimization ratio of sedimentation and calculating energy and exergy analyses.Table 6 Collectors with different percentages of nanofluids and porous media.

    In this research, an enhanced solar collector with both porous media and Nanofluid is investigated. In the present study, 0.1–0.5% CuO and Al2O3 concentration were added to the collector fully filled by porous media to achieve maximum energy and exergy efficiencies of solar CCHP systems. All steps of the investigation are shown in Table 6.

    Energy and exergy analyses of parabolic solar collectors and SCCHP systems are shown in Figures 6 and 7.

    Figure 6 Energy and exergy efficiencies of the PTC with porous media and nanofluid.

    Figure 7 Energy and exergy efficiency of the SCCHP.

    Results show that the highest energy and exergy efficiencies are 74.19% and 32.6%, respectively, that is achieved in Step 12 (parabolic collectors with filled porous media and 0.5% Al2O3). In the second step, the maximum energy efficiency of SCCHP systems with fourteen steps of simulation are shown in Figure 7.

    In the second step, where 0.1, −0.6% of the nanofluids were added, it is found that 0.5% leads to the highest energy and exergy efficiency enhancement in solar collectors and SCCHP systems. Using concentrations more than 0.5% leads to sediment in the solar collector’s pipe and a decrease of porosity in the pipe [73]. According to Figure 7, maximum energy and exergy efficiencies of SCCHP are achieved in Step 12. In this step energy efficiency is 54.49% and exergy efficiency is 18.29%. In steps 13 and 14, with increasing concentration of CUO and Al2O3 nanofluid solution in porous materials, decreasing of energy and exergy efficiency of PTC and SCCHP system at the same time happened. This decrease in efficiency is due to the formation of sediment in the porous material. Calculations and simulations have shown that porous materials more than 0.5% nanofluids inside the collector pipe cause sediment and disturb the porosity of porous materials and pressure drop and reduce the coefficient of performance of the cogeneration system. Most experience showed that CUO and AL2O3 nanofluids with less than 0.6% percent solution are used in the investigation on the solar collectors at low temperatures and discharges [74]. One of the important points of this research is that the best ratio of nanofluids in the solar collector with a low temperature is 0.5% (AL2O3 and CUO); with this replacement, the cost of solar collectors and SCCHP cycle is reduced.

    5. Conclusion and Future Directions

    In the present study, ways for increasing the efficiency of solar collectors in order to enhance the efficiency of the SCCHP cycle are examined. The research is aimed at adding both porous materials and nanofluids for estimating the best ratio of nanofluid for enhanced solar collector and protecting sedimentation in porous media. By adding porous materials (copper foam with porosity of 95%) and 0.5% nanofluids together, high efficiency in solar parabolic collectors can be achieved. The novelty in this research is the addition of both nanofluids and porous materials and calculating the best ratio for preventing sedimentation and pressure drop in solar collector’s pipe. In this study, it was observed that, by adding 0.5% of AL2O3 nanofluid in working fluids, the energy efficiency of PTC rises to 74.19% and exergy efficiency is grown up to 32.6%. In SCCHP cycle, energy efficiency is 54.49% and exergy efficiency is 18.29%.

    In this research, parabolic solar collectors fully filled by porous media (copper foam with a porosity of 95) are investigated. In the next step, parabolic solar collectors in the SCCHP cycle were simultaneously filled by porous media and different percentages of Al2O3 and CuO nanofluid. At this step, values of 0.1% to 0.6% of each nanofluid were added to the working fluid, and the efficiency of the energy and exergy of the collectors and the SCCHP cycle were determined. In this case, nanofluid and the porous media were used together in the solar collector and maximum efficiency achieved. 0.5% of both nanofluids were used to achieve the biggest efficiency enhancement.

    In the present study, as expected, the highest efficiency is for the parabolic solar collector fully filled by porous material (copper foam with a porosity of 95%) and 0.5% Al2O3. Results of the present study are as follows:(1)The average enhancement of collectors’ efficiency using porous media and nanofluids is 28%.(2)Solutions with 0.1 to 0.5% of nanofluids (CuO and Al2O3) are used to prevent collectors from sediment occurrence in porous media.(3)Collector of solar cogeneration cycles that is enhanced by both porous media and nanofluid has higher efficiency, and the stability of output temperature is more as well.(4)By using 0.6% of the nanofluids in the enhanced parabolic solar collectors with copper porous materials, sedimentation occurs and makes a high-pressure drop in the solar collector’s pipe which causes decrease in energy efficiency.(5)Average enhancement of SCCHP cycle efficiency is enhanced by both porous media and nanofluid 13%.


    :Solar radiation
    a:Heat transfer augmentation coefficient
    A:Solar collector area
    Bf:Basic fluid
    :Specific heat capacity of the nanofluid
    F:Constant of air dilution
    :Thermal conductivity of the nanofluid
    :Thermal conductivity of the basic fluid
    :Viscosity of the nanofluid
    :Viscosity of the basic fluid
    :Collector efficiency
    :Collector energy receives
    :Auxiliary boiler heat
    :Expander energy
    :Gas energy
    :Screw expander work
    :Cooling load, in kilowatts
    :Heating load, in kilowatts
    :Solar radiation energy on collector, in Joule
    :Sanitary hot water load
    :Energy efficiency
    :Heat exchanger efficiency
    :Sun exergy
    :Collector exergy
    :Natural gas exergy
    :Expander exergy
    :Cooling exergy
    :Heating exergy
    :Exergy efficiency
    :Steam mass flow rate
    :Hot water mass flow rate
    :Specific heat capacity of water
    :Power output form by the screw expander
    Tam:Average ambient temperature
    :Density of the mixture.

    Greek symbols

    ϕ:Nanoparticles volume fraction
    β:Ratio of the nanolayer thickness.


    CCHP:Combined cooling, heating, and power
    EES:Engineering equation solver.

    Data Availability

    For this study, data were generated by CARRIER software for the average electrical, heating, and cooling load of a residential building with 600 m2 in the city of Zahedan, Iran.

    Conflicts of Interest

    The authors declare that they have no conflicts of interest.


    This work was partially supported by the National Natural Science Foundation of China under Contract no. 71761030 and Natural Science Foundation of Inner Mongolia under Contract no. 2019LH07003.


    1. A. Fudholi and K. Sopian, “Review on solar collector for agricultural produce,” International Journal of Power Electronics and Drive Systems (IJPEDS), vol. 9, no. 1, p. 414, 2018.View at: Publisher Site | Google Scholar
    2. G. Yang and X. Zhai, “Optimization and performance analysis of solar hybrid CCHP systems under different operation strategies,” Applied Thermal Engineering, vol. 133, pp. 327–340, 2018.View at: Publisher Site | Google Scholar
    3. J. Wang, Z. Han, and Z. Guan, “Hybrid solar-assisted combined cooling, heating, and power systems: a review,” Renewable and Sustainable Energy Reviews, vol. 133, p. 110256, 2020.View at: Publisher Site | Google Scholar
    4. Y. Tian and C. Y. Zhao, “A review of solar collectors and thermal energy storage in solar thermal applications,” Applied Energy, vol. 104, pp. 538–553, 2013.View at: Publisher Site | Google Scholar
    5. J. M. Hassan, Q. J. Abdul-Ghafour, and M. F. Mohammed, “CFD simulation of enhancement techniques in flat plate solar water collectors,” Al-Nahrain Journal for Engineering Sciences, vol. 20, no. 3, pp. 751–761, 2017.View at: Google Scholar
    6. M. Jahangiri, O. Nematollahi, A. Haghani, H. A. Raiesi, and A. Alidadi Shamsabadi, “An optimization of energy cost of clean hybrid solar-wind power plants in Iran,” International Journal of Green Energy, vol. 16, no. 15, pp. 1422–1435, 2019.View at: Publisher Site | Google Scholar
    7. I. H. Yılmaz and A. Mwesigye, “Modeling, simulation and performance analysis of parabolic trough solar collectors: a comprehensive review,” Applied Energy, vol. 225, pp. 135–174, 2018.View at: Google Scholar
    8. F. Wang, J. Tan, and Z. Wang, “Heat transfer analysis of porous media receiver with different transport and thermophysical models using mixture as feeding gas,” Energy Conversion and Management, vol. 83, pp. 159–166, 2014.View at: Publisher Site | Google Scholar
    9. H. Zhai, Y. J. Dai, J. Y. Wu, and R. Z. Wang, “Energy and exergy analyses on a novel hybrid solar heating, cooling and power generation system for remote areas,” Applied Energy, vol. 86, no. 9, pp. 1395–1404, 2009.View at: Publisher Site | Google Scholar
    10. M. H. Abbasi, H. Sayyaadi, and M. Tahmasbzadebaie, “A methodology to obtain the foremost type and optimal size of the prime mover of a CCHP system for a large-scale residential application,” Applied Thermal Engineering, vol. 135, pp. 389–405, 2018.View at: Google Scholar
    11. R. Jiang, F. G. F. Qin, X. Yang, S. Huang, and B. Chen, “Performance analysis of a liquid absorption dehumidifier driven by jacket-cooling water of a diesel engine in a CCHP system,” Energy and Buildings, vol. 163, pp. 70–78, 2018.View at: Publisher Site | Google Scholar
    12. F. A. Boyaghchi and M. Chavoshi, “Monthly assessments of exergetic, economic and environmental criteria and optimization of a solar micro-CCHP based on DORC,” Solar Energy, vol. 166, pp. 351–370, 2018.View at: Publisher Site | Google Scholar
    13. F. A. Boyaghchi and M. Chavoshi, “Multi-criteria optimization of a micro solar-geothermal CCHP system applying water/CuO nanofluid based on exergy, exergoeconomic and exergoenvironmental concepts,” Applied Thermal Engineering, vol. 112, pp. 660–675, 2017.View at: Publisher Site | Google Scholar
    14. B. Su, W. Han, Y. Chen, Z. Wang, W. Qu, and H. Jin, “Performance optimization of a solar assisted CCHP based on biogas reforming,” Energy Conversion and Management, vol. 171, pp. 604–617, 2018.View at: Publisher Site | Google Scholar
    15. F. A. Al-Sulaiman, F. Hamdullahpur, and I. Dincer, “Performance assessment of a novel system using parabolic trough solar collectors for combined cooling, heating, and power production,” Renewable Energy, vol. 48, pp. 161–172, 2012.View at: Publisher Site | Google Scholar
    16. J. Wang, Y. Dai, L. Gao, and S. Ma, “A new combined cooling, heating and power system driven by solar energy,” Renewable Energy, vol. 34, no. 12, pp. 2780–2788, 2009.View at: Publisher Site | Google Scholar
    17. Y.-Y. Jing, H. Bai, J.-J. Wang, and L. Liu, “Life cycle assessment of a solar combined cooling heating and power system in different operation strategies,” Applied Energy, vol. 92, pp. 843–853, 2012.View at: Publisher Site | Google Scholar
    18. J.-J. Wang, Y.-Y. Jing, and C.-F. Zhang, “Optimization of capacity and operation for CCHP system by genetic algorithm,” Applied Energy, vol. 87, no. 4, pp. 1325–1335, 2010.View at: Publisher Site | Google Scholar
    19. L. Ali, “LDA–GA–SVM: improved hepatocellular carcinoma prediction through dimensionality reduction and genetically optimized support vector machine,” Neural Computing and Applications, vol. 87, pp. 1–10, 2020.View at: Google Scholar
    20. S. Hirasawa, R. Tsubota, T. Kawanami, and K. Shirai, “Reduction of heat loss from solar thermal collector by diminishing natural convection with high-porosity porous medium,” Solar Energy, vol. 97, pp. 305–313, 2013.View at: Publisher Site | Google Scholar
    21. E. Bellos, C. Tzivanidis, and Z. Said, “A systematic parametric thermal analysis of nanofluid-based parabolic trough solar collectors,” Sustainable Energy Technologies and Assessments, vol. 39, p. 100714, 2020.View at: Publisher Site | Google Scholar
    22. H. J. Jouybari, S. Saedodin, A. Zamzamian, M. E. Nimvari, and S. Wongwises, “Effects of porous material and nanoparticles on the thermal performance of a flat plate solar collector: an experimental study,” Renewable Energy, vol. 114, pp. 1407–1418, 2017.View at: Publisher Site | Google Scholar
    23. J. Subramani, P. K. Nagarajan, S. Wongwises, S. A. El-Agouz, and R. Sathyamurthy, “Experimental study on the thermal performance and heat transfer characteristics of solar parabolic trough collector using Al2O3 nanofluids,” Environmental Progress & Sustainable Energy, vol. 37, no. 3, pp. 1149–1159, 2018.View at: Publisher Site | Google Scholar
    24. E. Shojaeizadeh, F. Veysi, and A. Kamandi, “Exergy efficiency investigation and optimization of an Al2O3-water nanofluid based Flat-plate solar collector,” Energy and Buildings, vol. 101, pp. 12–23, 2015.View at: Publisher Site | Google Scholar
    25. A. K. Tiwari, P. Ghosh, and J. Sarkar, “Solar water heating using nanofluids–a comprehensive overview and environmental impact analysis,” International Journal of Emerging Technology and Advanced Engineering, vol. 3, no. 3, pp. 221–224, 2013.View at: Google Scholar
    26. D. R. Rajendran, E. Ganapathy Sundaram, P. Jawahar, V. Sivakumar, O. Mahian, and E. Bellos, “Review on influencing parameters in the performance of concentrated solar power collector based on materials, heat transfer fluids and design,” Journal of Thermal Analysis and Calorimetry, vol. 140, no. 1, pp. 33–51, 2020.View at: Publisher Site | Google Scholar
    27. M. Feizbahr, C. Kok Keong, F. Rostami, and M. Shahrokhi, “Wave energy dissipation using perforated and non perforated piles,” International Journal of Engineering, vol. 31, no. 2, pp. 212–219, 2018.View at: Google Scholar
    28. K. Khanafer and K. Vafai, “A critical synthesis of thermophysical characteristics of nanofluids,” International Journal of Heat and Mass Transfer, vol. 54, no. 19-20, pp. 4410–4428, 2011.View at: Publisher Site | Google Scholar
    29. K. Farhana, K. Kadirgama, M. M. Rahman et al., “Improvement in the performance of solar collectors with nanofluids – a state-of-the-art review,” Nano-Structures & Nano-Objects, vol. 18, p. 100276, 2019.View at: Publisher Site | Google Scholar
    30. M. Turkyilmazoglu, “Condensation of laminar film over curved vertical walls using single and two-phase nanofluid models,” European Journal of Mechanics-B/Fluids, vol. 65, pp. 184–191, 2017.View at: Publisher Site | Google Scholar
    31. X. Zhang, J. Wang, T. Wang, R. Jiang, J. Xu, and L. Zhao, “Robust feature learning for adversarial defense via hierarchical feature alignment,” Information Sciences, vol. 2020, 2020.View at: Google Scholar
    32. X. Zhang, T. Wang, W. Luo, and P. Huang, “Multi-level fusion and attention-guided CNN for image dehazing,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 1, 2020.View at: Publisher Site | Google Scholar
    33. X. Zhang, M. Fan, D. Wang, P. Zhou, and D. Tao, “Top-k feature selection framework using robust 0-1 integer programming,” IEEE Transactions on Neural Networks and Learning Systems, vol. 1, pp. 1–15, 2020.View at: Publisher Site | Google Scholar
    34. X. Zhang, D. Wang, Z. Zhou, and Y. Ma, “Robust low-rank tensor recovery with rectification and alignment,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 43, no. 1, pp. 238–255, 2019.View at: Google Scholar
    35. X. Zhang, R. Jiang, T. Wang, and J. Wang, “Recursive neural network for video deblurring,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 1, 2020.View at: Publisher Site | Google Scholar
    36. X. Zhang, T. Wang, J. Wang, G. Tang, and L. Zhao, “Pyramid channel-based feature attention network for image dehazing,” Computer Vision and Image Understanding, vol. 1, 2020.View at: Google Scholar
    37. M. Mirmozaffari, “Machine learning algorithms based on an optimization model,” 2020.View at: Google Scholar
    38. M. Mirmozaffari, M. Yazdani, A. Boskabadi, H. Ahady Dolatsara, K. Kabirifar, and N. Amiri Golilarz, “A novel machine learning approach combined with optimization models for eco-efficiency evaluation,” Applied Sciences, vol. 10, no. 15, p. 5210, 2020.View at: Publisher Site | Google Scholar
    39. M. Vosoogha and A. Addeh, “An intelligent power prediction method for wind energy generation based on optimized fuzzy system,” Computational Research Progress in Applied Science & Engineering (CRPASE), vol. 5, pp. 34–43, 2019.View at: Google Scholar
    40. A. Javadi, N. Mikaeilvand, and H. Hosseinzdeh, “Presenting a new method to solve partial differential equations using a group search optimizer method (GSO),” Computational Research Progress in Applied Science and Engineering, vol. 4, no. 1, pp. 22–26, 2018.View at: Google Scholar
    41. F. J. Golrokh, Gohar Azeem, and A. Hasan, “Eco-efficiency evaluation in cement industries: DEA malmquist productivity index using optimization models,” ENG Transactions, vol. 1, pp. 1–8, 2020.View at: Google Scholar
    42. H. Yu, “Dynamic Gaussian bare-bones fruit fly optimizers with abandonment mechanism: method and analysis,” Engineering with Computers, vol. 1, pp. 1–29, 2020.View at: Google Scholar
    43. C. Yu, “SGOA: annealing-behaved grasshopper optimizer for global tasks,” Engineering with Computers, vol. 1, pp. 1–28, 2021.View at: Google Scholar
    44. W. Shan, Z. Qiao, A. A. Heidari, H. Chen, H. Turabieh, and Y. Teng, “Double adaptive weights for stabilization of moth flame optimizer: balance analysis, engineering cases, and medical diagnosis,” Knowledge-Based Systems, vol. 1, p. 106728, 2020.View at: Google Scholar
    45. J. Tu, H. Chen, J. Liu et al., “Evolutionary biogeography-based whale optimization methods with communication structure: towards measuring the balance,” Knowledge-Based Systems, vol. 212, p. 106642, 2021.View at: Publisher Site | Google Scholar
    46. Y. Zhang, “Towards augmented kernel extreme learning models for bankruptcy prediction: algorithmic behavior and comprehensive analysis,” Neurocomputing, vol. 1, 2020.View at: Google Scholar
    47. Y. Zhang, R. Liu, X. Wang, H. Chen, and C. Li, “Boosted binary Harris hawks optimizer and feature selection,” Engineering with Computers, vol. 1, pp. 1–30, 2020.View at: Google Scholar
    48. H.-L. Chen, G. Wang, C. Ma, Z.-N. Cai, W.-B. Liu, and S.-J. Wang, “An efficient hybrid kernel extreme learning machine approach for early diagnosis of Parkinson’s disease,” Neurocomputing, vol. 184, pp. 131–144, 2016.View at: Publisher Site | Google Scholar
    49. L. Hu, G. Hong, J. Ma, X. Wang, and H. Chen, “An efficient machine learning approach for diagnosis of paraquat-poisoned patients,” Computers in Biology and Medicine, vol. 59, pp. 116–124, 2015.View at: Publisher Site | Google Scholar
    50. L. Shen, H. Chen, Z. Yu et al., “Evolving support vector machines using fruit fly optimization for medical data classification,” Knowledge-Based Systems, vol. 96, pp. 61–75, 2016.View at: Publisher Site | Google Scholar
    51. J. Xia, H. Chen, Q. Li et al., “Ultrasound-based differentiation of malignant and benign thyroid Nodules: an extreme learning machine approach,” Computer Methods and Programs in Biomedicine, vol. 147, pp. 37–49, 2017.View at: Publisher Site | Google Scholar
    52. C. Li, L. Hou, B. Y. Sharma et al., “Developing a new intelligent system for the diagnosis of tuberculous pleural effusion,” Computer Methods and Programs in Biomedicine, vol. 153, pp. 211–225, 2018.View at: Publisher Site | Google Scholar
    53. X. Zhao, X. Zhang, Z. Cai et al., “Chaos enhanced grey wolf optimization wrapped ELM for diagnosis of paraquat-poisoned patients,” Computational Biology and Chemistry, vol. 78, pp. 481–490, 2019.View at: Publisher Site | Google Scholar
    54. M. Wang and H. Chen, “Chaotic multi-swarm whale optimizer boosted support vector machine for medical diagnosis,” Applied Soft Computing Journal, vol. 88, 2020.View at: Publisher Site | Google Scholar
    55. X. Xu and H.-L. Chen, “Adaptive computational chemotaxis based on field in bacterial foraging optimization,” Soft Computing, vol. 18, no. 4, pp. 797–807, 2014.View at: Publisher Site | Google Scholar
    56. R. U. Khan, X. Zhang, R. Kumar, A. Sharif, N. A. Golilarz, and M. Alazab, “An adaptive multi-layer botnet detection technique using machine learning classifiers,” Applied Sciences, vol. 9, no. 11, p. 2375, 2019.View at: Publisher Site | Google Scholar
    57. A. Addeh, A. Khormali, and N. A. Golilarz, “Control chart pattern recognition using RBF neural network with new training algorithm and practical features,” ISA Transactions, vol. 79, pp. 202–216, 2018.View at: Publisher Site | Google Scholar
    58. N. Amiri Golilarz, H. Gao, R. Kumar, L. Ali, Y. Fu, and C. Li, “Adaptive wavelet based MRI brain image de-noising,” Frontiers in Neuroscience, vol. 14, p. 728, 2020.View at: Publisher Site | Google Scholar
    59. N. A. Golilarz, H. Gao, and H. Demirel, “Satellite image de-noising with Harris hawks meta heuristic optimization algorithm and improved adaptive generalized Gaussian distribution threshold function,” IEEE Access, vol. 7, pp. 57459–57468, 2019.View at: Publisher Site | Google Scholar
    60. M. Eisazadeh and J. Rezapour, “Multi-objective optimization of the composite sheets using PSO algorithm,” 2017.View at: Google Scholar
    61. I. Bargegol, M. Nikookar, R. V. Nezafat, E. J. Lashkami, and A. M. Roshandeh, “Timing optimization of signalized intersections using shockwave theory by genetic algorithm,” Computational Research Progress in Applied Science & Engineering, vol. 1, pp. 160–167, 2015.View at: Google Scholar
    62. B. Bai, Z. Guo, C. Zhou, W. Zhang, and J. Zhang, “Application of adaptive reliability importance sampling-based extended domain PSO on single mode failure in reliability engineering,” Information Sciences, vol. 546, pp. 42–59, 2021.View at: Publisher Site | Google Scholar
    63. J. Liu, C. Wu, G. Wu, and X. Wang, “A novel differential search algorithm and applications for structure design,” Applied Mathematics and Computation, vol. 268, pp. 246–269, 2015.View at: Publisher Site | Google Scholar
    64. X. Zhao, D. Li, B. Yang, C. Ma, Y. Zhu, and H. Chen, “Feature selection based on improved ant colony optimization for online detection of foreign fiber in cotton,” Applied Soft Computing, vol. 24, pp. 585–596, 2014.View at: Publisher Site | Google Scholar
    65. D. Zhao, “Chaotic random spare ant colony optimization for multi-threshold image segmentation of 2D Kapur entropy,” Knowledge-Based Systems, vol. 24, p. 106510, 2020.View at: Google Scholar
    66. H. Chen, A. A. Heidari, H. Chen, M. Wang, Z. Pan, and A. H. Gandomi, “Multi-population differential evolution-assisted Harris hawks optimization: framework and case studies,” Future Generation Computer Systems, vol. 111, pp. 175–198, 2020.View at: Publisher Site | Google Scholar
    67. J. Hu, H. Chen, A. A. Heidari et al., “Orthogonal learning covariance matrix for defects of grey wolf optimizer: insights, balance, diversity, and feature selection,” Knowledge-Based Systems, vol. 213, p. 106684, 2021.View at: Publisher Site | Google Scholar
    68. G. Sun, B. Yang, Z. Yang, and G. Xu, “An adaptive differential evolution with combined strategy for global numerical optimization,” Soft Computing, vol. 24, pp. 1–20, 2019.View at: Google Scholar
    69. G. Sun, C. Li, and L. Deng, “An adaptive regeneration framework based on search space adjustment for differential evolution,” Neural Computing and Applications, vol. 24, pp. 1–17, 2021.View at: Google Scholar
    70. A. Addeh and M. Iri, “Brain tumor type classification using deep features of MRI images and optimized RBFNN,” ENG Transactions, vol. 2, pp. 1–7, 2021.View at: Google Scholar
    71. F. J. Golrokh and A. Hasan, “A comparison of machine learning clustering algorithms based on the DEA optimization approach for pharmaceutical companies in developing countries,” Soft Computing, vol. 1, pp. 1–8, 2020.View at: Google Scholar
    72. H. Tyagi, P. Phelan, and R. Prasher, “Predicted efficiency of a low-temperature nanofluid-based direct absorption solar collector,” Journal of Solar Energy Engineering, vol. 131, no. 4, 2009.View at: Publisher Site | Google Scholar
    73. S. Rashidi, M. Bovand, and J. A. Esfahani, “Heat transfer enhancement and pressure drop penalty in porous solar heat exchangers: a sensitivity analysis,” Energy Conversion and Management, vol. 103, pp. 726–738, 2015.View at: Publisher Site | Google Scholar
    74. N. Akram, R. Sadri, S. N. Kazi et al., “A comprehensive review on nanofluid operated solar flat plate collectors,” Journal of Thermal Analysis and Calorimetry, vol. 139, no. 2, pp. 1309–1343, 2020.View at: Publisher Site | Google Scholar
    Fig. 11. Velocity vectors along x-direction through the center of the box culvert for B0, B30, B50, and B70 respectively.

