In this work, I present a comprehensive study on the optimization of the sand casting process for a spray pump volute housing, a critical component in nuclear power plant safety systems. The component is manufactured from ZG06Cr13Ni4Mo martensitic stainless steel, which exhibits complex phase transformation behavior. Through the integration of numerical simulation, orthogonal experimental design, and material property optimization, I successfully eliminated shrinkage porosity and shrinkage cavity defects that had plagued the original production process. Additionally, I conducted a preliminary investigation of the microstructure evolution using the CAFE method to understand the influence of pouring temperature on grain formation. The findings provide both practical manufacturing guidance and theoretical foundations for further research.
1. Introduction
The spray pump volute housing is an integral part of the containment spray system in nuclear power plants. This system is designed to inject borated water into the containment vessel during a loss-of-coolant accident, thereby limiting pressure buildup and preventing overpressure failure. The component directly or indirectly affects nuclear safety, demanding the highest quality standards. As a result, the castings must be free of defects such as shrinkage porosity, gas porosity, cold shuts, and slag inclusions. The material, ZG06Cr13Ni4Mo, is a low-carbon martensitic stainless steel widely used in hydroelectric, thermal, and nuclear power applications. Its balanced combination of strength, toughness, corrosion resistance, and weldability makes it suitable for critical components. However, its complex solidification path, involving ferrite, austenite, and martensite transformations, poses significant challenges during casting.
Sand casting, the chosen manufacturing route, offers exceptional flexibility in terms of alloy type, component size, and production volume. Its cost-effectiveness and adaptability make it the preferred method for producing large, complex steel castings. Nevertheless, sand casting of such intricate geometries with uneven wall thickness often leads to internal defects. Traditional trial-and-error methods are no longer sufficient to guarantee defect-free castings in a cost-effective manner. Hence, computer-aided engineering tools, particularly numerical simulation of the casting process, have become indispensable. In this study, I applied the AnyCasting software, based on the finite difference method, to simulate mold filling, temperature field evolution, and solidification of the volute housing. The simulation results allowed me to identify the root causes of shrinkage defects and to optimize the gating system, riser design, chills, process parameters, and even the alloy composition.
The key objectives of this research are summarized as follows:
- To analyze the original sand casting process and predict defect locations using numerical simulation.
- To propose and validate an improved process through simulation and actual production trials.
- To optimize the crucial process parameters via orthogonal experimentation.
- To optimize the chemical composition of ZG06Cr13Ni4Mo steel using JMatPro and orthogonal tests.
- To simulate the microstructure evolution during solidification using the CAFE method, focusing on the effect of pouring temperature.
This paper systematically presents all these efforts, concluding with practical recommendations for producing sound castings of this class.
2. Numerical Methods and Simulation Software
The simulation of casting processes relies on solving differential equations that govern fluid flow, heat transfer, and solidification. Among the various numerical techniques, the finite difference method (FDM) is widely adopted due to its simplicity and effectiveness for complex geometries. The fundamental concept of FDM is to replace derivatives with finite difference quotients, thereby converting differential equations into algebraic equations that can be solved iteratively. For a continuous function \( y=f(x) \), the derivative is defined as:
\[
\frac{dy}{dx} = \lim_{\Delta x \to 0} \frac{f(x+\Delta x)-f(x)}{\Delta x}
\]
In practice, \(\Delta x\) remains finite, leading to approximate differences: forward difference, backward difference, and central difference. The central difference offers second-order accuracy:
\[
\frac{dy}{dx} \approx \frac{f(x+\Delta x)-f(x-\Delta x)}{2\Delta x}
\]
The second-order derivative is approximated by:
\[
\frac{d^2y}{dx^2} \approx \frac{f(x+\Delta x)-2f(x)+f(x-\Delta x)}{\Delta x^2}
\]
These approximations are used to construct difference equations based on Taylor series expansions or the control volume method. In this work, I adopt the control volume approach because it ensures physical conservation of mass, momentum, and energy even on coarse grids, which is essential for complex castings.
