I studied the integral cast high manganese steel frog as a critical component of railway turnout systems. In my work, the central problem is not only the geometric complexity of the frog but also the strong tendency of high manganese steel to form shrinkage cavities and porosity during sand casting. Because the frog must carry repeated wheel impacts, rolling contact, and severe dynamic loads, its internal quality directly controls service safety and maintenance intervals. I therefore focused on a coupled numerical simulation approach that combines filling, solidification, defect prediction, grain structure prediction, and water toughening heat treatment. The main material I selected was Mn13 cast high manganese steel, and the main manufacturing route I examined was sand casting with gravity pouring.
From my perspective, the traditional trial-and-error method is no longer sufficient for a long, thick, and geometrically irregular casting such as an integral cast frog. A single unsuccessful sand casting trial consumes a large amount of molten steel, molding sand, energy, and labor. More importantly, internal shrinkage porosity may remain hidden until the frog is machined or placed into service. I used numerical simulation to make the sand casting process visible. I predicted the filling sequence, temperature gradient, solid fraction evolution, flow velocity, and Niyama distribution. I then used those predictions to design ordinary risers, insulating risers, chills, and a sand core. The final sand casting scheme produced a sound frog body with the shrinkage defect transferred into the riser system rather than remaining inside the frog.
I also investigated water toughening because Mn13 high manganese steel in the as-cast condition usually contains a large amount of grain-boundary carbide. This carbide network increases brittleness and reduces impact toughness. Water toughening dissolves the carbides into austenite at high temperature and then rapidly quenches the casting to obtain a predominantly supercooled austenitic structure. In my simulation, I used a heating stage, a 650 °C preheating hold, a 1050 °C austenitizing hold, and a rapid water quench. I analyzed the toe and the core of the frog separately because the toe is relatively simple and thin, while the core contains complex curved surfaces, thick sections, and hot spots. I found that the temperature evolution in both regions followed a stable trend, the maximum equivalent stress was 479 MPa in the toe and 749 MPa in the core, and the final deformation after water toughening approached zero. These results helped me verify that the water toughening simulation was physically reasonable and useful for practical production.
The frog geometry I used had an overall contour of approximately 6100 mm in length, 176 mm in width, and 440 mm in height. Its volume was about 0.472 m³, and its mass was about 1200 kg. The maximum wall thickness was 71 mm and the minimum wall thickness was 12 mm. This large difference in section size is one of the main reasons why sand casting of an integral frog is difficult. Thin sections cool rapidly, while thick sections remain liquid for a long time. If the feeding path is blocked before the thick section finishes solidifying, an isolated liquid region forms and later becomes shrinkage porosity. I therefore treated the thick and complex regions as the main targets for riser and chill design.
The chemical composition of Mn13 high manganese steel that I used is summarized in Table 1. The carbon and manganese contents are the most important because they control austenite stability, carbide formation, work hardening, and wear resistance. The silicon, phosphorus, and sulfur contents were kept low to reduce inclusions, hot tearing sensitivity, and grain-boundary embrittlement.
| Element | C | Si | Mn | P | S |
|---|---|---|---|---|---|
| Mass fraction, wt.% | 1.0 | 0.5 | 13.0 | 0.035 | 0.03 |
I calculated the thermophysical properties of Mn13 high manganese steel from its composition. The liquidus temperature was about 1497 °C and the solidus temperature was about 1418 °C. The linear contraction during solidification and cooling was in the range of 2.0% to 2.5%. The thermal conductivity increased with temperature, the density decreased as temperature increased, the solid fraction evolved nonlinearly through the mushy zone, and the enthalpy increased with temperature. These nonlinear properties are important in sand casting simulation because a constant-property assumption can seriously mispredict the location and severity of shrinkage.
