I am engaged in the research and engineering practice of large steel castings used in power generation equipment. In this study, I focus on a gas turbine compressor support ring casting, which is a typical medium-to-large steel casting with complex geometry, large wall thickness, and stringent service requirements under high temperature and high pressure. The presence of casting defects such as shrinkage cavities, porosity, gas entrapment, and inclusions significantly reduces the quality and reliability of the final product. To address these challenges, I employ the ProCAST simulation platform to numerically analyze the mold filling and solidification processes, predict the formation and distribution of casting defects, and optimize the casting process parameters through a systematic combination of process redesign and orthogonal experiments. This work demonstrates that the integration of computer simulation with engineering judgment effectively minimizes casting defects, improves casting soundness, and provides a reliable reference for industrial production.
1. Introduction and Research Background
Sand casting remains one of the most widely used manufacturing processes for producing large metallic components, especially for steel castings used in power plants, heavy machinery, and energy equipment. The advantages of sand casting include its low tooling cost, flexibility for producing complex shapes, and suitability for both small-lot and mass production. However, the casting process inherently involves complex physical phenomena such as turbulent flow, heat transfer, solidification shrinkage, and phase transformation. These phenomena can give rise to various casting defects, including gas porosity, shrinkage cavities, shrinkage porosity, hot tearing, and inclusions. Such casting defects not only impair the mechanical properties and service life of the castings but also increase the production cost due to scrap and rework. In the context of modern clean and efficient power generation, components like the compressor support ring must withstand elevated steam temperatures and pressures, which demands a high level of internal soundness and structural integrity. Therefore, the prediction and elimination of casting defects is a key technical challenge that must be solved before actual production.
The compressor support ring studied in this work is made of ZG13Cr9Mo2Co1NiVNbNB (often abbreviated as CB2), a newly developed 9%–12% Cr heat-resistant steel. CB2 steel exhibits excellent creep resistance and oxidation resistance at temperatures up to 600–620 °C and pressures up to 30 MPa, making it suitable for advanced ultra-supercritical steam turbine components. Table 1 lists the chemical composition requirements for this steel.
| Element | C | Si | Mn | P | S | Cr | Ni | Mo | Co | V | Nb | B | N | Al |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Spec (wt%) | 0.11–0.14 | 0.20–0.30 | 0.80–1.00 | ≤0.020 | ≤0.010 | 9.00–9.60 | 0.10–0.20 | 1.40–1.60 | 0.90–1.10 | 0.18–0.23 | 0.05–0.08 | 0.008–0.011 | 0.015–0.022 | ≤0.020 |
The room-temperature mechanical properties required for CB2 steel castings are summarized in Table 2.
| Property | Yield Strength RP0.2 (MPa) | Tensile Strength Rm (MPa) | Elongation A (%) | Reduction of Area Z (%) | Impact Energy AkV2 (J) |
|---|---|---|---|---|---|
| Required value | ≥500 | 630–750 | ≥15 | ≥40 | ≥30 (average of 3 specimens, min ≥24) |
The casting studied here has an overall envelope of 3070 mm × 1175 mm × 1462 mm, a weight of about 2.5 tons, and a maximum wall thickness of 368 mm. Owing to the large size and thick sections, the solidification process is prolonged and the feeding of liquid metal is difficult to control. In the original production attempts, shrinkage cavities and porosity were frequently observed in the thick-walled regions, leading to a low yield and a high cost of repair. In order to reduce the occurrence of such casting defects, I have adopted a numerical simulation approach based on the finite element method (FEM) using ProCAST software. This approach allows me to visualize the evolution of temperature, velocity, and solid fraction during mold filling and solidification, and to quantitatively assess the risk of casting defects using established criteria such as the Niyama criterion.
2. Sand Casting Process Design for the Compressor Support Ring
The entire sand casting process for large steel castings typically includes pattern making, molding, core making, melting, pouring, cooling, shakeout, cleaning, heat treatment, and inspection. In my research, I focus mainly on the design and optimization of the gating system, risers, and chillers, because these elements directly determine the flow behavior and solidification sequence, which are the root causes of many casting defects.
2.1 Selection of the Gating System
The gating system is the channel through which molten metal is introduced into the mold cavity. A well-designed gating system must ensure that the mold is filled completely within a suitable time, with smooth flow and minimal turbulence, while also providing favorable temperature gradients for directional solidification. Considering the large dimensions and the requirement for gentle filling, I selected a bottom gating system. In a bottom gating system, the ingates are placed at the bottom of the casting, so the molten metal rises steadily from the bottom upwards. This minimizes splashing, oxidation, and air entrapment, which are common sources of casting defects. In addition, a bottom gating system allows the runner to remain full during pouring, thereby improving slag trapping capability.
Based on the classification of gating systems by the cross-sectional area ratios, I adopted a semi-expansion type where the cross-sectional areas satisfy the relation:
$$ \sum F_{\text{sprue}} < \sum F_{\text{runner}} > \sum F_{\text{ingate}} \quad \text{and} \quad \sum F_{\text{sprue}} > \sum F_{\text{ingate}} $$
This configuration provides a balance between filling speed and flow stability, which is beneficial for large steel castings. The pouring cup was dimensioned with a height of 265 mm, and the sprue had a diameter of 80 mm. The system includes one main sprue, a runner, and three ingates that connect to the lower ring portion of the casting. The arrangement of the gating system is shown schematically in the overall casting assembly model used for simulation.
