Casting Defect and Process Optimization of Steel Crossing

Abstract

The casting quality of crossing in a railway turnout is required to be higher because of the significant impact load on the rail. The simulation results of using different fever risers, the temperature field, solidification process and casting defects, were obtained by the implicit finite element method based on the ProCAST software. To improve the authenticity of the visualization of the casting process, a numerical simulation assumption of the mass flow rate attenuation was proposed for the overflow of molten metal from the risers during tilt pouring. The result shows that the temperature field is more uniform when the risers with exothermic energy of 1200 kJ/kg were chosen and the defects converge to the risers in accordance with the principle of sequential solidification. Compared with insulation riser, shrinkage porosity proportion decreased from 23.10% to 17.01%, and the shrinkage cavity proportion decreased from 1.002% to 0.530%. However, changing the burned time has no obvious effect on the casting. At the same time, the process optimization scheme of risers was put forward in this study and the casting defects such as shrinkage cavity and porosity are predicted according to the ‘V type’ feeding area of the fever risers. This improvement has greatly improved the performance of the casting, and the passing gross tonnage could reach 300 million tons.

1. Introduction

Crossing, or frog, is a device required to switch trains from one track to another when the straight track and the branching track coincide in a railway system,¹⁾ as shown in Fig. 1. The wing rail and the nose rail are extremely vulnerable to damage due to repeated wheel impacts and uneven wheel-rail contact (the contact points of wheel and rail vary both horizontally and vertically). Safety is the most important factor in train operation because of the particularity of railway transportation. However, replacing the crossing that has reached the expected service life will waste a lot of time, which will cause the efficiency of railway transportation to decline.²⁾ Accordingly, a higher requirement is put forward to improving the quality of crossing and reducing the defects of casting, the first procedure of crossing production has become the top priority to improve its quality.

Fig. 1.

A casting steel crossing. (Online version in color.)

Characteristic failures in the region of the rail have spalling, head checks, squashing and horizontal crack,¹⁾ as shown in Fig. 2. To improve the quality of the nose rail and the wing rail in the actual production, direct external iron chills are placed on the surface of the nose and wing rails to ensure their preferential solidification. However, the pouring and solidification process of the sand mold casting is not visible, and the defect condition cannot be known before the knocking-out. Therefore, it is necessary to use finite element software to analyze the solidification results, predict the casting defects and optimize the process.³⁾

Fig. 2.

Damaged crossing. (Online version in color.)

After the casting model is determined, the size and type of the riser, the location of the chills, and the pouring method are the main factors controlling defects such as shrinkage cavity, shrinkage porosity, and sand hole. Many theoretical and numerical studies about the casting process and defects of high manganese steel crossing have been performed over the past few years. X. Y. Zhang et al., an open riser is arranged on the ingates, and the feeding is carried out by adding the heating agent after casting.⁴⁾ S. T. Zhang et al., for the larger shrinkage rate of high manganese steel, the thermal insulation riser is changed into an insulation exothermic riser to improve casting quality.⁵⁾ L. Y. Ma et al., giving a method to calculate the risers modulus and the feeding distance by high shrinkage rate of manganese steel and casting weight. Actually, a detailed study and analysis of crossing taking into account different fever risers and non-linear material properties have not been carried out so far.⁶⁾

Considering the nonlinear material characteristics, the matching relationship between open and blind risers and castings, a nonlinear finite element model of high manganese steel has been developed in this study. An innovative pouring method of velocity attenuation is proposed based on ProCAST Software to make the simulation closer to the real situation. The casting and solidification processes were simulated and its temperature field and solidification process with different risers with increasing heating and heating time parameters are analyzed and compared. Meanwhile, the casting defects of the crossing were predicted and optimized. This method can be used to visualize the pouring and solidification process and realize the defect prediction of casting and the casting quality was improved obviously after process optimization.^7,11) In addition, like other researchers, the total shrinkage porosity is adopted in this paper to directly reflect the severity of the defect without quantification. And the data in Table 2 used for quantitative analysis are obtained by ProCAST software. How to determine whether the crossing of high manganese steel is qualified according to the total shrinkage porosity has not been determined in other literatures and studies. Therefore, 3.10% in the paper is based on the experience in the actual production of the factory.