| Property | Trend with increasing temperature | Importance in sand casting |
|---|---|---|
| Thermal conductivity | Increases | Controls heat extraction and temperature gradient |
| Density | Decreases | Controls volumetric shrinkage and feeding demand |
| Solid fraction | Nonlinear decrease from liquid to solid | Defines mushy zone and feeding resistance |
| Enthalpy | Increases | Accounts for latent heat during solidification |
| Linear contraction | About 2.0% to 2.5% | Drives riser sizing and distortion risk |
For the filling simulation, I treated the molten steel as an incompressible fluid. The continuity equation I used is
$$ \frac{\partial u}{\partial x}+\frac{\partial v}{\partial y}+\frac{\partial w}{\partial z}=0 $$
where u, v, and w are the velocity components in the x, y, and z directions. The momentum conservation equation in the x direction can be written as
$$ \rho\left(\frac{\partial u}{\partial t}+u\frac{\partial u}{\partial x}+v\frac{\partial u}{\partial y}+w\frac{\partial u}{\partial z}\right)=-\frac{\partial p}{\partial x}+\rho g_x+\mu\nabla^2 u $$
and the y and z directions follow the same form. In these equations, ρ is density, p is pressure, μ is dynamic viscosity, and g is gravitational acceleration. I used these equations to evaluate whether the molten steel entered the mold cavity smoothly, whether the flow remained laminar, and whether the sand casting layout could cause splashing, gas entrainment, or sand erosion.
For heat transfer during solidification, I used the Fourier conduction equation with an internal source term:
$$ \rho c_p \frac{\partial T}{\partial t}=\nabla\cdot(\lambda\nabla T)+Q $$
where c_p is specific heat capacity, λ is thermal conductivity, T is temperature, t is time, and Q is the internal heat source. For latent heat release, I used the enthalpy method, which can be expressed as
$$ \lambda\left(\frac{\partial^2 T}{\partial x^2}+\frac{\partial^2 T}{\partial y^2}+\frac{\partial^2 T}{\partial z^2}\right)=\rho\left(c_p-\Delta H\frac{\partial f_s}{\partial T}\right)\frac{\partial T}{\partial t} $$
where ΔH is latent heat and f_s is solid fraction. I used this formulation because it handles both pure metals and alloys with a solidification interval. In the sand casting of a large frog, the solidification interval and the latent heat release are both significant, so this treatment is necessary.
I used the Niyama criterion to predict shrinkage cavity and porosity. The Niyama value is
$$ \mathrm{Niyama}=\frac{G}{\sqrt{R}} $$
where G is the local temperature gradient and R is the cooling rate. A low Niyama value indicates a high risk of shrinkage porosity. In my simulations, regions with Niyama values below 3 were treated as defect-prone. I also found that values in the range of 2.4 to 6.0 were associated with high-risk locations in the thick and complex sections of the frog. The temperature gradient was calculated from neighboring nodes, and the cooling rate was calculated from the temperature history:
$$ R=\frac{T^{m}-T^{m+1}}{\Delta t} $$
where T^m and T^{m+1} are temperatures at successive time steps. This criterion allowed me to compare the initial sand casting scheme with the optimized schemes quantitatively.
For the heat treatment simulation, I used the following transient heat conduction equation with phase-change enthalpy:
$$ \rho c_p \frac{\partial T}{\partial t}=\nabla\cdot(k\nabla T)+L\frac{\partial f}{\partial t} $$
where k is thermal conductivity, L is latent heat, and f is the transformed fraction. I also used diffusion-controlled carbide dissolution and austenite transformation kinetics. A simplified coupled form is
$$ \frac{\partial C}{\partial t}=D\nabla^2 C+\beta(C_{\mathrm{eq}}-C) $$
and
$$ \frac{\partial f}{\partial t}=g(T,C,f) $$
where C is carbon concentration, D is diffusion coefficient, β is a dissolution rate coefficient, and g is a transformation-rate function. For the mechanical response, I used the thermo-elasto-plastic relation
$$ \sigma=C(T,f):(\varepsilon_{\mathrm{total}}-\varepsilon_{\mathrm{plastic}}-\varepsilon_{\mathrm{thermal}}-\varepsilon_{\mathrm{phase}}) $$
where σ is stress, C is the stiffness tensor, and the strain components include plastic, thermal, and phase-transformation strains. This framework allowed me to simulate the temperature field, stress field, and strain field during water toughening.