2.1 AnyCasting Software
AnyCasting is a specialized simulation package developed for modeling various casting processes. It comprises five main modules:
| Module | Function |
|---|---|
| anyPRE | Pre-processing: geometry import, mesh generation, condition setup |
| anyMESH | Mesh editing and refinement |
| anyDBASE | Material database with standard and user-defined alloys |
| anySOLVER | Finite difference solver for flow, heat transfer, and solidification |
| anyPOST | Post-processing: visualization of temperature, velocity, pressure, and defect criteria |
The software uses a “real flow” solver for accurate mold filling analysis and applies several criteria for predicting shrinkage defects, such as temperature gradient, solid fraction gradient, and residual melt modulus. In this study, I use the residual melt modulus as the primary indicator for defect propensity.
2.2 Simulation Workflow
The simulation workflow consists of the following steps:
- Create three-dimensional models of the casting, gating system, risers, and chills using CAD software (e.g., UG).
- Export models in STL format and import into anyPRE.
- Generate finite difference meshes, using variable grid spacing for thin sections and complex features.
- Set material properties, boundary conditions, initial temperatures, and interface heat transfer coefficients.
- Run the solver for mold filling and solidification analysis.
- Post-process results to evaluate flow behavior, temperature distribution, and defect formation.
This workflow enables accurate prediction of casting defects before production, minimizing costly trial runs.
3. Original Casting Process and Defect Analysis
3.1 Component Description and Material Properties
The spray pump volute housing has a maximum outer diameter of 755 mm, an inner diameter of 315 mm, and a height of 338 mm. The wall thickness varies between 7 mm and 90 mm, resulting in significant thermal gradients during solidification. The casting weighs approximately 350 kg. Its geometry is roughly symmetrical, with a bottom flange containing recesses, various protrusions on the outer surface, and a central cylindrical bore surrounded by spiral cavities. These features make uniform cooling and feeding challenging.
The specified material, ZG06Cr13Ni4Mo, has the following nominal chemical composition:
| Element | C | Si | Mn | S | P | Cr | Ni | Mo |
|---|---|---|---|---|---|---|---|---|
| Content (wt.%) | ≤0.06 | 1.0 | 1.0 | 0.02 | 0.03 | 12.0–13.5 | 3.5–4.5 | 0.4–0.7 |
Its mechanical properties are given below:
| Yield Strength (MPa) | Tensile Strength (MPa) | Elongation (%) |
|---|---|---|
| ≥350 | ≥750 | ≥15 |
The steel undergoes a complex phase transformation sequence: liquid → ferrite → austenite + ferrite → austenite → martensite upon cooling. The final microstructure at room temperature is predominantly lath martensite, which provides the desired combination of strength and toughness.
3.2 Original Process Design
The original sand casting process was designed based on empirical rules and prior experience. The pouring position was chosen with the bottom flange downward to obtain a dense structure in the critical sealing area. The parting line was placed accordingly, and a bottom-gating system was used with refractory brick tubes. The gating system consisted of a sprue of diameter 70 mm, a runner of diameter 50 mm, and two ingates of diameter 50 mm. Two internal chills made of 45# steel were placed at the bottom flange to promote directional solidification. Riser design followed the modulus method and empirical charts. The top of the casting was fed by an open riser (1#) of dimensions φ110×165 mm, while the bottom flange protrusions were fed by three blind risers (2#, 3#, 4#) with dimensions φ100×150, φ120×180, and φ150×380 mm, respectively. The pouring temperature was 1580°C, pouring time 18 s, and the mold was at ambient temperature.
3.3 Numerical Simulation of the Original Process
I used AnyCasting to simulate the original process. The mesh generation used variable grid spacing to capture the thin walls while maintaining computational efficiency. The total number of cells was 1,662,648. The boundary conditions included heat transfer coefficients between the casting and the mold, which varied with temperature as shown in the software database. Table 1 lists key simulation parameters.
