As a casting engineer, I have long been challenged by the development of high-quality cast steel axle housings used in heavy-duty mining vehicles. These components are large in size, complex in geometry, and must withstand extreme service loads. The traditional trial-and-error approach to casting process design is both time-consuming and costly. Over the past few years, I have employed numerical simulation techniques, particularly using the ProCAST software, to optimize the casting process of such components. This article summarizes my work on the application of numerical simulation for the prediction of shrinkage cavities and porosity in steel casting, with a focus on a cast steel axle housing for a 70-ton mining dump truck.
Steel casting is a critical manufacturing route for heavy-duty structural components. However, the high pouring temperature, large solidification shrinkage, and complex mold filling behavior often lead to defects such as shrinkage cavity, porosity, hot tearing, and sand inclusion. In this project, the axle housing has an overall size of approximately 2116 mm × 596 mm × 364 mm and a mass of about 553 kg. The material is SCW550, a low-alloy cast steel with high strength and toughness. Because the wall thickness varies from 15 mm to 87 mm, and because the part contains bosses, spring seats, flanges, and a central differential housing, the solidification behavior is highly non-uniform. Without a well-designed feeding system, the last-solidifying regions tend to form shrinkage defects.
My primary objective was to shorten the development cycle and reduce the rejection rate by using casting simulation to evaluate and improve the initial process design. I combined CAD modeling with finite-element-based simulation to analyze both mold filling and solidification, and I applied several prediction criteria for shrinkage cavity and porosity. The work involved iterative process modifications, including the addition of risers, the use of insulating riser sleeves, and the substitution of core sand materials. Ultimately, the optimized process was verified through pilot production, and the actual casting results matched the simulation predictions.
1. Introduction
Axle housings are among the most safety-critical components of off-highway mining trucks. They support the vehicle weight and transmit driving and braking torques. A casting axle housing is preferred when the part is too large and complex to be fabricated by forging or stamping. However, the casting process must be carefully controlled to avoid internal defects that could compromise the component integrity. In the past, process engineers relied on empirical rules and physical trials, which consumed substantial time and material resources. With the rapid advancement of computer simulation, foundry engineers can now visualize the filling and solidification sequences and predict the locations where shrinkage porosity may occur. This technology, often called casting CAE, has become a powerful tool for modern steel casting foundries.
In this study, I focused on a specific cast steel axle housing. The initial casting process was designed based on standard principles for steel casting, including the gating system, sprue, runners, and ingates. I then built a three-dimensional model using CATIA and performed numerical simulations using ProCAST. The simulation results revealed that shrinkage defects were mainly located at the two end flanges, at the spring seats, and in the thick central section. To solve these problems, I adopted a stepwise optimization strategy: first, I added conventional risers; second, I replaced them with insulating riser sleeves; third, I changed the core sand from phenolic resin-bonded sand to sodium silicate sand with better collapsibility. Each modification was evaluated through simulation before being applied to actual production. The final optimized process produced castings that met both internal soundness and dimensional requirements.
2. Fundamentals of Casting Process Simulation and Shrinkage Prediction
2.1 Governing Equations for Mold Filling
The mold filling process for steel casting involves transient, turbulent flow of a high-temperature liquid metal. To simulate this behavior, I used the continuity equation, the Navier–Stokes equations, and the energy conservation equation. For an incompressible Newtonian fluid, the continuity equation can be written as:
$$ \frac{\partial u}{\partial x} + \frac{\partial v}{\partial y} + \frac{\partial w}{\partial z} = 0 $$ (1)
The momentum equations in three dimensions are:
$$ \rho \left( \frac{\partial u}{\partial t} + u\frac{\partial u}{\partial x} + v\frac{\partial u}{\partial y} + w\frac{\partial u}{\partial z} \right) = -\frac{\partial P}{\partial x} + \rho g_x + \mu \nabla^2 u $$ (2)
$$ \rho \left( \frac{\partial v}{\partial t} + u\frac{\partial v}{\partial x} + v\frac{\partial v}{\partial y} + w\frac{\partial v}{\partial z} \right) = -\frac{\partial P}{\partial y} + \rho g_y + \mu \nabla^2 v $$ (3)
$$ \rho \left( \frac{\partial w}{\partial t} + u\frac{\partial w}{\partial x} + v\frac{\partial w}{\partial y} + w\frac{\partial w}{\partial z} \right) = -\frac{\partial P}{\partial z} + \rho g_z + \mu \nabla^2 w $$ (4)
Here, $\rho$ is the density, $u$, $v$, $w$ are the velocity components, $P$ is the pressure, $\mu$ is the dynamic viscosity, and $g_x$, $g_y$, $g_z$ are the gravitational accelerations. In addition, the energy equation accounts for heat transfer during filling:
$$ \rho c_p \left( \frac{\partial T}{\partial t} + u\frac{\partial T}{\partial x} + v\frac{\partial T}{\partial y} + w\frac{\partial T}{\partial z} \right) = \frac{\partial}{\partial x}\left(k\frac{\partial T}{\partial x}\right) + \frac{\partial}{\partial y}\left(k\frac{\partial T}{\partial y}\right) + \frac{\partial}{\partial z}\left(k\frac{\partial T}{\partial z}\right) + S $$ (5)
where $c_p$ is the specific heat, $k$ is the thermal conductivity, and $S$ represents the internal heat source, including the latent heat of solidification. To handle turbulence, I applied the standard $k-\varepsilon$ two-equation model, which introduces additional transport equations for turbulent kinetic energy $k$ and its dissipation rate $\varepsilon$.