For the initial sand casting scheme, I used a bottom gating system with a 6° tilt. I selected resin-bonded sand as the mold material. The pouring temperature was 1600 °C and the filling time was about 20 s. The mesh contained about 3,130,795 elements. The interface heat transfer parameters are given in Table 2. I set the mold-to-casting and mold-to-runner interfaces as coincident, while the casting-to-runner interface was treated as equivalent material. This initial sand casting layout was intended to fill the mold smoothly and avoid excessive turbulence.
| Interface | Mold-casting | Mold-runner | Casting-runner |
|---|---|---|---|
| Material condition | Coincident | Coincident | Equivalent |
| Heat transfer coefficient, W/(m²·K) | 500 | 500 | Not set |
In the initial sand casting simulation, the filling process was smooth. The metal filled the toe first, then progressed toward the central region, and finally reached the area near the sprue. The maximum flow velocity was about 1.52 m/s, which was low enough to avoid severe sand erosion. However, the solidification sequence was not ideal. The toe cooled and solidified first because it was thin and simple. The region near the sprue also cooled relatively quickly because it had a simpler shape. The central core and the thick sections remained liquid for a longer time. As a result, the feeding path from the riser, which did not yet exist in the initial scheme, was absent. The central thick sections became isolated liquid regions, and shrinkage porosity formed.
The temperature curves I extracted from three positions showed that the toe cooled fastest, the core cooled more slowly, and the region near the sprue cooled at an intermediate rate. The solid fraction curves confirmed that the toe reached complete solidification before the core. This behavior violates the desired sequence for sound sand casting. In a proper sequence, the farthest thin sections should solidify first, the intermediate sections should solidify next, and the riser should solidify last. If this order is not achieved, the riser cannot feed the last liquid regions.
The defect prediction for the initial sand casting scheme showed shrinkage porosity concentrated in the thick central region and in complex curved areas. I extracted Niyama values at several points. The points located in the thick core and complex sections fell below the critical threshold. The lowest Niyama values appeared where the local temperature gradient was low and the cooling rate was also low, which is characteristic of a hot spot. This result confirmed that the initial sand casting scheme required optimization.
| Initial sand casting observation | Result | Consequence |
|---|---|---|
| Filling behavior | Smooth and laminar | Low risk of gas entrainment |
| Maximum flow velocity | About 1.52 m/s | Acceptable for sand erosion resistance |
| Solidification sequence | Toe first, core last | Feeding path blocked |
| Niyama risk zone | Thick core and complex curves | Shrinkage porosity |
| Defect location | Central and thick sections | Poor internal quality |
I then designed an ordinary riser and chill optimization. In this scheme, I placed ordinary risers at the thick and complex regions identified by the initial simulation. I also attached external chills to the frog surface to promote rapid local solidification and to help establish a directional temperature gradient. The mesh size increased to about 4,076,549 elements because the risers and chills added geometric complexity. The pouring temperature remained 1600 °C and the filling time remained about 20 s. The heat transfer coefficients are listed in Table 3.
| Interface | Mold-casting | Mold-runner | Casting-runner or riser | Chill-casting | Ordinary riser-mold |
|---|---|---|---|---|---|
| Material condition | Coincident | Coincident | Equivalent | Coincident | Coincident |
| Heat transfer coefficient, W/(m²·K) | 500 | 500 | Not set | 2000 | 200 |
In the ordinary riser scheme, the temperature field showed a more favorable gradient. The thin toe still cooled first, but the thick central section remained hotter than the surrounding thin sections. The riser remained liquid longer than the casting body, which allowed feeding to continue. The solid fraction curves showed that the riser solidified later than the adjacent casting sections. The flow field showed that the maximum velocity increased to about 2.293 m/s, but most of the flow remained below about 1.07 m/s. The filling remained orderly, and the riser filled after the main cavity. This sequence is desirable because the riser can supply liquid metal to the casting during the final stages of solidification.
The defect prediction for the ordinary riser scheme showed a clear improvement. Many shrinkage defects moved into the ordinary risers. The frog body contained fewer defects than in the initial sand casting scheme. However, two large shrinkage regions remained in the core and in the complex central area. I selected four points in these regions and plotted the Niyama history. The Niyama values remained between 2.4 and 6.0, which confirmed that these locations were still at high risk. The ordinary riser did not provide enough feeding volume or enough thermal insulation to keep the feeding path open long enough. I concluded that the feeding capacity of the ordinary riser was insufficient for the largest hot spots.