    Numerical investigation of scour characteristics downstream of blocked culverts

    막힌 암거 하류의 세굴 특성 수치 조사

    NesreenTahabMaged M.El-FekyaAtef A.El-SaiadaIsmailFathya
    aDepartment of Water and Water Structures Engineering, Faculty of Engineering, Zagazig University, Zagazig 44519, Egypt
    bLab Manager, Faculty of Engineering, Zagazig University, Zagazig 44519, Egypt


    횡단 구조물을 통한 막힘은 안정성을 위협하는 위험한 문제 중 하나입니다. 암거의 막힘 형상 및 하류 세굴 특성에 미치는 영향에 관한 연구는 거의 없습니다.

    이 연구의 목적은 수면과 세굴 모두에서 상자 암거를 통한 막힘의 작용을 수치적으로 논의하는 것입니다. 이를 위해 FLOW 3D v11.1.0을 사용하여 퇴적물 수송 모델을 조사했습니다.

    상자 암거를 통한 다양한 차단 비율이 연구되었습니다. FLOW 3D 모델은 실험 데이터로 보정되었습니다. 결과는 FLOW 3D 프로그램이 세굴 다운스트림 상자 암거를 정확하게 시뮬레이션할 수 있음을 나타냅니다.

    막힌 경우에 대한 속도 분포, 최대 세굴 깊이 및 수심을 플롯하고 비차단된 사례(기본 사례)와 비교했습니다.

    그 결과 암거 높이의 70% 차단율은 상류의 수심을 암거 높이의 2.3배 증가시키고 평균 유속은 기본 경우보다 3배 더 증가시키는 것으로 입증되었다. 막힘 비율의 함수로 상대 최대 세굴 깊이를 추정하는 방정식이 만들어졌습니다.

    Blockage through crossing structures is one of the dangerous problems that threaten its stability. There are few researches concerned with blockage shape in culverts and its effect on characteristics of scour downstream it.

    The study’s purpose is to discuss the action of blockage through box culvert on both water surface and scour numerically. A sediment transport model has been investigated for this purpose using FLOW 3D v11.1.0. Different ratios of blockage through box culvert have been studied. The FLOW 3D model was calibrated with experimental data.

    The results present that the FLOW 3D program was capable to simulate accurately the scour downstream box culvert. The velocity distribution, maximum scour depth and water depths for blocked cases have been plotted and compared with the non-blocked case (base case).

    The results proved that the blockage ratio 70% of culvert height makes the water depth upstream increases by 2.3 times of culvert height and mean velocity increases by 3 times more than in the base case. An equation has been created to estimate the relative maximum scour depth as a function of blockage ratio.

    1. Introduction

    Local scour is the removal of granular bed material by the action of hydrodynamic forces. As the depth of scour hole increases, the stability of the foundation of the structure may be endangered, with a consequent risk of damage and failure [1]. So the prediction and control of scour is considered to be very important for protecting the water structures from failure. Most previous studies were designed to study the different factors that impact on scour and their relationship with scour hole dimensions like fluid characteristics, flow conditions, bed properties, and culvert geometry. Many previous researches studied the effect of flow rate on scour hole by information Froude number or modified Froude number [2][3][4][5][6]. Cesar Mendoza [6] found a good correlation between the scour depth and the discharge Intensity (Qg−.5D−2.5). Breusers and Raudkiv [7] used shear velocity in the outlet-scour prediction procedure. Ali and Lim [8] used the densimetric Froude number in estimation of the scour depth [1][8][9][10][11][12][13][14]. “The densimetric Froude number presents the ratio of the tractive force on sediment particle to the submerged specific weight of the sediment” [15](1)Fd=uρsρ-1gD50

    Ali and Lim [8] pointed to the consequence of tailwater depth on scour behavior [1][2][8][13]. Abida and Townsend [2] indicated that the maximum depth of local scour downstream culvert was varying with the tailwater depth in three ways: first, for very shallow tailwater depths, local scouring decreases with a decrease in tailwater depth; second, when the ratio of tailwater depth to culvert height ranged between 0.2 and 0.7, the scour depth increases with decreasing tailwater depth; and third for a submerged outlet condition. The tailwater depth has only a marginal effect on the maximum depth of scour [2]. Ruff et al. [16] observed that for materials having similar mean grain sizes (d50) but different standard deviations (σ). As (σ) increased, the maximum scour hole depth decreased. Abt et al. [4] mentioned to role of soil type of maximum scour depth. It was noticed that local scour was more dangerous for uniform sands than for well-graded mixtures [1][2][4][9][17][18]. Abt et al [3][19] studied the culvert shape effect on scour hole. The results evidenced that the culvert shape has a limited effect on outlet scour. Under equivalent discharge conditions, it was noted that a square culvert with height equal to the diameter of a circular culvert would reduce scour [16][20]. The scour hole dimension was also effected by the culvert slope. Abt et al. [3][21] showed that the culvert slope is a key element in estimating the culvert flow velocity, the discharge capacity, and sediment transport capability. Abt et al. [21][22] tested experimentally culvert drop height effect on maximum scour depth. It was observed that as the drop height was increasing, the depth of scour was also increasing. From the previous studies, it could have noticed that the most scour prediction formula downstream unblocked culvert was the function of densimetric Froude number, soil properties (d50, σ), tailwater depth and culvert opening size. Blockage is the phenomenon of plugging water structures due to the movement of water flow loaded with sediment and debris. Water structures blockage has a bad effect on water flow where it causes increasing of upstream water level that may cause flooding around the structure and increase of scour rate downstream structures [23][24]. The blockage phenomenon through was studied experimentally and numerical [15][25][26][27][28][29][30][31][32][33]. Jaeger and Lucke [33] studied the debris transport behavior in a natural channel in Australia. Froude number scale model of an existing culvert was used. It was noticed that through rainfall event, the mobility of debris was impressed by stream shape (depth and width). The condition of the vegetation (size and quantities) through the catchment area was the main factor in debris transport. Rigby et al. [26] reported that steep slope was increasing the ability to mobilize debris that form field data of blocked culverts and bridges during a storm in Wollongong city.

    Streftaris et al. [32] studied the probability of screen blockage by debris at trash screens through a numerical model to relate between the blockage probability and nature of the area around. Recently, many commercial computational fluid programs (CFD) such as SSIIM, Fluent, and FLOW 3D are used in the analysis of the scour process. Scour and sediment transport numerical model need to validate by using experimental data or field data [34][35][36][37][38]. Epely-Chauvin et al. [36] investigated numerically the effect of a series of parallel spur diked. The experimental data were compared by SSIIM and FLOW 3D program. It was found that the accuracy of calibrated FLOW 3D model was better than SSIIM model. Nielsen et al. [35] used the physical model and FLOW 3D model to analyze the scour process around the pile. The soil around the pile was uniform coarse stones in the physical models that were simulated by regular spheres, porous media, and a mixture of them. The calibrated porous media model can be used to determine the bed shear stress. In partially blocked culverts, there aren’t many studies that explain the blockage impact on scour dimensions. Sorourian et al. [14][15] studied the effect of inlet partial blockage on scour characteristics downstream box culvert. It resulted that the partial blockage at the culvert inlet could be the main factor in estimating the depth of scour. So, this study is aiming to investigate the effects of blockage through a box culvert on flow and scour characteristics by different blockage ratios and compares the results with a non-blocked case. Create a dimensionless equation relates the blockage ratio of the culvert with scour characteristics downstream culvert.

    2. Experimental data

    The experimental work of the study was conducted in the Hydraulics and Water Engineering Laboratory, Faculty of Engineering, Zagazig University, Egypt. The flume had a rectangular cross-section of 66 cm width, 65.5 cm depth, and 16.2 m long. A rectangular culvert was built with 0.2 m width, 0.2 m height and 3.00 m long with θ = 25° gradually outlet and 0.8 m fixed apron. The model was located on the mid-point of the channel. The sediment part was extended for a distance 2.20 m with 0.66 m width and 0.20 m depth of coarse sand with specific weight 1.60 kg/cm3, d50 = 2.75 mm and σ (d90/d50) = 1.50. The particle size distribution was as shown in Fig. 1. The experimental model was tested for different inlet flow (Q) of 25, 30, 34, 40 l/s for different submerged ratio (S) of 1.25, 1.50, 1.75.

    3. Dimensional analysis

    A dimensional analysis has been used to reduce the number of variables which affecting on the scour pattern downstream partial blocked culvert. The main factors affecting the maximum scour depth are:(2)ds=f(ρ.ρ

    Fig. 2 shows a definition sketch of the experimental model. The maximum scour depth can be written in a dimensionless form as:(3)dsh=f(B.Fd.S)where the ds/h is the relative maximum scour depth.

    4. Numerical work

    The FLOW 3D is (CFD) program used by many researchers and appeared high accuracy in solving hydrodynamic and sediment transport models in the three dimensions. Numerical simulation with FLOW 3D was performed to study the impacts of blockage ratio through box culvert on shear stress, velocity distribution and the sediment transport in terms of the hydrodynamic features (water surface, velocity and shear stress) and morphological parameters (scour depth and sizes) conditions in accurately and efficiently. The renormalization group (RNG) turbulence model was selected due to its high ability to predict the velocity profiles and turbulent kinetic energy for the flow through culvert [39]. The one-fluid incompressible mode was used to simulate the water surface. Volume of fluid (VOF) method was employed in FLOW 3D to tracks a liquid interface through arbitrary deformations and apply the correct boundary conditions at the interface [40].1.

    Governing equations

    Three-dimensional Reynolds-averaged Navier Stokes (RANS) equation was applied for incompressible viscous fluid motion. The continuity equation is as following:(4)VF∂ρ∂t+∂∂xρuAx+∂∂yρvAy+∂∂zρwAz=RDIF(5)∂u∂t+1VFuAx∂u∂x+vAy∂u∂y+ωAz∂u∂z=-1ρ∂P∂x+Gx+fx(6)∂v∂t+1VFuAx∂v∂x+vAy∂v∂y+ωAz∂v∂z=-1ρ∂P∂y+Gy+fy(7)∂ω∂t+1VFuAx∂ω∂x+vAy∂ω∂y+ωAz∂ω∂z=-1ρ∂P∂z+Gz+fz

    ρ is the fluid density,

    VF is the volume fraction,

    (x,y,z) is the Cartesian coordinates,

    (u,v,w) are the velocity components,

    (Ax,Ay,Az) are the area fractions and

    RDIF is the turbulent diffusion.

    P is the average hydrodynamic pressure,

    (Gx, Gy, Gz) are the body accelerations and

    (fx, fy, fz) are the viscous accelerations.

    The motion of sediment transport (suspended, settling, entrainment, bed load) is estimated by predicting the erosion, advection and deposition process as presented in [41].

    The critical shields parameter is (θcr) is defined as the critical shear stress τcr at which sediments begin to move on a flat and horizontal bed [41]:(8)θcr=τcrgd50(ρs-ρ)

    The Soulsby–Whitehouse [42] is used to predict the critical shields parameter as:(9)θcr=0.31+1.2d∗+0.0551-e(-0.02d∗)(10)d∗=d50g(Gs-1ν3where:

    d* is the dimensionless grain size

    Gs is specific weight (Gs = ρs/ρ)

    The entrainment coefficient (0.005) was used to scale the scour rates and fit the experimental data. The settling velocity controls the Soulsby deposition equation. The volumetric sediment transport rate per width of the bed is calculated using Van Rijn [43].2.

    Meshing and geometry of model

    After many trials, it was found that the uniform cell size with 0.03 m cell size is the closest to the experimental results and takes less time. As shown in Fig. 3. In x-direction, the total model length in this direction is 700 cm with mesh planes at −100, 0, 300, 380 and 600 cm respectively from the origin point, in y-direction, the total model length in this direction is 66 cm at distances 0, 23, 43 and 66 cm respectively from the origin point. In z-direction, the total model length in this direction is 120 cm. with mesh planes at −20, 0, 20 and 100 cm respectively.3.

    Boundary condition

    As shown in Fig. 4, the boundary conditions of the model have been defined to simulate the experimental flow conditions accurately. The upstream boundary was defined as the volume flow rate with a different flow rate. The downstream boundary was defined as specific pressure with different fluid elevation. Both of the right side, the left side, and the bottom boundary were defined as a wall. The top boundary defined as specified pressure with pressure value equals zero.

    5. Validation of experimental results and numerical results

    The experimental results investigated the flow and scour characteristics downstream culvert due to different flow conditions. The measured value of maximum scour depth is compared with the simulated depth from FLOW 3D model as shown in Fig. 5. The scour results show that the simulated results from the numerical model is quite close to the experimental results with an average error of 3.6%. The water depths in numerical model results is so close to the experimental results as shown in Fig. 6 where the experiment and numerical results are compared at different submerged ratios and flow rates. The results appear maximum error percentage in water depths upstream and downstream the culvert is about 2.37%. This indicated that the FLOW 3D is efficient for the prediction of maximum scour depth and the flow depths downstream box culvert.

    6. Computation time

    The run time was chosen according to reaching to the stability limit. Hydraulic stability was achieved after 50 s, where the scour development may still go on. For run 1, the numerical simulation was run for 1000 s as shown in Fig. 7 where it mostly reached to scour stability at 800 s. The simulation time was taken 500 s at about 95% of scour stability.

    7. Analysis and discussions

    Fig. 8 shows the study sections where sec 1 represents to upstream section, sec2 represents to inside section and sec3 represents to downstream stream section. Table 1 indicates the scour hole dimensions at different blockage case. The symbol (B) represents to blockage and the number points to blockage ratio. B0 case signifies to the non-blocked case, B30 is that blockage height is 30% to the culvert height and so on.

    Table 1. The scour results of different blockage ratio.

    Casehb cmB = hb/hQ lit/sSFdd50 mmds/h measuredls/hdd/hld/hds/h estimated

    7.1. Scour hole geometry

    The scour hole geometry mainly depends on the properties of soil of the bed downstream the fixed apron. From Table 1, the results show that the maximum scour depth in B0 case is about 0.58 of culvert height while the maximum deposition in B0 is 0.27 culvert height. There is a symmetric scour hole as shown in Fig. 9 in B0 case. An asymmetric scour hole is created in B50 and B70 due to turbulences that causes the deviation of the jet direction from the center of the flume where appear in Fig. 11 and Fig. 19.

    7.2. Flow water surface

    Fig. 10 presents the relative free surface water (hw/h) along the x-direction at center of the box culvert. From the mention Figure, it is easy to release the effect of different blockage ratios. The upstream water level rises by increasing the blockage ratio. Increasing upstream water level may cause flooding over the banks of the waterway. In the 70% blockage case, the upstream water level rises to 2.3 times of culvert height more than the non-blocked case at the same discharge and submerged ratio. The water surface profile shows an increase in water level upstream the culvert due to a decrease in transverse velocity. Because of decreasing velocity downstream culvert, there is an increase in water level before it reaches its uniform depth.

    7.3. Velocity vectors

    Scour downstream hydraulic structures mainly affects by velocities distribution and bed shear stress. Fig. 11 shows the velocity vectors and their magnitude in xz plane at the same flow conditions. The difference in the upstream water level due to the different blockage ratios is so clear. The maximum water level is in B70 and the minimum level is in B0. The inlet mean velocity value is about 0.88 m/s in B0 increases to 2.86 m/s in B70. As the blockage ratio increases, the inlet velocity increases. The outlet velocity in B0 case makes downward jet causes scour hole just after the fixed apron in the middle of the bed while the blockage causes upward water flow that appears clearly in B70. The upward jet decreases the scour depth to 0.13 culvert height less than B0 case. After the scour hole, the velocity decreases and the flow becomes uniform.

    7.4. Velocity distribution

    Fig. 12 represents flow velocity (Vx) distribution along the vertical depth (z/hu) upstream the inlet for the different blockage ratios at the same flow conditions. From the Figure, the maximum velocity creates closed to bed in B0 while in blocked case, the maximum horizontal velocity creates at 0.30 of relative vertical depth (z/hu). Fig. 13 shows the (Vz) distribution along the vertical depth (z/hu) upstream culvert at sec 1. From the mentioned Figure, it is easy to note that the maximum vertical is in B70 which appears that as the blockage ratio increases the vertical ratio also increases. In the non-blocked case. The vertical velocity (Vz) is maximum at (z/hu) equals 0.64. At the end of the fixed apron (sec 3), the horizontal velocity (Vx) is slowly increasing to reach the maximum value closed to bed in B0 and B30 while the maximum horizontal velocity occurs near to the top surface in B50 and B70 as shown in Fig. 14. The vertical velocity component along the vertical depth (z/hd) is presented in Fig. 15. The vertical velocity (Vz) is maximum in B0 at vertical depth (z/hd) 0.3 with value 0.45 m/s downward. Figs. 16 and 17 observe velocity components (Vx, Vz) along the vertical depth just after the end of blockage length at the centerline of the culvert barrel. It could be noticed the uniform velocity distribution in B0 case with horizontal velocity (Vx) closed to 1.0 m/s and vertical velocity closed to zero. In the blocked case, the maximum horizontal velocity occurs in depth more than the blockage height.

    7.5. Bed velocity distribution

    Fig. 18 presents the x-velocity vectors at 1.5 cm above the bed for different blockage ratios from the velocity vectors distribution and magnitude, it is easy to realize the position of the scour hole and deposition region. In B0 and B30, the flow is symmetric so that the scour hole is created around the centerline of flow while in B50 and B70 cases, the flow is asymmetric and the scour hole creates in the right of flow direction in B50. The maximum scour depth is found in the left of flow direction in B70 case where the high velocity region is found.

    8. Maximum scour depth prediction

    Regression analysis is used to estimate maximum scour depth downstream box culvert for different ratios of blockage by correlating the maximum relative scour by other variables that affect on it in one formula. An equation is developed to predict maximum scour depth for blocked and non-blocked. As shown in the equation below, the relative maximum scour depth(ds/hd) is a function of densimetric Froude number (Fd), blockage ratio (B) and submerged ratio (S)(11)dsh=0.56Fd-0.20B+0.45S-1.05

    In this equation the coefficient of correlation (R2) is 0.82 with standard error equals 0·08. The developed equation is valid for Fd = [0.9 to 2.10] and submerged ratio (S) ≥ 1.00. Fig. 19 shows the comparison between relative maximum scour depths (ds/h) measured and estimated for different blockage ratios. Fig. 20 clears the comparison between residuals and ds/h estimated for the present study. From these figures, it could be noticed that there is a good agreement between the measured and estimated relative scour depth.

    9. Comparison with previous scour equations

    Many previous scour formulae have been produced for calculation the maximum scour depth downstream non-blockage culvert. These equations have been included the effect of flow regime, culvert shape, soil properties and the flow rate on maximum scour depth. Two of previous experimental studies data have been chosen to be compared with the present study results in non-blocked study data. Table 2 shows comparison of culvert shape, densmetric Froude number, median particle size and scour equations for these previous studies. By applying the present study data in these studies scour formula as shown in Fig. 21, it could be noticed that there are a good agreement between present formula results and others empirical equations results. Where that Lim [44] and Abt [4] are so closed to the present study data.

    Table 2. Comparison of some previous scour formula.

    ResearchersFdCulvert shaped50(mm)Proposed equationSubmerged ratio
    Present study0.9–2.11square2.75dsh=0.56Fd-0.20B+0.45S-1.051.25–1.75
    Lim [44]1–10Circular1.65dsh=0.45Fd0.47
    Abt [4]Fd ≥ 1Circular0.22–7.34-dsh=3.67Fd0.57∗D500.4∗σ-0.4

    10. Conclusions

    The present study has shown that the FLOW 3D model can accurately simulate water surface and the scour hole characteristics downstream the box culvert with error percentage in water depths does not exceed 2.37%. Velocities distribution through and outlets culvert barrel helped on understanding the scour hole shape.

    The blockage through culvert had caused of increasing of water surface upstream structure where the upstream water level in B70 was 2.3 of culvert height more than non-blocked case at the same discharge that could be dangerous on the stability of roads above. The depth averaged velocity through culvert barrel increased by 3 times its value in non-blocked case.

    On the other hand, blockage through culvert had a limited effect on the maximum scour depth. The little effect of blockage on maximum scour depth could be noticed in Fig. 11. From this Figure, it could be noted that the residual part of culvert barrel after the blockage part had made turbulences. These turbulences caused the deviation of the flow resulting in the formation of asymmetric scour hole on the side of channel. This not only but in B70 the blockage height caused upward jet which made a wide far scour hole as cleared from the results in Table 1.

    An empirical equation was developed from the results to estimate the maximum scour depth relative to culvert height function of blockage ratio (B), submerged ratio (S), and densimetric Froude number (Fd). The equation results was compared with some scour formulas at the same densimetric Froude number rang where the present study results was in between the other equations results as shown in Fig. 21.

    Declaration of Competing Interest

    The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.