| Parameter | Value |
|---|---|
| Casting material | ZG06Cr13Ni4Mo |
| Pouring temperature | 1580 °C |
| Mold material | Furan resin sand |
| Pouring time | 18 s |
| Pouring stream diameter | 50 mm |
| Heat transfer coefficient (casting-mold) | Temperature-dependent (41.87–3000 W/m²K) |
| Heat transfer coefficient (casting-chill) | 3000 W/m²K |
The filling simulation indicated that the molten metal entered the cavity smoothly from the bottom, rising in a planar front. However, at the thin spiral inner walls, there was potential for flow merging and gas entrapment, likely leading to gas porosity. The temperature field at the end of filling showed a progressive decrease from the bottom to the top, with higher temperatures near the ingates. During solidification, the simulation predicted isolated liquid regions in three areas:
- At the junction of thick walls in the middle of the casting (1# isolated liquid)
- At the contact regions between the top open risers and the casting (2# isolated liquid)
- Inside the bottom blind risers (benign, not causing defects)
These isolated regions were not adequately fed, resulting in shrinkage porosity. The residual melt modulus criterion clearly indicated defects at the top casting-riser interface, at the middle thick-wall junction, and inside the bottom flange protrusions. Actual production trials by the factory confirmed these simulation results: ultrasonic inspection and machining revealed shrinkage cavities at the bottom protrusion threaded holes. This validation confirmed the accuracy of the simulation model.
4. Process Optimization
Based on the defect analysis, I developed a comprehensive optimization strategy encompassing riser redesign, gating system improvement, additional chills, process parameter optimization, and alloy composition refinement. Each measure is described below.
4.1 Riser Improvement
The original open riser at the top was replaced by an insulating exothermic riser. A 25 mm thick insulating sleeve was added around the riser circumference, and an exothermic topping compound was applied after pouring. This extended the solidification time of the riser, enhancing its feeding capacity. Moreover, a new blind riser (5#) was introduced at the center of the casting bore. This riser was also equipped with a 35 mm thick insulating sleeve and had dimensions of φ160×240 mm. The new riser provided feeding to the inner wall regions that previously suffered from shrinkage. The improved riser system is illustrated schematically in Figure 1 (not reproduced here; refer to the original layout).
4.2 Gating System Improvement
Despite the acceptable filling behavior, the gating system was modified to accommodate the additional riser and to improve feeding efficiency. The number of ingates was increased from two to three by adding an ingate at the base of the new central riser. Consequently, the runner system was rearranged to supply all three ingates uniformly. The sprue diameter remained unchanged at 70 mm. The modified system promoted more uniform temperature distribution and ensured that the central riser was filled with hot metal, maximizing its feeding efficiency.
4.3 Additional Chills
To eliminate hot spots and refine grain structure, three types of external chills were added:
- 2# chills: curved chills of dimensions R400×10.5°×40 (two pieces), placed inside the protrusions corresponding to the 3# chills.
- 3# chills: round steel chills of diameter 32 mm (two pieces), located at the symmetrical inner cavity arc positions near the large sand core to reduce the local overheating observed in the original simulation.
- 4# chills: curved chills of dimensions R430×10.5°×40 (two pieces), placed between the blind risers and the corresponding protrusions on the casting.
These chills accelerated solidification in critical areas, promoted directional solidification, and reduced the risk of shrinkage defects.
4.4 Process Parameter Optimization Using Orthogonal Experiments
I selected three key process parameters—pouring temperature (A), pouring time (B), and mold preheating temperature (C)—as factors, each at three levels. The residual melt modulus, an indicator of shrinkage defect tendency, was chosen as the response variable. The factor levels are listed in Table 2.
| Level | A: Pouring Temperature (°C) | B: Pouring Time (s) | C: Mold Preheat Temperature (°C) |
|---|---|---|---|
| 1 | 1550 | 14 | 0 |
| 2 | 1565 | 18 | 12.5 |
| 3 | 1580 | 22 | 25 |
Using the \( L_9(3^4) \) orthogonal array, nine simulation runs were performed. The results, including the residual melt modulus, are summarized in Table 3.
| Trial No. | A (°C) | B (s) | C (°C) | Residual Melt Modulus |
|---|---|---|---|---|
| 1 | 1550 | 14 | 0 | 3.1785 |
| 2 | 1550 | 18 | 12.5 | 1.7204 |
| 3 | 1550 | 22 | 25 | 1.7054 |
| 4 | 1565 | 14 | 12.5 | 1.7115 |
| 5 | 1565 | 18 | 25 | 1.7187 |
| 6 | 1565 | 22 | 0 | 1.7045 |
| 7 | 1580 | 14 | 25 | 1.7138 |
| 8 | 1580 | 18 | 0 | 1.7191 |
| 9 | 1580 | 22 | 12.5 | 1.7058 |
Range analysis was conducted to determine the significance of each factor. The results are presented in Table 4.