| Ordinary riser scheme | Observation | Interpretation |
|---|---|---|
| Temperature gradient | Improved but not sufficient | Hot spots remained in core |
| Solidification sequence | Riser later than adjacent casting | Feeding path partially effective |
| Flow field | Maximum about 2.293 m/s | No severe sand erosion |
| Defect location | Mostly in risers, but large defects in core | Insufficient feeding volume |
| Niyama range in core | 2.4 to 6.0 | High risk remained |
I therefore replaced the ordinary risers in the critical locations with insulating risers. The purpose of the insulating riser was to reduce heat loss from the riser top and sides, delay riser solidification, and maintain a liquid feeding channel for a longer time. I placed an insulating sleeve around the riser. I set the heat transfer coefficient between the insulating sleeve and the sand mold to 100 W/(m²·K). The casting-to-insulating-riser interface was treated as equivalent material because the riser and casting were both Mn13 steel. The chill-to-casting heat transfer coefficient remained 2000 W/(m²·K). The optimized sand casting model included the insulating risers, the chills, and the gating system.
| Interface | Mold-casting | Insulating sleeve-mold | Casting-insulating riser | Chill-casting |
|---|---|---|---|---|
| Material condition | Coincident | Coincident | Equivalent | Coincident |
| Heat transfer coefficient, W/(m²·K) | 500 | 100 | Not set | 2000 |
In the insulating riser scheme, the filling time was about 19.6 s, and the solidification start time was about 10.2 s. Complete solidification was reached at about 1733 s, with a total process time of about 1780 s. This was longer than the ordinary riser scheme, which completed solidification at about 1404 s with a total time of about 1451 s. The longer solidification time was not a disadvantage in this case because it indicated that the insulating riser remained liquid and continued to feed the casting. I compared the process times in Table 5.
| Scheme | Filling complete, s | Solidification start, s | Solidification complete, s | Total time, s |
|---|---|---|---|---|
| Ordinary riser | 21.7 | 13.2 | 1404 | 1451 |
| Insulating riser | 19.6 | 10.2 | 1733 | 1780 |
The temperature curves for the insulating riser scheme showed that the bottom of the frog cooled fastest, the region near the insulating riser cooled more slowly, and the insulating riser itself cooled most slowly. This is the desired sequence for sand casting. The solid fraction curves showed the same trend: the riser remained liquid after the adjacent casting had solidified. The defect prediction showed that the large defect near the sprue was greatly reduced or eliminated. Most of the remaining shrinkage was located in the insulating risers. However, two large shrinkage regions still remained in the core. I selected two points in those regions and plotted the Niyama history. Both points remained below the safe threshold, which indicated that the local hot spot was still not fully fed.
| Insulating riser scheme | Observation | Interpretation |
|---|---|---|
| Filling time | About 19.6 s | Stable filling |
| Solidification time | About 1780 s | Riser stayed liquid longer |
| Temperature gradient | Bottom cooled first, riser last | Improved directional solidification |
| Defect transfer | Most defects in risers | Riser feeding effective |
| Remaining defects | Two large core defects | Local feeding still insufficient |
For the final optimization, I combined three measures. First, I placed a chrome ore sand core directly against the frog surface at the second remaining defect location. Chrome ore sand has much higher thermal conductivity than silica sand, and it also has high refractoriness and good chilling capacity. This sand core increased the local cooling rate, reduced the hot spot, and promoted directional solidification. Second, I increased the volume of the insulating riser at the other remaining defect location. The larger riser provided more liquid steel and a longer feeding time. Third, I retained the chills and the optimized gating system. The heat transfer coefficients for the final scheme are given in Table 7.
| Interface | Sand core-casting | Insulating sleeve-mold | Casting-insulating riser | Ordinary riser-mold |
|---|---|---|---|---|
| Material condition | Coincident | Coincident | Equivalent | Coincident |
| Heat transfer coefficient, W/(m²·K) | 1000 | 100 | Not set | 200 |
In the final sand casting simulation, the filling was complete at about 18.6 s. The liquidus temperature was about 1391.8 °C and the solidus temperature was about 1124.4 °C in the simulation database. The temperature field showed that the thin sections cooled first, the thick sections cooled later, and the insulating riser remained hot for a long time. At 3600 s, the insulating riser still contained liquid metal, which provided continued feeding. The solid fraction distribution showed that the frog followed the desired sequence: the two ends solidified first, then the intermediate sections, then the core, and finally the risers. At 282 s, the casting was about 50% solid, and the sequence was already clearly directional. At 3600 s, the casting was about 85% solid, and the new insulating riser still contained liquid. The final defect prediction showed that the internal shrinkage cavity and porosity were eliminated. The shrinkage defects were transferred to the insulating risers, and the frog body was dense and uniform.