    [1]P. Sarathi, M. Faruque, R. BalachandarInfluence of tailwater depth, sediment size and densimetric Froude number on scour by submerged square wall jetsJ. Hydraul. Res., 46 (2) (2008), pp. 158-175CrossRefView Record in ScopusGoogle Scholar[2]H. Abida, R. TownsendLocal scour downstream of box-culvert outletsJ. Irrig. Drain. Eng., 117 (3) (1991), pp. 425-440CrossRefView Record in ScopusGoogle Scholar[3]S.R. Abt, C.A. Donnell, J.F. Ruff, F.K. DoehringCulvert Slope and Shape Effects on Outlet ScourTransp. Res. Rec., 1017 (1985), pp. 24-30View Record in ScopusGoogle Scholar[4]S.R. Abt, R.L. Kloberdanz, C. MendozaUnified culvert scour determinationJ. Hydraul. Eng., 110 (10) (1984), pp. 1475-1479CrossRefView Record in ScopusGoogle Scholar[5]J.P. Bohan, Erosion And Riprap Requirements At Culvert And Storm-Drain Outlets, ARMY ENGINEER WATERWAYS EXPERIMENT STATION VICKSBURG MISS1970.Google Scholar[6]C. Mendoza, S.R. Abt, J.F. RuffHeadwall influence on scour at culvert outletsJ. Hydraul. Eng., 109 (7) (1983), pp. 1056-1060CrossRefView Record in ScopusGoogle Scholar[7]H. Breusers, A. Raudkivi, Scouring, hydraulic structures design manual, vol. 143, IAHR, AA Balkema, Rotterdam, 1991.Google Scholar[8]K. Ali, S. LimLocal scour caused by submerged wall jetsProc. Inst. Civ. Eng., 81 (4) (1986), pp. 607-645CrossRefView Record in ScopusGoogle Scholar[9]O. Aderibigbe, N. RajaratnamEffect of sediment gradation on erosion by plane turbulent wall jetsJ. Hydraul. Eng., 124 (10) (1998), pp. 1034-1042View Record in ScopusGoogle Scholar[10]F.W. Blaisdell, C.L. AndersonA comprehensive generalized study of scour at cantilevered pipe outletsJ. Hydraul. Res., 26 (4) (1988), pp. 357-376CrossRefView Record in ScopusGoogle Scholar[11]Y.-M. Chiew, S.-Y. LimLocal scour by a deeply submerged horizontal circular jetJ. Hydraul. Eng., 122 (9) (1996), pp. 529-532View Record in ScopusGoogle Scholar[12]R.A. Day, S.L. Liriano, W.R. WhiteEffect of tailwater depth and model scale on scour at culvert outletsProc. Instit. Civil Eng. – Water Marit. Eng., 148 (3) (2001), pp. 189-198 Record in ScopusGoogle Scholar[13]S. Emami, A.J. SchleissPrediction of localized scour hole on natural mobile bed at culvert outletsScour and Erosion (2010), pp. 844-853CrossRefView Record in ScopusGoogle Scholar[14]S. Sorourian, A. Keshavarzi, J. Ball, B. SamaliStudy of Blockage Effect on Scouring Pattern Downstream of a Box Culvert under Unsteady FlowAustr. J Water Resor. (2013)Google Scholar[15]S. Sorourian, Turbulent Flow Characteristics At The Outlet Of Partially Blocked Box Culverts, in: 36th IAHR World Congress, The Hague, the Netherlands, 2015.Google Scholar[16]J. Ruff, S. Abt, C. Mendoza, A. Shaikh, R. KloberdanzScour at culvert outlets in mixed bed materialsUnited States. Federal Highway Administration. Office of Research and Development (1982)Google Scholar[17]S.A. Ansari, U.C. Kothyari, K.G.R. RajuInfluence of cohesion on scour under submerged circular vertical jetsJ. Hydraul. Eng., 129 (12) (2003), pp. 1014-1019View Record in ScopusGoogle Scholar[18]B. Crookston B. Tullis, Scour and Riprap Protection in a Bottomless Arch Culvert, in: World Environmental and Water Resources Congress 2008: Ahupua’A, 2008, pp. 1–10.Google Scholar[19]S.R. Abt, J. Ruff, F. Doehring, C. DonnellInfluence of culvert shape on outlet scourJ. Hydraul. Eng., 113 (3) (1987), pp. 393-400View Record in ScopusGoogle Scholar[20]Y.H. Chen, Scour at outlets of box culverts, Colorado State University, 1970.Google Scholar[21]S. Abt, P. Thompson, T. LewisEnhancement of the culvert outlet scour estimation equationsTransp. Res. Rec. J. Transp. Res. Board, 1523 (1996), pp. 178-185View Record in ScopusGoogle Scholar[22]F.K. Doehring, S.R. AbtDrop height influence on outlet scourJ. Hydraul. Eng., 120 (12) (1994), pp. 1470-1476CrossRefView Record in ScopusGoogle Scholar[23]W. Weeks, A. Barthelmess, E. Rigby, G. Witheridge, R. Adamson, Australian rainfall and runoff revison project 11: blockage of hydraulic structures, 2009.Google Scholar[24]W. Weeks, G. Witheridge, E. Rigby, A. BarthelmessProject 11: blockage of hydraulic structuresEngineers Australia (2013)Google Scholar[25]S.R. Abt, T.E. Brisbane, D.M. Frick, C.A. McKnightTrash rack blockage in supercritical flowJ. Hydraul. Eng., 118 (12) (1992), pp. 1692-1696View Record in ScopusGoogle Scholar[26]E. Rigby, M. Boyd, S. Roso, P. Silveri, A. Davis, Causes and effects of culvert blockage during large storms, in: Global solutions for urban drainage, 2002, pp. 1–16.Google Scholar[27]S. Roso, M. Boyd, E. Rigby, R. VanDrie“Prediction of increased flooding in urban catchments due to debris blockage and flow diversionsProceedings Novatech (2004), pp. 8-13View Record in ScopusGoogle Scholar[28]C.-D. Jan, C.-L. ChenDebris flows caused by Typhoon Herb in Taiwanin Debris-Flow Hazards and Related Phenomena, Springer (2005), pp. 539-563CrossRefGoogle Scholar[29]L.W. Zevenbergen, P.F. Lagasse, P.E. Clopper, Effects of debris on bridge pier scour, in: World Environmental and Water Resources Congress 2007: Restoring Our Natural Habitat, 2007, pp. 1–10.Google Scholar[30]A. Barthelmess, E. Rigby, Estimating Culvert and Bridge Blockages-a Simplified Procedure, in: Proceedings of the 34th World Congress of the International Association for Hydro-Environment Research and Engineering: 33rd Hydrology and Water Resources Symposium and 10th Conference on Hydraulics in Water Engineering, Engineers Australia, 2011, pp. 39.Google Scholar[31]E. Rigby, A. Barthelmess, Culvert Blockage Mechanisms and their Impact on Flood Behaviour, in: Proceedings of the 34th World Congress of the International Association for Hydro-Environment Research and Engineering: 33rd Hydrology and Water Resources Symposium and 10th Conference on Hydraulics in Water Engineering, Engineers Australia, 2011, pp. 380.Google Scholar[32]G. Streftaris, N. Wallerstein, G. Gibson, S. ArthurModeling probability of blockage at culvert trash screens using Bayesian approachJ. Hydraul. Eng., 139 (7) (2012), pp. 716-726Google Scholar[33]R. Jaeger, T. LuckeInvestigating the relationship between rainfall intensity, catchment vegetation and debris mobilityInt. J. GEOMATE, 12 (33) (2017), pp. 22-29 Download PDFView Record in ScopusGoogle Scholar[34]S. Amiraslani, J. Fahimi, H. Mehdinezhad, The Numerical Investigation of Free Falling Jet’s Effect On the Scour of Plunge Pool, in: XVIII International conference on water resources, Tehran University, Iran, 2008.Google Scholar[35]A.W. Nielsen, X. Liu, B.M. Sumer, J. FredsøeFlow and bed shear stresses in scour protections around a pile in a currentCoast. Eng., 72 (2013), pp. 20-38ArticleDownload PDFView Record in ScopusGoogle Scholar[36]G. Epely-Chauvin, G. De Cesare, S. SchwindtNumerical modelling of plunge pool scour evolution in non-cohesive sedimentsEng. Appl. Comput. Fluid Mech., 8 (4) (2014), pp. 477-487 Download PDFCrossRefView Record in ScopusGoogle Scholar[37]H. Karami, H. Basser, A. Ardeshir, S.H. HosseiniVerification of numerical study of scour around spur dikes using experimental dataWater Environ. J., 28 (1) (2014), pp. 124-134CrossRefView Record in ScopusGoogle Scholar[38]S.-H. Oh, K.S. Lee, W.-M. JeongThree-dimensional experiment and numerical simulation of the discharge performance of sluice passageway for tidal power plantRenew. Energy, 92 (2016), pp. 462-473ArticleDownload PDFView Record in ScopusGoogle Scholar[39]M.A. Khodier, B.P. TullisExperimental and computational comparison of baffled-culvert hydrodynamics for fish passageJ. Appl. Water Eng. Res. (2017), pp. 1-9CrossRefView Record in ScopusGoogle Scholar[40]F.S. Inc., FLOW-3D user’s manual, Flow Science, Inc., 2009.Google Scholar[41]G. Wei, J. Brethour, M. Grünzner, J. BurnhamSedimentation scour modelFlow Science Report, 7 (2014), pp. 1-29View Record in ScopusGoogle Scholar[42]R. Soulsby, R. Whitehouse, Threshold of sediment motion in coastal environments, in: Pacific Coasts and Ports’ 97: Proceedings of the 13th Australasian Coastal and Ocean Engineering Conference and the 6th Australasian Port and Harbour Conference, vol. 1, Centre for Advanced Engineering, University of Canterbury, 1997, pp. 145.Google Scholar[43]L.C.v. RijnSediment transport, part II: suspended load transportJ. Hydraul. Eng., 110 (11) (1984), pp. 1613-1641View Record in ScopusGoogle Scholar[44]S Y LIMScour below unsubmerged full-flowing culvert outletsProc. Instit. Civil Eng. – Water Marit. Energy, 112 (2) (1995), pp. 136-149 Record in ScopusGoogle Scholar

    Peer review under responsibility of Faculty of Engineering, Alexandria University.

    Figure 1- Schematic diagram of pooled stepped spillway conducted by Felder et al. (2012A): Notes: h step height (10 cm): w pool height (3.1 cm): l horizontal step length (20 cm): lw pool weir length (1.5 cm): d' is the water depth above the crest; y' is the distance normal to the crest invert

    Study of inception point, void fraction and pressure over pooled stepped spillways using Flow-3D

    Khosro Morovati , Afshin Eghbalzadeh 
    International Journal of Numerical Methods for Heat & Fluid Flow

    ISSN: 0961-5539

    Article publication date: 3 April 2018


    많은 계단식 배수로 지오메트리 설계 지침이 평평한 단계를 위해 개발되었지만 통합 단계를 설계하는 것이 더 효율적으로 작동하는 배수로에 대한 적절한 대안이 될 수 있습니다.

    이 논문은 POOL의 다른 높이에서 공기 연행과 보이드 비율의 시작점을 다루는 것을 목표로 합니다. 그 후, FLOW-3D 소프트웨어를 사용하여 POOL과 경사면의 높이를 다르게 하여 폭기된 지역과 폭기되지 않은 지역에서 압력 분포를 평가했습니다.

    얻어진 수치 결과와 실험 결과의 비교는 본 연구에 사용된 모든 방류에 대해 잘 일치했습니다. POOL 높이는 시작 지점 위치에 미미한 영향을 미쳤습니다. 공극률의 값은 높은 방류에 비해 낮은 방전에서 더 많은 영향을 받았습니다.

    여수로의 마루(통기되지 않은 지역)에서는 음압이 나타나지 않았으며 각 방류에서 마루를 따라 높이가 15cm인 수영장에서 최대 압력 값이 얻어졌습니다.

    모든 사면에서 웅덩이 및 평평한 계단형 여수로의 계단층 부근에서는 음압이 형성되지 않았습니다. 그러나 평단식 여수로에 비해 평단식 여수로의 수직면 부근에서 음압이 더 많이 형성되어 평단식 슈트에서 캐비테이션 현상이 발생할 확률이 증가하였습니다.

    Study of inception point, void fraction and pressure over pooled
    stWhile many stepped spillways geometry design guidelines were developed for flat steps, designing pooled steps might be an appropriate alternative to spillways working more efficiency. This paper aims to deal with the inception point of air-entrainment and void fraction in the different height of the pools. Following that, pressure distribution was evaluated in aerated and non-aerated regions under the effect of different heights of the pools and slopes through the use of the FLOW-3D software. Comparison of obtained numerical results with experimental ones was in good agreement for all discharges used in this study. Pools height had the insignificant effect on the inception point location. The value of void fraction was more affected in lower discharges in comparison with higher ones. Negative pressure was not seen over the crest of spillway (non-aerated region), and the maximum pressure values were obtained for pools with 15 cm height along the crest in each discharge. In all slopes, negative pressure was not formed near the step bed in the pooled and flat stepped spillways. However, negative pressure was formed in more area near the vertical face in the flat stepped spillway compared with the pooled stepped spillway which increases the probability of cavitation phenomenon in the flat stepped chute.


    압력, 공극률 및 시작점을 평가하기 위해 POOL된 계단식 여수로가 사용되었습니다. 또한 POOL의 다른 높이가 사용되었습니다. 이 연구의 수치 시뮬레이션은 Flow-3D 소프트웨어를 통해 수행되었습니다. 얻어진 결과는 풀이 압력, 공극률 및 시작점을 포함한 2상 유동 특성에 영향을 미칠 수 있음을 나타냅니다.


    마루 위에는 음압이 보이지 않았습니다. 압력 값은 사용된 모든 높이와 15cm 높이에서 얻은 최대 값에 대해 다릅니다. 또한, 풀링 스텝은 플랫 케이스에 비해 음압점 감소에 더 효과적인 역할을 하였습니다. 시작 지점 위치는 특히 9 및 15cm 높이에 대해 스키밍 흐름 영역과 비교하여 낮잠 및 전환 흐름 영역에서 더 많은 영향을 받았습니다.



    Morovati, K. and Eghbalzadeh, A. (2018), “Study of inception point, void fraction and pressure over pooled stepped spillways using Flow-3D”, International Journal of Numerical Methods for Heat & Fluid Flow, Vol. 28 No. 4, pp. 982-998.

    Figure 1- Schematic diagram of pooled stepped spillway conducted by Felder et al. (2012A): Notes: h  step height (10 cm): w pool height (3.1 cm): l horizontal step length (20 cm): lw pool weir length (1.5 cm):  d' is the water depth above the crest; y' is the distance normal to the crest invert
    Figure 1- Schematic diagram of pooled stepped spillway conducted by Felder et al. (2012A): Notes: h step height (10 cm): w pool height (3.1 cm): l horizontal step length (20 cm): lw pool weir length (1.5 cm): d’ is the water depth above the crest; y’ is the distance normal to the crest invert
    Figure 2- meshing domain and distribution of blocks
    Figure 2- meshing domain and distribution of blocks
    Figure 3- Comparison of numerical simulation with experimental data by Felder et al. (2012A);  mesh convergence analysis; pooled stepped spillway (slope: 26.6 0 )
    Figure 3- Comparison of numerical simulation with experimental data by Felder et al. (2012A); mesh convergence analysis; pooled stepped spillway (slope: 26.6 0 )
    Figure 4- Comparison of numerical simulation with experimental data by Felder et al. (2012A);  Flat stepped spillway (slope: 0 26 6. )
    Figure 4- Comparison of numerical simulation with experimental data by Felder et al. (2012A); Flat stepped spillway (slope: 0 26 6. )
    Figure 5-Comparison of numerical simulation with experimental data by Felder et al. (2012B); pooled  and flat stepped spillways (slope: 0 9.8 )
    Figure 5-Comparison of numerical simulation with experimental data by Felder et al. (2012B); pooled and flat stepped spillways (slope: 0 9.8 )
    Figure 6- TKE distribution on steps 8, 9 and 10 for four different mesh numbers: 261252 (model 1),  288941 (model 2), 323578 (model 3) and 343154 (model 4)
    Figure 6- TKE distribution on steps 8, 9 and 10 for four different mesh numbers: 261252 (model 1), 288941 (model 2), 323578 (model 3) and 343154 (model 4)
    Figure 7- Comparison of obtained Void fraction distribution on step 10 in numerical simulation with  experimental work conducted by Felder et al. (2012A); (slope 26.60 )
    Figure 7- Comparison of obtained Void fraction distribution on step 10 in numerical simulation with experimental work conducted by Felder et al. (2012A); (slope 26.60 )
    Figure 8- Results of inception point of air entrainment in different height of the pools: comparison with  empirical correlations (Eqs 8-9), experimental (Felder et al. (2012A)) and numerical data
    Figure 8- Results of inception point of air entrainment in different height of the pools: comparison with empirical correlations (Eqs 8-9), experimental (Felder et al. (2012A)) and numerical data
    Figure 9- Void fraction distribution for different pool heights on steps 10; slope 26.6 0
    Figure 9- Void fraction distribution for different pool heights on steps 10; slope 26.6 0
    Figure 10- Comparison of pressure distribution between numerical simulation and experimental work  conducted by Zhang and Chanson (2016); flat stepped spillway (slope: 0 45 )
    Figure 10- Comparison of pressure distribution between numerical simulation and experimental work conducted by Zhang and Chanson (2016); flat stepped spillway (slope: 0 45 )
    Figure 11- A comparison of the pressure distribution above the crest of the spillway; B comparison of the  free surface profile along the crest of the spillway.  Note: x' indicates the longitudinal distance from the starting point of the crest.
    Figure 11- A comparison of the pressure distribution above the crest of the spillway; B comparison of the free surface profile along the crest of the spillway. Note: x’ indicates the longitudinal distance from the starting point of the crest.
    Figure 12- pressure distribution along crest of spillway in different discharges; slope 26.6
    Figure 12- pressure distribution along crest of spillway in different discharges; slope 26.6
    Figure 13- Pressure distribution near the last step bed for different slopes and discharges: x'' indicatesthe  longitudinal distance from the intersection of the horizontal and vertical faces of step 10; y" is the distance from the intersection of the horizontal and vertical faces in the vertical direction
    Figure 13- Pressure distribution near the last step bed for different slopes and discharges: x” indicatesthe longitudinal distance from the intersection of the horizontal and vertical faces of step 10; y” is the distance from the intersection of the horizontal and vertical faces in the vertical direction
    Figure 14- Pressure distribution adjacent the vertical face of step 9 for different discharges and slopes
    Figure 14- Pressure distribution adjacent the vertical face of step 9 for different discharges and slopes
    Table1- Used discharges for assessments of mesh convergence analysis and hydraulic  characteristics
    Table1- Used discharges for assessments of mesh convergence analysis and hydraulic characteristics


    본 연구에서는 자유표면을 모사하기 위해 VOF 방법과 k -ε (RNG) 난류 모델을 활용하여 FLOW-3D 소프트웨어를 사용하였고, 계단식 배수로의 유동을 모사하기 위한 목적으로 난류 특성을 모사하였다. 얻은 결과는 수치 모델이 시작점 위치, 보이드 비율 및 압력을 적절하게 시뮬레이션했음을 나타냅니다. 풀의 높이는 공기 유입 위치에 미미한 영향을 미치므로 얻은 결과는 이 문서에서 제시된 상관 관계와 잘 일치했습니다. 즉, 사용 가능한 상관 관계를 서로 다른 풀 높이에 사용할 수 있습니다. 공극률의 결과는 스텝 풀 근처의 나프 유동 영역에서 공극율 값이 다른 배출보다 더 큰 것으로 나타났다. 더욱이 고방출량 .0 113m3/s에서 수영장 높이를 변경해도 수영장 표면 근처의 공극률 값에는 영향을 미치지 않았습니다.

    낮잠 및 전환 체제의 압력 분포에 대한 0 및 3cm 높이의 수영장 효과는 많은 지점에서 대부분 유사했습니다. 더욱이 조사된 모든 높이에서 여수로의 마루를 따라 부압이 없었습니다. 여수로 끝단의 바닥 부근의 압력 결과는 평평하고 고인 경우 부압이 발생하지 않았음을 나타냅니다. 수직면 부근의 음압은 웅덩이에 비해 평평한 계단형 여수로의 깊이(w=0 cm)의 대부분에서 발생하였다. 또한 더 큰 사면에 대한 풀링 케이스에서 음압이 제거되었습니다. 평단식 여수로에서는 계단의 수직면에 인접한 더 넓은 지역에서 음압이 발생하였기 때문에 이 여수로에서는 고형단식여수로보다 캐비테이션 현상이 발생할 가능성이 더 큽니다.

    In this study, the FLOW-3D software was used through utilizing the VOF method and k −ε (RNG) turbulence model in order to simulate free surface, and turbulence characteristics for the purpose of simulating flow over pooled stepped spillway. The results obtained indicated that the numerical model properly simulated the inception point location, void fraction, and pressure. The height of the pools has the insignificant effect on the location of air entrainment, so that obtained results were in good agreement with the correlations presented in this paper. In other words, available correlations can be used for different pool heights. The results of void fraction showed that the void fraction values in nappe flow regime near the step pool were more than the other discharges. Furthermore in high discharge, 0.113m3/s, altering pool height had no effect on the value of void fraction near the pool surface.

    The effect of the pools with 0 and 3 cm heights over the pressure distribution in nappe and transition regimes was mostly similar in many points. Furthermore, in all examined heights there was no negative pressure along the crest of the spillway. The pressure results near the bed of the step at the end of the spillway indicated that negative pressure did not occur in the flat and pooled cases. Negative pressure near the vertical face occurred in the most part of the depth in the flat stepped spillway (w=0 cm) in comparison with the pooled case. Also, the negative pressure was eliminated in the pooled case for the larger slopes. Since negative pressure occurred in a larger area adjacent the vertical face of the steps in the flat stepped spillways, it is more likely that cavitation phenomenon occurs in this spillway rather than the pooled stepped spillways.