| Statistic | A | B | C |
|---|---|---|---|
| K1 | 6.6043 | 6.6038 | 6.6021 |
| K2 | 5.1347 | 5.1582 | 5.1377 |
| K3 | 5.1387 | 5.1157 | 5.1379 |
| k1 | 2.2014 | 2.2013 | 2.2007 |
| k2 | 1.7116 | 1.7194 | 1.7126 |
| k3 | 1.7129 | 1.7052 | 1.7126 |
| Range R | 0.4961 | 0.4898 | 0.4881 |
Based on the range values, the factor significance order is A > B > C, meaning pouring temperature has the greatest influence, followed by pouring time and mold preheat temperature. The optimal combination is A2B3C3, i.e., pouring temperature 1565°C, pouring time 22 s, and mold preheat temperature 25°C. These settings minimize the residual melt modulus, thereby reducing shrinkage defect risk.
4.5 Alloy Composition Optimization
The chemical composition of ZG06Cr13Ni4Mo was optimized to ensure a fully martensitic microstructure at room temperature, minimizing residual austenite and delta ferrite. The key parameter is the martensite start temperature (MS), which can be calculated using the empirical formula:
\[
M_S(\text{°C}) = 561 – 474[\%C] – 33[\%Mn] – 17[\%Si] – 17[\%Cr] – 21[\%Mo]
\]
To achieve a high MS (above room temperature), certain elements must be controlled. I used JMatPro software to compute the MS temperature for various compositions. Five elements—C, Cr, Ni, Si, and Mo—were selected as factors, each with four levels. The orthogonal array \( L_{16}(4^5) \) was employed. The factor levels are shown in Table 5.
| Level | C (%) | Cr (%) | Ni (%) | Si (%) | Mo (%) |
|---|---|---|---|---|---|
| 1 | 0.015 | 12.00 | 3.45 | 0.20 | 0.40 |
| 2 | 0.030 | 12.50 | 3.80 | 0.40 | 0.50 |
| 3 | 0.045 | 13.00 | 4.15 | 0.60 | 0.60 |
| 4 | 0.060 | 13.50 | 4.50 | 0.80 | 0.70 |
The experiments were designed to maximize MS. The computed MS values for each trial are listed in Table 6.
| Trial | C | Cr | Ni | Si | Mo | MS (°C) |
|---|---|---|---|---|---|---|
| 1 | 0.015 | 12.00 | 3.45 | 0.20 | 0.40 | 250.1 |
| 2 | 0.015 | 12.50 | 3.80 | 0.40 | 0.50 | 239.3 |
| 3 | 0.015 | 13.00 | 4.15 | 0.60 | 0.60 | 227.4 |
| 4 | 0.015 | 13.50 | 4.50 | 0.80 | 0.70 | 213.5 |
| 5 | 0.030 | 12.00 | 3.80 | 0.60 | 0.70 | 233.0 |
| 6 | 0.030 | 12.50 | 3.45 | 0.80 | 0.60 | 240.2 |
| 7 | 0.030 | 13.00 | 4.50 | 0.20 | 0.50 | 216.1 |
| 8 | 0.030 | 13.50 | 4.15 | 0.40 | 0.40 | 223.1 |
| 9 | 0.045 | 12.00 | 4.15 | 0.80 | 0.50 | 221.1 |
| 10 | 0.045 | 12.50 | 4.50 | 0.60 | 0.40 | 212.5 |
| 11 | 0.045 | 13.00 | 3.45 | 0.40 | 0.70 | 233.8 |
| 12 | 0.045 | 13.50 | 3.80 | 0.20 | 0.60 | 225.3 |
| 13 | 0.060 | 12.00 | 4.50 | 0.40 | 0.60 | 208.2 |
| 14 | 0.060 | 12.50 | 4.15 | 0.20 | 0.70 | 214.0 |
| 15 | 0.060 | 13.00 | 3.80 | 0.80 | 0.40 | 221.8 |
| 16 | 0.060 | 13.50 | 3.45 | 0.60 | 0.50 | 228.7 |
Range analysis (Table 7) indicates that Ni has the greatest effect on MS, followed by C, Cr, Mo, and Si. The optimal composition to maximize MS is C=0.015%, Cr=12.5%, Ni=3.8%, Si=0.4%, Mo=0.5%. This composition also maintains a Cr/Ni equivalent ratio within the range of 0.36–0.44, which is favorable for obtaining a fully martensitic structure.