Table 2. The proportion of defects corresponding to the parameters of different risers.

Tab-Figure number	Ignition temperature (°C)	Burned time (s)	Exothermic energy (kJ/kg)	Shrinkage porosity proportion	Shrinkage cavity proportion
Fig. 10-b1	0	0	0	23.10%	1.002%
Fig. 10-b2	1000	240	400	19.69%	0.670%
Fig. 9-a2	1000	240	1200	16.51%	0.514%
Fig. 10-b3	1000	240	2000	15.84%	0.499%
Fig. 9-a1	1000	120	1200	17.01%	0.530%
Fig. 9-a3	1000	360	1200	16.17%	0.504%

2. Establishment of Model

2.1. Geometric Models

The metal shrinkage rate of high manganese steel is as high as ε=8.5%, so reasonable use of retracement equipment is necessary. According to the actual production experience, the riser with module M_R=4.33 cm was selected, and its size parameters were shown in Table 1. In the actual casting, the nose rail and wing rail side towards the ground to fill the mold first, so that the casting defects up to the risers and the bottom of the crossing. And the direct external chillers are placed on the surface of the nose and wing rails to ensure their preferential solidification.^9,10)

Table 1. Parameters values of risers.

The cast length of the crossing in the research is 4100 mm and the sand mold length is 7100 mm. High manganese steel is selected as the integral casting crossing material because of its good wear resistance and high hardness, especially its characteristics of shock hardening. But as far as casting science is concerned, high manganese steel has the fatal defect of very high shrinkage rate, which will make the casting produce more shrinkage porosity and other defects.⁸⁾ The whole casting model including the crossing, risers, chills, gating system and tilted sand mold was established, as shown in Fig. 3(1), and the red line is the cut part after casting. And the 12 mm surface layer of the crossing has been improved to ensure the accuracy of numerical simulation, in order to obtain reliable simulation results. Meanwhile, to reduce the calculation time, the mesh size of the sand mold surface was determined as 100 mm without affecting the casting heat transfer. And the mesh size of the nose rail and wing rail surface as the working area is divided into 5 mm. After local mesh refinement, the model has a total of 3657 221 elements, of which more than 1.5 million are concentrated in the interesting crossing part, as shown in Fig. 3(2).

Fig. 3.

(1) The three-dimensional model of integral casting crossing. (2) The finite element mesh model. (Online version in color.)

2.2. Mathematics Models

The process of pouring and solidification, as well as the temperature field and casting defects, are analyzed using ProCAST software. However, the tilted casting method is adopted in this study, which is different from the usual sand mold casting in that there is a large amount of metal liquid overflow. And according to the production site experience, raise the gating end by 2.85 degrees, as shown in Fig. 3(1). To get more practical simulation results, this paper gives the following assumptions for the judgment basis of shrinkage porosity, the state of metal liquid and equivalent pressure modeling,

(1) In this study, only the macroscopic movement and heat transfer law of molten metal are studied, without considering the influence of its microstructure.¹²⁾ In order to obtain the simulation results closer to the real casting situation, it is assumed that the molten metal is a continuous Newtonian fluid, which satisfies the Continuity Equation. Moreover, the density between Liquidus and Solidus changes non-linearly with temperature, and the density and Newtonian viscosity change curves with temperature are shown in Fig. 4.

Fig. 4.

The density and Newtonian viscosity of high manganese steel change with time at 1000 C–1600 C. (Online version in color.)

