Numerical Simulation of Casting Solidification

In my research, I focused on the numerical simulation of the solidification process of sand casting parts. Sand casting parts are widely used in industrial applications because they can be produced with complex geometries and at relatively low cost. However, these castings often suffer from defects such as shrinkage porosity, hot tearing, and distortion. My central aim was to develop a practical finite-element-based simulation workflow for the temperature field of sand casting parts, and then to use the calculated thermal history to predict shrinkage, deformation, and hot tearing. The entire procedure is built on three main pillars: accurate heat transfer modeling, reliable defect criteria, and coupled thermal-stress analysis.

1. Introduction

The numerical simulation of casting solidification has become one of the most important computer-aided engineering tools in modern foundries. For sand casting parts, the solidification stage determines not only the final microstructure but also the soundness of the component. In the past, foundry engineers relied on empirical rules and costly trial-and-error experiments. With the development of computational heat transfer, it is now possible to predict the evolution of temperature in a casting and its mold before production begins. This is particularly valuable for sand casting parts, where the mold is disposable and the thermal behavior of the sand mold strongly controls the cooling rate.

The history of solidification simulation begins in the 1940s, when analog computers were first used to study heat flow in simple castings. In the 1960s, finite difference methods were applied to one-dimensional heat conduction problems. By the 1970s, researchers in several countries began to use finite difference and direct difference methods for two-dimensional and three-dimensional casting solidification. In the 1980s, finite element methods became more popular because they are better suited to complex geometry and to coupling thermal and mechanical analyses. Since then, commercial software packages have made numerical simulation available to foundries producing sand casting parts of all sizes.

The content of solidification simulation can be summarized as follows. First, a mathematical model of heat transfer is established for the casting-mold system. Second, the computational domain is discretized into finite elements or finite differences. Third, the initial conditions, boundary conditions, and material properties are specified. Fourth, the temperature field is solved as a function of time. Finally, defect-prediction criteria are applied to the temperature field to identify regions that may contain shrinkage, hot tearing, or excessive deformation. In my work, I followed this general flow and integrated the entire process into a computer-aided analysis technology for sand casting parts.

Several challenges remain in this field. The thermal properties of molding sand and cast alloys at high temperature are often uncertain. The heat transfer coefficient at the casting-mold interface varies with time because of air-gap formation. Latent heat release during solidification must be treated carefully. Moreover, shrinkage and hot tearing are not purely thermal phenomena; they also depend on stress and strain in the semisolid state. Therefore, I combined temperature field simulation with thermal-stress coupling so that the predictions could be applied to real sand casting parts with greater confidence.

2. Heat Transfer and Numerical Methods

Heat transfer in a casting system occurs by three basic mechanisms: conduction, convection, and radiation. In sand casting parts, conduction is usually the dominant mode because the sand mold is in direct contact with the solidifying metal. Convection occurs in the liquid metal and at the external surfaces of the mold. Radiation becomes important at high temperatures, especially between the exposed surfaces of risers and the environment. In my simulation, I considered conduction and convection explicitly and treated radiation indirectly through effective heat transfer coefficients.

The basic law of heat conduction is Fourier’s law. For a unit area and unit time, the heat flux is proportional to the temperature gradient:

$$
q = -\lambda \frac{\partial T}{\partial n}
$$

Here, \( q \) is the heat flux, \( \lambda \) is the thermal conductivity, \( T \) is temperature, and \( n \) is the direction normal to the isothermal surface. In a three-dimensional Cartesian coordinate system, the transient heat conduction equation with internal heat generation can be written as:

$$
\rho c \frac{\partial T}{\partial t} =
\frac{\partial}{\partial x}\left(\lambda \frac{\partial T}{\partial x}\right) +
\frac{\partial}{\partial y}\left(\lambda \frac{\partial T}{\partial y}\right) +
\frac{\partial}{\partial z}\left(\lambda \frac{\partial T}{\partial z}\right) + \dot{q}_v
$$

In this equation, \( \rho \) is density, \( c \) is specific heat, \( t \) is time, and \( \dot{q}_v \) is the volumetric heat source. During solidification, the latent heat of fusion acts as an internal heat source. If the material properties depend on temperature, the equation becomes nonlinear and must be solved incrementally.

