Introduction
The lost foam casting (LFC) process, also known as evaporative pattern casting, is recognized as a near-net-shape manufacturing technology with significant advantages in reducing production steps, lowering costs, and achieving high dimensional accuracy for complex components. This process has been widely adopted in the automotive, aerospace, and general machinery industries. However, the industrial application of LFC for iron castings still faces challenges related to process design and defect prediction. Traditional trial-and-error methods are time-consuming and costly, and analytical techniques for gating and riser design remain underdeveloped. Numerical simulation has therefore become an indispensable tool for understanding the mould filling, solidification, and defect formation mechanisms in lost foam casting. In this work, a comprehensive simulation and process design methodology for iron castings produced by lost foam casting is developed, employing the commercial finite element software ProCAST integrated with JMatPro for material property calculations. The study addresses the critical issue that the filling and solidification stages are usually treated separately in conventional simulations, which leads to inaccurate shrinkage porosity predictions. A unified modelling approach is presented here, enabling the filling and solidification of iron castings in a single computational model. Furthermore, a quantitative design method for gating/riser systems is proposed, based on thermal modulus analysis, section ratio design, and the theory of balanced solidification. Finally, the preliminary application of the CAFE method for simulating grain nucleation and growth in grey iron lost foam castings is discussed. The simulation results are validated by experimental trials, demonstrating the feasibility of the proposed simulation framework for industrial application.
Numerical Simulation Fundamentals of Lost Foam Casting
Mathematical Models for Mould Filling
The filling process in lost foam casting is characterized by the complex interaction between the molten metal, the decomposing foam pattern, and the permeable sand mould. The fluid flow is governed by the conservation of mass, momentum, and energy. For an incompressible Newtonian fluid, the continuity equation is expressed as:
$$ \frac{\partial u}{\partial x} + \frac{\partial v}{\partial y} + \frac{\partial w}{\partial z} = 0 $$ (1)
where \(u\), \(v\), and \(w\) are the velocity components in the \(x\), \(y\), and \(z\) directions, respectively. The momentum equations, i.e., the Navier–Stokes equations, are written as:
$$ \frac{\partial u}{\partial t} + u\frac{\partial u}{\partial x} + v\frac{\partial u}{\partial y} + w\frac{\partial u}{\partial z} = -\frac{1}{\rho}\frac{\partial P}{\partial x} + \nu \nabla^2 u + F_x $$ (2)
$$ \frac{\partial v}{\partial t} + u\frac{\partial v}{\partial x} + v\frac{\partial v}{\partial y} + w\frac{\partial v}{\partial z} = -\frac{1}{\rho}\frac{\partial P}{\partial y} + \nu \nabla^2 v + F_y $$ (3)
$$ \frac{\partial w}{\partial t} + u\frac{\partial w}{\partial x} + v\frac{\partial w}{\partial y} + w\frac{\partial w}{\partial z} = -\frac{1}{\rho}\frac{\partial P}{\partial z} + \nu \nabla^2 w + F_z $$ (4)
where \(\rho\) is the density, \(P\) the pressure, \(\nu\) the kinematic viscosity, and \(F_x, F_y, F_z\) are body force components. In the lost foam process, the thermal interaction between the molten metal and the foam pattern is accounted for by an energy equation with a source term related to the degradation of the foam:
$$ \rho C_p \frac{\partial T}{\partial t} + \rho C_p \left( u\frac{\partial T}{\partial x} + v\frac{\partial T}{\partial y} + w\frac{\partial T}{\partial z} \right) = \nabla \cdot (k \nabla T) + \rho L \frac{\partial f_s}{\partial t} $$ (5)
where \(C_p\) is the specific heat, \(k\) the thermal conductivity, \(L\) the latent heat of solidification, and \(f_s\) the solid fraction. The free surface boundary condition at the metal–foam interface is treated using a heat balance that includes the latent heat of foam vaporisation.