    1. André, S. (2004), “High velocity aerated flows on stepped chutes with macro-roughness elements.” Ph.D. thesis,
      Laboratoire de Constructions Hydraulics (LCH), EPFL, Lausanne, Switzerland, 272 pages.
    2. Attarian, A. Hosseini, Kh. Abdi, H and Hosseini, M. (2014), “The Effect of the Step Height on Energy
      Dissipation in Stepped Spillways Using Numerical Simulation”. Arabian Journal for Science and
      Engineering, 39(4), 2587-2594.
    3. Bombardelli, F.A. Meireles. I. Matos, J. (2011), “Laboratory measurements and multi-block numerical
      simulations of the mean flow and turbulence in the non-aerated skimming flow region of steep stepped
      spillways”. Environmental fluid mechanics, 11(3) 263-288.
    4. Chakib, B. (2013), “Numerical Computation of Inception Point Location for Flat-sloped Stepped Spillway”.
      International Journal of Hydraulic Engineering; 2(3): 47-52.
    5. Chakib, B. Mohammed, H. (2015), “Numerical Simulation of Air Entrainment for Flat-Sloped Stepped Spillway.
      Journal of computational multiphase flows”, Volume 7. Number 1.
    6. Chanson, H. Toombes, L. (2002), “Air–water flows down stepped chutes: turbulence and flow structure
      observations”. International Journal of Multiphase Flow, 28(11) 1737-1761
    7. Chen, Q. Dai, G. Liu, H. (2002), “Volume of Fluid Model for Turbulence Numerical Simulation
      of Stepped Spillway Overflow”. DOI: 10.1061/(ASCE)0733-9429128:7(683).
    8. Cheng, X. Chen, Y. Luo, L. (2006), “Numerical simulation of air-water two-phase flow over stepped spillways”.
      Science in China Series E: Technological Sciences, 49(6), 674-684.
    9. Cheng, X. Luo, L. Zhao, W. (2004), “Study of aeration in the water flow over stepped spillway”. In: Proceedings
      of the world water congress.
    10. Chinnarasri, Ch. Kositgittiwong, D. Julien, Y. (2013), “Model of flow over spillways by computational fluid
      dynamics”. Proceedings of the ICE – Water Management, Volume 167(3) 164 –175.
    11. Dastgheib, A. Niksokhan, M.H. and Nowroozpour, A.R. (2012), “Comparing of Flow Pattern and Energy
      Dissipation over different forms of Stepped Spillway”. World Environmental and Water Resources
      Congress ASCE.
    12. Eghbalzadeh, A. Javan, M. (2012), “Comparison of mixture and VOF models for numerical simulation of air
      entrainment in skimming flow over stepped spillway”. Procedia Engineering, 28. 657-660.
    13. Felder, S, Chanson, H. (2012), “Free-surface Profiles, Velocity and Pressure Distributions on a
      Broad-Crested Weir: a Physical study “Free-surface Profiles, Velocity and Pressure Distributions on a
      Broad-Crested Weir: a Physical study
    14. Felder, S. Fromm, Ch. Chanson, H. (2012B), “Air entrainment and energy dissipation on a 8.9 slope stepped
      spillway with flat and pooled steps”, School of Civil Engineering, The University of Queensland,.
      Brisbane, Australia.
    15. Felder, S. Chanson, H. (2014A), Triple decomposition technique in air–water flows: application to instationary
      flows on a stepped spillway. International Journal of Multiphase Flow, 58, 139-153.
    16. Felder, S. Chanson, H. (2014B), Effects of step pool porosity upon flow aeration and energy dissipation on
      pooled stepped spillways. Journal of Hydraulic Engineering, 140(4), 04014002.
    17. Felder, S. Chanson, H. (2013A), “Air entrainment and energy dissipation on porous pooled stepped spillways”.
      Paper presented at the International Workshop on Hydraulic Design of Low-Head Structures.
    18. Felder, S. Chanson, H. (2013B), “Aeration, flow instabilities, and residual energy on pooled stepped spillways of
      embankment dams”. Journal of irrigation and drainage engineering, 139(10) 880-887.
    19. Felder, S. Guenther, Ph. Chanson, H. (2012A). “Air-water flow properties and energy dissipation on stepped
      spillways: a physical study of several pooled stepped configurations”, School of Civil Engineering, The
      University of Queensland,. Brisbane, Australia.
    20. Flow Science, (2013). “FLOW-3D user’s manual”, version 10.1. Flow Science, Inc, Los Alamos.
    21. Frizell, K.W. Renna, F.M. Matos, J. (2012), “Cavitation potential of flow on stepped spillways”. Journal of
      Hydraulic Engineering, 139(6), 630-636.
    22. Gonzalez, C. (2005), “An experimental study of free-surface aeration on embankment stepped chutes”,
      department of civil engineering, Brisbane, Australia, Phd thesis.
    23. Gonzalez, C.A. Chanson, H. (2008), “Turbulence manipulation in air–water flows on a stepped chute: An
      experimental study”. European Journal of Mechanics-B/Fluids, 27(4), 388-408.
    24. Guenther, Ph.. Felder, S. Chanson, H. (2013), “Flow aeration, cavity processes and energy dissipation on flat and
      pooled stepped spillways for embankments”. Environmental fluid mechanics, 13(5) 503-525.
    25. Hamedi, A. Mansoori, A. Malekmohamadi, I. Roshanaei, H. (2011), “Estimating Energy Dissipation in Stepped
      Spillways with Reverse Inclined Steps and End Sill”. World Environmental and Water Resources
      Congress, ASCE.
    26. Hirt, C.W. (2003), “Modeling Turbulent Entrainment of Air at a Free Surface”. Flow Science Inc.
    27. Hunt, S.L. Kadavy, K.C. (2013), “Inception point for enbankment dam stepped spillway”. J. Hydraul. Eng.,
      139(1), 60–64.
    28. Hunt, S.L. Kadavy, K.C. (2010), “Inception Point Relationship for Flat-Sloped Stepped
      Spillways”. DOI: 10.1061/ASCEHY.1943-7900.0000297.
    29. Matos, J. Quintela, A. (2000), “Air entrainment and safety against cavitation damage in stepped spillways over
      RCC dams. In: Proceeding Intl. Workshop on Hydraulics of Stepped Spillways”, VAW, ETH-Zurich, H.E.
      Minor and W.H. Hager. Balkema. 69–76.
    30. Meireles, I. Matos, J. (2009), “Skimming flow in the nonaerated region of stepped spillways over embankment
      dams”. J. Hydraul. Eng., 135(8), 685–689.
    31. Miang-liang, ZH. Yong-ming, SH. (2008), “Three dimentional simulation of meandering river basin on 3-D
      RNG k − ε turbulence model”. Journal of hydrodynamics, 20(4): 448-455.
    32. Morovati, Kh. Eghbalzadeh, A. Javan, M. (2015), “Numerical investigation of the configuration of the pools on
      the flowPattern passing over pooled stepped spillway in skimming flow regime. Acta Mech, DOI
    33. Morovati, Kh. Eghbalzadeh, A. Soori, S. (2016), “Numerical Study of Energy Dissipation of Pooled Stepped
      spillway”. Civil Engineering Journal. Vol. 2, No. 5.
    34. Nikseresht, A.H. Talebbeydokhti, N. and Rezaei, M.J. (2013), “Numerical simulation of two-phase flow on steppool spillways”. Scientia Iranica, A 20 (2), 222–230.
    35. Peyras, L. Royet, P. Degoutte, G. (1990), “Flow and energy dissipation over stepped gabion weirs”. ASCE
    36. Qun, Ch. Guang-qing, D. Feu-qing, Zh. Qing, Y. (2004). “Three-dimensional turbulence numerical simulation of
      a stepped spillway overflow”. Journal of hydrodynamics, Ser. B, 1, 74-79.
    37. Relvas, A. T. Pinheiro, A. N. (2008), Inception point and air concentration in flows on stepped chutes lined with
      wedge-shaped concrete blocks. Journal of Hydraulic Engineering, 134(8), 1042-1051
    38. Sanchez, M. (2000), “Pressure field in skimming flow over a stepped spillways”. In: Proceeding Intl. Workshop
      on Hydraulics of Stepped Spillways, VAW, ETH-Zurich, H.E. Minor and W.H. Hager. Balkema,
    39. Sarfaraz, M. Attari, J. Pfister, M. (2012), “Numerical Computation of Inception Point Location for Steeply
      Sloping Stepped Spillways”. 9th International Congress on Civil Engineering, May 8-10. Isfahan
      University of Technology (IUT), Isfahan, Iran.
    40. Savage, Bruce M. Michael C. Johnson. (2001), “Flow over ogee spillway: Physical and numerical model case
      study.” Journal of Hydraulic Engineering 127.8:640-649.
    41. Shahhedari, H. Jafari Nodoshan, E. Barati, R. Azhdary moghadam, M. (2014). “Discharge coeficient and energy
      dissipation over stepped spillway under skimming flow regime”. KSCE Journal of Civil Engineering, DOI
    42. Tabbara, M. Chatila, J. Awwad, R. (2005), “Computational simulation of flow over stepped spillways”.
      Computers & structures, 83(27) 2215-2224.
    43. Thorwarth, J. (2008), “Hydraulisches Verhalten der Treppengerinne mit eingetieften Stufen—Selbstinduzierte
      Abflussinstationaritäten und Energiedissipation” [Hydraulics of pooled stepped spillways— Self-induced
      unsteady flow and energy dissipation]. Ph.D. thesis, Univ. of Aachen, Aachen, Germany (in German).
    44. WeiLin, XU. ShuJing, LUO, QiuWen, ZH. Jing, LUO. (2015), “Experimental study on pressure and aeration
      characteristics in stepped chute flows. SCIENCE CHINA. Vol.58 No.4: 720–726. doi: 10.1007/s11431-015-
    45. Xiangju, Ch. Yongcan, C. Lin, L. (2006), “Numerical simulation of air-water two-phase flow over stepped
      spillways”. Science in China Series E: Technological Sciences, 49(6), 674-684.
    46. Zare, K.H. Doering, J.C. (2012), “Inception Point of Air Entrainment and Training Wall
      Characteristics of Baffles and Sills on Stepped Spillways”. DOI: 10.1061/(ASCE)HY
    47. Zhan, J. Zhang, J. Gong, Y. (2016), “Numerical investigation of air-entrainment in skimming flow over stepped
      spillways”. Theoretical and Applied Mechanics Letters. Volume 6. Pages 139–142.
    48. Zhang, G. Chanson, H. (2016), Hydraulics of the developing flow region of stepped spillways. II: Pressure and
      velocity fields. Journal of Hydraulic Engineering, 142(7).
    49. Zhenwei, M. Zhiyan, Zh. Tao, Zh. (2012), “Numerical Simulation of 3-D Flow Field of Spillway based on VOF
      Method”. Procedia Engineering, 28, 808-812.
    50. Zhi-yong, D. Hun-wei, L.J. (2006), “Numerical simulation of skimming flow over mild stepped channel”.
      Journal of Hydrodynamics, Ser. B, 18(3) 367-371.
    51. ZhongDong, Q. XiaoQing, H. WenXin, H. António, A. (2009), “Numerical simulation and analysis of water
      flow over stepped spillways”. Science in China Series E: Technological Sciences, 52(7) 1958-1965.
    Fig. 1. (a) Dimensions of the casting with runners (unit: mm), (b) a melt flow simulation using Flow-3D software together with Reilly's model[44], predicted that a large amount of bifilms (denoted by the black particles) would be contained in the final casting. (c) A solidification simulation using Pro-cast software showed that no shrinkage defect was contained in the final casting.

    AZ91 합금 주물 내 연행 결함에 대한 캐리어 가스의 영향

    Effect of carrier gases on the entrainment defects within AZ91 alloy castings

    Tian Liab J.M.T.Daviesa Xiangzhen Zhuc
    aUniversity of Birmingham, Birmingham B15 2TT, United Kingdom
    bGrainger and Worrall Ltd, Bridgnorth WV15 5HP, United Kingdom
    cBrunel Centre for Advanced Solidification Technology, Brunel University London, Kingston Ln, London, Uxbridge UB8 3PH, United Kingdom


    An entrainment defect (also known as a double oxide film defect or bifilm) acts a void containing an entrapped gas when submerged into a light-alloy melt, thus reducing the quality and reproducibility of the final castings. Previous publications, carried out with Al-alloy castings, reported that this trapped gas could be subsequently consumed by the reaction with the surrounding melt, thus reducing the void volume and negative effect of entrainment defects. Compared with Al-alloys, the entrapped gas within Mg-alloy might be more efficiently consumed due to the relatively high reactivity of magnesium. However, research into the entrainment defects within Mg alloys has been significantly limited. In the present work, AZ91 alloy castings were produced under different carrier gas atmospheres (i.e., SF6/CO2, SF6/air). The evolution processes of the entrainment defects contained in AZ91 alloy were suggested according to the microstructure inspections and thermodynamic calculations. The defects formed in the different atmospheres have a similar sandwich-like structure, but their oxide films contained different combinations of compounds. The use of carrier gases, which were associated with different entrained-gas consumption rates, affected the reproducibility of AZ91 castings.


    Magnesium alloyCastingOxide film, Bifilm, Entrainment defect, Reproducibility

    연행 결함(이중 산화막 결함 또는 이중막 결함이라고도 함)은 경합금 용융물에 잠길 때 갇힌 가스를 포함하는 공극으로 작용하여 최종 주물의 품질과 재현성을 저하시킵니다. Al-합금 주조로 수행된 이전 간행물에서는 이 갇힌 가스가 주변 용융물과의 반응에 의해 후속적으로 소모되어 공극 부피와 연행 결함의 부정적인 영향을 줄일 수 있다고 보고했습니다. Al-합금에 비해 마그네슘의 상대적으로 높은 반응성으로 인해 Mg-합금 내에 포집된 가스가 더 효율적으로 소모될 수 있습니다. 그러나 Mg 합금 내 연행 결함에 대한 연구는 상당히 제한적이었습니다. 현재 작업에서 AZ91 합금 주물은 다양한 캐리어 가스 분위기(즉, SF 6 /CO2 , SF 6 / 공기). AZ91 합금에 포함된 엔트레인먼트 결함의 진화 과정은 미세조직 검사 및 열역학적 계산에 따라 제안되었습니다. 서로 다른 분위기에서 형성된 결함은 유사한 샌드위치 구조를 갖지만 산화막에는 서로 다른 화합물 조합이 포함되어 있습니다. 다른 동반 가스 소비율과 관련된 운반 가스의 사용은 AZ91 주물의 재현성에 영향을 미쳤습니다.


    마그네슘 합금주조Oxide film, Bifilm, Entrainment 불량, 재현성

    1 . 소개

    지구상에서 가장 가벼운 구조용 금속인 마그네슘은 지난 수십 년 동안 가장 매력적인 경금속 중 하나가 되었습니다. 결과적으로 마그네슘 산업은 지난 20년 동안 급속한 발전을 경험했으며 [1 , 2] , 이는 전 세계적으로 Mg 합금에 대한 수요가 크게 증가했음을 나타냅니다. 오늘날 Mg 합금의 사용은 자동차, 항공 우주, 전자 등의 분야에서 볼 수 있습니다. [3 , 4] . Mg 금속의 전 세계 소비는 특히 자동차 산업에서 앞으로 더욱 증가할 것으로 예측되었습니다. 기존 자동차와 전기 자동차 모두의 에너지 효율성 요구 사항이 설계를 경량화하도록 더욱 밀어붙이기 때문입니다 [3 , 56] .

    Mg 합금에 대한 수요의 지속적인 성장은 Mg 합금 주조의 품질 및 기계적 특성 개선에 대한 광범위한 관심을 불러일으켰습니다. Mg 합금 주조 공정 동안 용융물의 표면 난류는 소량의 주변 대기를 포함하는 이중 표면 필름의 포획으로 이어질 수 있으므로 동반 결함(이중 산화막 결함 또는 이중막 결함이라고도 함)을 형성합니다. ) [7] , [8] , [9] , [10] . 무작위 크기, 수량, 방향 및 연행 결함의 배치는 주조 특성의 변화와 관련된 중요한 요인으로 널리 받아들여지고 있습니다 [7] . 또한 Peng et al. [11]AZ91 합금 용융물에 동반된 산화물 필름이 Al 8 Mn 5 입자에 대한 필터 역할을 하여 침전될 때 가두는 것을 발견했습니다 . Mackie et al. [12]는 또한 동반된 산화막이 금속간 입자를 트롤(trawl)하는 작용을 하여 입자가 클러스터링되어 매우 큰 결함을 형성할 수 있다고 제안했습니다. 금속간 화합물의 클러스터링은 비말동반 결함을 주조 특성에 더 해롭게 만들었습니다.

    연행 결함에 관한 이전 연구의 대부분은 Al-합금에 대해 수행되었으며 [7 , [13] , [14] , [15] , [16] , [17] , [18] 몇 가지 잠재적인 방법이 제안되었습니다. 알루미늄 합금 주물의 품질에 대한 부정적인 영향을 줄이기 위해. Nyahumwa et al., [16] 은 연행 결함 내의 공극 체적이 열간 등방압 압축(HIP) 공정에 의해 감소될 수 있음을 보여줍니다. Campbell [7] 은 결함 내부의 동반된 가스가 주변 용융물과의 반응으로 인해 소모될 수 있다고 제안했으며, 이는 Raiszedeh와 Griffiths [19]에 의해 추가로 확인되었습니다 ..혼입 가스 소비가 Al-합금 주물의 기계적 특성에 미치는 영향은 [8 , 9]에 의해 조사되었으며 , 이는 혼입 가스의 소비가 주조 재현성의 개선을 촉진함을 시사합니다.

    Al-합금 내 결함에 대한 조사와 비교하여 Mg-합금 내 연행 결함에 대한 연구는 상당히 제한적입니다. 연행 결함의 존재는 Mg 합금 주물 [20 , 21] 에서 입증 되었지만 그 거동, 진화 및 연행 가스 소비는 여전히 명확하지 않습니다.

    Mg 합금 주조 공정에서 용융물은 일반적으로 마그네슘 점화를 피하기 위해 커버 가스로 보호됩니다. 따라서 모래 또는 매몰 몰드의 공동은 용융물을 붓기 전에 커버 가스로 세척해야 합니다 [22] . 따라서, Mg 합금 주물 내의 연행 가스는 공기만이 아니라 주조 공정에 사용되는 커버 가스를 포함해야 하며, 이는 구조 및 해당 연행 결함의 전개를 복잡하게 만들 수 있습니다.

    SF 6 은 Mg 합금 주조 공정에 널리 사용되는 대표적인 커버 가스입니다 [23] , [24] , [25] . 이 커버 가스는 유럽의 마그네슘 합금 주조 공장에서 사용하도록 제한되었지만 상업 보고서에 따르면 이 커버는 전 세계 마그네슘 합금 산업, 특히 다음과 같은 글로벌 마그네슘 합금 생산을 지배한 국가에서 여전히 인기가 있습니다. 중국, 브라질, 인도 등 [26] . 또한, 최근 학술지 조사에서도 이 커버가스가 최근 마그네슘 합금 연구에서 널리 사용된 것으로 나타났다 [27] . SF 6 커버 가스 의 보호 메커니즘 (즉, 액체 Mg 합금과 SF 6 사이의 반응Cover gas)에 대한 연구는 여러 선행연구자들에 의해 이루어졌으나 표면 산화막의 형성과정이 아직 명확하게 밝혀지지 않았으며, 일부 발표된 결과들도 상충되고 있다. 1970년대 초 Fruehling [28] 은 SF 6 아래에 형성된 표면 피막이 주로 미량의 불화물과 함께 MgO 임을 발견 하고 SF 6 이 Mg 합금 표면 피막에 흡수 된다고 제안했습니다 . Couling [29] 은 흡수된 SF 6 이 Mg 합금 용융물과 반응하여 MgF 2 를 형성함을 추가로 확인했습니다 . 지난 20년 동안 아래에 자세히 설명된 것처럼 Mg 합금 표면 필름의 다양한 구조가 보고되었습니다.(1)

    단층 필름 . Cashion [30 , 31] 은 X선 광전자 분광법(XPS)과 오제 분광법(AES)을 사용하여 표면 필름을 MgO 및 MgF 2 로 식별했습니다 . 그는 또한 필름의 구성이 두께와 전체 실험 유지 시간에 걸쳐 일정하다는 것을 발견했습니다. Cashion이 관찰한 필름은 10분에서 100분의 유지 시간으로 생성된 단층 구조를 가졌다.(2)

    이중층 필름 . Aarstad et. al [32] 은 2003년에 이중층 표면 산화막을 보고했습니다. 그들은 예비 MgO 막에 부착된 잘 분포된 여러 MgF 2 입자를 관찰 하고 전체 표면적의 25-50%를 덮을 때까지 성장했습니다. 외부 MgO 필름을 통한 F의 내부 확산은 진화 과정의 원동력이었습니다. 이 이중층 구조는 Xiong의 그룹 [25 , 33] 과 Shih et al. 도 지지했습니다 . [34] .(삼)

    트리플 레이어 필름 . 3층 필름과 그 진화 과정은 Pettersen [35]에 의해 2002년에 보고되었습니다 . Pettersen은 초기 표면 필름이 MgO 상이었고 F의 내부 확산에 의해 점차적으로 안정적인 MgF 2 상 으로 진화한다는 것을 발견했습니다 . 두꺼운 상부 및 하부 MgF 2 층.(4)

    산화물 필름은 개별 입자로 구성 됩니다. Wang et al [36] 은 Mg-alloy 표면 필름을 SF 6 커버 가스 하에서 용융물에 교반 한 다음 응고 후 동반된 표면 필름을 검사했습니다. 그들은 동반된 표면 필름이 다른 연구자들이 보고한 보호 표면 필름처럼 계속되지 않고 개별 입자로 구성된다는 것을 발견했습니다. 젊은 산화막은 MgO 나노 크기의 산화물 입자로 구성되어 있는 반면, 오래된 산화막은 한쪽 면에 불화물과 질화물이 포함된 거친 입자(평균 크기 약 1μm)로 구성되어 있습니다.

    Mg 합금 용융 표면의 산화막 또는 동반 가스는 모두 액체 Mg 합금과 커버 가스 사이의 반응으로 인해 형성되므로 Mg 합금 표면막에 대한 위에서 언급한 연구는 진화에 대한 귀중한 통찰력을 제공합니다. 연행 결함. 따라서 SF 6 커버 가스 의 보호 메커니즘 (즉, Mg-합금 표면 필름의 형성)은 해당 동반 결함의 잠재적인 복잡한 진화 과정을 나타냅니다.

    그러나 Mg 합금 용융물에 표면 필름을 형성하는 것은 용융물에 잠긴 동반된 가스의 소비와 다른 상황에 있다는 점에 유의해야 합니다. 예를 들어, 앞서 언급한 연구에서 표면 성막 동안 충분한 양의 커버 가스가 담지되어 커버 가스의 고갈을 억제했습니다. 대조적으로, Mg 합금 용융물 내의 동반된 가스의 양은 유한하며, 동반된 가스는 완전히 고갈될 수 있습니다. Mirak [37] 은 3.5% SF 6 /기포를 특별히 설계된 영구 금형에서 응고되는 순수한 Mg 합금 용융물에 도입했습니다. 기포가 완전히 소모되었으며, 해당 산화막은 MgO와 MgF 2 의 혼합물임을 알 수 있었다.. 그러나 Aarstad [32] 및 Xiong [25 , 33]에 의해 관찰된 MgF 2 스팟 과 같은 핵 생성 사이트 는 관찰되지 않았습니다. Mirak은 또한 조성 분석을 기반으로 산화막에서 MgO 이전에 MgF 2 가 형성 되었다고 추측했는데 , 이는 이전 문헌에서 보고된 표면 필름 형성 과정(즉, MgF 2 이전에 형성된 MgO)과 반대 입니다. Mirak의 연구는 동반된 가스의 산화막 형성이 표면막의 산화막 형성과 상당히 다를 수 있음을 나타내었지만 산화막의 구조와 진화에 대해서는 밝히지 않았습니다.

    또한 커버 가스에 캐리어 가스를 사용하는 것도 커버 가스와 액체 Mg 합금 사이의 반응에 영향을 미쳤습니다. SF 6 /air 는 용융 마그네슘의 점화를 피하기 위해 SF 6 /CO 2 운반 가스 [38] 보다 더 높은 함량의 SF 6을 필요로 하여 다른 가스 소비율을 나타냅니다. Liang [39] 은 CO 2 가 캐리어 가스로 사용될 때 표면 필름에 탄소가 형성된다고 제안했는데 , 이는 SF 6 /air 에서 형성된 필름과 다릅니다 . Mg 연소 [40]에 대한 조사 에서 Mg 2 C 3 검출이 보고되었습니다.CO 2 연소 후 Mg 합금 샘플 에서 이는 Liang의 결과를 뒷받침할 뿐만 아니라 이중 산화막 결함에서 Mg 탄화물의 잠재적 형성을 나타냅니다.

    여기에 보고된 작업은 다양한 커버 가스(즉, SF 6 /air 및 SF 6 /CO 2 )로 보호되는 AZ91 Mg 합금 주물에서 형성된 연행 결함의 거동과 진화에 대한 조사 입니다. 이러한 캐리어 가스는 액체 Mg 합금에 대해 다른 보호성을 가지며, 따라서 상응하는 동반 가스의 다른 소비율 및 발생 프로세스와 관련될 수 있습니다. AZ91 주물의 재현성에 대한 동반 가스 소비의 영향도 연구되었습니다.