| Final sand casting scheme | Result | Benefit |
|---|---|---|
| Filling complete | About 18.6 s | Smooth mold filling |
| Solidification sequence | Ends, intermediate, core, riser | Directional solidification |
| Riser condition at 3600 s | Still liquid | Continuous feeding |
| Internal defects | Eliminated | Sound frog body |
| Defect location | Riser system | Easy removal during cleaning |
I also used the CAFE module to simulate grain growth during solidification. I used a surface nucleation maximum density of 1.6 × 10^5 m^-2, a volume nucleation density of 5.2 × 10^7 m^-3, a surface nucleation mean undercooling of 1 K, a surface nucleation standard deviation of 0.1 K, a volume nucleation mean undercooling of 10 K, and a volume nucleation standard deviation of 0.1 K. The growth kinetic parameters were α₂ = 5.6 × 10^-7 and α₃ = 2.8 × 10^-7. The alloy elements and their growth parameters are listed in Table 9.
| Element | Mass fraction, % | Liquidus slope | Equilibrium partition coefficient | Self-diffusion coefficient |
|---|---|---|---|---|
| Mn | 13 | -3.81 | 0.7776 | 3 × 10^9 |
| C | 1 | -69.33 | 0.3600 | 3 × 10^9 |
| Si | 0.5 | -23.51 | 0.6129 | 3 × 10^9 |
| P | 0.035 | -38.37 | 0.0579 | 3 × 10^9 |
| S | 0.03 | -29.46 | 0.0340 | 3 × 10^9 |
The CAFE simulation showed that a very thin fine equiaxed grain zone formed at the surface during the first stage of solidification. This happened because the sand mold was at room temperature and the molten steel was at high temperature, creating a large initial undercooling at the mold wall. After that, columnar grains grew inward along the direction of the decreasing temperature gradient. In the central region, equiaxed grains nucleated and grew, and some dendrites formed because of the local temperature gradient. The final structure of the frog body was fine and dense equiaxed grains, while the riser contained columnar grains. The fine equiaxed structure is beneficial because it increases grain boundary area, improves strength, and distributes strain more uniformly. The grain simulation therefore supported the final sand casting design.
After solidification, I studied the water toughening heat treatment of the integral cast frog. The as-cast Mn13 structure contains austenite, carbides, and a small amount of pearlite. The carbide network is harmful because it embrittles the grain boundaries. Water toughening dissolves the carbides at high temperature and then quenches rapidly to suppress carbide reprecipitation. I used the following process: slow heating to 650 °C for preheating, continued heating to 1050 °C, holding for 2 h, and then quenching in water at about 30 °C. The cooling rate was set to about 30 °C/s, and the cooling time was about 34 s. The preheating step reduces thermal gradients and prevents cracking. The 1050 °C hold allows carbon and manganese to diffuse and the carbides to dissolve. The rapid quench produces a single supercooled austenitic structure.
I extracted temperature-dependent material properties from a material property calculation tool. The properties included Poisson’s ratio, thermal expansion coefficient, Young’s modulus, specific heat capacity, thermal conductivity, and the stress-strain curve. The room-temperature yield strength was about 445 MPa. I used these properties in the DEFORM simulation. I divided the frog into a toe model and a core model to reduce computation time while preserving the representative geometry. The toe model had about 80,000 mesh elements and the core model also had about 80,000 mesh elements. A mesh sensitivity check showed that 50,000 elements produced a rough curved surface, while 80,000 elements produced acceptable geometric accuracy and reasonable computation cost.
| Model | Mesh elements | Nodes | Elements | Polyhedra | Endpoints |
|---|---|---|---|---|---|
| Toe | 80,000 | 17,269 | 68,572 | 36,330 | 18,165 |
| Core | 80,000 | 11,452 | 43,955 | 53,100 | 26,534 |
In the toe water toughening simulation, the heating and cooling temperature fields were stable. During preheating to 650 °C, the two ends of the toe heated faster than the central thick section. This is consistent with the smaller section size and simpler shape at the ends. As the furnace temperature increased to 1050 °C, the entire toe reached the target temperature. During quenching, the edge regions cooled faster than the interior. The temperature difference gradually decreased, and the final temperature approached room temperature. The temperature histories at four selected points showed almost the same trend, with no large fluctuation. This indicates that the heating and cooling were uniform enough to avoid inconsistent phase transformation.