Continuity Equation:   

$$\frac{\partial \rho }{\partial \text{t}}+\frac{\partial \left(\rho {\text{u}}_{\text{x}}\right)}{\partial \text{x}}+\frac{\partial \left(\rho {\text{u}}_{\text{y}}\right)}{\partial \text{y}}+\frac{\partial \left(\rho {\text{u}}_{\text{z}}\right)}{\partial \text{z}}=0$$

Where, u_x, u_y, u_z are the three orthogonal velocity components of the fluid particle, ρ is the density, t is the pouring time.

(2) In practice, the whole casting of crossing is done by tilted casting method, so the condition that the molten metal overflows the sand from the risers will occur. However, this complex fluid overflow cannot be simulated in ProCAST software. Therefore, a finite element method for establishing the equivalent pressure model is proposed for this problem, and the assumption of mass flow rate attenuation is made for this method. The model of allowable gas passing through the hole of risers and no molten metal overflow is consistent with the actual pressure, because the molten metal overflow from riser will not produce any pressure on the castings. On the other hand, the mass flow rate of molten metal containing overflow is often carried out at a constant speed of 50 kg/s in actual production, so the change of pouring speed needs also to be considered. If the overflow molten metal is neglected, the pouring speed will be attenuated rather than constant. To be closer to the actual situation of the casting process, a negative pouring velocity increment should be introduced to offset this effect as shown in Fig. 5.

Fig. 5.

Intercepting the mass flow rate attenuation superposition curve from pouring to the third riser. (Online version in color.)

Where, u_i is the mass flow attenuation curve caused only by the i^th riser, and v_i is the superimposed curve of mass flow attenuation caused by the first (i-1) risers. In actual production, the mass flow rate curve is a constant line v₀ as shown in Fig. 5. However, at the time of t₀ in this numerical simulation, the first riser at the far end of the gating system is poured, which produces molten metal overflow. According to the actual production of pouring time and the weight of the molten metal, the non-linear algebra equations with indeterminate coefficients are as follows,   

$$\{\begin{array}{l}\text{S}={\int }_0^{22.56}\text{F}\left(\text{t}\right)\text{dt}\hfill \\ {\text{V}}_{\text{i}}=\text{F}\left({\text{t}}_{\text{i}}\right)\hfill \\ {\text{V}}_0=\text{F}\left({\text{t}}_0\right)\hfill \end{array}$$

Where, S is the weight of the molten metal, i.e. the area of the non-shaded part, V_i is the mass flow rate at the time of t_i, F(t) is the mass flow rate attenuation superposition curve from.

By solving the undetermined coefficient, the mass flow rate expression to remove the overflow weight of the molten metal can be modified to the following equation,   

$${\text{u}}_1=-2.07{\text{e}}^{0.12\left(\text{t}-{\text{t}}_0\right)}+52.07$$

The shaded area above the u₁ curve represents the weight of the molten metal overflow from the first riser. The second-order overflow occurs when the molten metal is poured to the second riser at the time of t₁, and the overflow weight also attenuates as the mass flow rate attenuates.^13,14,15) According to the attenuation coefficient f=0.98 measured by the simulation results, the mass flow rate at the second riser can be modified into the following equation,   

$${\text{u}}_2=0.98\left(2.07{\text{e}}^{0.12\left(\text{t}-{\text{t}}_1\right)}+52.07\right)$$

The red shaded area shown in Fig. 5 shows the weight of the molten metal at the second riser. Besides, the v₂ curve obtained by u₁, u₂ superposition is the mass flow rate attenuation curve before pouring to the third riser (t₂), and the shadow part above is the overflow weight of molten metal. By iterating to the final riser, the eighth order molten metal overflow was finally obtained, and the mass flow rate attenuation equation of the whole numerical simulation is also obtained.

3. Results and Discussion

3.1. Visualization of Casting Process

The casting properties of the crossing put into production were tested firstly. The shrinkage porosity defect of casting is characterized by the insufficient density of metal and decreased bearing capacity, as shown in Fig. 6(a). The shrinkage cavity defect of casting is characterized as a large hole appearing in metal, and stress concentration is easy to damage frog, as shown in Fig. 6(b). The sand hole defect caused by plastic sand washing is characterized as more impurities mixed in the metal, affecting the local performance of the crossing, as shown in Fig. 6(c).