    2 . 실험

    2.1 . 용융 및 주조

    3kg의 AZ91 합금을 700 ± 5 °C의 연강 도가니에서 녹였습니다. AZ91 합금의 조성은 표 1 에 나타내었다 . 가열하기 전에 잉곳 표면의 모든 산화물 스케일을 기계가공으로 제거했습니다. 사용 된 커버 가스는 0.5 %이었다 SF 6 / 공기 또는 0.5 % SF 6 / CO 2 (부피. %) 다른 주물 6L / 분의 유량. 용융물은 15분 동안 0.3L/min의 유속으로 아르곤으로 가스를 제거한 다음 [41 , 42] , 모래 주형에 부었습니다. 붓기 전에 샌드 몰드 캐비티를 20분 동안 커버 가스로 플러싱했습니다 [22] . 잔류 용융물(약 1kg)이 도가니에서 응고되었습니다.

    표 1 . 본 연구에 사용된 AZ91 합금의 조성(wt%).


    그림 1 (a)는 러너가 있는 주물의 치수를 보여줍니다. 탑 필링 시스템은 최종 주물에서 연행 결함을 생성하기 위해 의도적으로 사용되었습니다. Green과 Campbell [7 , 43] 은 탑 필링 시스템이 바텀 필링 시스템에 비해 주조 과정에서 더 많은 연행 현상(즉, 이중 필름)을 유발한다고 제안했습니다. 이 금형의 용융 흐름 시뮬레이션(Flow-3D 소프트웨어)은 연행 현상에 관한 Reilly의 모델 [44] 을 사용하여 최종 주조에 많은 양의 이중막이 포함될 것이라고 예측했습니다( 그림 1 에서 검은색 입자로 표시됨) . NS).

    그림 1

    수축 결함은 또한 주물의 기계적 특성과 재현성에 영향을 미칩니다. 이 연구는 주조 품질에 대한 이중 필름의 영향에 초점을 맞추었기 때문에 수축 결함이 발생하지 않도록 금형을 의도적으로 설계했습니다. ProCAST 소프트웨어를 사용한 응고 시뮬레이션은 그림 1c 와 같이 최종 주조에 수축 결함이 포함되지 않음을 보여주었습니다 . 캐스팅 건전함도 테스트바 가공 전 실시간 X-ray를 통해 확인했다.

    모래 주형은 1wt를 함유한 수지 결합된 규사로 만들어졌습니다. % PEPSET 5230 수지 및 1wt. % PEPSET 5112 촉매. 모래는 또한 억제제로 작용하기 위해 2중량%의 Na 2 SiF 6 을 함유했습니다 .. 주입 온도는 700 ± 5 °C였습니다. 응고 후 러너바의 단면을 Sci-Lab Analytical Ltd로 보내 H 함량 분석(LECO 분석)을 하였고, 모든 H 함량 측정은 주조 공정 후 5일째에 실시하였다. 각각의 주물은 인장 강도 시험을 위해 클립 신장계가 있는 Zwick 1484 인장 시험기를 사용하여 40개의 시험 막대로 가공되었습니다. 파손된 시험봉의 파단면을 주사전자현미경(SEM, Philips JEOL7000)을 이용하여 가속전압 5~15kV로 조사하였다. 파손된 시험 막대, 도가니에서 응고된 잔류 Mg 합금 및 주조 러너를 동일한 SEM을 사용하여 단면화하고 연마하고 검사했습니다. CFEI Quanta 3D FEG FIB-SEM을 사용하여 FIB(집속 이온 빔 밀링 기술)에 의해 테스트 막대 파괴 표면에서 발견된 산화막의 단면을 노출했습니다. 분석에 필요한 산화막은 백금층으로 코팅하였다. 그런 다음 30kV로 가속된 갈륨 이온 빔이 산화막의 단면을 노출시키기 위해 백금 코팅 영역을 둘러싼 재료 기판을 밀링했습니다. 산화막 단면의 EDS 분석은 30kV의 가속 전압에서 FIB 장비를 사용하여 수행되었습니다.

    2.2 . 산화 세포

    전술 한 바와 같이, 몇몇 최근 연구자들은 마그네슘 합금의 용탕 표면에 형성된 보호막 조사 [38 , 39 , [46] , [47] , [48] , [49] , [50] , [51] , [52 ] . 이 실험 동안 사용된 커버 가스의 양이 충분하여 커버 가스에서 불화물의 고갈을 억제했습니다. 이 섹션에서 설명하는 실험은 엔트레인먼트 결함의 산화막의 진화를 연구하기 위해 커버 가스의 공급을 제한하는 밀봉된 산화 셀을 사용했습니다. 산화 셀에 포함된 커버 가스는 큰 크기의 “동반된 기포”로 간주되었습니다.

    도 2에 도시된 바와 같이 , 산화셀의 본체는 내부 길이가 400mm, 내경이 32mm인 폐쇄형 연강관이었다. 수냉식 동관을 전지의 상부에 감았습니다. 튜브가 가열될 때 냉각 시스템은 상부와 하부 사이에 온도 차이를 만들어 내부 가스가 튜브 내에서 대류하도록 했습니다. 온도는 도가니 상단에 위치한 K형 열전대로 모니터링했습니다. Nieet al. [53] 은 Mg 합금 용융물의 표면 피막을 조사할 때 SF 6 커버 가스가 유지로의 강철 벽과 반응할 것이라고 제안했습니다 . 이 반응을 피하기 위해 강철 산화 전지의 내부 표면(그림 2 참조)) 및 열전대의 상반부는 질화붕소로 코팅되었습니다(Mg 합금은 질화붕소와 ​​접촉하지 않았습니다).

    그림 2

    실험 중에 고체 AZ91 합금 블록을 산화 셀 바닥에 위치한 마그네시아 도가니에 넣었습니다. 전지는 1L/min의 가스 유속으로 전기 저항로에서 100℃로 가열되었다. 원래의 갇힌 대기(즉, 공기)를 대체하기 위해 셀을 이 온도에서 20분 동안 유지했습니다. 그런 다음, 산화 셀을 700°C로 더 가열하여 AZ91 샘플을 녹였습니다. 그런 다음 가스 입구 및 출구 밸브가 닫혀 제한된 커버 가스 공급 하에서 산화를 위한 밀폐된 환경이 생성되었습니다. 그런 다음 산화 전지를 5분 간격으로 5분에서 30분 동안 700 ± 10°C에서 유지했습니다. 각 유지 시간이 끝날 때 세포를 물로 켄칭했습니다. 실온으로 냉각한 후 산화된 샘플을 절단하고 연마한 다음 SEM으로 검사했습니다.

    3 . 결과

    3.1 . SF 6 /air 에서 형성된 엔트레인먼트 결함의 구조 및 구성

    0.5 % SF의 커버 가스 하에서 AZ91 주물에 형성된 유입 결함의 구조 및 조성 6 / 공기는 SEM 및 EDS에 의해 관찰되었다. 결과는 그림 3에 스케치된 엔트레인먼트 결함의 두 가지 유형이 있음을 나타냅니다 . (1) 산화막이 전통적인 단층 구조를 갖는 유형 A 결함 및 (2) 산화막이 2개 층을 갖는 유형 B 결함. 이러한 결함의 세부 사항은 다음에 소개되었습니다. 여기에서 비말동반 결함은 생물막 또는 이중 산화막으로도 알려져 있기 때문에 B형 결함의 산화막은 본 연구에서 “다층 산화막” 또는 “다층 구조”로 언급되었습니다. “이중 산화막 결함의 이중층 산화막”과 같은 혼란스러운 설명을 피하기 위해.

    그림 3

    그림 4 (ab)는 약 0.4μm 두께의 조밀한 단일층 산화막을 갖는 Type A 결함을 보여줍니다. 이 필름에서 산소, 불소, 마그네슘 및 알루미늄이 검출되었습니다( 그림 4c). 산화막은 마그네슘과 알루미늄의 산화물과 불화물의 혼합물로 추측됩니다. 불소의 검출은 동반된 커버 가스가 이 결함의 형성에 포함되어 있음을 보여주었습니다. 즉, Fig. 4 (a)에 나타난 기공 은 수축결함이나 수소기공도가 아니라 연행결함이었다. 알루미늄의 검출은 Xiong과 Wang의 이전 연구 [47 , 48] 와 다르며 , SF 6으로 보호된 AZ91 용융물의 표면 필름에 알루미늄이 포함되어 있지 않음을 보여주었습니다.커버 가스. 유황은 원소 맵에서 명확하게 인식할 수 없었지만 해당 ESD 스펙트럼에서 S-피크가 있었습니다.

    그림 4

    도 5 (ab)는 다층 산화막을 갖는 Type B 엔트레인먼트 결함을 나타낸다. 산화막의 조밀한 외부 층은 불소와 산소가 풍부하지만( 그림 5c) 상대적으로 다공성인 내부 층은 산소만 풍부하고(즉, 불소가 부족) 부분적으로 함께 성장하여 샌드위치 모양을 형성합니다. 구조. 따라서 외층은 불화물과 산화물의 혼합물이며 내층은 주로 산화물로 추정된다. 황은 EDX 스펙트럼에서만 인식될 수 있었고 요소 맵에서 명확하게 식별할 수 없었습니다. 이는 커버 가스의 작은 S 함량(즉, SF 6 의 0.5% 부피 함량 때문일 수 있음)커버 가스). 이 산화막에서는 이 산화막의 외층에 알루미늄이 포함되어 있지만 내층에서는 명확하게 검출할 수 없었다. 또한 Al의 분포가 고르지 않은 것으로 보입니다. 결함의 우측에는 필름에 알루미늄이 존재하지만 그 농도는 매트릭스보다 높은 것으로 식별할 수 없음을 알 수 있다. 그러나 결함의 왼쪽에는 알루미늄 농도가 훨씬 높은 작은 영역이 있습니다. 이러한 알루미늄의 불균일한 분포는 다른 결함(아래 참조)에서도 관찰되었으며, 이는 필름 내부 또는 아래에 일부 산화물 입자가 형성된 결과입니다.

    그림 5

    무화과 도 4 및 5 는 SF 6 /air 의 커버 가스 하에 주조된 AZ91 합금 샘플에서 형성된 연행 결함의 횡단면 관찰을 나타낸다 . 2차원 단면에서 관찰된 수치만으로 연행 결함을 특성화하는 것만으로는 충분하지 않습니다. 더 많은 이해를 돕기 위해 테스트 바의 파단면을 관찰하여 엔트레인먼트 결함(즉, 산화막)의 표면을 더 연구했습니다.

    Fig. 6 (a)는 SF 6 /air 에서 생산된 AZ91 합금 인장시험봉의 파단면을 보여준다 . 파단면의 양쪽에서 대칭적인 어두운 영역을 볼 수 있습니다. 그림 6 (b)는 어두운 영역과 밝은 영역 사이의 경계를 보여줍니다. 밝은 영역은 들쭉날쭉하고 부서진 특징으로 구성되어 있는 반면, 어두운 영역의 표면은 비교적 매끄럽고 평평했습니다. 또한 EDS 결과( Fig. 6 c-d 및 Table 2) 불소, 산소, 황 및 질소는 어두운 영역에서만 검출되었으며, 이는 어두운 영역이 용융물에 동반된 표면 보호 필름임을 나타냅니다. 따라서 어두운 영역은 대칭적인 특성을 고려할 때 연행 결함이라고 제안할 수 있습니다. Al-합금 주조물의 파단면에서 유사한 결함이 이전에 보고되었습니다 [7] . 질화물은 테스트 바 파단면의 산화막에서만 발견되었지만 그림 1과 그림 4에 표시된 단면 샘플에서는 검출되지 않았습니다 4 및 5 . 근본적인 이유는 이러한 샘플에 포함된 질화물이 샘플 연마 과정에서 가수분해되었을 수 있기 때문입니다 [54] .

    그림 6

    표 2 . EDS 결과(wt.%)는 그림 6에 표시된 영역에 해당합니다 (커버 가스: SF 6 /공기).

    그림 6 (b)의 어두운 영역3.481.3279.130.4713.630.570.080.73
    그림 6 (b)의 밝은 영역3.5884.4811.250.68

    도 1 및 도 2에 도시된 결함의 단면 관찰과 함께 도 4 및 도 5 를 참조하면, 인장 시험봉에 포함된 연행 결함의 구조를 도 6 (e) 와 같이 스케치하였다 . 결함에는 산화막으로 둘러싸인 동반된 가스가 포함되어 있어 테스트 바 내부에 보이드 섹션이 생성되었습니다. 파괴 과정에서 결함에 인장력이 가해지면 균열이 가장 약한 경로를 따라 전파되기 때문에 보이드 섹션에서 균열이 시작되어 연행 결함을 따라 전파됩니다 [55] . 따라서 최종적으로 시험봉이 파단되었을 때 Fig. 6 (a) 와 같이 시험봉의 양 파단면에 연행결함의 산화피막이 나타났다 .

    3.2 . SF 6 /CO 2 에 형성된 연행 결함의 구조 및 조성

    SF 6 /air 에서 형성된 엔트레인먼트 결함과 유사하게, 0.5% SF 6 /CO 2 의 커버 가스 아래에서 형성된 결함 도 두 가지 유형의 산화막(즉, 단층 및 다층 유형)을 가졌다. 도 7 (a)는 다층 산화막을 포함하는 엔트레인먼트 결함의 예를 도시한다. 결함에 대한 확대 관찰( 그림 7b )은 산화막의 내부 층이 함께 성장하여 SF 6 /air 의 분위기에서 형성된 결함과 유사한 샌드위치 같은 구조를 나타냄을 보여줍니다 ( 그림 7b). 5 나 ). EDS 스펙트럼( 그림 7c) 이 샌드위치형 구조의 접합부(내층)는 주로 산화마그네슘을 함유하고 있음을 보여주었다. 이 EDS 스펙트럼에서는 불소, 황, 알루미늄의 피크가 확인되었으나 그 양은 상대적으로 적었다. 대조적으로, 산화막의 외부 층은 조밀하고 불화물과 산화물의 혼합물로 구성되어 있습니다( 그림 7d-e).

    그림 7

    Fig. 8 (a)는 0.5%SF 6 /CO 2 분위기에서 제작된 AZ91 합금 인장시험봉의 파단면의 연행결함을 보여준다 . 상응하는 EDS 결과(표 3)는 산화막이 불화물과 산화물을 함유함을 보여주었다. 황과 질소는 검출되지 않았습니다. 게다가, 확대 관찰(  8b)은 산화막 표면에 반점을 나타내었다. 반점의 직경은 수백 나노미터에서 수 마이크론 미터까지 다양했습니다.

    그림 8

    산화막의 구조와 조성을 보다 명확하게 나타내기 위해 테스트 바 파단면의 산화막 단면을 FIB 기법을 사용하여 현장에서 노출시켰다( 그림 9 ). 도 9a에 도시된 바와 같이 , 백금 코팅층과 Mg-Al 합금 기재 사이에 연속적인 산화피막이 발견되었다. 그림 9 (bc)는 다층 구조( 그림 9c 에서 빨간색 상자로 표시)를 나타내는 산화막에 대한 확대 관찰을 보여줍니다 . 바닥층은 불소와 산소가 풍부하고 불소와 산화물의 혼합물이어야 합니다 . 5 와 7, 유일한 산소가 풍부한 최상층은 도 1 및 도 2에 도시 된 “내층”과 유사하였다 5 및 7 .

    그림 9

    연속 필름을 제외하고 도 9 에 도시된 바와 같이 연속 필름 내부 또는 하부에서도 일부 개별 입자가 관찰되었다 . 그림 9( b) 의 산화막 좌측에서 Al이 풍부한 입자가 검출되었으며, 마그네슘과 산소 원소도 풍부하게 함유하고 있어 스피넬 Mg 2 AlO 4 로 추측할 수 있다 . 이러한 Mg 2 AlO 4 입자의 존재는 Fig. 5 와 같이 관찰된 필름의 작은 영역에 높은 알루미늄 농도와 알루미늄의 불균일한 분포의 원인이 된다 .(씨). 여기서 강조되어야 할 것은 연속 산화막의 바닥층의 다른 부분이 이 Al이 풍부한 입자보다 적은 양의 알루미늄을 함유하고 있지만, 그림 9c는 이 바닥층의 알루미늄 양이 여전히 무시할 수 없는 수준임을 나타냅니다 . , 특히 필름의 외층과 비교할 때. 도 9b에 도시된 산화막의 우측 아래에서 입자가 검출되어 Mg와 O가 풍부하여 MgO인 것으로 추측되었다. Wang의 결과에 따르면 [56], Mg 용융물과 Mg 증기의 산화에 의해 Mg 용융물의 표면에 많은 이산 MgO 입자가 형성될 수 있다. 우리의 현재 연구에서 관찰된 MgO 입자는 같은 이유로 인해 형성될 수 있습니다. 실험 조건의 차이로 인해 더 적은 Mg 용융물이 기화되거나 O2와 반응할 수 있으므로 우리 작업에서 형성되는 MgO 입자는 소수에 불과합니다. 또한 필름에서 풍부한 탄소가 발견되어 CO 2 가 용융물과 반응하여 탄소 또는 탄화물을 형성할 수 있음을 보여줍니다 . 이 탄소 농도는 표 3에 나타낸 산화막의 상대적으로 높은 탄소 함량 (즉, 어두운 영역) 과 일치하였다 . 산화막 옆 영역.

    표 3 . 도 8에 도시된 영역에 상응하는 EDS 결과(wt.%) (커버 가스: SF 6 / CO 2 ).

    그림 8 (a)의 어두운 영역7.253.6469.823.827.030.86
    그림 8 (a)의 밝은 영역2.100.4482.8313.261.36

    테스트 바 파단면( 도 9 ) 에서 산화막의 이 단면 관찰은 도 6 (e)에 도시된 엔트레인먼트 결함의 개략도를 추가로 확인했다 . SF 6 /CO 2 와 SF 6 /air 의 서로 다른 분위기에서 형성된 엔트레인먼트 결함 은 유사한 구조를 가졌지만 그 조성은 달랐다.

    3.3 . 산화 전지에서 산화막의 진화

    섹션 3.1 및 3.2 의 결과 는 SF 6 /air 및 SF 6 /CO 2 의 커버 가스 아래에서 AZ91 주조에서 형성된 연행 결함의 구조 및 구성을 보여줍니다 . 산화 반응의 다른 단계는 연행 결함의 다른 구조와 조성으로 이어질 수 있습니다. Campbell은 동반된 가스가 주변 용융물과 반응할 수 있다고 추측했지만 Mg 합금 용융물과 포획된 커버 가스 사이에 반응이 발생했다는 보고는 거의 없습니다. 이전 연구자들은 일반적으로 개방된 환경에서 Mg 합금 용융물과 커버 가스 사이의 반응에 초점을 맞췄습니다 [38 , 39 , [46] , [47][48] , [49] , [50] , [51] , [52] , 이는 용융물에 갇힌 커버 가스의 상황과 다릅니다. AZ91 합금에서 엔트레인먼트 결함의 형성을 더 이해하기 위해 엔트레인먼트 결함의 산화막의 진화 과정을 산화 셀을 사용하여 추가로 연구했습니다.

    .도 10 (a 및 d) 0.5 % 방송 SF 보호 산화 셀에서 5 분 동안 유지 된 표면 막 (6) / 공기. 불화물과 산화물(MgF 2 와 MgO) 로 이루어진 단 하나의 층이 있었습니다 . 이 표면 필름에서. 황은 EDS 스펙트럼에서 검출되었지만 그 양이 너무 적어 원소 맵에서 인식되지 않았습니다. 이 산화막의 구조 및 조성은 도 4 에 나타낸 엔트레인먼트 결함의 단층막과 유사하였다 .

    그림 10

    10분의 유지 시간 후, 얇은 (O,S)가 풍부한 상부층(약 700nm)이 예비 F-농축 필름에 나타나 그림 10 (b 및 e) 에서와 같이 다층 구조를 형성했습니다 . ). (O, S)가 풍부한 최상층의 두께는 유지 시간이 증가함에 따라 증가했습니다. Fig. 10 (c, f) 에서 보는 바와 같이 30분간 유지한 산화막도 다층구조를 가지고 있으나 (O,S)가 풍부한 최상층(약 2.5μm)의 두께가 10분 산화막의 그것. 도 10 (bc) 에 도시 된 다층 산화막 은 도 5에 도시된 샌드위치형 결함의 막과 유사한 외관을 나타냈다 .

    도 10에 도시된 산화막의 상이한 구조는 커버 가스의 불화물이 AZ91 합금 용융물과의 반응으로 인해 우선적으로 소모될 것임을 나타내었다. 불화물이 고갈된 후, 잔류 커버 가스는 액체 AZ91 합금과 추가로 반응하여 산화막에 상부 (O, S)가 풍부한 층을 형성했습니다. 따라서 도 1 및 도 3에 도시된 연행 결함의 상이한 구조 및 조성 4 와 5 는 용융물과 갇힌 커버 가스 사이의 진행 중인 산화 반응 때문일 수 있습니다.

    이 다층 구조는 Mg 합금 용융물에 형성된 보호 표면 필름에 관한 이전 간행물 [38 , [46] , [47] , [48] , [49] , [50] , [51] 에서 보고되지 않았습니다 . . 이는 이전 연구원들이 무제한의 커버 가스로 실험을 수행했기 때문에 커버 가스의 불화물이 고갈되지 않는 상황을 만들었기 때문일 수 있습니다. 따라서 엔트레인먼트 결함의 산화피막은 도 10에 도시된 산화피막과 유사한 거동특성을 가지나 [38 ,[46] , [47] , [48] , [49] , [50] , [51] .

    SF 유지 산화막와 마찬가지로 6 / 공기, SF에 형성된 산화물 막 (6) / CO 2는 또한 세포 산화 다른 유지 시간과 다른 구조를 가지고 있었다. .도 11 (a)는 AZ91 개최 산화막, 0.5 %의 커버 가스 하에서 SF 표면 용융 도시 6 / CO 2, 5 분. 이 필름은 MgF 2 로 이루어진 단층 구조를 가졌다 . 이 영화에서는 MgO의 존재를 확인할 수 없었다. 30분의 유지 시간 후, 필름은 다층 구조를 가졌다; 내부 층은 조밀하고 균일한 외관을 가지며 MgF 2 로 구성 되고 외부 층은 MgF 2 혼합물및 MgO. 0.5%SF 6 /air 에서 형성된 표면막과 다른 이 막에서는 황이 검출되지 않았다 . 따라서, 0.5%SF 6 /CO 2 의 커버 가스 내의 불화물 도 막 성장 과정의 초기 단계에서 우선적으로 소모되었다. SF 6 /air 에서 형성된 막과 비교하여 SF 6 /CO 2 에서 형성된 막에서 MgO 는 나중에 나타났고 황화물은 30분 이내에 나타나지 않았다. 이는 SF 6 /air 에서 필름의 형성과 진화 가 SF 6 /CO 2 보다 빠르다 는 것을 의미할 수 있습니다 . CO 2 후속적으로 용융물과 반응하여 MgO를 형성하는 반면, 황 함유 화합물은 커버 가스에 축적되어 반응하여 매우 늦은 단계에서 황화물을 형성할 수 있습니다(산화 셀에서 30분 후).

    그림 11

    4 . 논의

    4.1 . SF 6 /air 에서 형성된 연행 결함의 진화

    Outokumpu HSC Chemistry for Windows( )의 HSC 소프트웨어를 사용하여 갇힌 기체와 액체 AZ91 합금 사이에서 발생할 수 있는 반응을 탐색하는 데 필요한 열역학 계산을 수행했습니다. 계산에 대한 솔루션은 소량의 커버 가스(즉, 갇힌 기포 내의 양)와 AZ91 합금 용융물 사이의 반응 과정에서 어떤 생성물이 가장 형성될 가능성이 있는지 제안합니다.