I also divided the toe water toughening process into preheating, austenitizing, and quenching stages. In the preheating stage, the two end points heated slightly faster than the central points. In the austenitizing stage, the temperature curves at all four points converged toward 1050 °C. In the quenching stage, the point near the thin end cooled fastest, while the point in the thicker central region cooled more slowly. The heat transfer direction was mainly from the interior to the two ends, and then from the surface to the water. This behavior is consistent with the physical response of a long sand casting during water toughening.
The toe stress field showed a maximum equivalent stress of about 479 MPa. The maximum stress was located at the bent region of the wing rail, where multiple complex curved surfaces meet. This is a stress concentration area. The stress distribution on the surface was relatively uniform, and no abnormal high-stress region appeared at the ends or on the outer surfaces. The stress increased gradually from the outer surface toward the interior. The stress histories at four points showed that the stress first increased, then decreased during preheating and holding, and finally increased again during quenching. The maximum stress remained below the level that would cause cracking. The stress field therefore indicated that the water toughening process was reasonable for the toe.
The toe strain field showed that the largest deformation occurred in the middle of the toe model, again near the bent wing rail region. This region has complex curvature and is prone to uneven heating and cooling. The deformation along the Y axis reached about -2.49 at the point of maximum deformation. However, after cooling was complete, the Y-axis deformation at all four points approached zero. This means that the toe experienced stress during water toughening but did not retain significant permanent deformation. The uniform heating, appropriate holding time, and suitable cooling medium all contributed to this result.
| Toe water toughening result | Value or observation | Interpretation |
|---|---|---|
| Maximum equivalent stress | 479 MPa | Below cracking risk |
| Stress concentration location | Bent wing rail region | Complex curved surfaces |
| Maximum Y-axis deformation | -2.49 | Transient deformation during treatment |
| Final deformation after cooling | Approaches zero | No significant permanent distortion |
| Temperature uniformity | Stable curves | Uniform phase transformation |
In the core water toughening simulation, the temperature field was also stable. During preheating to 650 °C, the temperature distribution was layered, with the edge slightly hotter than the center. The temperature uniformity was good. During the 1050 °C hold, the temperature was concentrated near 1050 °C with no obvious low-temperature region. During quenching, a significant temperature difference appeared between the interior and the surface. The internal temperature remained higher, while the surface cooled rapidly. After quenching, the core temperature approached room temperature, with only a small central region remaining slightly warmer. This indicates that the quench was uniform and thorough. The temperature histories at four points in the core followed the same trend as the toe. This consistency supports the validity of the water toughening simulation.
The core stress field showed a maximum equivalent stress of about 749 MPa. The maximum stress was again located near the bent wing rail region, where multiple complex curved surfaces meet. The stress distribution was relatively uniform on the surface, and the stress increased from the surface toward the interior. The core stress values were generally higher than the toe stress values because the core has a more complex shape and larger wall thickness. The stress histories at four points showed that the stress increased sharply at the beginning of preheating, then decreased during holding, and finally increased again during quenching. The maximum stress remained below the cracking threshold but was high enough to require careful process control.
The core strain field showed that the largest deformation occurred in the middle of the core and near the bent wing rail region. These are the same regions identified as stress concentration zones. The maximum Y-axis deformation was about -2.63, which is slightly larger than the toe deformation. After cooling, the Y-axis deformation at all four points approached zero. This means that the core also did not retain significant permanent deformation. The uniform temperature field, appropriate heating rate, and suitable cooling medium helped to minimize residual stress and distortion.
| Core water toughening result | Value or observation | Interpretation |
|---|---|---|
| Maximum equivalent stress | 749 MPa | Higher than toe due to complex thickness |
| Stress concentration location | Bent wing rail region | Complex curved surfaces |
| Maximum Y-axis deformation | -2.63 | Slightly larger than toe |
| Final deformation after cooling | Approaches zero | No significant permanent distortion |
| Temperature uniformity | Stable curves | Uniform phase transformation |
I compared the toe and core results. The temperature histories were nearly identical in trend. The stress concentration locations were the same, although the core reached a higher maximum stress. The deformation histories were also similar, and both approached zero after cooling. These comparisons show that the toe and core simulations provide consistent information about the water toughening process. They also show that the most critical regions are the bent wing rail and the complex curved surfaces. These regions should receive special attention in actual production, including careful control of heating rate, holding time, and quench uniformity.