Fig. 6.

Electron microscope image of casting defects, (a) shrinkage porosity defect, (b) shrinkage cavity defect, (c) the sand hole defect caused by plastic sand washing.

Figure 6 shows that during the process of solidification, more shrinkage holes, shrinkage loosening, and other defects occur in the casting, which directly affects the quality of frog and reduces its service life. Therefore, to analyze the causes of defects and to predict the occurrence of defects, it is urgent to manifest the invisible sand casting process.^9,16)

In this study, the casting process of crossing is visualized and divided into five stages. (1) Filling stage, the filling speed of metal liquid was controlled by the time-casting velocity curve and the pouring was completed at 22.56 s, during which the risers were burned. (2) Burning stage, the metal liquid higher than the ignition point of the fever risers enters it successively from the far end of the gate, and the risers are in a continuous burning state. (3) The solidification stage, after pouring the rail surface near the chillers solidifies first, and its crystal quality is the best. (4) Feeding stage, the crystals at the chillers end and the left and right walls grow towards the middle of the casting, and the high-temperature metal liquid at the riser and the middle of the casting is fed downward. (5) Defect generation stage, the metal liquid in the risers solidifies from the top and sleeve to the core of the casting, so that the shrinkage is concentrated at the riser. However, shrinkage cavity, shrinkage porosity, and other casting defects occur in the areas where feeding cannot be completed in time.

3.2. Analysis of Temperature Field and Casting Defect

The filling process was completed in 22.56 second, and its solidification time of casting as shown in Fig. 7(a). The crystal character and purity of the chiller-end casting are better than that of the late crystallization position, so the quality of the working area of crossing can be improved effectively.^17,18) The metal liquid temperature of crossing core is close to the solidus around 2500 seconds and loses its fluidity basically, while the metal liquid in the risers has just dropped to the liquidus and has enough fluidity to ensure its ability to fill.

Fig. 7.

Numerical simulation results of temperature field, (a) local solidification time of casting and six-point selection diagram, (b) the temperature graphs at different nodes. (Online version in color.)

In Fig. 7(b), b₁ and b₂ are the temperature graphs at two points in contact with the direct external chillers. Once the chiller is touched, a rapid temperature drop will occur in the metal liquid, and the temperature immediately drops to near the solidus. The metal liquid begins to crystallize, and small temperature rise will occur around 170 s. c₁ and c₂ are the temperature graphs of the crossing core at two points. After the completion of pouring, the metal liquid in the core of the casting produces a small sharp temperature drop, and the metal temperature falls rapidly between the liquidus and the solidus. As the temperature approached the solidus, the solidification rate slowed down, and this region cooled simultaneously with the nose rail and wing rail. a₁ and a₂ are the temperature graphs at two points inside the fever risers. Different from the crossing inner metal liquid, the inner metal fluid did not have rapid temperature drop along with the continuous burned of the risers, and the solidification process is obviously gentle compared with other positions. When the burning of the riser is finished, the solidification rate slowly increases, so the temperature graphs are superior convex. However, the temperature graphs of the metal liquid inside the insulating risers shown in a₃ shows only slow cooling without superior convex characteristic. When the metal of crossing core basically solidified, the liquid flow in the risers was seriously insufficient, thus the ability of feeding was lacking. Figure 7 shows that the fever risers have the function of slowing down the solidification of the metal liquid and increasing the feeding distance, and the shrinkage can present the “reverse ripple” aggregation to the risers consistent with the temperature field.