    실험에서 압력은 1기압으로, 온도는 700°C로 설정했습니다. 커버 가스의 사용량은 7 × 10으로 가정 하였다 -7  약 0.57 cm의 양으로 kg 3 (3.14 × 10 -6  0.5 % SF위한 kmol) 6 / 공기, 0.35 cm (3) (3.12 × 10 – 8  kmol) 0.5%SF 6 /CO 2 . 포획된 가스와 접촉하는 AZ91 합금 용융물의 양은 모든 반응을 완료하기에 충분한 것으로 가정되었습니다. SF 6 의 분해 생성물 은 SF 5 , SF 4 , SF 3 , SF 2 , F 2 , S(g), S 2(g) 및 F(g) [57] , [58] , [59] , [60] .

    그림 12 는 AZ91 합금과 0.5%SF 6 /air 사이의 반응에 대한 열역학적 계산의 평형 다이어그램을 보여줍니다 . 다이어그램에서 10 -15  kmol 미만의 반응물 및 생성물은 표시되지 않았습니다. 이는 존재 하는 SF 6 의 양 (≈ 1.57 × 10 -10  kmol) 보다 5배 적 으므로 영향을 미치지 않습니다. 실제적인 방법으로 과정을 관찰했습니다.

    그림 12

    이 반응 과정은 3단계로 나눌 수 있다.

    1단계 : 불화물의 형성. AZ91 용융물은 SF 6 및 그 분해 생성물과 우선적으로 반응하여 MgF 2 , AlF 3 및 ZnF 2 를 생성 합니다. 그러나 ZnF 2 의 양 이 너무 적어서 실제적으로 검출되지  않았을 수 있습니다(  MgF 2 의 3 × 10 -10 kmol에 비해 ZnF 2 1.25 × 10 -12 kmol ). 섹션 3.1 – 3.3에 표시된 모든 산화막 . 한편, 잔류 가스에 황이 SO 2 로 축적되었다 .

    2단계 : 산화물의 형성. 액체 AZ91 합금이 포획된 가스에서 사용 가능한 모든 불화물을 고갈시킨 후, Mg와의 반응으로 인해 AlF 3 및 ZnF 2 의 양이 빠르게 감소했습니다. O 2 (g) 및 SO 2 는 AZ91 용융물과 반응하여 MgO, Al 2 O 3 , MgAl 2 O 4 , ZnO, ZnSO 4 및 MgSO 4 를 형성 합니다. 그러나 ZnO 및 ZnSO 4 의 양은 EDS에 의해 실제로 발견되기에는 너무 적었을 것입니다(예: 9.5 × 10 -12  kmol의 ZnO, 1.38 × 10 -14  kmol의 ZnSO 4 , 대조적으로 4.68 × 10−10  kmol의 MgF 2 , X 축의 AZ91 양 이 2.5 × 10 -9  kmol일 때). 실험 사례에서 커버 가스의 F 농도는 매우 낮고 전체 농도 f O는 훨씬 높습니다. 따라서 1단계와 2단계, 즉 불화물과 산화물의 형성은 반응 초기에 동시에 일어나 그림 1과 2와 같이 불화물과 산화물의 가수층 혼합물이 형성될 수 있다 . 4 및 10 (a). 내부 층은 산화물로 구성되어 있지만 불화물은 커버 가스에서 F 원소가 완전히 고갈된 후에 형성될 수 있습니다.

    단계 1-2는 도 10 에 도시 된 다층 구조의 형성 과정을 이론적으로 검증하였다 .

    산화막 내의 MgAl 2 O 4 및 Al 2 O 3 의 양은 도 4에 도시된 산화막과 일치하는 검출하기에 충분한 양이었다 . 그러나, 도 10 에 도시된 바와 같이, 산화셀에서 성장된 산화막에서는 알루미늄의 존재를 인식할 수 없었다 . 이러한 Al의 부재는 표면 필름과 AZ91 합금 용융물 사이의 다음 반응으로 인한 것일 수 있습니다.(1)

    Al 2 O 3  + 3Mg + = 3MgO + 2Al, △G(700°C) = -119.82 kJ/mol(2)

    Mg + MgAl 2 O 4  = MgO + Al, △G(700°C) = -106.34 kJ/mol이는 반응물이 서로 완전히 접촉한다는 가정 하에 열역학적 계산이 수행되었기 때문에 HSC 소프트웨어로 시뮬레이션할 수 없었습니다. 그러나 실제 공정에서 AZ91 용융물과 커버 가스는 보호 표면 필름의 존재로 인해 서로 완전히 접촉할 수 없습니다.

    3단계 : 황화물과 질화물의 형성. 30분의 유지 시간 후, 산화 셀의 기상 불화물 및 산화물이 고갈되어 잔류 가스와 용융 반응을 허용하여 초기 F-농축 또는 (F, O )이 풍부한 표면 필름, 따라서 그림 10 (b 및 c)에 표시된 관찰된 다층 구조를 생성합니다 . 게다가, 질소는 모든 반응이 완료될 때까지 AZ91 용융물과 반응했습니다. 도 6 에 도시 된 산화막 은 질화물 함량으로 인해 이 반응 단계에 해당할 수 있다. 그러나, 그 결과는 도 1 및 도 5에 도시 된 연마된 샘플에서 질화물이 검출되지 않음을 보여준다. 4 와 5, 그러나 테스트 바 파단면에서만 발견됩니다. 질화물은 다음과 같이 샘플 준비 과정에서 가수분해될 수 있습니다 [54] .(삼)

    Mg 3 N 2  + 6H 2 O = 3Mg(OH) 2  + 2NH 3 ↑(4)

    AlN+ 3H 2 O = Al(OH) 3  + NH 3 ↑

    또한 Schmidt et al. [61] 은 Mg 3 N 2 와 AlN이 반응하여 3원 질화물(Mg 3 Al n N n+2, n=1, 2, 3…) 을 형성할 수 있음을 발견했습니다 . HSC 소프트웨어에는 삼원 질화물 데이터베이스가 포함되어 있지 않아 계산에 추가할 수 없습니다. 이 단계의 산화막은 또한 삼원 질화물을 포함할 수 있습니다.

    4.2 . SF 6 /CO 2 에서 형성된 연행 결함의 진화

    도 13 은 AZ91 합금과 0.5%SF 6 /CO 2 사이의 열역학적 계산 결과를 보여준다 . 이 반응 과정도 세 단계로 나눌 수 있습니다.

    그림 13

    1단계 : 불화물의 형성. SF 6 및 그 분해 생성물은 AZ91 용융물에 의해 소비되어 MgF 2 , AlF 3 및 ZnF 2 를 형성했습니다 . 0.5% SF 6 /air 에서 AZ91의 반응에서와 같이 ZnF 2 의 양 이 너무 작아서 실제적으로 감지되지  않았습니다( 2.67 x 10 -10  kmol의 MgF 2 에 비해 ZnF 2 1.51 x 10 -13 kmol ). S와 같은 잔류 가스 트랩에 축적 유황 2 (g) 및 (S)의 일부분 (2) (g)가 CO와 반응하여 2 SO 형성하는 2및 CO. 이 반응 단계의 생성물은 도 11 (a)에 도시된 필름과 일치하며 , 이는 불화물만을 함유하는 단일 층 구조를 갖는다.

    2단계 : 산화물의 형성. ALF 3 및 ZnF 2 MgF로 형성 용융 AZ91 마그네슘의 반응 2 , Al 및 Zn으로한다. SO 2 는 소모되기 시작하여 표면 필름에 산화물을 생성 하고 커버 가스에 S 2 (g)를 생성했습니다. 한편, CO 2 는 AZ91 용융물과 직접 반응하여 CO, MgO, ZnO 및 Al 2 O 3 를 형성 합니다. 도 1에 도시 된 산화막 9 및 11 (b)는 산소가 풍부한 층과 다층 구조로 인해 이 반응 단계에 해당할 수 있습니다.

    커버 가스의 CO는 AZ91 용융물과 추가로 반응하여 C를 생성할 수 있습니다. 이 탄소는 온도가 감소할 때(응고 기간 동안) Mg와 추가로 반응하여 Mg 탄화물을 형성할 수 있습니다 [62] . 이것은 도 4에 도시된 산화막의 탄소 함량이 높은 이유일 수 있다 8 – 9 . Liang et al. [39] 또한 SO 2 /CO 2 로 보호된 AZ91 합금 표면 필름에서 탄소 검출을 보고했습니다 . 생성된 Al 2 O 3 는 MgO와 더 결합하여 MgAl 2 O [63]를 형성할 수 있습니다 . 섹션 4.1 에서 논의된 바와 같이, 알루미나 및 스피넬은 도 11 에 도시된 바와 같이 표면 필름에 알루미늄 부재를 야기하는 Mg와 반응할 수 있다 .

    3단계 : 황화물의 형성. AZ91은 용융물 S 소비하기 시작 2 인 ZnS와 MGS 형성 갇힌 잔류 가스 (g)를. 이러한 반응은 반응 과정의 마지막 단계까지 일어나지 않았으며, 이는 Fig. 7 (c)에 나타난 결함의 S-함량 이 적은 이유일 수 있다 .

    요약하면, 열역학적 계산은 AZ91 용융물이 커버 가스와 반응하여 먼저 불화물을 형성한 다음 마지막에 산화물과 황화물을 형성할 것임을 나타냅니다. 다른 반응 단계에서 산화막은 다른 구조와 조성을 가질 것입니다.

    4.3 . 운반 가스가 동반 가스 소비 및 AZ91 주물의 재현성에 미치는 영향

    SF 6 /air 및 SF 6 /CO 2 에서 형성된 연행 결함의 진화 과정은 4.1절 과 4.2  에서 제안되었습니다 . 이론적인 계산은 실제 샘플에서 발견되는 해당 산화막과 관련하여 검증되었습니다. 연행 결함 내의 대기는 Al-합금 시스템과 다른 시나리오에서 액체 Mg-합금과의 반응으로 인해 효율적으로 소모될 수 있습니다(즉, 연행된 기포의 질소가 Al-합금 용융물과 효율적으로 반응하지 않을 것입니다 [64 , 65] 그러나 일반적으로 “질소 연소”라고 하는 액체 Mg 합금에서 질소가 더 쉽게 소모될 것입니다 [66] ).

    동반된 가스와 주변 액체 Mg-합금 사이의 반응은 동반된 가스를 산화막 내에서 고체 화합물(예: MgO)로 전환하여 동반 결함의 공극 부피를 감소시켜 결함(예: 공기의 동반된 가스가 주변의 액체 Mg 합금에 의해 고갈되면 용융 온도가 700 °C이고 액체 Mg 합금의 깊이가 10 cm라고 가정할 때 최종 고체 제품의 총 부피는 0.044가 됩니다. 갇힌 공기가 취한 초기 부피의 %).

    연행 결함의 보이드 부피 감소와 해당 주조 특성 사이의 관계는 알루미늄 합금 주조에서 널리 연구되었습니다. Nyahumwa와 Campbell [16] 은 HIP(Hot Isostatic Pressing) 공정이 Al-합금 주물의 연행 결함이 붕괴되고 산화물 표면이 접촉하게 되었다고 보고했습니다. 주물의 피로 수명은 HIP 이후 개선되었습니다. Nyahumwa와 Campbell [16] 도 서로 접촉하고 있는 이중 산화막의 잠재적인 결합을 제안했지만 이를 뒷받침하는 직접적인 증거는 없었습니다. 이 결합 현상은 Aryafar et.al에 의해 추가로 조사되었습니다. [8], 그는 강철 튜브에서 산화물 스킨이 있는 두 개의 Al-합금 막대를 다시 녹인 다음 응고된 샘플에 대해 인장 강도 테스트를 수행했습니다. 그들은 Al-합금 봉의 산화물 스킨이 서로 강하게 결합되어 용융 유지 시간이 연장됨에 따라 더욱 강해짐을 발견했으며, 이는 이중 산화막 내 동반된 가스의 소비로 인한 잠재적인 “치유” 현상을 나타냅니다. 구조. 또한 Raidszadeh와 Griffiths [9 , 19] 는 연행 가스가 반응하는 데 더 긴 시간을 갖도록 함으로써 응고 전 용융 유지 시간을 연장함으로써 Al-합금 주물의 재현성에 대한 연행 결함의 부정적인 영향을 성공적으로 줄였습니다. 주변이 녹습니다.

    앞서 언급한 연구를 고려할 때, Mg 합금 주물에서 혼입 가스의 소비는 다음 두 가지 방식으로 혼입 결함의 부정적인 영향을 감소시킬 수 있습니다.

    (1) 이중 산화막의 결합 현상 . 도 5 및 도 7 에 도시 된 샌드위치형 구조 는 이중 산화막 구조의 잠재적인 결합을 나타내었다. 그러나 산화막의 결합으로 인한 강도 증가를 정량화하기 위해서는 더 많은 증거가 필요합니다.

    (2) 연행 결함의 보이드 체적 감소 . 주조품의 품질에 대한 보이드 부피 감소의 긍정적인 효과는 HIP 프로세스 [67]에 의해 널리 입증되었습니다 . 섹션 4.1 – 4.2 에서 논의된 진화 과정과 같이 , 동반된 가스와 주변 AZ91 합금 용융물 사이의 지속적인 반응으로 인해 동반 결함의 산화막이 함께 성장할 수 있습니다. 최종 고체 생성물의 부피는 동반된 기체에 비해 상당히 작았다(즉, 이전에 언급된 바와 같이 0.044%).

    따라서, 혼입 가스의 소모율(즉, 산화막의 성장 속도)은 AZ91 합금 주물의 품질을 향상시키는 중요한 매개변수가 될 수 있습니다. 이에 따라 산화 셀의 산화막 성장 속도를 추가로 조사했습니다.

    도 14 는 상이한 커버 가스(즉, 0.5%SF 6 /air 및 0.5%SF 6 /CO 2 ) 에서의 표면 필름 성장 속도의 비교를 보여준다 . 필름 두께 측정을 위해 각 샘플의 15개의 임의 지점을 선택했습니다. 95% 신뢰구간(95%CI)은 막두께의 변화가 가우시안 분포를 따른다는 가정하에 계산하였다. 0.5%SF 6 /air 에서 형성된 모든 표면막이 0.5%SF 6 /CO 2 에서 형성된 것보다 빠르게 성장함을 알 수 있다 . 다른 성장률은 0.5%SF 6 /air 의 연행 가스 소비율 이 0.5%SF 6 /CO 2 보다 더 높음 을 시사했습니다., 이는 동반된 가스의 소비에 더 유리했습니다.

    그림 14

    산화 셀에서 액체 AZ91 합금과 커버 가스의 접촉 면적(즉, 도가니의 크기)은 많은 양의 용융물과 가스를 고려할 때 상대적으로 작았다는 점에 유의해야 합니다. 결과적으로, 산화 셀 내에서 산화막 성장을 위한 유지 시간은 비교적 길었다(즉, 5-30분). 하지만, 실제 주조에 함유 된 혼입 결함은 (상대적으로 매우 적은, 즉, 수 미크론의 크기에 도시 된 바와 같이 ,도 3. – 6 및 [7]), 동반된 가스는 주변 용융물로 완전히 둘러싸여 상대적으로 큰 접촉 영역을 생성합니다. 따라서 커버 가스와 AZ91 합금 용융물의 반응 시간은 비교적 짧을 수 있습니다. 또한 실제 Mg 합금 모래 주조의 응고 시간은 몇 분일 수 있습니다(예: Guo [68] 은 직경 60mm의 Mg 합금 모래 주조가 응고되는 데 4분이 필요하다고 보고했습니다). 따라서 Mg-합금 용융주조 과정에서 포획된 동반된 가스는 특히 응고 시간이 긴 모래 주물 및 대형 주물의 경우 주변 용융물에 의해 쉽게 소모될 것으로 예상할 수 있습니다.

    따라서, 동반 가스의 다른 소비율과 관련된 다른 커버 가스(0.5%SF 6 /air 및 0.5%SF 6 /CO 2 )가 최종 주물의 재현성에 영향을 미칠 수 있습니다. 이 가정을 검증하기 위해 0.5%SF 6 /air 및 0.5%SF 6 /CO 2 에서 생산된 AZ91 주물 을 기계적 평가를 위해 테스트 막대로 가공했습니다. Weibull 분석은 선형 최소 자승(LLS) 방법과 비선형 최소 자승(비 LLS) 방법을 모두 사용하여 수행되었습니다 [69] .

    그림 15 (ab)는 LLS 방법으로 얻은 UTS 및 AZ91 합금 주물의 연신율의 전통적인 2-p 선형 Weibull 플롯을 보여줍니다. 사용된 추정기는 P= (i-0.5)/N이며, 이는 모든 인기 있는 추정기 중 가장 낮은 편향을 유발하는 것으로 제안되었습니다 [69 , 70] . SF 6 /air 에서 생산된 주물 은 UTS Weibull 계수가 16.9이고 연신율 Weibull 계수가 5.0입니다. 대조적으로, SF 6 /CO 2 에서 생산된 주물의 UTS 및 연신 Weibull 계수는 각각 7.7과 2.7로, SF 6 /CO 2 에 의해 보호된 주물의 재현성이 SF 6 /air 에서 생산된 것보다 훨씬 낮음을 시사합니다. .

    그림 15

    또한 저자의 이전 출판물 [69] 은 선형화된 Weibull 플롯의 단점을 보여주었으며, 이는 Weibull 추정 의 더 높은 편향과 잘못된 2 중단을 유발할 수 있습니다 . 따라서 그림 15 (cd) 와 같이 Non-LLS Weibull 추정이 수행되었습니다 . SF 6 /공기주조물 의 UTS Weibull 계수 는 20.8인 반면, SF 6 /CO 2 하에서 생산된 주조물의 UTS Weibull 계수는 11.4로 낮아 재현성에서 분명한 차이를 보였다. 또한 SF 6 /air elongation(El%) 데이터 세트는 SF 6 /CO 2 의 elongation 데이터 세트보다 더 높은 Weibull 계수(모양 = 5.8)를 가졌습니다.(모양 = 3.1). 따라서 LLS 및 Non-LLS 추정 모두 SF 6 /공기 주조가 SF 6 /CO 2 주조 보다 더 높은 재현성을 갖는다고 제안했습니다 . CO 2 대신 공기를 사용 하면 혼입된 가스의 더 빠른 소비에 기여하여 결함 내의 공극 부피를 줄일 수 있다는 방법을 지원합니다 . 따라서 0.5%SF 6 /CO 2 대신 0.5%SF 6 /air를 사용 하면(동반된 가스의 소비율이 증가함) AZ91 주물의 재현성이 향상되었습니다.

    그러나 모든 Mg 합금 주조 공장이 현재 작업에서 사용되는 주조 공정을 따랐던 것은 아니라는 점에 유의해야 합니다. Mg의 합금 용탕 본 작업은 탈기에 따라서, 동반 가스의 소비에 수소의 영향을 감소 (즉, 수소 잠재적 동반 가스의 고갈 억제, 동반 된 기체로 확산 될 수있다 [7 , 71 , 72] ). 대조적으로, 마그네슘 합금 주조 공장에서는 마그네슘을 주조할 때 ‘가스 문제’가 없고 따라서 인장 특성에 큰 변화가 없다고 널리 믿어지기 때문에 마그네슘 합금 용융물은 일반적으로 탈기되지 않습니다 [73] . 연구에 따르면 Mg 합금 주물의 기계적 특성에 대한 수소의 부정적인 영향 [41 ,42 , 73] , 탈기 공정은 마그네슘 합금 주조 공장에서 여전히 인기가 없습니다.

    또한 현재 작업에서 모래 주형 공동은 붓기 전에 SF 6 커버 가스 로 플러싱되었습니다 [22] . 그러나 모든 Mg 합금 주조 공장이 이러한 방식으로 금형 캐비티를 플러싱한 것은 아닙니다. 예를 들어, Stone Foundry Ltd(영국)는 커버 가스 플러싱 대신 유황 분말을 사용했습니다. 그들의 주물 내의 동반된 가스 는 보호 가스라기 보다는 SO 2 /공기일 수 있습니다 .

    따라서 본 연구의 결과는 CO 2 대신 공기를 사용 하는 것이 최종 주조의 재현성을 향상시키는 것으로 나타났지만 다른 산업용 Mg 합금 주조 공정과 관련하여 캐리어 가스의 영향을 확인하기 위해서는 여전히 추가 조사가 필요합니다.

    7 . 결론


    AZ91 합금에 형성된 연행 결함이 관찰되었습니다. 그들의 산화막은 단층과 다층의 두 가지 유형의 구조를 가지고 있습니다. 다층 산화막은 함께 성장하여 최종 주조에서 샌드위치 같은 구조를 형성할 수 있습니다.2.

    실험 결과와 이론적인 열역학적 계산은 모두 갇힌 가스의 불화물이 황을 소비하기 전에 고갈되었음을 보여주었습니다. 이중 산화막 결함의 3단계 진화 과정이 제안되었습니다. 산화막은 진화 단계에 따라 다양한 화합물 조합을 포함했습니다. SF 6 /air 에서 형성된 결함 은 SF 6 /CO 2 에서 형성된 것과 유사한 구조를 갖지만 산화막의 조성은 달랐다. 엔트레인먼트 결함의 산화막 형성 및 진화 과정은 이전에 보고된 Mg 합금 표면막(즉, MgF 2 이전에 형성된 MgO)의 것과 달랐다 .삼.

    산화막의 성장 속도는 SF하에 큰 것으로 입증되었다 (6) / SF보다 공기 6 / CO 2 손상 봉입 가스의 빠른 소비에 기여한다. AZ91 합금 주물의 재현성은 SF 6 /CO 2 대신 SF 6 /air를 사용할 때 향상되었습니다 .

    감사의 말

    저자는 EPSRC LiME 보조금 EP/H026177/1의 자금 지원 과 WD Griffiths 박사와 Adrian Carden(버밍엄 대학교)의 도움을 인정합니다. 주조 작업은 University of Birmingham에서 수행되었습니다.