2.2 Solidification and Heat Transfer Modeling
After the mold cavity is filled, the molten steel starts to solidify. The solidification process is governed by a three-dimensional transient heat conduction equation:
$$ \rho c_p \frac{\partial T}{\partial t} = \frac{\partial}{\partial x}\left(\lambda_x \frac{\partial T}{\partial x}\right) + \frac{\partial}{\partial y}\left(\lambda_y \frac{\partial T}{\partial y}\right) + \frac{\partial}{\partial z}\left(\lambda_z \frac{\partial T}{\partial z}\right) + Q_L $$ (6)
Here, $\lambda_x$, $\lambda_y$, $\lambda_z$ are the directional thermal conductivities, and $Q_L$ is the latent heat released during phase transformation. I used the Lever rule to compute the solid fraction as a function of temperature, which is a common approach for steel casting simulations. The solid fraction is given by:
$$ f_s = \sqrt{\frac{T_L – T}{T_L – T_S}} $$ (7)
where $T_L$ is the liquidus temperature, $T_S$ is the solidus temperature, and $T$ is the local temperature. The latent heat affects the effective specific heat, which can be written as:
$$ C_{eff} = f_s C_S + (1-f_s)C_L – L \frac{\partial f_s}{\partial T} $$ (8)
where $C_S$ and $C_L$ are the specific heats of the solid and liquid phases, respectively, and $L$ is the latent heat of fusion.
2.3 Shrinkage Cavity and Porosity Prediction Criteria
Shrinkage defects in steel casting occur because the volume of solid metal is smaller than that of the liquid metal. When the feeding path is blocked, the last-solidifying liquid contracts and leaves voids. Several criteria are available to predict shrinkage porosity. In my simulation work, I used the following methods:
- Temperature gradient method: A low thermal gradient in the mushy zone indicates poor feeding.
- Niyama criterion: This criterion combines temperature gradient $G$ and cooling rate $R$ into a dimensionless parameter $G/\sqrt{R}$. When this value is less than a critical threshold, porosity is likely to form.
- HOTSPOT criterion: This is used to identify the last-solidifying regions, which are potential defect locations.
- Porosity criterion: The ProCAST software provides a porosity module that quantitatively predicts the percentage of micro-porosity.
The Niyama criterion is described by:
$$ NIYAMA = \frac{G}{\sqrt{R}} $$ (9)
where $G$ is the local temperature gradient and $R$ is the cooling rate. In practice, a threshold value around 0.8 to 1.0 is used for steel castings, depending on the section size. I combined the HOTSPOT and Porosity criteria to evaluate the shrinkage tendency in the axle housing.
3. Modeling and Simulation Setup
3.1 CAD Modeling and Mesh Generation
I created a full three-dimensional model of the casting with the gating system using CATIA V5. The model included the axle housing, the sprue, the runner, the ingates, and later the risers. To reduce the computation time while maintaining accuracy, I performed local mesh refinement in the thin-wall regions and in areas where high thermal gradients were expected. The mesh was generated using the Visual-Mesh module of ProCAST. First, I checked the geometry for errors such as gaps and intersections. I used the automatic repair function whenever possible. After generating the surface mesh, I checked its quality, and then generated a conforming volume mesh. For the initial process design, the final mesh consisted of approximately 30,562 surface nodes and 129,338 volume elements. The mesh quality was acceptable for a finite-element thermal-fluid simulation.
3.2 Virtual Mold Design
Instead of explicitly drawing the sand mold and cores, I used a virtual mold technique in ProCAST. The virtual mold automatically fills the void space between the casting geometry and an outer boundary, which greatly simplifies the pre-processing work. I defined the mold material as phenolic resin-bonded sand for the initial simulations. The sand mold dimensions were chosen to be about twice the casting size, which is typical for this class of steel casting. The virtual mold boundary conditions were set to represent natural air convection on the outer surface.