| Statistic | C | Cr | Ni | Si | Mo |
|---|---|---|---|---|---|
| k1 | 232.575 | 228.100 | 238.200 | 226.375 | 226.875 |
| k2 | 228.100 | 226.500 | 229.850 | 226.100 | 226.300 |
| k3 | 223.175 | 224.775 | 221.400 | 225.400 | 225.275 |
| k4 | 218.175 | 222.650 | 212.575 | 224.150 | 223.575 |
| Range | 14.400 | 5.450 | 25.625 | 2.225 | 3.300 |
4.6 Simulation of the Optimized Process
I created a new simulation model with the improved gating, riser, and chill system, using the optimized process parameters and alloy composition. The mesh contained 1,720,128 cells. The filling simulation confirmed smooth, quiescent filling without turbulence or gas entrapment. The solidification sequence was strictly directional: from the thin walls and chill regions toward the risers, with the new central riser being the last to solidify. No isolated liquid regions appeared within the casting after 20% solidification. The defect prediction graphs showed no shrinkage porosity or cavities in any section. The simulated temperature at 96% solidification indicated that all regions except the risers had fully solidified, confirming sound feeding.
4.7 Production Validation
Based on the optimized process, the factory produced a trial batch of spray pump volute housings. The castings exhibited smooth surfaces, no visible defects, and excellent internal soundness. Ultrasonic inspection confirmed the absence of shrinkage defects, meeting all customer requirements. The success of this validation demonstrates the efficacy of the systematic optimization approach.
5. Microstructure Simulation of ZG06Cr13Ni4Mo Steel
5.1 Phase Transformation Analysis
To simulate the microstructure evolution, I first analyzed the phase transformation sequence of ZG06Cr13Ni4Mo using the vertical section phase diagram of the Fe-C-12.5Cr-4.5Ni-0.5Si-0.6Mn system. As the alloy cools from the pouring temperature:
- Above the liquidus temperature (1490°C), the metal is fully liquid.
- Between liquidus and about 1435°C, primary ferrite precipitates, resulting in a mixture of liquid and ferrite.
- At approximately 1435°C, ferrite begins to transform to austenite. This continues until about 1305°C, where the structure is fully austenitic.
- Below 845°C, carbides such as Cr23C6 may precipitate from austenite.
- When cooling below the martensite start temperature (around 280°C), austenite transforms to martensite upon rapid cooling (air cooling is sufficient due to high hardenability).
At room temperature, the microstructure is predominantly lath martensite. This complex transformation must be accounted for in the simulation model, particularly regarding latent heat release and solid fraction evolution.
5.2 CAFE Model
I used the CAFE (Cellular Automaton Finite Element) method coupled with ProCAST software (rather than AnyCasting) for microstructure simulation. The CAFE model couples macroscopic heat flow (computed by FE) with a cellular automaton algorithm that simulates nucleation and growth of dendrites. The key components are described below.
Macroscopic Heat Transfer Model
The transient temperature field is governed by the heat conduction equation:
\[
\rho C_p \frac{\partial T}{\partial t} = \nabla \cdot (\lambda \nabla T) + \Delta H \frac{\partial f_s}{\partial t}
\]
where \( \rho \) is density, \( C_p \) is specific heat, \( \lambda \) is thermal conductivity, \( T \) is temperature, \( \Delta H \) is latent heat, and \( f_s \) is solid fraction.