    MK McNutt , SALAZAR K.
    마그네슘, 화합물 및 금속, 미국 지질 조사국 및 미국 내무부
    레 스톤 , 버지니아 ( 2013 )
    Google 학술검색
    화합물 및 금속, 미국 지질 조사국 및 미국 내무부
    ( 1996 )
    Google 학술검색
    I. Ostrovsky , Y. Henn
    ASTEC’07 International Conference-New Challenges in Aeronautics , Moscow ( 2007 ) , pp. 1 – 5
    8월 19-22일
    Scopus에서 레코드 보기Google 학술검색
    Y. Wan , B. Tang , Y. Gao , L. Tang , G. Sha , B. Zhang , N. Liang , C. Liu , S. Jiang , Z. Chen , X. Guo , Y. Zhao
    액타 메이터. , 200 ( 2020 ) , 274 – 286 페이지
    기사PDF 다운로드Scopus에서 레코드 보기
    JTJ Burd , EA Moore , H. Ezzat , R. Kirchain , R. Roth
    적용 에너지 , 283 ( 2021 ) , 제 116269 조
    기사PDF 다운로드Scopus에서 레코드 보기
    AM 루이스 , JC 켈리 , 조지아주 Keoleian
    적용 에너지 , 126 ( 2014 ) , pp. 13 – 20
    기사PDF 다운로드Scopus에서 레코드 보기
    J. 캠벨
    버터워스-하이네만 , 옥스퍼드 ( 2004 )
    Google 학술검색
    M. Aryafar , R. Raiszadeh , A. Shalbafzadeh
    J. 메이터. 과학. , 45 ( 2010 년 ) , PP. (3041) – 3051
    교차 참조Scopus에서 레코드 보기
    R. 라이자데 , WD 그리피스
    메탈. 메이터. 트랜스. B-프로세스 메탈. 메이터. 프로세스. 과학. , 42 ( 2011 ) , 133 ~ 143페이지
    교차 참조Scopus에서 레코드 보기
    R. 라이자데 , WD 그리피스
    J. 합금. Compd. , 491 ( 2010 ) , 575 ~ 580 쪽
    기사PDF 다운로드Scopus에서 레코드 보기
    L. Peng , G. Zeng , TC Su , H. Yasuda , K. Nogita , CM Gourlay
    JOM , 71 ( 2019 ) , pp. 2235 – 2244
    교차 참조Scopus에서 레코드 보기
    S. Ganguly , AK Mondal , S. Sarkar , A. Basu , S. Kumar , C. Blawert
    코로스. 과학. , 166 ( 2020 )
    GE Bozchaloei , N. Varahram , P. Davami , SK 김
    메이터. 과학. 영어 A-구조체. 메이터. 소품 Microstruct. 프로세스. , 548 ( 2012 ) , 99 ~ 105페이지
    Scopus에서 레코드 보기
    S. 폭스 , J. 캠벨
    Scr. 메이터. , 43 ( 2000 ) , PP. 881 – 886
    기사PDF 다운로드Scopus에서 레코드 보기
    M. 콕스 , RA 하딩 , J. 캠벨
    메이터. 과학. 기술. , 19 ( 2003 ) , 613 ~ 625페이지
    Scopus에서 레코드 보기
    C. Nyahumwa , NR Green , J. Campbell
    메탈. 메이터. 트랜스. A-Phys. 메탈. 메이터. 과학. , 32 ( 2001 ) , 349 ~ 358 쪽
    Scopus에서 레코드 보기
    A. Ardekhani , R. Raiszadeh
    J. 메이터. 영어 공연하다. , 21 ( 2012 ) , pp. 1352 – 1362
    교차 참조Scopus에서 레코드 보기
    X. Dai , X. Yang , J. Campbell , J. Wood
    메이터. 과학. 기술. , 20 ( 2004 ) , 505 ~ 513 쪽
    Scopus에서 레코드 보기
    EM 엘갈라드 , MF 이브라힘 , HW 도티 , FH 사무엘
    필로스. 잡지. , 98 ( 2018 ) , PP. 1337 – 1359
    교차 참조Scopus에서 레코드 보기
    WD 그리피스 , NW 라이
    메탈. 메이터. 트랜스. A-Phys. 메탈. 메이터. 과학. , 38A ( 2007 ) , PP. 190 – 196
    교차 참조Scopus에서 레코드 보기
    AR Mirak , M. Divandari , SMA Boutorabi , J. 캠벨
    국제 J. 캐스트 만났습니다. 해상도 , 20 ( 2007 ) , PP. 215 – 220
    교차 참조Scopus에서 레코드 보기
    C. 칭기
    주조공학 연구실
    Helsinki University of Technology , Espoo, Finland ( 2006 )
    Google 학술검색
    Y. Jia , J. Hou , H. Wang , Q. Le , Q. Lan , X. Chen , L. Bao
    J. 메이터. 프로세스. 기술. , 278 ( 2020 ) , 제 116542 조
    기사PDF 다운로드Scopus에서 레코드 보기
    S. Ouyang , G. Yang , H. Qin , S. Luo , L. Xiao , W. Jie
    메이터. 과학. 영어 A , 780 ( 2020 ) , 제 139138 조
    기사PDF 다운로드Scopus에서 레코드 보기
    에스엠. Xiong , X.-F. 왕
    트랜스. 비철금속 사회 중국 , 20 ( 2010 ) , pp. 1228 – 1234
    기사PDF 다운로드Scopus에서 레코드 보기
    그랜드뷰 리서치
    ( 2018 )
    Google 학술검색
    T. 리 , J. 데이비스
    메탈. 메이터. 트랜스. , 51 ( 2020 ) , PP. 5,389 – (5400)
    교차 참조Scopus에서 레코드 보기
    JF Fruehling, 미시간 대학, 1970.
    Google 학술검색
    S. 쿨링
    제36회 세계 마그네슘 연례 회의 , 노르웨이 ( 1979 ) , pp. 54 – 57
    Scopus에서 레코드 보기Google 학술검색
    S. Cashion , N. Ricketts , P. Hayes
    J. 가벼운 만남. , 2 ( 2002 ) , 43 ~ 47페이지
    기사PDF 다운로드Scopus에서 레코드 보기
    S. Cashion , N. Ricketts , P. Hayes
    J. 가벼운 만남. , 2 ( 2002 ) , PP. 37 – 42
    기사PDF 다운로드Scopus에서 레코드 보기
    K. Aarstad , G. Tranell , G. Pettersen , TA Engh
    SF6에 의해 보호되는 마그네슘의 표면을 연구하는 다양한 기술
    TMS ( 2003년 )
    Google 학술검색
    에스엠 Xiong , X.-L. 리우
    메탈. 메이터. 트랜스. , 38 ( 2007 년 ) , PP. (428) – (434)
    교차 참조Scopus에서 레코드 보기
    T.-S. 시 , J.-B. Liu , P.-S. 웨이
    메이터. 화학 물리. , 104 ( 2007 ) , 497 ~ 504페이지
    기사PDF 다운로드Scopus에서 레코드 보기
    G. Pettersen , E. Øvrelid , G. Tranell , J. Fenstad , H. Gjestland
    메이터. 과학. 영어 , 332 ( 2002 ) , PP. (285) – (294)
    기사PDF 다운로드Scopus에서 레코드 보기
    H. Bo , LB Liu , ZP Jin
    J. 합금. Compd. , 490 ( 2010 ) , 318 ~ 325 쪽
    기사PDF 다운로드Scopus에서 레코드 보기
    A. 미락 , C. 데이비슨 , J. 테일러
    코로스. 과학. , 52 ( 2010 ) , PP. 1992 년 – 2000
    기사PDF 다운로드Scopus에서 레코드 보기
    BD 리 , UH 부리 , KW 리 , GS 한강 , JW 한
    메이터. 트랜스. , 54 ( 2013 ) , 66 ~ 73페이지
    Scopus에서 레코드 보기
    WZ Liang , Q. Gao , F. Chen , HH Liu , ZH Zhao
    China Foundry , 9 ( 2012 ) , pp. 226 – 230
    교차 참조Scopus에서 레코드 보기
    UI 골드슐레거 , EY 샤피로비치
    연소. 폭발 충격파 , 35 ( 1999 ) , 637 ~ 644페이지
    Scopus에서 레코드 보기
    A. Elsayed , SL Sin , E. Vandersluis , J. Hill , S. Ahmad , C. Ravindran , S. Amer Foundry
    트랜스. 오전. 파운드리 Soc. , 120 ( 2012 ) , 423 ~ 429페이지
    Scopus에서 레코드 보기
    E. Zhang , GJ Wang , ZC Hu
    메이터. 과학. 기술. , 26 ( 2010 ) , 1253 ~ 1258페이지
    Scopus에서 레코드 보기
    NR 그린 , J. 캠벨
    메이터. 과학. 영어 A-구조체. 메이터. 소품 Microstruct. 프로세스. , 173 ( 1993 ) , 261 ~ 266 쪽
    기사PDF 다운로드Scopus에서 레코드 보기
    C 라일리 , MR 졸리 , NR 그린
    MCWASP XII 논문집 – 주조, 용접 및 고급 Solidifcation 프로세스의 12 모델링 , 밴쿠버, 캐나다 ( 2009 )
    Google 학술검색
    HE Friedrich, BL Mordike, Springer, 독일, 2006.
    Google 학술검색
    C. Zheng , BR Qin , XB Lou
    기계, 산업 및 제조 기술에 관한 2010 국제 회의 , ASME ( 2010 ) , pp. 383 – 388
    2010년 미트
    교차 참조Scopus에서 레코드 보기Google 학술검색
    SM Xiong , XF 왕
    트랜스. 비철금속 사회 중국 , 20 ( 2010 ) , pp. 1228 – 1234
    기사PDF 다운로드Scopus에서 레코드 보기
    SM Xiong , XL Liu
    메탈. 메이터. 트랜스. A-Phys. 메탈. 메이터. 과학. , 38A ( 2007 ) , PP. (428) – (434)
    교차 참조Scopus에서 레코드 보기
    TS Shih , JB Liu , PS Wei
    메이터. 화학 물리. , 104 ( 2007 ) , 497 ~ 504페이지
    기사PDF 다운로드Scopus에서 레코드 보기
    K. Aarstad , G. Tranell , G. Pettersen , TA Engh
    매그. 기술. ( 2003 ) , PP. (5) – (10)
    Scopus에서 레코드 보기
    G. Pettersen , E. Ovrelid , G. Tranell , J. Fenstad , H. Gjestland
    메이터. 과학. 영어 A-구조체. 메이터. 소품 Microstruct. 프로세스. , 332 ( 2002 ) , 285 ~ 294페이지
    기사PDF 다운로드Scopus에서 레코드 보기
    XF 왕 , SM Xiong
    코로스. 과학. , 66 ( 2013 ) , PP. 300 – 307
    기사PDF 다운로드Scopus에서 레코드 보기
    SH Nie , SM Xiong , BC Liu
    메이터. 과학. 영어 A-구조체. 메이터. 소품 Microstruct. 프로세스. , 422 ( 2006 ) , 346 ~ 351페이지
    기사PDF 다운로드Scopus에서 레코드 보기
    C. Bauer , A. Mogessie , U. Galovsky
    Zeitschrift 모피 Metallkunde , 97 ( 2006 ) , PP. (164) – (168)
    교차 참조Scopus에서 레코드 보기
    QG 왕 , D. Apelian , DA Lados
    J. 가벼운 만남. , 1 ( 2001 ) , PP. (73) – 84
    기사PDF 다운로드Scopus에서 레코드 보기
    S. Wang , Y. Wang , Q. Ramasse , Z. Fan
    메탈. 메이터. 트랜스. , 51 ( 2020 ) , PP. 2957 – 2974
    교차 참조Scopus에서 레코드 보기
    S. Hayashi , W. Minami , T. Oguchi , HJ Kim
    카그. 코그. 론분슈 , 35 ( 2009 ) , 411 ~ 415페이지
    교차 참조Scopus에서 레코드 보기
    K. 아르스타드
    노르웨이 과학 기술 대학교
    ( 2004년 )
    Google 학술검색
    RL 윌킨스
    J. Chem. 물리. , 51 ( 1969 ) , p. 853
    Scopus에서 레코드 보기
    O. Kubaschewski , K. Hesselemam
    무기물의 열화학적 성질
    Springer-Verlag , 벨린 ( 1991 )
    Google 학술검색
    R. Schmidt , M. Strobele , K. Eichele , HJ Meyer
    유로 J. Inorg. 화학 ( 2017 ) , PP. 2727 – 2735
    교차 참조Scopus에서 레코드 보기
    B. Hu , Y. Du , H. Xu , W. Sun , WW Zhang , D. Zhao
    제이민 메탈. 분파. B-금속. , 46 ( 2010 ) , 97 ~ 103페이지
    Scopus에서 레코드 보기
    O. Salas , H. Ni , V. Jayaram , KC Vlach , CG Levi , R. Mehrabian
    J. 메이터. 해상도 , 6 ( 1991 ) , 1964 ~ 1981페이지
    Scopus에서 레코드 보기
    SSS Kumari , UTS Pillai , BC 빠이
    J. 합금. Compd. , 509 ( 2011 ) , pp. 2503 – 2509
    기사PDF 다운로드Scopus에서 레코드 보기
    H. Scholz , P. Greil
    J. 메이터. 과학. , 26 ( 1991 ) , 669 ~ 677 쪽
    Scopus에서 레코드 보기
    P. Biedenkopf , A. Karger , M. Laukotter , W. Schneider
    매그. 기술. , 2005년 ( 2005년 ) , 39 ~ 42 쪽
    Scopus에서 레코드 보기
    HV 앳킨슨 , S. 데이비스
    메탈. 메이터. 트랜스. , 31 ( 2000 ) , PP. 2981 – 3000
    교차 참조Scopus에서 레코드 보기
    EJ Guo , L. Wang , YC Feng , LP Wang , YH Chen
    J. 썸. 항문. 칼로리. , 135 ( 2019 ) , PP. 2001 년 – 2008 년
    교차 참조Scopus에서 레코드 보기
    T. Li , WD Griffiths , J. Chen
    메탈. 메이터. 트랜스. A-Phys. 메탈. 메이터. 과학. , 48A ( 2017 ) , PP. 5516 – 5528
    교차 참조Scopus에서 레코드 보기
    M. Tiryakioglu , D. Hudak는
    J. 메이터. 과학. , 42 ( 2007 ) , pp. 10173 – 10179
    교차 참조Scopus에서 레코드 보기
    Y. Yue , WD Griffiths , JL Fife , NR Green
    제1회 3d 재료과학 국제학술대회 논문집 ( 2012 ) , pp. 131 – 136
    교차 참조Scopus에서 레코드 보기Google 학술검색
    R. 라이자데 , WD 그리피스
    메탈. 메이터. 트랜스. B-프로세스 메탈. 메이터. 프로세스. 과학. , 37 ( 2006 ) , PP. (865) – (871)
    Scopus에서 레코드 보기
    ZC Hu , EL Zhang , SY Zeng
    메이터. 과학. 기술. , 24 ( 2008 ) , 1304 ~ 1308페이지
    교차 참조Scopus에서 레코드 보기

    Conflict resolution in the multi-stakeholder stepped spillway design under uncertainty by machine learning techniques

    기계 학습 기술에 의한 불확실성 하에서 다중 이해 관계자 계단형 배수로 설계의 충돌 해결

    Conflict resolution in the multi-stakeholder stepped spillway design under uncertainty by machine learning techniques

    Mehrdad GhorbaniMooseluaMohammad RezaNikoobParnian HashempourBakhtiaribNooshin BakhtiariRayanicAzizallahIzadyd
    aDepartment of Engineering Sciences, University of Agder, Norway
    bDepartment of Civil and Environmental Engineering, Shiraz University, Shiraz, Iran
    cSchool of Engineering, Department of Civil and Environmental Engineering, Shiraz University, Shiraz, IrandWater Research Center, Sultan Qaboos University, Muscat, Oman


    The optimal spillway design is of great significance since these structures can reduce erosion downstream of the dams. This study proposes a risk-based optimization framework for a stepped spillway to achieve an economical design scenario with the minimum loss in hydraulic performance. Accordingly, the stepped spillway was simulated in the FLOW-3D® model, and the validated model was repeatedly performed for various geometric states.

    The results were used to form a Multilayer Perceptron artificial neural network (MLP-ANN) surrogate model. Then, a risk-based optimization model was formed by coupling the MLP-ANN and NSGA-II. The concept of conditional value at risk (CVaR) was utilized to reduce the risk of the designed spillway malfunctions in high flood flow rates, while minimizing the construction cost and the loss in hydraulic performance.

    Lastly, given the conflicting objectives of stakeholders, the non-cooperative graph model for conflict resolution (GMCR) was applied to achieve a compromise on the Pareto optimal solutions. Applicability of the suggested approach in the Jarreh Dam, Iran, resulted in a practical design scenario, which simultaneously minimizes the loss in hydraulic performance and the project cost and satisfies the priorities of decision-makers.


    Stepped spillway, FLOW-3D® ,CVaR-based optimization model, GMCR-plus, NSGA-II

    최적의 배수로 설계는 이러한 구조가 댐 하류의 침식을 줄일 수 있기 때문에 매우 중요합니다. 본 연구에서는 유압 성능 손실을 최소화하면서 경제적인 설계 시나리오를 달성하기 위해 계단형 여수로에 대한 위험 기반 최적화 프레임워크를 제안합니다. 따라서 FLOW-3D® 모델에서 계단식 배수로를 시뮬레이션하고 다양한 기하학적 상태에 대해 검증된 모델을 반복적으로 수행했습니다.

    결과는 다층 퍼셉트론 인공 신경망(MLP-ANN) 대리 모델을 형성하는 데 사용되었습니다. 그런 다음 MLP-ANN과 NSGA-II를 결합하여 위험 기반 최적화 모델을 구성했습니다. 위험 조건부 값(CVaR)의 개념은 높은 홍수 유량에서 설계된 방수로 오작동의 위험을 줄이는 동시에 건설 비용과 수리 성능 손실을 최소화하기 위해 활용되었습니다.

    마지막으로 이해관계자의 상충되는 목표를 고려하여 파레토 최적해에 대한 절충안을 달성하기 위해 갈등 해결을 위한 비협조적 그래프 모델(GMCR)을 적용하였다. 이란 Jarreh 댐에서 제안된 접근 방식의 적용 가능성은 수력 성능 손실과 프로젝트 비용을 동시에 최소화하고 의사 결정자의 우선 순위를 만족시키는 실용적인 설계 시나리오로 귀결되었습니다.

    FLOW-3D AM

    flow3d AM-product
    FLOW-3D AM-product

    와이어 파우더 기반 DED | Wire Powder Based DED

    일부 연구자들은 부품을 만들기 위해 더 넓은 범위의 처리 조건을 사용하여 하이브리드 와이어 분말 기반 DED 시스템을 찾고 있습니다. 예를 들어, 이 시뮬레이션은 다양한 분말 및 와이어 이송 속도를 가진 하이브리드 시스템을 살펴봅니다.

    와이어 기반 DED | Wire Based DED

    와이어 기반 DED는 분말 기반 DED보다 처리량이 높고 낭비가 적지만 재료 구성 및 증착 방향 측면에서 유연성이 떨어집니다. FLOW-3D AM 은 와이어 기반 DED의 처리 결과를 이해하는데 유용하며 최적화 연구를 통해 빌드에 대한 와이어 이송 속도 및 직경과 같은 최상의 처리 매개 변수를 찾을 수 있습니다.

    FLOW-3D AM은 레이저 파우더 베드 융합 (L-PBF), 바인더 제트 및 DED (Directed Energy Deposition)와 같은 적층 제조 공정 ( additive manufacturing )을 시뮬레이션하고 분석하는 CFD 소프트웨어입니다. FLOW-3D AM 의 다중 물리 기능은 공정 매개 변수의 분석 및 최적화를 위해 분말 확산 및 압축, 용융 풀 역학, L-PBF 및 DED에 대한 다공성 형성, 바인더 분사 공정을 위한 수지 침투 및 확산에 대해 매우 정확한 시뮬레이션을 제공합니다.

    3D 프린팅이라고도하는 적층 제조(additive manufacturing)는 일반적으로 층별 접근 방식을 사용하여, 분말 또는 와이어로 부품을 제조하는 방법입니다. 금속 기반 적층 제조 공정에 대한 관심은 지난 몇 년 동안 시작되었습니다. 오늘날 사용되는 3 대 금속 적층 제조 공정은 PBF (Powder Bed Fusion), DED (Directed Energy Deposition) 및 바인더 제트 ( Binder jetting ) 공정입니다.  FLOW-3D  AM  은 이러한 각 프로세스에 대한 고유 한 시뮬레이션 통찰력을 제공합니다.

    파우더 베드 융합 및 직접 에너지 증착 공정에서 레이저 또는 전자 빔을 열원으로 사용할 수 있습니다. 두 경우 모두 PBF용 분말 형태와 DED 공정용 분말 또는 와이어 형태의 금속을 완전히 녹여 융합하여 층별로 부품을 형성합니다. 그러나 바인더 젯팅(Binder jetting)에서는 결합제 역할을 하는 수지가 금속 분말에 선택적으로 증착되어 층별로 부품을 형성합니다. 이러한 부품은 더 나은 치밀화를 달성하기 위해 소결됩니다.

    FLOW-3D AM 의 자유 표면 추적 알고리즘과 다중 물리 모델은 이러한 각 프로세스를 높은 정확도로 시뮬레이션 할 수 있습니다. 레이저 파우더 베드 융합 (L-PBF) 공정 모델링 단계는 여기에서 자세히 설명합니다. DED 및 바인더 분사 공정에 대한 몇 가지 개념 증명 시뮬레이션도 표시됩니다.

    레이저 파우더 베드 퓨전 (L-PBF)

    LPBF 공정에는 유체 흐름, 열 전달, 표면 장력, 상 변화 및 응고와 같은 복잡한 다중 물리학 현상이 포함되어 공정 및 궁극적으로 빌드 품질에 상당한 영향을 미칩니다. FLOW-3D AM 의 물리적 모델은 질량, 운동량 및 에너지 보존 방정식을 동시에 해결하는 동시에 입자 크기 분포 및 패킹 비율을 고려하여 중규모에서 용융 풀 현상을 시뮬레이션합니다.

    FLOW-3D DEM FLOW-3D WELD 는 전체 파우더 베드 융합 공정을 시뮬레이션하는 데 사용됩니다. L-PBF 공정의 다양한 단계는 분말 베드 놓기, 분말 용융 및 응고,이어서 이전에 응고 된 층에 신선한 분말을 놓는 것, 그리고 다시 한번 새 층을 이전 층에 녹이고 융합시키는 것입니다. FLOW-3D AM  은 이러한 각 단계를 시뮬레이션하는 데 사용할 수 있습니다.

    파우더 베드 부설 공정

    FLOW-3D DEM을 통해 분말 크기 분포, 재료 특성, 응집 효과는 물론 롤러 또는 블레이드 움직임 및 상호 작용과 같은 기하학적 효과와 관련된 분말 확산 및 압축을 이해할 수 있습니다. 이러한 시뮬레이션은 공정 매개 변수가 후속 인쇄 공정에서 용융 풀 역학에 직접적인 영향을 미치는 패킹 밀도와 같은 분말 베드 특성에 어떻게 영향을 미치는지에 대한 정확한 이해를 제공합니다.

    다양한 파우더 베드 압축을 달성하는 한 가지 방법은 베드를 놓는 동안 다양한 입자 크기 분포를 선택하는 것입니다. 아래에서 볼 수 있듯이 세 가지 크기의 입자 크기 분포가 있으며, 이는 가장 높은 압축을 제공하는 Case 2와 함께 다양한 분말 베드 압축을 초래합니다.

    파우더 베드 분포 다양한 입자 크기 분포
    세 가지 다른 입자 크기 분포를 사용하여 파우더 베드 배치
    파우더 베드 압축 결과
    세 가지 다른 입자 크기 분포를 사용한 분말 베드 압축

    입자-입자 상호 작용, 유체-입자 결합 및 입자 이동 물체 상호 작용은 FLOW-3D DEM을 사용하여 자세히 분석 할 수도 있습니다 . 또한 입자간 힘을 지정하여 분말 살포 응용 분야를 보다 정확하게 연구 할 수도 있습니다.

    FLOW-3D AM  시뮬레이션은 이산 요소 방법 (DEM)을 사용하여 역 회전하는 원통형 롤러로 인한 분말 확산을 연구합니다. 비디오 시작 부분에서 빌드 플랫폼이 위로 이동하는 동안 분말 저장소가 아래로 이동합니다. 그 직후, 롤러는 분말 입자 (초기 위치에 따라 색상이 지정됨)를 다음 층이 녹고 구축 될 준비를 위해 구축 플랫폼으로 펼칩니다. 이러한 시뮬레이션은 저장소에서 빌드 플랫폼으로 전송되는 분말 입자의 선호 크기에 대한 추가 통찰력을 제공 할 수 있습니다.