| Comparison item | Toe | Core | Conclusion |
|---|---|---|---|
| Temperature trend | Stable | Stable | Uniform heating and cooling |
| Maximum equivalent stress | 479 MPa | 749 MPa | Core is more critical |
| Stress concentration | Bent wing rail | Bent wing rail | Same critical geometry |
| Maximum Y-axis deformation | -2.49 | -2.63 | Core deformation slightly larger |
| Final deformation | Near zero | Near zero | No permanent distortion |
The optimization process I followed can be summarized as a sequence of increasingly accurate sand casting designs. The initial scheme established the baseline filling and solidification behavior. The ordinary riser and chill scheme introduced feeding and local cooling. The insulating riser scheme extended the feeding time and reduced the defect volume. The final scheme combined a chrome ore sand core, a larger insulating riser, chills, and an optimized gating system. The final sand casting simulation showed that the shrinkage defects were eliminated from the frog body and transferred to the riser system. The CAFE simulation showed that the final frog body had a fine equiaxed grain structure, which is beneficial for strength and toughness. The water toughening simulation showed that the heat treatment produced uniform temperature fields, acceptable stress levels, and negligible final deformation.
From a practical point of view, my results suggest that sand casting of an integral high manganese steel frog should not rely on a single riser or a single chill. The feeding system must be designed according to the local hot spot size, the local cooling rate, and the distance from the feeding path. The Niyama criterion is a useful quantitative tool for identifying risky regions. A Niyama value below 3 is a strong indicator of shrinkage. In my case, values between 2.4 and 6.0 were associated with high-risk locations in the thick and complex sections. The use of insulating risers is especially effective because it delays riser solidification and maintains a liquid feeding channel. The use of chrome ore sand cores is effective because it increases the local cooling rate and reduces the hot spot. The combination of these methods creates a more favorable temperature gradient and promotes directional solidification.
I also conclude that water toughening is not merely a final step but an integral part of the manufacturing chain. The as-cast carbide network must be dissolved to obtain the required toughness and wear resistance. The heating rate must be controlled to avoid thermal shock. The 650 °C preheating stage is important for reducing temperature gradients. The 1050 °C holding stage is important for carbide dissolution and austenite homogenization. The water quench must be fast enough to suppress carbide reprecipitation but uniform enough to avoid excessive residual stress. My simulation showed that the temperature curves at different locations remained stable, the maximum equivalent stress was below the cracking threshold, and the final deformation approached zero. These results support the use of numerical simulation for designing and verifying the water toughening process.
In my final assessment, the integral cast high manganese steel frog can be produced with good internal quality if the sand casting process is designed with a combination of insulating risers, ordinary risers where appropriate, external chills, and a chrome ore sand core. The final sand casting scheme eliminates internal shrinkage cavity and porosity, promotes a fine equiaxed grain structure in the frog body, and transfers residual defects to the riser system. The water toughening simulation confirms that the heat treatment can dissolve carbides, produce a supercooled austenitic structure, and minimize residual stress and distortion. The coupled use of filling simulation, solidification simulation, defect prediction, grain growth simulation, and heat treatment simulation provides a complete digital framework for manufacturing the integral cast high manganese steel frog.
$$ \mathrm{Niyama}=\frac{G}{\sqrt{R}}<3 \quad \text{indicates shrinkage risk} $$
$$ \frac{\partial f_s}{\partial t}<0 \quad \text{and} \quad \frac{\partial T}{\partial t}<0 \quad \text{define solidification progress} $$
$$ \sigma_{\max,\mathrm{toe}}=479\ \mathrm{MPa},\quad \sigma_{\max,\mathrm{core}}=749\ \mathrm{MPa} $$
$$ \Delta Y_{\mathrm{toe}}=-2.49,\quad \Delta Y_{\mathrm{core}}=-2.63,\quad \Delta Y_{\mathrm{final}}\approx 0 $$
These quantitative results and qualitative observations guided my process optimization. I believe that the final sand casting and water toughening scheme is technically sound, and that the simulation framework can be transferred to other large, complex, high manganese steel castings. The key is to maintain a clear understanding of the relationship between temperature gradient, cooling rate, solid fraction, feeding distance, and stress evolution. When these factors are controlled together, the integral cast high manganese steel frog can achieve the required internal quality, dimensional stability, and service performance.