The boundary conditions for the casting-mold system can be classified into three types. The first type specifies the temperature on the boundary:

$$
T = T_w \quad \text{on } \Gamma_1
$$

The second type specifies the heat flux on the boundary:

$$
-\lambda \frac{\partial T}{\partial n} = q_w \quad \text{on } \Gamma_2
$$

The third type specifies convective heat exchange with the environment:

$$
-\lambda \frac{\partial T}{\partial n} = h\left(T – T_\infty\right) \quad \text{on } \Gamma_3
$$

Here, \( h \) is the heat transfer coefficient, and \( T_\infty \) is the ambient temperature. In foundry simulation of sand casting parts, the third boundary condition is the most common because the sand mold surface exchanges heat with the surrounding air. At the casting-mold interface, an effective heat transfer coefficient is often used to account for the air gap that forms during solidification.

I used the finite element method to solve the heat conduction equation. The finite element method is especially suitable for sand casting parts because it allows unstructured meshes and can accurately represent complex curved surfaces. In the finite element formulation, the temperature within each element is interpolated from nodal values using shape functions:

$$
T^e(x,y,z,t) = \sum_{i=1}^{n} N_i(x,y,z) T_i(t)
$$

After applying the Galerkin weighted-residual method, the spatial discretization leads to a system of first-order ordinary differential equations:

$$
\mathbf{C} \dot{\mathbf{T}} + \mathbf{K} \mathbf{T} = \mathbf{F}
$$

Here, \( \mathbf{C} \) is the heat capacity matrix, \( \mathbf{K} \) is the heat conduction matrix, \( \mathbf{F} \) is the thermal load vector, \( \mathbf{T} \) is the nodal temperature vector, and \( \dot{\mathbf{T}} \) is the time derivative of the nodal temperature vector. For a steady-state problem, the term involving \( \dot{\mathbf{T}} \) vanishes and the equation reduces to:

$$
\mathbf{K} \mathbf{T} = \mathbf{F}
$$

I compared the finite element method with finite difference methods during my research. Finite difference methods are easy to derive and computationally efficient, but they usually require regular grids. Finite element methods require more data preparation but provide better geometric flexibility and are more convenient for thermal-stress coupling. Since sand casting parts often have ribs, bosses, and varying wall thickness, the finite element method is the more practical choice.

Numerical Method Advantages Disadvantages
Finite Difference Method Simple formulation; low computational cost Regular mesh required; difficult for complex geometry
Finite Element Method Handles complex geometry; good for stress coupling More preprocessing and larger memory requirement
Boundary Element Method Only boundary discretization required Difficult for variable properties and nonlinear problems

3. Temperature Field Simulation During Solidification

The temperature field simulation is the foundation for all subsequent defect analysis. In sand casting parts, the cooling process begins with pouring of liquid metal at a high temperature and ends when the casting reaches the surrounding ambient temperature. During this process, the metal releases superheat, latent heat, and sensible heat. The sand mold absorbs heat continuously, and its surface temperature rises rapidly after pouring. To obtain accurate temperatures, I had to specify variable material properties, proper boundary conditions, and a reliable latent heat treatment.

3.1 Variable Thermophysical Properties

The thermal conductivity, specific heat, and density of both the metal and the sand mold change with temperature. In my finite element model, I treated these properties as functions of temperature. The thermal properties of sand are especially important because the mold occupies a large volume in the computational domain. The sand mold has a relatively low thermal conductivity, which creates steep temperature gradients in the metal and influences the solidification morphology of sand casting parts.

For the cast steel used in the plate hammer example, I adopted the following typical properties. The values are not universal; they must be calibrated for each alloy and molding material.

Property Symbol Typical Value Unit
Thermal conductivity \( \lambda \) 25–45 W/(m·K)
Specific heat \( c \) 460–750 J/(kg·K)
Density \( \rho \) 7400–7800 kg/m³
Latent heat \( L \) 250–270 kJ/kg
Liquidus temperature \( T_l \) 1490–1520 °C
Solidus temperature \( T_s \) 1420–1450 °C

In the finite element calculation, I used the temperature computed at the previous time step to evaluate the properties for the next time step. This method is simple and stable when the time step is sufficiently small. For many sand casting parts, the temperature changes rapidly immediately after pouring and more slowly during the late stages of solidification, so an adaptive time step is recommended.

3.2 Boundary Conditions and Heat Transfer Coefficients

The boundary conditions for sand casting parts include the casting-sand mold interface, the sand mold-air interface, the riser top, and sometimes the core surfaces. The heat transfer coefficient at the casting-mold interface is particularly difficult to determine. In the early stage of solidification, the metal is in close contact with the sand mold. As the casting cools and solidifies, contraction creates an air gap, which reduces the heat flow. I used effective heat transfer coefficients that change with time to represent this behavior.