Molten metal solidification will bring volume shrinkage, but high manganese steel has a higher shrinkage rate, which makes it as casting liquid will produce a large number of shrinkage due to the failure to timely feed, thus reducing casting density. This study indicated that there is a “V” type feeding area at the bottom of each fever riser for high manganese steel castings, which is related to the distance between the two risers as shown in Fig. 8. Moreover, there is an arc area between two fever risers that cannot be effectively fed. The “V” type feeding area of the edge riser has a serious trend of inward contraction and converges with the “arc area” of the outside to extend to the bottom of the riser. Figure 8 shows the shrinkage cloud diagram of the integral casting with insulation riser, which together with Fig. 7 reveals the relationship between temperature field and the formation of casting defects. When the temperature is between the liquidus and the solidus, the solidification time cloud chart at the bottom of the riser rises in a “V” shape, and the crystal grows from the bottom to the top. The molten metal in a high-temperature state fills in the reduced volume downward, to achieve the feeding effect. The molten metal crystallizes in a similar “reverse ripple” pattern to the region where the solidification time cloud chart rises inward, accompanied by a reduction in volume.¹⁹⁾ At this time there is no additional metal liquid as a supplement of the area in the “corrugated” core will form shrinkage porosity or even shrinkage cavity. Therefore, it can improve casting quality by setting riser outside the working area edge of casting.

Fig. 8.

Numerical simulation results of shrinkage cavity and porosity of casting with insulation risers. (Online version in color.)

3.3. Performance Contrast Test of Fever Risers

Risers, as the most important part to reduce casting defects, play a role in controlling the final solidification of molten metal in the casting during cooling and the casting defect is reduced by this directional solidification. Figure 8 shows that the casting defect of the crossing with insulating risers is extended from the riser to the inner rail and the casting has a large number of shrinkage and even has the tendency of serious defects, whose shrinkage porosity and shrinkage cavity account for 23.10% and 1.002% respectively. Therefore, it is necessary to select the fever risers on the casting process. However, there is no exact calculation method to determine the two key parameters of the fever risers, exothermic energy and burned time.²⁰⁾ When the exothermic energy is too little or the burned time is insufficient, the inner molten metal of risers will solidify before crystallization of the crossing core, and thus cannot be effectively fed. On the contrary, increasing exothermic energy or extending burned time can only help reduce shrinkage within the feeding distance, but once the distance is exceeded, resource waste and cost increase will be caused, and the temperature drop of the riser is very slow, which affects the knocking-out time. In order to study the influence of heating riser parameters on this study, control single variable method was adopted for testing. The parameters of the risers to be tested are listed in Table 1.

Figure 9 shows local defects with the same exothermic energy and different burned time, corresponding heating time is 120 s, 240 s, and 360 s respectively. When the exothermic energy of risers is constant, the change of burned time below 360 s has little effect on the casting result, as shown in Fig. 9 and the ratio of shrinkage porosity and cavity with different heating time also confirms this situation. According to the Table 2, the shrinkage porosity and cavity was 16.51% and 0.514% respectively when the burned time was 240 s, while the defect change rate of 120 s and 360 s was less than 3% compared with that of 240 s. The difference of the burned time of the risers (from a few seconds to a few hundred seconds) can be ignored because the average cooling rate of the casting is very low, and the numerical simulation time is all over 6000 s.

Fig. 9.

Numerical simulation results of shrinkage porosity and cavity at a different burned time under the exothermic energy of risers at 1200 kJ/kg, (a₁) the burned time was 120 s, (a₂) the burned time was 240 s, (a₃) the burned time was 360 s. (Online version in color.)