3.3 Material Properties
The casting material is SCW550, a low-carbon cast steel. Its chemical composition is listed in the table below. I used the material database in ProCAST, supplemented with data from foundry handbooks for the thermal properties. The liquidus and solidus temperatures were calculated using empirical equations and then cross-checked with the ProCAST database. The calculated values were within 2 °C of the software values, which gave me confidence in the simulation input.
| Element | C | Si | Mn | P | S | Mo | Cr | Ni | Fe |
|---|---|---|---|---|---|---|---|---|---|
| Content | 0.20 | 0.50 | 1.00 | 0.03 | 0.03 | 0.05 | 0.20 | 0.50 | Balance |
The density of SCW550 was treated as 7.6 g/cm³ in the solid state and 7.4 g/cm³ in the two-phase region. I obtained the thermal conductivity and specific heat curves as functions of temperature from the ProCAST database. The solidus and liquidus temperatures were 1457 °C and 1504 °C, respectively, as determined by the software. The latent heat was set to 310 kJ/kg, a typical value for low-carbon steel. These thermal parameters are crucial for accurate solidification simulation because they directly affect the cooling curves and the predicted shrinkage locations.
3.4 Heat Transfer Boundary Conditions
The interface heat transfer coefficient between the steel casting and the sand mold plays a significant role in the solidification pattern. In the early stage, the liquid metal contacts the mold surface intimately, and the heat transfer coefficient is high. Once an air gap forms due to solidification contraction, the coefficient decreases sharply. I set the casting-to-mold interface coefficient to 750 W/(m²·K), which is within the commonly used range. For the outer surface of the virtual mold, I applied a natural convection coefficient of 25 W/(m²·K). These values were kept constant for all simulations unless otherwise noted.
3.5 Initial and Running Conditions
The pouring temperature was set to 1640 °C, the mold initial temperature was 25 °C, and the pouring time was set by a mass flow rate of 15 kg/s. The gravity direction was specified as +Z (vertical downward). The stopping temperature for the solidification calculation was set to 700 °C. In the ProCAST run parameters, I enabled the porosity, hot spot, and Niyama criteria. The filling was set to 100% so that the entire cavity would be filled before the solidification stage began. These settings are summarized in the table below.
| Parameter | Value |
|---|---|
| Pouring temperature (°C) | 1640 |
| Pouring rate (kg/s) | 15 |
| Gravity direction | +Z |
| Initial casting temperature (°C) | 1640 |
| Initial mold temperature (°C) | 25 |
| Casting-mold heat transfer coefficient (W/(m²·K)) | 750 |
| Mold-air heat transfer coefficient (W/(m²·K)) | 25 |
| Stop temperature (°C) | 700 |
| Filling percentage | 100% |
4. Initial Simulation Results and Defect Analysis
The initial casting process did not include any risers. The gating system was designed as a bottom-filling system with a single sprue and two ingates. After running the simulation, I obtained the filling sequence and the solidification time. The filling process was relatively stable, with no severe turbulence. However, the solidification time distribution showed that the last-solidifying regions were located at the two end flanges, at the spring seats, and in the thick central section. These regions corresponded exactly to the shrinkage predictions from the Porosity and Hotspot criteria. I observed a large shrinkage cavity at the upper central part, which indicated that the gating system was not capable of feeding the casting during the final stage of solidification.
The shrinkage porosity maps showed three main clusters of defects: at the left and right axle-end flanges, at the spring seats, and in the middle bridge section. The predicted porosity percentage reached as high as 5% in some areas. These defects would certainly cause the casting to be rejected during ultrasonic inspection. Therefore, I concluded that the initial process design was inadequate and needed to be improved by adding external risers and by modifying the core sand material.
5. Process Optimization through Simulation
5.1 Optimization Strategy 1: Addition of Conventional Riser
Riser design is the most common method to eliminate shrinkage defects in steel casting. The riser must solidify after the casting section it feeds, must contain enough liquid metal to compensate for the volume contraction, and must maintain a clear feeding path. Based on the simulation results, I placed one conventional riser at each end flange and one riser near the ingate system. The riser dimensions were determined using the modulus method, considering the local thermal modulus of the casting. With these risers added, the filling and solidification simulations were repeated. The new solidification time map showed that the last-solidifying regions shifted from the casting body to the risers themselves. The porosity defects in the end flanges were significantly reduced, but the central rail seat area still showed some porosity. Also, the conventional sand risers did not remain liquid long enough to feed the thick central section completely.
5.2 Optimization Strategy 2: Insulating Riser Sleeves
To improve the feeding efficiency of the risers, I replaced the conventional sand risers with insulating riser sleeves. These sleeves are made of a low thermal conductivity material, typically with a density of about 500 kg/m³, and they slow down the heat loss from the riser surface. I used the ProCAST material database to define the thermal conductivity and specific heat of the insulating material as functions of temperature. In addition, I applied an exothermic topping compound on the open top of the risers to keep the liquid steel hot. This is a common practice in steel casting to maintain the riser liquid level and to ensure that the last portion of the riser remains molten.