    Melting | 파우더 베드 용해

    DEM 시뮬레이션에서 파우더 베드가 생성되면 STL 파일로 추출됩니다. 다음 단계는 CFD를 사용하여 레이저 용융 공정을 시뮬레이션하는 것입니다. 여기서는 레이저 빔과 파우더 베드의 상호 작용을 모델링 합니다. 이 프로세스를 정확하게 포착하기 위해 물리학에는 점성 흐름, 용융 풀 내의 레이저 반사 (광선 추적을 통해), 열 전달, 응고, 상 변화 및 기화, 반동 압력, 차폐 가스 압력 및 표면 장력이 포함됩니다. 이 모든 물리학은 이 복잡한 프로세스를 정확하게 시뮬레이션하기 위해 TruVOF 방법을 기반으로 개발되었습니다.

    레이저 출력 200W, 스캔 속도 3.0m / s, 스폿 반경 100μm에서 파우더 베드의 용융 풀 분석.

    용융 풀이 응고되면 FLOW-3D AM  압력 및 온도 데이터를 Abaqus 또는 MSC Nastran과 같은 FEA 도구로 가져와 응력 윤곽 및 변위 프로파일을 분석 할 수도 있습니다.

    Multilayer | 다층 적층 제조

    용융 풀 트랙이 응고되면 DEM을 사용하여 이전에 응고된 층에 새로운 분말 층의 확산을 시뮬레이션 할 수 있습니다. 유사하게, 레이저 용융은 새로운 분말 층에서 수행되어 후속 층 간의 융합 조건을 분석 할 수 있습니다.

    해석 진행 절차는 첫 번째 용융층이 응고되면 입자의 두 번째 층이 응고 층에 증착됩니다. 새로운 분말 입자 층에 레이저 공정 매개 변수를 지정하여 용융 풀 시뮬레이션을 다시 수행합니다. 이 프로세스를 여러 번 반복하여 연속적으로 응고된 층 간의 융합, 빌드 내 온도 구배를 평가하는 동시에 다공성 또는 기타 결함의 형성을 모니터링 할 수 있습니다.

    다층 적층 적층 제조 시뮬레이션

    LPBF의 키홀 링 | Keyholing in LPBF

    키홀링 중 다공성은 어떻게 형성됩니까? 이것은 TU Denmark의 연구원들이 FLOW-3D AM을 사용하여 답변한 질문이었습니다. 레이저 빔의 적용으로 기판이 녹으면 기화 및 상 변화로 인한 반동 압력이 용융 풀을 압박합니다. 반동 압력으로 인한 하향 흐름과 레이저 반사로 인한 추가 레이저 에너지 흡수가 공존하면 폭주 효과가 발생하여 용융 풀이 Keyholing으로 전환됩니다. 결국, 키홀 벽을 따라 온도가 변하기 때문에 표면 장력으로 인해 벽이 뭉쳐져서 진행되는 응고 전선에 의해 갇힐 수 있는 공극이 생겨 다공성이 발생합니다. FLOW-3D AM 레이저 파우더 베드 융합 공정 모듈은 키홀링 및 다공성 형성을 시뮬레이션 하는데 필요한 모든 물리 모델을 보유하고 있습니다.

    바인더 분사 (Binder jetting)

    Binder jetting 시뮬레이션은 모세관 힘의 영향을받는 파우더 베드에서 바인더의 확산 및 침투에 대한 통찰력을 제공합니다. 공정 매개 변수와 재료 특성은 증착 및 확산 공정에 직접적인 영향을 미칩니다.

    Scan Strategy | 스캔 전략

    스캔 전략은 온도 구배 및 냉각 속도에 영향을 미치기 때문에 미세 구조에 직접적인 영향을 미칩니다. 연구원들은 FLOW-3D AM 을 사용하여 결함 형성과 응고된 금속의 미세 구조에 영향을 줄 수 있는 트랙 사이에서 발생하는 재 용융을 이해하기 위한 최적의 스캔 전략을 탐색하고 있습니다. FLOW-3D AM 은 하나 또는 여러 레이저에 대해 시간에 따른 방향 속도를 구현할 때 완전한 유연성을 제공합니다.

    Beam Shaping | 빔 형성

    레이저 출력 및 스캔 전략 외에도 레이저 빔 모양과 열유속 분포는 LPBF 공정에서 용융 풀 역학에 큰 영향을 미칩니다. AM 기계 제조업체는 공정 안정성 및 처리량에 대해 다중 코어 및 임의 모양의 레이저 빔 사용을 모색하고 있습니다. FLOW-3D AM을 사용하면 멀티 코어 및 임의 모양의 빔 프로파일을 구현할 수 있으므로 생산량을 늘리고 부품 품질을 개선하기 위한 최상의 구성에 대한 통찰력을 제공 할 수 있습니다.

    이 영역에서 수행 된 일부 작업에 대해 자세히 알아 보려면 “The Next Frontier of Metal AM”웨비나를 시청하십시오.

    Multi-material Powder Bed Fusion | 다중 재료 분말 베드 융합

    이 시뮬레이션에서 스테인리스 강 및 알루미늄 분말은 FLOW-3D AM 이 용융 풀 역학을 정확하게 포착하기 위해 추적하는 독립적으로 정의 된 온도 의존 재료 특성을 가지고 있습니다. 시뮬레이션은 용융 풀에서 재료 혼합을 이해하는 데 도움이됩니다.

    다중 재료 용접 사례 연구

    이종 금속의 레이저 키홀 용접에서 금속 혼합 조사

    GM과 University of Utah의 연구원들은 FLOW-3D WELD 를 사용 하여 레이저 키홀 용접을 통한 이종 금속의 혼합을 이해했습니다. 그들은 반동 압력 및 Marangoni 대류와 관련하여 구리와 알루미늄의 혼합 농도에 대한 레이저 출력 및 스캔 속도의 영향을 조사했습니다. 그들은 시뮬레이션을 실험 결과와 비교했으며 샘플 내의 절단 단면에서 재료 농도 사이에 좋은 일치를 발견했습니다.

    이종 금속의 레이저 키홀 용접에서 금속 혼합 조사
    이종 금속의 레이저 키홀 용접에서 금속 혼합 조사
    참조 : Wenkang Huang, Hongliang Wang, Teresa Rinker, Wenda Tan, 이종 금속의 레이저 키홀 용접에서 금속 혼합 조사 , Materials & Design, Volume 195, (2020).
    참조 : Wenkang Huang, Hongliang Wang, Teresa Rinker, Wenda Tan, 이종 금속의 레이저 키홀 용접에서 금속 혼합 조사 , Materials & Design, Volume 195, (2020).

    방향성 에너지 증착

    FLOW-3D AM 의 내장 입자 모델 을 사용하여 직접 에너지 증착 프로세스를 시뮬레이션 할 수 있습니다. 분말 주입 속도와 고체 기질에 입사되는 열유속을 지정함으로써 고체 입자는 용융 풀에 질량, 운동량 및 에너지를 추가 할 수 있습니다. 다음 비디오에서 고체 금속 입자가 용융 풀에 주입되고 기판에서 용융 풀의 후속 응고가 관찰됩니다.

    Figure 6. Evolution of melt pool in the overhang region (θ = 45°, P = 100 W, v = 1000 mm/s, the streamlines are shown by arrows).

    Experimental and numerical investigation of the origin of surface roughness in laser powder bed fused overhang regions

    레이저 파우더 베드 융합 오버행 영역에서 표면 거칠기의 원인에 대한 실험 및 수치 조사

    Shaochuan Feng,Amar M. Kamat,Soheil Sabooni &Yutao PeiPages S66-S84 | Received 18 Jan 2021, Accepted 25 Feb 2021, Published online: 10 Mar 2021


    Surface roughness of laser powder bed fusion (L-PBF) printed overhang regions is a major contributor to deteriorated shape accuracy/surface quality. This study investigates the mechanisms behind the evolution of surface roughness (Ra) in overhang regions. The evolution of surface morphology is the result of a combination of border track contour, powder adhesion, warp deformation, and dross formation, which is strongly related to the overhang angle (θ). When 0° ≤ θ ≤ 15°, the overhang angle does not affect Ra significantly since only a small area of the melt pool boundaries contacts the powder bed resulting in slight powder adhesion. When 15° < θ ≤ 50°, powder adhesion is enhanced by the melt pool sinking and the increased contact area between the melt pool boundary and powder bed. When θ > 50°, large waviness of the overhang contour, adhesion of powder clusters, severe warp deformation and dross formation increase Ra sharply.

    레이저 파우더 베드 퓨전 (L-PBF) 프린팅 오버행 영역의 표면 거칠기는 형상 정확도 / 표면 품질 저하의 주요 원인입니다. 이 연구 는 오버행 영역에서 표면 거칠기 (Ra ) 의 진화 뒤에 있는 메커니즘을 조사합니다 . 표면 형태의 진화는 오버행 각도 ( θ ) 와 밀접한 관련이있는 경계 트랙 윤곽, 분말 접착, 뒤틀림 변형 및 드로스 형성의 조합의 결과입니다 . 0° ≤  θ  ≤ 15° 인 경우 , 용융풀 경계의 작은 영역 만 분말 베드와 접촉하여 약간의 분말 접착이 발생하기 때문에 오버행 각도가 R a에 큰 영향을 주지 않습니다 . 15° < θ 일 때  ≤ 50°, 용융 풀 싱킹 및 용융 풀 경계와 분말 베드 사이의 증가된 접촉 면적으로 분말 접착력이 향상됩니다. θ  > 50° 일 때 오버행 윤곽의 큰 파형, 분말 클러스터의 접착, 심한 휨 변형 및 드 로스 형성이 Ra 급격히 증가 합니다.

    KEYWORDS: Laser powder bed fusion (L-PBF), melt pool dynamics, overhang region, shape deviation, surface roughness

    1. Introduction

    레이저 분말 베드 융합 (L-PBF)은 첨단 적층 제조 (AM) 기술로, 집중된 레이저 빔을 사용하여 금속 분말을 선택적으로 융합하여 슬라이스 된 3D 컴퓨터 지원에 따라 층별로 3 차원 (3D) 금속 부품을 구축합니다. 설계 (CAD) 모델 (Chatham, Long 및 Williams 2019 ; Tan, Zhu 및 Zhou 2020 ). 재료가 인쇄 층 아래에 ​​존재하는지 여부에 따라 인쇄 영역은 각각 솔리드 영역 또는 돌출 영역으로 분류 될 수 있습니다. 따라서 오버행 영역은 고체 기판이 아니라 분말 베드 바로 위에 건설되는 특수 구조입니다 (Patterson, Messimer 및 Farrington 2017). 오버행 영역은지지 구조를 포함하거나 포함하지 않고 구축 할 수 있으며, 지지대가있는 돌출 영역의 L-PBF는 지지체가 더 낮은 밀도로 구축된다는 점을 제외 하고 (Wang and Chou 2018 ) 고체 기판의 공정과 유사합니다 (따라서 기계적 강도가 낮기 때문에 L-PBF 공정 후 기계적으로 쉽게 제거 할 수 있습니다. 따라서지지 구조로 인쇄 된 오버행 영역은 L-PBF 공정 후 지지물 제거, 연삭 및 연마와 같은 추가 후 처리 단계가 필요합니다.

    수평 내부 채널의 제작과 같은 일부 특정 경우에는 공정 후 지지대를 제거하기가 어려우므로 채널 상단 절반의 돌출부 영역을 지지대없이 건설해야합니다 (Hopkinson and Dickens 2000 ). 수평 내부 채널에 사용할 수없는지지 구조 외에도 내부 표면, 특히 등각 냉각 채널 (Feng, Kamat 및 Pei 2021 ) 에서 발생하는 복잡한 3D 채널 네트워크의 경우 표면 마감 프로세스를 구현하는 것도 어렵습니다 . 결과적으로 오버행 영역은 (i) 잔류 응력에 의한 변형, (ii) 계단 효과 (Kuo et al. 2020 ; Li et al. 2020 )로 인해 설계된 모양에서 벗어날 수 있습니다 .) 및 (iii) 원하지 않는 분말 소결로 인한 향상된 표면 거칠기; 여기서, 앞의 두 요소는 일반적으로 mm 길이 스케일에서 ‘매크로’편차로 분류되고 후자는 일반적으로 µm 길이 스케일에서 ‘마이크로’편차로 인식됩니다.

    열 응력에 의한 변형은 오버행 영역에서 발생하는 중요한 문제입니다 (Patterson, Messimer 및 Farrington 2017 ). 국부적 인 용융 / 냉각은 용융 풀 내부 및 주변에서 큰 온도 구배를 유도하여 응고 된 층에 집중적 인 열 응력을 유발합니다. 열 응력에 의한 뒤틀림은 고체 영역을 현저하게 변형하지 않습니다. 이러한 영역은 아래의 여러 레이어에 의해 제한되기 때문입니다. 반면에 오버행 영역은 구속되지 않고 공정 중 응력 완화로 인해 상당한 변형이 발생합니다 (Kamat 및 Pei 2019 ). 더욱이 용융 깊이는 레이어 두께보다 큽니다 (이전 레이어도 재용 해되어 빌드 된 레이어간에 충분한 결합을 보장하기 때문입니다 [Yadroitsev et al. 2013 ; Kamath et al.2014 ]),응고 된 두께가 설계된 두께보다 크기 때문에형태 편차 (예 : 드 로스 [Charles et al. 2020 ; Feng et al. 2020 ])가 발생합니다. 마이크로 스케일에서 인쇄 된 표면 (R a 및 S a ∼ 10 μm)은 기계적으로 가공 된 표면보다 거칠다 (Duval-Chaneac et al. 2018 ; Wen et al. 2018 ). 이 문제는고형화 된 용융 풀의 가장자리에 부착 된 용융되지 않은 분말의 결과로 표면 거칠기 (R a )가 일반적으로 약 20 μm인 오버행 영역에서 특히 심각합니다 (Mazur et al. 2016 ; Pakkanen et al. 2016 ).

    오버행 각도 ( θ , 빌드 방향과 관련하여 측정)는 오버행 영역의 뒤틀림 편향과 표면 거칠기에 영향을 미치는 중요한 매개 변수입니다 (Kamat and Pei 2019 ; Mingear et al. 2019 ). θ ∼ 45 ° 의 오버행 각도 는 일반적으로지지 구조없이 오버행 영역을 인쇄 할 수있는 임계 값으로 합의됩니다 (Pakkanen et al. 2016 ; Kadirgama et al. 2018 ). θ 일 때이 임계 값보다 크면 오버행 영역을 허용 가능한 표면 품질로 인쇄 할 수 없습니다. 오버행 각도 외에도 레이저 매개 변수 (레이저 에너지 밀도와 관련된)는 용융 풀의 모양 / 크기 및 용융 풀 역학에 영향을줌으로써 오버행 영역의 표면 거칠기에 영향을줍니다 (Wang et al. 2013 ; Mingear et al . 2019 ).

    용융 풀 역학은 고체 (Shrestha 및 Chou 2018 ) 및 오버행 (Le et al. 2020 ) 영역 모두에서 수행되는 L-PBF 공정을 포함한 레이저 재료 가공의 일반적인 물리적 현상입니다 . 용융 풀 모양, 크기 및 냉각 속도는 잔류 응력으로 인한 변형과 ​​표면 거칠기에 모두 영향을 미치므로 처리 매개 변수와 표면 형태 / 품질 사이의 다리 역할을하며 용융 풀을 이해하기 위해 수치 시뮬레이션을 사용하여 추가 조사를 수행 할 수 있습니다. 거동과 표면 거칠기에 미치는 영향. 현재까지 고체 영역의 L-PBF 동안 용융 풀 동작을 시뮬레이션하기 위해 여러 연구가 수행되었습니다. 유한 요소 방법 (FEM)과 같은 시뮬레이션 기술 (Roberts et al. 2009 ; Du et al.2019 ), 유한 차분 법 (FDM) (Wu et al. 2018 ), 전산 유체 역학 (CFD) (Lee and Zhang 2016 ), 임의의 Lagrangian-Eulerian 방법 (ALE) (Khairallah and Anderson 2014 )을 사용하여 증발 반동 압력 (Hu et al. 2018 ) 및 Marangoni 대류 (Zhang et al. 2018 ) 현상을포함하는 열 전달 (온도 장) 및 물질 전달 (용융 흐름) 프로세스. 또한 이산 요소법 (DEM)을 사용하여 무작위 분산 분말 베드를 생성했습니다 (Lee and Zhang 2016 ; Wu et al. 2018 ). 이 모델은 분말 규모의 L-PBF 공정을 시뮬레이션했습니다 (Khairallah et al. 2016) 메조 스케일 (Khairallah 및 Anderson 2014 ), 단일 트랙 (Leitz et al. 2017 )에서 다중 트랙 (Foroozmehr et al. 2016 ) 및 다중 레이어 (Huang, Khamesee 및 Toyserkani 2019 )로.

    그러나 결과적인 표면 거칠기를 결정하는 오버행 영역의 용융 풀 역학은 문헌에서 거의 관심을받지 못했습니다. 솔리드 영역의 L-PBF에 대한 기존 시뮬레이션 모델이 어느 정도 참조가 될 수 있지만 오버행 영역과 솔리드 영역 간의 용융 풀 역학에는 상당한 차이가 있습니다. 오버행 영역에서 용융 금속은 분말 입자 사이의 틈새로 아래로 흘러 용융 풀이 다공성 분말 베드가 제공하는 약한 지지체 아래로 가라 앉습니다. 이것은 중력과 표면 장력의 영향이 용융 풀의 결과적인 모양 / 크기를 결정하는 데 중요하며, 결과적으로 오버행 영역의 마이크로 스케일 형태의 진화에 중요합니다. 또한 분말 입자 사이의 공극, 열 조건 (예 : 에너지 흡수,2019 ; Karimi et al. 2020 ; 노래와 영 2020 ). 표면 거칠기는 (마이크로) 형상 편차를 증가시킬뿐만 아니라 주기적 하중 동안 미세 균열의 시작 지점 역할을함으로써 기계적 강도를 저하시킵니다 (Günther et al. 2018 ). 오버행 영역의 높은 표면 거칠기는 (마이크로) 정확도 / 품질에 대한 엄격한 요구 사항이있는 부품 제조에서 L-PBF의 적용을 제한합니다.

    본 연구는 실험 및 시뮬레이션 연구를 사용하여 오버행 영역 (지지물없이 제작)의 미세 형상 편차 형성 메커니즘과 표면 거칠기의 기원을 체계적이고 포괄적으로 조사합니다. 결합 된 DEM-CFD 시뮬레이션 모델은 경계 트랙 윤곽, 분말 접착 및 뒤틀림 변형의 효과를 고려하여 오버행 영역의 용융 풀 역학과 표면 형태의 형성 메커니즘을 나타 내기 위해 개발되었습니다. 표면 거칠기 R의 시뮬레이션 및 단일 요인 L-PBF 인쇄 실험을 사용하여 오버행 각도의 함수로 연구됩니다. 용융 풀의 침몰과 관련된 오버행 영역에서 분말 접착의 세 가지 메커니즘이 식별되고 자세히 설명됩니다. 마지막으로, 인쇄 된 오버행 영역에서 높은 표면 거칠기 문제를 완화 할 수 있는 잠재적 솔루션에 대해 간략하게 설명합니다.

    The shape and size of the L-PBF printed samples are illustrated in Figure 1
    The shape and size of the L-PBF printed samples are illustrated in Figure 1
    Figure 2. Borders in the overhang region depending on the overhang angle θ
    Figure 2. Borders in the overhang region depending on the overhang angle θ
    Figure 3. (a) Profile of the volumetric heat source, (b) the model geometry of single-track printing on a solid substrate (unit: µm), and (c) the comparison of melt pool dimensions obtained from the experiment (right half) and simulation (left half) for a calibrated optical penetration depth of 110 µm (laser power 200 W and scan speed 800 mm/s, solidified layer thickness 30 µm, powder size 10–45 µm).
    Figure 3. (a) Profile of the volumetric heat source, (b) the model geometry of single-track printing on a solid substrate (unit: µm), and (c) the comparison of melt pool dimensions obtained from the experiment (right half) and simulation (left half) for a calibrated optical penetration depth of 110 µm (laser power 200 W and scan speed 800 mm/s, solidified layer thickness 30 µm, powder size 10–45 µm).
    Figure 4. The model geometry of an overhang being L-PBF processed: (a) 3D view and (b) right view.
    Figure 4. The model geometry of an overhang being L-PBF processed: (a) 3D view and (b) right view.
    Figure 5. The cross-sectional contour of border tracks in a 45° overhang region.
    Figure 5. The cross-sectional contour of border tracks in a 45° overhang region.
    Figure 6. Evolution of melt pool in the overhang region (θ = 45°, P = 100 W, v = 1000 mm/s, the streamlines are shown by arrows).
    Figure 6. Evolution of melt pool in the overhang region (θ = 45°, P = 100 W, v = 1000 mm/s, the streamlines are shown by arrows).
    Figure 7. The overhang contour is contributed by (a) only outer borders when θ ≤ 60° (b) both inner borders and outer borders when θ > 60°.
    Figure 7. The overhang contour is contributed by (a) only outer borders when θ ≤ 60° (b) both inner borders and outer borders when θ > 60°.
    Figure 8. Schematic of powder adhesion on a 45° overhang region.
    Figure 8. Schematic of powder adhesion on a 45° overhang region.
    Figure 9. The L-PBF printed samples with various overhang angle (a) θ = 0° (cube), (b) θ = 30°, (c) θ = 45°, (d) θ = 55° and (e) θ = 60°.
    Figure 9. The L-PBF printed samples with various overhang angle (a) θ = 0° (cube), (b) θ = 30°, (c) θ = 45°, (d) θ = 55° and (e) θ = 60°.
    Figure 10. Two mechanisms of powder adhesion related to the overhang angle: (a) simulation-predicted, θ = 45°; (b) simulation-predicted, θ = 60°; (c, e) optical micrographs, θ = 45°; (d, f) optical micrographs, θ = 60°. (e) and (f) are partial enlargement of (c) and (d), respectively.
    Figure 10. Two mechanisms of powder adhesion related to the overhang angle: (a) simulation-predicted, θ = 45°; (b) simulation-predicted, θ = 60°; (c, e) optical micrographs, θ = 45°; (d, f) optical micrographs, θ = 60°. (e) and (f) are partial enlargement of (c) and (d), respectively.
    Figure 11. Simulation-predicted surface morphology in the overhang region at different overhang angle: (a) θ = 15°, (b) θ = 30°, (c) θ = 45°, (d) θ = 60° and (e) θ = 80° (Blue solid lines: simulation-predicted contour; red dashed lines: the planar profile of designed overhang region specified by the overhang angles).
    Figure 11. Simulation-predicted surface morphology in the overhang region at different overhang angle: (a) θ = 15°, (b) θ = 30°, (c) θ = 45°, (d) θ = 60° and (e) θ = 80° (Blue solid lines: simulation-predicted contour; red dashed lines: the planar profile of designed overhang region specified by the overhang angles).
    Figure 12. Effect of overhang angle on surface roughness Ra in overhang regions
    Figure 12. Effect of overhang angle on surface roughness Ra in overhang regions
    Figure 13. Surface morphology of L-PBF printed overhang regions with different overhang angle: (a) θ = 15°, (b) θ = 30°, (c) θ = 45° and (d) θ = 60° (overhang border parameters: P = 100 W, v = 1000 mm/s).
    Figure 13. Surface morphology of L-PBF printed overhang regions with different overhang angle: (a) θ = 15°, (b) θ = 30°, (c) θ = 45° and (d) θ = 60° (overhang border parameters: P = 100 W, v = 1000 mm/s).