Figure 10(b₁) shows that the effect of using insulation risers as the method of feeding is poor and the serious defect extends from the risers to the crossing. Figure 10(b₂) shows that the feeding effect of fever risers is significantly better than that of insulation risers. It controls the final solidification of the molten metal in the risers by raising the temperature so that the parts with serious defects can be moved up and the shrinkage porosity and cavity proportion decreased to 19.69% and 0.670% respectively. However, some shrinkage at the bottom of the riser extends to the crossing due to insufficient exothermic energy of the riser. Figure 9(a₂) shows that the bottom of the riser with exothermic energy of 1200 kJ/kg has no defect extension, which has good feeding effect and the proportion of shrinkage porosity and the cavity is 16.51% and 0.514% respectively. Figure 10(b₃) shows that the feeding capacity of this exothermic energy is basically consistent with that of 1200 kJ/kg and the increment of defects is less than 2% compared to 240 s. Therefore, it is considered that the risers of 1200 kJ/kg exothermic energy reach the feeding limit of the casting, and only increasing the exothermic energy will waste resources and even affect the knocking-out time.

Fig. 10.

Numerical simulation results of shrinkage porosity and cavity at a different exothermic energy under the burned time of risers at 240 s, (b₁) insulating riser, (b₂) the exothermic energy was 400 kJ/kg, (b₃) the exothermic energy was 2000 kJ/kg. (Online version in color.)

4. Process Optimizations

The crossing has a transition area of nearly 800 mm between the open riser at the gating system end and the first fever riser, which has no feeding module established during casting. Due to the too-long distance between the risers on both sides, there are many defects and a large area in the “arc area” that cannot be effectively fed as shown in Fig. 11(c₁). And the casting quality in this area cannot be improved by changing the risers. Therefore, based on the above research, the process optimization scheme of adding a riser is proposed, and the feeding distance is predicted according to the “V” type feeding area of fever risers as shown in Fig. 11(c₁). To verify the reliability of the prediction, a riser is added in the three-dimensional model and the numerical simulation is carried out by ProCAST software.²¹⁾

Fig. 11.

Comparison of numerical simulation results before and after optimization of risers process, (c₁) prediction of simulation results, (c₂) simulation results. (Online version in color.)

Figure 11(c₂) shows that the casting defects in this “arc region” are greatly reduced due to the feeding of risers, and the proportion of local shrinkage porosity decreases from 36.17% to 22.11%, while the cavity decreases from 1.127% to 0.686%. At the same time, the feeding distance and feeding ability of this process optimization are consistent with the predicted “V” type shrinkage result, and this method provides a new idea for the prediction of fever risers’ feeding ability at crossing casting. In addition, an interesting fact was found that, although the model height in the region a and b in Fig. 11(c₂) is different, all the defects outside the “V” type feeding region tend to move upward due to the adding riser. The defect in region b moves up to the upper part of the bolt hole, because the height of the rail body H is 80 mm higher than that of the region a, and even in region c located outside the edge riser shows an upward trend. Moreover, moving defect towards the top of casting (the bottom side of the crossing) means that the quality of the working area is improved. This process optimization provides a new idea of improving the crossing quality.

By adjusting the riser parameters and optimizing the fever riser, the defects such as shrinkage porosity and cavity are reduced, and the performance of casting crossing is greatly improved. Figure 12 shows a significant reduction in casting defects with this excellent riser, and the passing gross tonnage could reach 300 million tons.

Fig. 12.

The defect of crossing after optimization, (a) light microscope, (b) longitudinal section micrographs in the SEM. (Online version in color.)

5. Conclusion

In this paper, an innovative finite element simulation method of the mass flow rate attenuation is used to draw the following conclusions:

(1) The fever risers with exothermic energy of 1200 kJ/kg have the least number of defects for the 75–852 crossing. Compared with insulation risers, shrinkage porosity decreased from 23.10% to 16.51%, while the cavity decreased from 1.002% to 0.514%.

(2) Compared with the same type of fever risers, the feeding capacity of exothermic energy less than 1200 kJ/kg cannot meet the requirements, and the defects are seriously increased. On the contrary, if the exothermic energy is further increased, the defects will not be reduced obviously but will lead to a waste of resources, and even affect the overall cooling rate of castings, thus affecting the knocking-out time.

(3) For large castings such as crossing, the burned time of risers is a secondary factor, whose influence is far less than the total exothermic energy.