Boundary Interface Heat Transfer Coefficient \( h \) (W/m²K)
Steel casting–sand mold with close contact 400–800
Steel casting–sand mold after air-gap formation 100–250
Sand mold–air natural convection 5–20
Riser top with insulating cover 2–10

I used a convective boundary condition at the external surfaces of the sand mold. The heat flux from the mold surface to the air is given by Newton’s law of cooling:

$$
q = h\left(T_{surface} – T_{air}\right)
$$

In many foundry simulations, radiation is also important. The radiation heat flux is proportional to the fourth power of the absolute temperature:

$$
q_r = \varepsilon \sigma \left(T^4 – T_{air}^4\right)
$$

where \( \varepsilon \) is the surface emissivity and \( \sigma \) is the Stefan-Boltzmann constant. Because this equation is nonlinear, I linearized the radiation term in some cases and combined it with the convection coefficient into an effective heat transfer coefficient.

3.3 Latent Heat Treatment

Latent heat is the energy released when liquid metal transforms into solid. This release significantly slows the cooling rate and must be included in the simulation of sand casting parts. In my finite element model, I compared three common methods: the temperature recovery method, the equivalent specific heat method, and the enthalpy method.

The temperature recovery method assumes that the latent heat released during a small time step is used to raise the temperature of the same control volume. The temperature increment can be written as:

$$
\Delta T = \frac{L}{c_p}
$$

This method is suitable for pure metals and eutectic alloys that solidify at a constant temperature. However, it is less convenient for alloys with a wide freezing range, so I did not use it as the primary method.

The equivalent specific heat method distributes the latent heat uniformly over the solidification interval:

$$
c_{eq} = c_p + \frac{L}{T_l – T_s}
$$

Here, \( T_l \) is the liquidus temperature and \( T_s \) is the solidus temperature. This method is easy to implement, but it requires care when the solidification interval is very narrow.

The enthalpy method is the most general approach. The enthalpy \( H \) is defined as the integral of heat capacity plus the latent heat contribution:

$$
H(T) = \int_{T_0}^{T} c_p(T’) \, dT’ + L\left(1 – f_s(T)\right)
$$

where \( f_s \) is the solid fraction. Substituting this into the heat conduction equation gives:

$$
\frac{\partial H}{\partial t} = \nabla \cdot \left(\lambda \nabla T\right)
$$

I used the enthalpy method in my simulation of the plate hammer casting. To verify the latent heat treatment, I constructed a simple two-dimensional model and compared the temperature curve with and without latent heat. The result clearly showed a temperature plateau, or even a slight temperature recovery, when the enthalpy method was used. This plateau is physically expected during the solidification of sand casting parts and confirms that the latent heat treatment is correct.

4. Shrinkage and Porosity Prediction

Shrinkage defects are among the most common internal defects in sand casting parts. As liquid metal solidifies, it undergoes liquid contraction and solidification contraction. If the contraction is not compensated by liquid feeding from the riser, a void will form in the last solidifying region. Large and concentrated voids are called shrinkage cavities, while small and dispersed voids are called shrinkage porosity.

The formation of shrinkage defects is controlled by several factors. First, the pressure in the liquid metal decreases as it flows through the narrow channels between dendrites. Second, the gas solubility in the liquid decreases with temperature, so dissolved gases may form bubbles. Third, the solid network restricts feeding in the late stages of solidification. In my work, I did not solve the full liquid flow equations. Instead, I used thermal criteria based on the temperature field, which are simpler and have been successfully applied to many sand casting parts.

4.1 Isothermal Curve Method

The isothermal curve method is based on the principle of directional solidification. If the solidification front remains open to the riser until the end of solidification, no shrinkage will occur. If the isotherms form closed loops, then the enclosed liquid is isolated from the riser and shrinkage will form inside that region. In my simulation, I plotted the isotherms on the center plane of the casting. A closed isotherm of the solidus temperature indicates an isolated liquid pool, which is a strong indicator of shrinkage.

This method is intuitive and easy to implement. However, it only gives a qualitative prediction. It does not account for the feeding distance or the pressure drop in the mushy zone. Therefore, I also used the temperature gradient method.

4.2 Temperature Gradient Method

The temperature gradient method assumes that a positive temperature gradient toward the riser is necessary for feeding. When the temperature gradient is small, the feeding channel is narrow and may close prematurely. The local temperature gradient at the end of solidification is defined as:

$$
G = \left. \frac{\partial T}{\partial n} \right|_{T = T_s}
$$

If this gradient is smaller than a critical value \( G_{crit} \), the region is predicted to have shrinkage porosity. In my analysis of sand casting parts, I calculated the temperature gradient for each element when the element reached the solidus temperature. Then I compared the gradient with the calibrated critical value for the alloy.