The gas pressure in the gap between the molten metal and the decomposing foam is critical. The gas pressure at the \(i+1\) time step can be derived from the mass conservation of the pyrolysis gases:
$$ P^{i+1} = P_0 + \frac{\Delta t V_P (T_m – T_i) \alpha_P}{L_P T_i S} – \frac{\Delta t F K (P_i – P_0)}{x_c T_i S} – \frac{S \Delta t (P_i \delta_i – P_0 \delta_0)}{T_i S} $$ (6)
where \(P_0\) is the atmospheric pressure, \(\Delta t\) the time step, \(V_P\) the volume of decomposed foam, \(L_P\) the latent heat of foam degradation, \(S\) the effective area, \(F\) the perimeter, \(K\) the coating permeability, \(x_c\) the coating thickness, \(\alpha_P\) the heat transfer coefficient, and \(\delta\) is the gap size. The gap size at the next time step is expressed as:
$$ \delta_{i+1} = \delta_i + \left( u – \frac{\alpha_P (T_m – T_i)}{\rho_P L_P} \right) \Delta t $$ (7)
where \(u\) is the filling velocity and \(\rho_P\) the foam density.
Solidification and Heat Transfer Models
During solidification, the heat conduction in the mould and casting is described by the transient heat conduction equation:
$$ \rho C_p \frac{\partial T}{\partial t} = \frac{\partial}{\partial x}\left(k\frac{\partial T}{\partial x}\right) + \frac{\partial}{\partial y}\left(k\frac{\partial T}{\partial y}\right) + \frac{\partial}{\partial z}\left(k\frac{\partial T}{\partial z}\right) + Q $$ (8)
where \(Q\) is the internal heat generation rate due to latent heat release. The latent heat is handled by the enthalpy method:
$$ H(T) = \int_{T_0}^{T} C_p dT + L (1 – f_s) $$ (9)
Thus the heat conduction equation becomes:
$$ \rho \frac{\partial H}{\partial t} = \nabla \cdot (k \nabla T) $$ (10)
The solid fraction as a function of temperature is obtained from the Scheil equation:
$$ f_s = 1 – \left( \frac{T – T_L}{T_m – T_L} \right)^{1/(k_p – 1)} $$ (11)
where \(T_L\) is the liquidus temperature, \(T_m\) the melting point of pure metal, and \(k_p\) the equilibrium partition coefficient.
Shrinkage Porosity Prediction
The Niyama criterion and its generalisations are commonly used to predict microporosity. The criterion can be expressed as:
$$ G / \sqrt{R} < C_{crit} $$ (12)
where \(G\) is the temperature gradient, \(R\) the cooling rate, and \(C_{crit}\) a material-dependent constant. In addition, a multi-hot-spot and isolated-domain method is implemented in this work. The total shrinkage volume in an isolated hot spot is calculated as:
$$ \Delta V = \sum_{i=1}^{n} \left( \beta f_{si} + \alpha_l (1 – f_{si}) \right) V_i $$ (13)
where \(\beta\) is the solidification shrinkage coefficient, \(\alpha_l\) the liquid thermal expansion coefficient, \(f_{si}\) the solid fraction increment in element \(i\), and \(V_i\) the element volume.
Simulation Procedure and Material Property Determination
General Pre-processing in ProCAST
The simulation is performed with ProCAST (ESI Group) integrated in the Visual Environment platform. The main work flow is illustrated by the following steps:
- Solid geometry creation in CAD and import as Parasolid format.
- Mesh generation using Visual-Mesh with tetrahedral elements.
- PreCAST setup: material assignment, boundary and initial conditions, interfacial heat transfer coefficients, and process parameters.
- Running the solver with the “Lost Foam” option, which automatically sets parameters for foam degradation and gas effects.
- Post-processing in Visual-Viewer to extract temperature, velocity, solid fraction, and porosity fields.
For lost foam simulations, the volume fractions are defined as shown in Table 1, where the casting region is initially empty (Empty=YES), the foam is modelled as a solid material (Empty=NO), and the sand mould is also solid (Empty=NO).
| Region | Type | Material | Empty | Initial temperature (°C) |
|---|---|---|---|---|
| Casting | Casting | HT150 / QT600-3 | YES | 1450–1480 |
| Foam pattern | Foam | EPS | NO | 25 |
| Sand mould | Sand | Silica sand | NO | 25 |
The interface between the casting and the foam is set to “EQUIV”, while the casting–sand interface is “COINC” with a heat transfer coefficient of \(150~\mathrm{W/(m^2\cdot K)}\). The vacuum pressure is set to \(0.05~\mathrm{MPa}\) on the outer boundary of the sand mould.