Solute Diffusion Model
For the martensitic stainless steel, solute back-diffusion in the solid is significant. I selected the Lever rule model:
\[
C_L = C_0 \left[ 1 – (1-k) f_s \right]^{-1}
\]
where \( C_L \) is the liquid concentration at the interface, \( C_0 \) is the initial alloy concentration, and \( k \) is the equilibrium partition coefficient.
Nucleation Model
I used the continuous nucleation model based on a Gaussian distribution of nucleation sites:
\[
\frac{dn}{d(\Delta T)} = \frac{n_{\max}}{\Delta T_\sigma \sqrt{2\pi}} \exp\left[ -\frac{1}{2}\left( \frac{\Delta T – \Delta T_N}{\Delta T_\sigma} \right)^2 \right]
\]
Here, \( n_{\max} \) is the maximum nucleus density, \( \Delta T_N \) is the mean nucleation undercooling, and \( \Delta T_\sigma \) is the standard deviation.
Dendrite Tip Growth Model
For growth kinetics, I applied the KGT (Kurz-Giovanola-Trivedi) model, which relates undercooling \( \Delta T \) to growth velocity \( v \). For multicomponent alloys, the model was extended using effective solute coefficients. The resulting empirical relation is:
\[
v(\Delta T) = a_2 \Delta T^2 + a_3 \Delta T^3
\]
Using the ProCAST CAFE module, the coefficients were computed from the alloy’s thermophysical data. The parameters used are:
- Gibbs-Thomson coefficient: \( \Gamma = 1.8707 \times 10^{-7} \) m·K
- \( a_2 = 0 \)
- \( a_3 = 1.8707 \times 10^{-5} \) m/s/K³
Coupled CA-FE Calculation
The temperature at each CA cell is interpolated from the FE nodes using shape functions. The latent heat released by nucleation and growth is fed back to the FE thermal solver. This iterative coupling continues until solidification is complete.
5.3 Simulation Setup
I selected a small rectangular area (40 mm × 10 mm × 5 mm) on the top flange of the casting for detailed microstructure analysis. This area was discretized with fine cells. Three pouring temperatures were investigated: 1550°C (Process A), 1565°C (Process B), and 1580°C (Process C). The mold was at ambient temperature, and the interface heat transfer coefficients were the same as in the optimized casting process. The Gaussian nucleation parameters for the volume and surface were set as follows:
| Parameter | Volume | Surface |
|---|---|---|
| Maximum nucleus density \( n_{\max} \) | 3×10¹¹ m⁻³ | 3×10⁹ m⁻² |
| Mean undercooling \( \Delta T_N \) | 15 K | 2 K |
| Standard deviation \( \Delta T_\sigma \) | 4.8 K | 1 K |
The simulation also included the calculation of grain orientation, assigning a random crystallographic orientation to each nucleus. The <100> direction was taken as the preferred growth direction.
5.4 Microstructure Simulation Results
Figure 1 shows the simulated grain structures at the selected cross-section for the three pouring temperatures. The images reveal a columnar-to-equiaxed transition from the mold wall toward the interior. The grains are generally columnar near the chilled surface and become more equiaxed toward the center, with varying orientations relative to the heat flow direction.
Figure 1: Simulated grain structures at different pouring temperatures (left to right: Process A, B, C).
The statistical analysis of the simulated grains is presented in Table 9. With increasing pouring temperature, the number of grains decreases, while the average grain radius and average grain area increase. The maximum grain area shows a non-monotonic trend, initially increasing from Process A to B, then decreasing in Process C. This behavior can be attributed to the competing effects of nucleation rate and growth time. Lower pouring temperatures provide greater undercooling, enhancing nucleation, while higher temperatures allow existing grains to grow larger before impingement.
| Process | Pouring Temp (°C) | Grain Count | Average Radius (m) | Max Grain Area (m²) | Average Grain Area (m²) |
|---|---|---|---|---|---|
| A | 1550 | 204 | 1.0777×10⁻³ | 7.16×10⁻⁶ | 1.9802×10⁻⁶ |
| B | 1565 | 193 | 1.08497×10⁻³ | 7.95×10⁻⁶ | 2.07254×10⁻⁶ |
| C | 1580 | 185 | 1.11056×10⁻³ | 6.64×10⁻⁶ | 2.13904×10⁻⁶ |
The orientation deviation from the ideal <100> direction was also analyzed. The average deviation angles were 31.857° for Process A, 32.491° for Process B, and 31.957° for Process C. These values are similar, indicating that pouring temperature has a minor effect on the texture spread. However, the grain size distribution shifted toward larger grains with increased pouring temperature, and the shape factor distribution revealed a slightly higher percentage of near-spherical grains in Process C, but also the presence of some very elongated grains (maximum shape factor 17.783), suggesting less uniformity.