The critical temperature gradient depends on the casting shape, size, and alloy composition. Therefore, it must be calibrated for each foundry. The table below summarizes common shrinkage prediction criteria that I reviewed during my research.

Criterion Mathematical Parameter Advantage Limitation
Isothermal Curve Method Closed isotherms Simple; visual Qualitative only
Temperature Gradient Method \( G = \partial T/\partial n \) Practical; easy to compute Critical value depends on geometry
Niyama Criterion \( N = G/\sqrt{\dot{T}} \) More geometry-independent Needs local cooling rate
Solid Fraction Gradient Method \( G_{fs} = \partial f_s/\partial n \) Applicable to eutectic alloys Needs solid fraction model
Pressure Gradient Method \( \nabla P \) Directly considers feeding Requires flow simulation

In addition to the temperature gradient, I also used the Niyama criterion for steel castings. The Niyama criterion is defined as:

$$
N = \frac{G}{\sqrt{\dot{T}}}
$$

where \( \dot{T} \) is the cooling rate at the end of solidification. A low Niyama value indicates that microporosity is likely. I found this criterion to be useful for sand casting parts made of steel because it reduces the influence of casting geometry compared with the simple temperature gradient.

5. Thermal-Stress Coupling and Hot Tearing Prediction

Hot tearing is a serious defect that occurs during the final stage of solidification. In sand casting parts, hot tearing usually appears in hot spots where the contraction of the solidifying shell is constrained by the sand mold or by other parts of the casting. The phenomenon involves both heat transfer and mechanical behavior. Therefore, I coupled the temperature field with a finite element stress analysis to predict deformation and hot tearing.

The thermal strain caused by a temperature change is:

$$
\boldsymbol{\varepsilon}_t = \alpha \left(T – T_{ref}\right) \mathbf{I}
$$

where \( \alpha \) is the coefficient of thermal expansion, \( T_{ref} \) is the reference temperature, and \( \mathbf{I} \) is the identity tensor. In an incremental finite element formulation, the total strain increment is the sum of the elastic strain increment, plastic strain increment, and thermal strain increment:

$$
d\boldsymbol{\varepsilon} = d\boldsymbol{\varepsilon}_e + d\boldsymbol{\varepsilon}_p + d\boldsymbol{\varepsilon}_t
$$

At each time step, I read the temperature change from the solidification simulation and applied it as a thermal load to the mechanical model. The stress increment was then calculated using the elastic-plastic constitutive relation:

$$
d\boldsymbol{\sigma} = \mathbf{D}_{ep}\left(d\boldsymbol{\varepsilon} – d\boldsymbol{\varepsilon}_t\right)
$$

Here, \( \mathbf{D}_{ep} \) is the elastic-plastic material matrix. The yield condition depends on temperature and accumulated plastic strain:

$$
F = \bar{\sigma} – \sigma_y\left(\bar{\varepsilon}_p, T\right) = 0
$$

where \( \bar{\sigma} \) is the equivalent stress and \( \sigma_y \) is the yield strength of the material at the current temperature. In the semisolid range, the yield strength is very small, so even a small amount of constraint can lead to high local strain.

For hot tearing prediction, I evaluated the tensile stress and the accumulated plastic strain in the last solidifying region. A simple criterion is that hot tearing occurs when the local equivalent plastic strain exceeds a critical value in the coherent temperature range. Another criterion uses the maximum principal stress:

$$
\tau_{max} = \frac{\sigma_1 – \sigma_3}{2} < \tau_{crit}
$$

where \( \sigma_1 \) and \( \sigma_3 \) are the maximum and minimum principal stresses. If the maximum shear stress remains below the critical shear stress, no hot tear is expected.

Model Description Typical Application
Thermal Elastic Model Linear elastic stress from thermal strain Quick estimation of residual stress
Thermal Elastoplastic Model Elastic-plastic strain with temperature-dependent yield Predicting permanent deformation and hot tearing
Thermal Elasto-Viscoplastic Model Includes creep and strain-rate effects High-temperature sand casting parts with complex stress states

During the solidification of sand casting parts, the sand mold is not a rigid body. It can deform and provide some yielding. To include this effect, I used contact elements at the casting-mold interface. This allowed the mold to resist the contraction of the casting without creating unrealistically high stress. The contact condition was updated at each time step based on the current gap between the casting and the mold.

6. Application to Plate Hammer Production

To validate my numerical simulation technology, I applied it to a typical sand casting part: a plate hammer used in crushing equipment. The plate hammer has a relatively thick central section and two thinner ends, which makes it susceptible to shrinkage at the center. My analysis followed a complete workflow: three-dimensional geometry modeling, mesh generation, temperature field simulation, shrinkage prediction, thermal stress analysis, and process optimization.