Determination of Thermophysical Properties for Cast Iron
Accurate thermophysical data are essential for reliable simulation. In this work, the composition of the iron is obtained from ladle analysis, and then the solidification path and physical properties are calculated using JMatPro software. The composition of HT250 is given in Table 2.
| Fe | Mn | Si | C | P | S |
|---|---|---|---|---|---|
| 93.936 | 0.81 | 2.0 | 3.24 | 0.014 | 0.009 |
JMatPro predicts the phase evolution during solidification. The solidification range of HT250 is from \(1152\,^\circ\mathrm{C}\) to \(1218\,^\circ\mathrm{C}\). At the end of solidification, the mass fractions of austenite, graphite, and MnS are \(98.33\%\), \(1.65\%\), and \(0.02\%\), respectively. The density variation with temperature is obtained as shown in Figure 1 (conceptually). The overall density first increases, then decreases, and later increases again after complete solidification. The two density decreases are attributed to graphite precipitation. These data are directly imported into ProCAST to replace the built-in alloy database, enabling the simultaneous simulation of filling and solidification for iron castings.
A critical issue is that, for iron castings, the density curve below the solidus temperature must be modified to account for graphite expansion. This is implemented by setting the density values from JMatPro as a tabulated function of temperature. The resulting material file also includes liquidus and solidus temperatures, latent heat, thermal conductivity, and viscosity.
Validation of the Unified Filling and Solidification Simulation for Iron LFC
Mould Filling Simulation and Experimental Comparison
To validate the filling model, a plate-shaped casting with a thickness of 10 mm and made of HT150 was simulated. The geometry is shown in Figure 2 (conceptual), including the casting, foam, and sand mould. The simulation conditions are summarised in Table 3.
| Parameter | Value |
|---|---|
| Number of elements | 120,000 |
| Material | HT150 |
| Heat transfer coefficient (casting/mould) | 150 W/(m²·K) |
| Vacuum pressure | 15 mbar |
| Initial temperature | 1450 °C |
| Solver option | LOST FOAM |
The simulated filling sequence is shown in Figure 3 (conceptual). The entire filling time was calculated as \(3.5~\mathrm{s}\), while the experimental measured filling time was \(3.38~\mathrm{s}\). The filling pattern exhibited a relatively smooth flow front with a slight turbulence at the junction between the downsprue and the plate. The simulation also reproduced the formation of a vortex near the far end of the plate, which was observed in the experimental video. The air gap evolution is shown in Figure 4 (conceptual): the gap initially grows, reaches a maximum, and then decreases. The simulated gap volume curve agrees well with the experimental measurements reported in the literature. These results confirm that the present model can accurately capture the characteristic phenomena of lost foam casting, including the progressive foam degradation and the transient gas gap.
Solidification Simulation of a Ductile Iron Fork-shaped Casting
Since ductile iron exhibits significant graphitic expansion during eutectic solidification, it provides a rigorous test for the density-modification method. A fork-shaped casting made of QT600-3 was selected. The casting dimensions are \(375~\mathrm{mm}\times 142~\mathrm{mm}\times 552~\mathrm{mm}\). Two gating systems were evaluated: a stepped (bottom) gating system and a top gating system. The simulation conditions are listed in Table 4.
| Parameter | Value |
|---|---|
| Number of elements | 1,470,000 |
| Material | QT600-3 |
| Heat transfer coefficient | 150 W/(m²·K) |
| Vacuum pressure | 0.05 MPa |
| Initial temperature | 1480 °C |
| Solver option | LOST FOAM |
The solidification process was simulated with the filling and solidification stages coupled in one model. Figure 5 (conceptual) shows the temperature distribution at different solid fractions. For the stepped gating system, the melt level in the riser dropped rapidly during the first 3% of solidification, indicating liquid feeding. From 3% to 10% solidification, the melt level remained almost constant, suggesting that the solidification shrinkage was balanced by the expansion due to graphite nucleation. After about 40% solidification, the melt level in the sprue rose slightly, confirming that the internal graphitic expansion pushed still-liquid metal back into the feeding system, but the rise was limited, implying that the expansion pressure was effectively used for self-feeding. The porosity prediction shows no internal shrinkage defects, which was confirmed by sectioning the trial casting.
For the top gating system, the temperature distribution was asymmetric: the top remained hotter than the bottom, leading to a last-solidifying region near the top. The melt level in the sprue descended during the first 9.6% of solidification, then rose after about 20% solidification due to graphitic expansion. However, the final porosity prediction indicated shrinkage defects near the top region, and the trial casting exhibited corresponding defects in the same location. The comparison demonstrates that the unified simulation approach, with the modified density curve, can accurately predict both the occurrence and location of shrinkage porosity in iron LFC castings.