(4) The results are consistent with the predicted “v-shaped” feeding area of the riser after optimization design and process parameter adjustment. The performance of casting crossing is greatly improved by adjusting parameters and optimizing riser and the passing gross tonnage could reach 300 million tons.

Acknowledgements

This work was supported by the Key R&D Program of Hebei Province (19211018D), Tianjin Science and Technology Project (19JCTPJC41700), and the PhD Foundation Project of Yanshan University (BL19007).

References

• 1)   X.  Yue,  L.  Li,  R.  Zhang and  F.-C.  Zhang: Mater. Charact., 60 (2009), 1541.
• 2)   L.-Y.  Ma: Foundry Technol., 39 (2018), 91 (in Chinese).
• 3)   J.-L.  Li,  R.-S.  Chen,  Y.-Q.  Ma and  W.  Ke: J. Mater. Sci. Technol., 30 (2014), No. 10, 991.
• 4)   X.-Y.  Zhang: Modern Manufacturing Technology Equipment, (2018), No. 09, 173 (in Chinese).
• 5)   S.-T.  Zhang: Railw. Qual. Control, 37 (2009), No. 09, 30 (in Chinese).
• 6)   L.-Y.  Ma: Foundry Technol., 39 (2018), 91 (in Chinese).
• 7)   S.-L.  Lu,  F.-R.  Xiao,  S.-J.  Zhang,  Y.-W.  Mao and  B.  Lv: Appl. Therm. Eng., 73 (2014), 512.
• 8)  Foundry Institution of Chinese Mechanical Engineering Society: Casting Manual, China Machine Press, Beijing, (2003), 29.
• 9)   Z.-N.  Yang,  F.-C.  Zhang,  C.-L.  Zheng,  M.  Zhang. B. Lv and L. Qu: Mater. Des., 66 (2015), 258.
• 10)   R.-D.  Li,  Z.-H.  Zhang,  Q.-C.  Xiang,  X.-D.  Qu,  G.-X.  Yang and  K.-Q.  Qiu: Foundry Technol., 36 (2015), 1196 (in Chinese).
• 11)   L.-W.  Pan,  L.-J.  Zheng,  H.  Zhang,  W.-L.  Gao: J. Beijing Univ. Aeronaut. Astronaut., 37 (2011), 428.
• 12)   L.-M.  Zhu: Fluid Mechanics, Tongji University Press, Shanghai, (2009), 20.
• 13)   D.  Bolster,  R. E.  Hershberger and  R. J.  Donnelly: Phys. Today, 64 (2011), 42.
• 14)   D.  Wang,  Q.  Zhou and  K.-L.  Cheung: J. Sound Vib., 395 (2017), 393.
• 15)   S.  Aradhya,  H.  Taofeeq and  M.  Al-Dahhan: Int. J. Multiphase Flow, 94 (2017), 209.
• 16)   L.-J.  Zhao,  L.-H.  Qian,  J.-Y.  Meng,  Q.  Zhou and  F.-C.  Zhang: Scr. Mater., 112 (2016), 96.
• 17)   F.  Inegbedion and  J. A.  Akpobi: Niger. J. Technol., 37 (2018), 696.
• 18)   Y.-G.  Li,  F.-C.  Zhang,  C.  Chen,  B.  Lv,  Z.-N.  Yang and  C.-L.  Zheng: Mater. Sci. Eng. A, 651 (2016), 945.
• 19)   H.  Tiwery,  W.  Hoziefa,  A. B.  El-Shabasy and  I.  El-Mahallawi: TMS Annual Meeting & Exhibition, Light Metals 2018, Springer, Cham, (2018), 1105.
• 20)   S.  Yao,  L.  Zhang,  J.  Xu,  Z.  Yu and  Z.  Lu: Energy Fuels, 31 (2017), 12093.
• 21)   M.  Brůna,  D.  Bolibruchová and  R.  Pastirčák: Procedia Eng., 177 (2017), 488.