I first created a three-dimensional solid model of the plate hammer and the surrounding sand mold. The mold was represented as a separate solid body with its own thermal properties. I then generated a finite element mesh with smaller elements in the casting and larger elements in the sand mold. The initial temperature of the liquid steel was set to the pouring temperature, and the initial temperature of the sand mold was set to room temperature. The top riser was included in the model to feed the casting during solidification.

The latent heat was treated with the enthalpy method. I used the temperature-dependent material properties for cast steel and sand. The heat transfer coefficient at the casting-mold interface was specified according to the time-dependent values discussed earlier. After solving the transient heat conduction equation, I imported the temperature history into the stress solver.

6.1 Temperature Field Results

The simulated temperature field showed that the plate hammer cooled first at the thinner ends and last at the thick central section. The riser remained hot longer than the casting, which is desirable for feeding. However, when I examined the isothermal curves at the end of solidification, I observed that the liquid region in the center of the plate hammer had become isolated from the riser. This closed isotherm indicated that the feeding channel was cut off before solidification was complete. Therefore, I predicted that a large shrinkage cavity would form in the center of the casting.

I also calculated the temperature gradient in the central region at the solidus temperature. The minimum temperature gradient was lower than the critical value for cast steel. This result was consistent with the isothermal curve method and confirmed that shrinkage was likely. The predicted shrinkage shape was almost continuous through the center of the plate hammer, which would make the casting unacceptable for service.

6.2 Deformation and Hot Tearing Results

After the thermal stress analysis, I examined the displacement and stress distribution in the plate hammer. The overall deformation was relatively small and remained within the dimensional tolerance of the component. There was no obvious stress concentration in the casting, and the maximum shear stress was below the critical value for hot tearing. Therefore, I concluded that the main defect risk was shrinkage, not hot tearing or permanent deformation.

These results demonstrate the importance of applying multiple defect criteria to sand casting parts. A purely thermal analysis can predict shrinkage, while a coupled thermal-stress analysis is necessary to evaluate hot tearing and deformation.

6.3 Process Optimization

Based on the simulation results, I evaluated three process modifications for the plate hammer. The first modification was to enlarge the riser. A larger riser would provide more liquid metal and would stay open longer, improving the feeding of the center section. The second modification was to add a chill near the center or at the thick section. A chill would increase the local cooling rate and change the solidification sequence. The third modification was to use an insulating riser sleeve, which would reduce the cooling rate in the riser and allow it to feed the casting for a longer time.

Process Modification Mechanism Expected Effect
Enlarge riser More liquid metal supply Reduces shrinkage cavity size
Add chill Faster local cooling Changes solidification sequence
Insulating riser sleeve Slower riser cooling Improves feeding efficiency

All three modifications were effective in reducing the tendency for shrinkage. After considering production cost and practical convenience, I selected the enlarged riser as the final solution. The modified process was implemented in the foundry, and the resulting plate hammer was sound. The actual production result matched my simulation prediction, confirming that the numerical simulation technology is reliable for optimizing sand casting parts.

7. Conclusions

In this research, I established a finite-element-based computer-aided engineering technology for the temperature field simulation and defect prediction of sand casting parts. The main conclusions of my work are summarized below.

First, the finite element method is well suited for simulating the solidification of sand casting parts with complex shapes. It allows for accurate representation of curved boundaries and provides a natural platform for coupling thermal and mechanical analyses. The transient heat conduction equation, combined with the enthalpy method for latent heat, gives a reliable temperature field.

Second, the prediction of shrinkage defects can be performed using thermal criteria such as the isothermal curve method and the temperature gradient method. These methods are computationally efficient and provide useful guidance for foundry engineers. For the plate hammer, the simulation predicted a central shrinkage cavity, which was confirmed by the original production result.

Third, thermal-stress coupling is necessary for predicting hot tearing and distortion. By applying the temperature history as a thermal load, I was able to calculate stress, strain, and displacement in the casting. The plate hammer showed no significant hot tearing risk, and the deformation was within acceptable limits.

Fourth, the process optimization study showed that enlarging the riser is the most effective and economical way to eliminate shrinkage in the plate hammer. The final production trial produced sand casting parts without shrinkage, hot tearing, or deformation defects. The agreement between simulation and production confirms that this technology can be used to guide real foundry processes.

In the future, I plan to extend this work to microstructure prediction and to the automatic optimization of gating and risering systems. The current approach provides a solid foundation for the digital design of high-quality sand casting parts, and further development will make the simulation process faster, more accurate, and more accessible to foundry engineers.

Scroll to Top