Quantitative Gating/Riser System Design for Grey Iron Castings
Thermal Modulus Analysis
A bearing pedestal made of HT150 was selected for process design. The casting has a main wall thickness of 25–50 mm and is mass-produced. The goal is to design a gating system that also serves as an effective riser, i.e., a “gating-riser” system. The first step is to determine the thermal modulus distribution of the casting. The Chvorinov modulus is calculated by extracting solidification parameters from ProCAST:
$$ M = \frac{V}{A} \approx \frac{ \left( \frac{k_{mold,in} \rho_{mold,in} c_{p,mold,in}}{\pi} \right)^{1/2} (T_{al,sol} – T_{mold,in}) }{ \rho_{al,sol} \Delta H_{al} } \sqrt{t_{sol}} $$ (14)
where \(V\) and \(A\) are the volume and heat-transferring surface area of the local region, \(T_{al,sol}\) is the solidus temperature of the alloy, \(T_{mold,in}\) the initial mould temperature, \(\rho_{al,sol}\) the alloy density at the solidus, \(\Delta H_{al}\) the enthalpy change from initial temperature to solidus, \(k_{mold,in}\), \(\rho_{mold,in}\), and \(c_{p,mold,in}\) are the mould thermal conductivity, density, and specific heat at the initial temperature, and \(t_{sol}\) is the local solidification time. By solving this parameter field, the high-modulus regions are identified. For the bearing pedestal, the maximum modulus is about \(2.0~\mathrm{cm}\) in the central ring area. The volume of the region with a modulus greater than or equal to \(2.0~\mathrm{cm}\) is \(1368~\mathrm{cm^3}\), corresponding to a mass of \(9.3~\mathrm{kg}\).
Design of the Gating-Riser Dimensions
Based on the modulus results, the riser neck modulus is determined using the balanced solidification method. The key equations are:
– The modulus of the hot spot region: \(M_c = 2.0~\mathrm{cm}\)
– The mass-to-modulus ratio (shape factor): \(Q = G / M_c^{3} = 9.3 / 2.0^{3} = 1.1625~\mathrm{kg/cm^3}\)
– The contraction time fraction: \(P = 0.5 \exp(0.011 Q) = 0.36\)
– The contraction modulus coefficient: \(f_2 = \sqrt{P} = 0.6\)
– The flow resistance factor: \(f_p = 0.5\)
– The neck length factor: \(f_4 = 0.8\)
– The riser neck modulus: \(M_N = f_p f_2 f_4 M_c = 0.5 \times 0.6 \times 0.8 \times 2.0 = 0.48~\mathrm{cm}\)
– When one riser feeds multiple castings, a correction factor \(f^* = 1.2\) is applied. The riser body modulus is then:
$$ M_R = f_1 f_2 f_3 M_c $$ (15)
with \(f_1 = 1.1\), \(f_2 = 0.6\), \(f_3 = 1.1\), giving:
$$ M_R = 1.1 \times 0.6 \times 1.1 \times 2.0 = 1.452~\mathrm{cm} $$
The gating system is designed as a rectangular runner with an aspect ratio of 2.5, leading to a runner cross-section of \(2~\mathrm{cm} \times 5~\mathrm{cm}\), which also serves as the riser neck.
For the pouring time, the empirical formula for grey iron castings under 450 kg is:
$$ t = S \sqrt{G_L} $$ (16)
where \(G_L\) is the total mass of metal in the mould (124 kg) and \(S\) is a coefficient depending on wall thickness (taken as 2.2). Thus:
$$ t = 2.2 \sqrt{124} \approx 24.5~\mathrm{s} $$
The cross-sectional area of the ingates is calculated using the orifice flow equation:
$$ A_{in} = \frac{G_L}{\rho_L \mu t \sqrt{2 g h_p}} $$ (17)
where \(\rho_L = 0.31~\mathrm{kg/cm^2}\) (density factor), \(\mu = 0.6\), \(g = 981~\mathrm{cm/s^2}\), and the static head height \(h_p\) is obtained from the four-element system as:
$$ h_p = \frac{k_2^2}{1 + k_1^2 + k_2^2} H_p $$ (18)
With the section ratio \(A_{direct}:A_{runner}:A_{ingate} = 1:1.1:1.1\), the coefficients are \(k_1 \approx 1.0\) and \(k_2 \approx 1.1\). The average static head height \(H_p = 27~\mathrm{cm}\), giving \(h_p \approx 10~\mathrm{cm}\). Thus:
$$ A_{in} = \frac{124}{0.31 \times 0.6 \times 24.5 \times \sqrt{2 \times 981 \times 10}} \approx 21~\mathrm{cm^2} $$
Since the gating-riser system is used, this area is divided between two ingates, each with a cross-section of \(10~\mathrm{cm^2}\). The final design uses a runner with dimensions \(5 \times 4 \times 5~\mathrm{cm^3}\), and a downsprue diameter of about \(5~\mathrm{cm}\).