Overall, Process B (1565°C) provided the best balance between grain size, number, and uniformity, which is consistent with the optimized pouring temperature obtained from the macroscopic defect minimization. This suggests that the same parameter set yields both sound castings and a desirable microstructure.
5.5 Discussion
The microstructure simulation demonstrates that pouring temperature significantly influences the as-cast grain structure of ZG06Cr13Ni4Mo steel. Elevated pouring temperatures reduce the thermal undercooling in the melt, leading to a lower nucleation rate. Simultaneously, the temperature gradient at the solidification front becomes shallower, which reduces the driving force for columnar growth and may promote equiaxed growth after some time. The net effect is a coarser grain structure. To achieve a finer, more uniform microstructure, lower pouring temperatures are generally preferred, provided that mold filling remains adequate and no cold shuts occur. The optimized temperature of 1565°C strikes a compromise between fluidity, defect prevention, and grain refinement.
This preliminary simulation sets the stage for future research involving the prediction of mechanical properties based on the simulated microstructure. It also highlights the importance of using the CAFE method to optimize casting parameters not only for macro-defect control but also for microstructural quality.
6. Conclusions
Through this comprehensive study, I successfully addressed the shrinkage defect problem in the sand casting of a spray pump volute housing made of ZG06Cr13Ni4Mo stainless steel. The main conclusions are as follows:
- Numerical simulation using AnyCasting accurately predicted the location and extent of shrinkage defects in the original process. The simulated defect patterns matched the actual production failures, verifying the reliability of the simulation method.
- The optimized process, which incorporated insulating exothermic risers, an additional central blind riser, modified gating, and strategically placed chills, eliminated shrinkage porosity and cavities. Production validation confirmed casting soundness.
- Orthogonal experiments identified pouring temperature as the most influential process parameter, followed by pouring time and mold preheat temperature. The optimal combination was pouring temperature 1565°C, pouring time 22 s, and mold preheat temperature 25°C.
- Composition optimization via JMatPro and orthogonal tests determined an optimal chemistry of C=0.015%, Cr=12.5%, Ni=3.8%, Si=0.4%, Mo=0.5%, ensuring a fully martensitic structure with a high MS temperature.
- CAFE simulations of microstructure showed that increasing pouring temperature reduces grain count and increases grain size. The pouring temperature of 1565°C yielded the most uniform grain structure, reinforcing the process optimization results.
This work provides a practical framework for optimizing sand casting processes for complex, safety-critical steel castings, integrating macro-level defect simulation, statistical design of experiments, and microstructural modeling.
7. Future Work
While the macroscopic simulation is now well established in industry, the application of microstructure simulation to actual production needs further development. Future research should focus on:
- Expanding the material database of ZG06Cr13Ni4Mo steel to include accurate phase transformation and nucleation data over a wider range of cooling rates.
- Coupling the microstructure simulation with thermal-stress analysis to predict hot tearing and residual stresses.
- Developing user-friendly interfaces to allow foundry engineers to run CAFE simulations without specialized training.
- Investigating the effect of other process parameters, such as mold material and coating, on grain structure.
- Integrating the optimized composition and process into the production of other martensitic stainless steel castings with similar complexity.
The ultimate goal is to establish a digital twin of the casting process, enabling virtual prototyping and rapid optimization of not only macroscopic defects but also microscopic material properties.
Acknowledgments
I express my gratitude to the foundry engineers who provided invaluable support during the production trials and offered practical insights into the casting defects. Their expertise significantly enhanced the quality of this research. I also thank my research group for their constructive discussions and technical assistance.
(Note: As per the instruction, the author names and institutional affiliations have been omitted. The image link has been inserted into the manuscript at the appropriate location.)