The simulation with insulating sleeves showed a further reduction in the shrinkage porosity. The hot spots were now clearly confined to the risers, and the feeding path to the spring seats remained open for a longer time. However, the central section of the axle housing, which is the thickest area around the differential housing, still contained some isolated porosity. This was because the sand core in the central cavity created a localized high-temperature region, and the feeding distance from the risers was too long. Thus, I realized that additional measures were needed to address the central shrinkage.
5.3 Optimization Strategy 3: Replacement of Core Sand Material
The third strategy was to change the core sand used in the central cavity from phenolic resin-bonded sand to sodium silicate (water glass) sand, which has better collapsibility and lower high-temperature strength. The rationale was that a more collapsible core would impose less restraint during solidification, thereby reducing hot tears and also altering the local heat transfer. More importantly, the thermal conductivity and heat capacity of sodium silicate sand differ from those of resin sand, which changes the cooling rate of the adjacent steel casting. Using a sand with higher thermal diffusivity would accelerate the cooling of the thick central section, reduce the thermal gradient, and minimize the formation of isolated hot spots.
I looked up the thermal properties of water glass sand from the ProCAST database. The density was around 1520 kg/m³. I then created a new mold model with a hybrid core: the outer parts still used resin sand, while the central core was defined as water glass sand. The simulation parameters were kept the same as those of the previous run. After solving, the temperature history in the central bridge section showed a significantly faster cooling rate than before. The porosity map indicated that the central shrinkage porosity was almost eliminated. The combination of insulating risers at the ends and the water glass sand core in the middle provided a comprehensive solution for the defects.
6. Pilot Production Verification
After the final optimized process was established, I proceeded to pilot production. The molds and cores were produced using the same resin sand system for the main sand and water glass sand for the central core. The core was assembled by bonding two half-cores with core adhesive. I applied a zircon-based refractory coating to the core surfaces and an alumina-based coating to the mold surfaces to prevent sand burn-on. The coating density was maintained between 1.8 and 2.1 g/ml for brushing and 1.6 to 1.8 g/ml for flow coating. After coating, the mold and cores were dried by ignition to remove any residual moisture.
The pouring parameters are listed in the table below. The pouring temperature was 1640 °C, the pouring time was about 45 seconds, and the mold temperature at pouring was around 20 °C. After solidification and cooling, the castings were knocked out and cleaned. I then sectioned the critical areas of several castings to inspect for internal defects.
| Parameter | Value |
|---|---|
| Pouring temperature (°C) | 1640 |
| Pouring time (s) | 45 |
| Mold initial temperature (°C) | 20 |
The cross-sectional photographs of the key regions showed that the shrinkage cavities and porosity, which were present in the initially designed castings, were no longer visible in the optimized castings. The end flanges, spring seats, and central bridge section all appeared sound. The actual defect locations and sizes closely matched the predictions from the ProCAST simulation. This agreement confirmed the reliability of the numerical simulation approach for steel casting process optimization.
7. Conclusions and Outlook
Through this project, I have demonstrated that numerical simulation is an invaluable tool for optimizing the casting process of a large steel casting component. The main conclusions are summarized as follows:
- The initial design of the cast steel axle housing exhibited significant shrinkage porosity in the end flanges, spring seats, and central section. The ProCAST simulation successfully predicted these defect locations.
- Adding conventional risers at the ends and at the ingate system improved the feeding and reduced the end-flange defects.
- Replacing conventional risers with insulating riser sleeves kept the riser liquid longer and further reduced porosity. The use of insulating sleeves proved to be more effective than ordinary sand risers.
- Changing the central core material from phenolic resin-bonded sand to sodium silicate sand with improved collapsibility and thermal properties significantly reduced the central shrinkage porosity and eliminated the last isolated hot spots.
- Pilot production verified the simulation results: the optimized castings were free of internal shrinkage defects and met the required quality standards.
For future work, I plan to investigate the effects of pouring temperature and pouring time on the filling pattern and on the final porosity using the same simulation methodology. Additionally, I aim to extend the material database in ProCAST by conducting experiments on other steel grades used in my foundry. The combination of casting simulation and physical verification will continue to drive cost reduction and quality improvement in steel casting production.
In the broader context of steel casting, this work shows that a careful balance between riser design, insulating materials, and core sand selection can eliminate most shrinkage-related defects. With the increasing availability of powerful simulation tools, steel casters can now achieve “first-time-right” process development, thereby reducing the time to market and enhancing product reliability. I believe that the methodology presented here can be readily applied to other complex steel castings.
Finally, I would like to express my gratitude to my colleagues who assisted in the molding, pouring, and inspection activities. This project was a team effort, and its success demonstrates the power of combining engineering experience with modern computational tools. The lessons learned have already been incorporated into the foundry’s standard process development procedure for new steel casting products.