Simulation Validation of the Designed Process
The designed gating-riser system was modelled in ProCAST, as shown in Figure 6 (conceptual). The filling simulation reveals that the entire filling process takes about 20 s. At the initial stage (0.64 s), there is a risk of reverse flow due to insufficient static pressure; however, once the sprue is filled, the flow becomes stable. The filling rate curve, derived from the filling fraction data and analysed in MATLAB, shows a “slow–fast–slow” pattern, with inflection points at \(t=2.23~\mathrm{s}\) (when the metal enters the casting) and \(t=8.45~\mathrm{s}\) (when the metal reaches the top of the casting). These inflection points correspond to abrupt changes in the cross-sectional area of the flow path. Therefore, the pouring operation should be controlled with an initially slow pour, followed by a fast pour, and then a slow pour at the end to avoid overflowing.
The solidification results are shown in Figure 7 (conceptual). The temperature distribution indicates that the last solidifying regions are the gating-riser areas. The melt level in the pouring cup descends continuously during solidification from 3% to 31%, demonstrating effective liquid feeding. After 31%, the melt level remains essentially constant, indicating that graphitic expansion compensates for solidification shrinkage. The porosity field shows a small shrinkage region appearing at 57% solidification, which disappears by 68% solidification due to graphitic self-feeding. This confirms that the gating-riser system successfully prevents macro-shrinkage defects. Trial production with the designed parameters yielded high-quality castings without visible porosity, satisfying the mechanical requirements for the bearing pedestal.
Microstructural Simulation of Grey Iron LFC Using the CAFE Model
Theoretical Framework
The mechanical properties of cast components are largely controlled by the grain structure formed during solidification. To predict grain nucleation and growth, the CAFE (Cellular Automaton Finite Element) method is employed. Nucleation is described by the continuous nucleation model proposed by Rappaz, using a Gaussian distribution of nucleation sites:
$$ \frac{dn}{d(\Delta T)} = \frac{n_{max}}{\sqrt{2\pi} \Delta T_\sigma} \exp\left[ -\frac{1}{2} \left( \frac{\Delta T – \Delta T_{max}}{\Delta T_\sigma} \right)^2 \right] $$ (19)
where \(n_{max}\) is the maximum nucleation density, \(\Delta T_{max}\) the mean undercooling, and \(\Delta T_\sigma\) the standard deviation. The grain growth velocity is calculated from the KGT model, which relates the undercooling \(\Delta T\) to the growth velocity \(v\):
$$ v = \alpha \Delta T^2 + \beta \Delta T^3 $$ (20)
The coefficients \(\alpha\) and \(\beta\) depend on the alloy composition. For HT250, the values are \(\alpha = 3.82458 \times 10^{-8}~\mathrm{m/(s\cdot K^2)}\) and \(\beta = 4.13428 \times 10^{-9}~\mathrm{m/(s\cdot K^3)}\). For a typical aluminium alloy, \(\alpha = 4.94606 \times 10^{-7}~\mathrm{m/(s\cdot K^2)}\) and \(\beta = 1.51145 \times 10^{-7}~\mathrm{m/(s\cdot K^3)}\). Thus, at the same undercooling, grey iron grows much slower than aluminium.
The CAFE coupling is performed by first solving the macroscopic thermal field in ProCAST, then mapping the temperature history onto a fine CA grid. The CA grid size is chosen small enough to resolve dendritic growth. In this study, the CAFE simulation is limited to a critical region of a lifting platform frame casting made of HT150.
Simulation Setup
The casting is a rectangular frame with the dimensions \(770~\mathrm{mm}\times 640~\mathrm{mm}\times 60~\mathrm{mm}\), weight about 114 kg. The CAFE region is located at a thin rod section in the middle of the casting. The mould filling and solidification are first simulated with the lost foam parameters. The thermal history at the CAFE region is extracted as a boundary condition for the CAFE calculation. The CAFE parameters are listed in Table 5.
| Parameter | Surface nucleation | Volume nucleation |
|---|---|---|
| \(\Delta T_{max}\) (K) | 2 | 2 |
| \(\Delta T_\sigma\) (K) | 1 | 0.1 |
| \(n_{max}\) (m\(^{-2}\) / m\(^{-3}\)) | \(1 \times 10^{8}\) | \(1 \times 10^{7}\) |
The CAFE simulation is performed using the ProCAST microstructure module, and the results are post-processed to visualise the grain structure.
Results and Discussion
Figure 8 (conceptual) shows the temperature distribution in the casting at different solidification times. The thin rod section cools rapidly and solidifies earlier than the thicker sections. The CAFE results, shown in Figure 9 (conceptual), reveal that the grains first nucleate at the four corners of the rod cross-section. This preferential nucleation at the L-shaped corners can be explained by two factors: (i) the cooling rate is higher at the corners because they have two heat extraction surfaces, leading to a greater undercooling and thus a higher nucleation rate; (ii) the energy barrier for heterogeneous nucleation is lower at a corner than on a flat surface due to the reduction in the total surface free energy. For a given nucleus volume, the contact area with the mould is the same for a corner and a planar surface, but the surface area exposed to the liquid is halved at a corner. Therefore, the total free energy change required for nucleation is lower, promoting earlier nucleation.
As solidification proceeds, equiaxed grains form in the interior of the rod, while a fine chilled layer appears near the mould wall. No columnar grains are observed, which is typical for grey iron with a relatively uniform cooling condition. The final simulated grain structure is compared with metallographic observations of a section from an actual LFC casting. The comparison shows good agreement: the core consists of equiaxed grains, and the edges exhibit a fine-grained region. This confirms that the CAFE method with the described parameter settings can reliably predict the grain morphology of grey iron lost foam castings.
Conclusions and Outlook
In this work, a comprehensive simulation and process design framework for iron castings produced by lost foam casting has been established using ProCAST and JMatPro. The main conclusions are summarised as follows:
- Through ladle analysis and JMatPro calculations, the density variation of grey iron during solidification was accurately obtained. The density curve was incorporated into ProCAST, enabling the filling and solidification stages to be simulated in a single computational model. This approach significantly improves the accuracy of shrinkage porosity prediction compared to conventional separate simulations.
- The mould filling simulation of a plate casting reproduced the experimentally observed filling time (3.5 s vs. 3.38 s), the filling morphology, and the evolution of the gas gap. The gap volume first increases and then decreases, in agreement with experimental measurements.
- For a ductile iron fork-shaped casting, the unified simulation predicted the melt level changes in the riser and correctly predicted the formation or absence of shrinkage porosity for both stepped and top gating systems. The results demonstrate that the simulation method can capture the interplay between liquid feeding and graphitic expansion.
- A quantitative design method for a gating-riser system was proposed, based on thermal modulus analysis, section ratio design, and balanced solidification theory. The method was applied to a grey iron bearing pedestal. The filling simulation indicated that a “slow–fast–slow” pouring rhythm should be adopted. The solidification simulation showed that liquid feeding occurs during the first 3–31% of solidification, followed by graphitic expansion that eliminates all internal shrinkage defects. The actual production of the casting confirmed the validity of the design.
- The CAFE method was applied to simulate grain nucleation and growth in a grey iron LFC casting. The simulation revealed that grains nucleate preferentially at L-shaped corners due to higher cooling rates and reduced energy barriers. The final grain structure, consisting of equiaxed grains in the interior and a fine-grained region near the surface, matched experimental observations. This demonstrates the feasibility of using numerical simulation for microstructural prediction in lost foam casting.
Future work should focus on fully coupling the macro-scale thermal-fluid simulation with the CAFE microstructural model to achieve a more accurate representation of the local solidification conditions. Additionally, the effect of vibration and pressure during lost foam casting on grain refinement can be studied using the developed simulation framework. The ultimate goal is to integrate the simulation methodology into industrial practice to optimise the quality of iron castings produced by the lost foam process.