2.2 Riser Design and Feeding System
Risers are reservoirs of molten metal that compensate for volumetric shrinkage during solidification. The design of risers must ensure that the riser solidifies later than the region it feeds, and that there is a sufficient supply of liquid metal to compensate for both liquid contraction and solidification shrinkage. I used the modulus method to design the risers. The modulus of a casting or riser is defined as
$$ M = \frac{V}{S} $$
where \(V\) is the volume and \(S\) is the cooling surface area. For steel castings, the modulus conditions are commonly expressed as:
$$ M_r \ge 1.2 M_c $$
where \(M_r\) is the modulus of the riser and \(M_c\) is the modulus of the casting section being fed. In the original scheme, two rectangular (腰圆形) conventional sand risers with dimensions 750 mm × 250 mm were placed on the thick-walled upper regions, and one cylindrical riser with a diameter of 450 mm was placed near the curved lower section. However, as I will show later, the simulation of the original scheme revealed that these risers were insufficient to eliminate casting defects, because the solidification sequence was not properly controlled. I therefore later replaced the conventional risers with exothermic insulating risers, which significantly enhance the feeding efficiency and allow a better directional solidification.
2.3 Use of Chills
Chills are metallic bodies placed in the mold to locally accelerate the cooling rate of the casting. By increasing the local cooling rate, chills help to reduce or eliminate shrinkage porosity in regions that are difficult to feed, and they also enlarge the effective feeding distance of risers. In the optimized process, I placed two pairs of external chills on both sides of the middle riser region. The chill dimensions were 740 mm × 210 mm × 130 mm and 217 mm × 200 mm × 156 mm. The chills were made of low-carbon steel and were positioned to promote the desired temperature gradient and to eliminate the shrinkage porosity that appeared near the central boss in the original design.
3. Theoretical Basis of Numerical Simulation
ProCAST is a comprehensive finite element based casting simulation software that can handle coupled fluid flow, heat transfer, and stress analysis. The governing equations for the mold filling process include the continuity equation, the momentum equation (Navier–Stokes), and the energy equation. For an incompressible Newtonian fluid, these equations are written as follows.
3.1 Continuity Equation
The mass conservation for a constant-density fluid is:
$$ \frac{\partial u}{\partial x} + \frac{\partial v}{\partial y} + \frac{\partial w}{\partial z} = 0 $$
3.2 Momentum Equations
The Navier–Stokes equations in the three coordinate directions are:
$$ \rho \left( \frac{\partial u}{\partial t} + u \frac{\partial u}{\partial x} + v \frac{\partial u}{\partial y} + w \frac{\partial u}{\partial z} \right) = -\frac{\partial P}{\partial x} + \rho g_x + \mu \nabla^2 u $$
$$ \rho \left( \frac{\partial v}{\partial t} + u \frac{\partial v}{\partial x} + v \frac{\partial v}{\partial y} + w \frac{\partial v}{\partial z} \right) = -\frac{\partial P}{\partial y} + \rho g_y + \mu \nabla^2 v $$
$$ \rho \left( \frac{\partial w}{\partial t} + u \frac{\partial w}{\partial x} + v \frac{\partial w}{\partial y} + w \frac{\partial w}{\partial z} \right) = -\frac{\partial P}{\partial z} + \rho g_z + \mu \nabla^2 w $$
Here, \(P\) is the pressure, \(g_x, g_y, g_z\) are the components of gravitational acceleration, and \(\mu\) is the dynamic viscosity.
3.3 Energy Equation
The temperature field is governed by the energy conservation equation:
$$ \rho c_p \frac{\partial T}{\partial t} + \rho c_p u \frac{\partial T}{\partial x} + \rho c_p v \frac{\partial T}{\partial y} + \rho c_p w \frac{\partial T}{\partial z} = \frac{\partial}{\partial x} \left( k \frac{\partial T}{\partial x} \right) + \frac{\partial}{\partial y} \left( k \frac{\partial T}{\partial y} \right) + \frac{\partial}{\partial z} \left( k \frac{\partial T}{\partial z} \right) + S $$
where \(c_p\) is the specific heat at constant pressure, \(k\) is the thermal conductivity, \(T\) is the temperature, and \(S\) is the internal heat source term that accounts for the latent heat of solidification.
3.4 Latent Heat Treatment Using the Enthalpy Method
During solidification, the release of latent heat significantly affects the temperature field and solidification behavior. I used the enthalpy method, in which the enthalpy \(H\) is defined as:
$$ H = H_0 + \int_0^T c_p \, dT + (1 – f_s) L $$
where \(f_s\) is the solid fraction, and \(L\) is the latent heat of fusion. This expression is incorporated into the energy equation to correctly capture the evolution of temperature during phase change.
3.5 Boundary Conditions and Interface Heat Transfer
The definition of boundary conditions is crucial for the accuracy of simulation. In ProCAST, I set the interface heat transfer coefficients between different material pairs according to the empirical values listed in Table 3.
| Interface type | Heat transfer coefficient (W/m²·K) |
|---|---|
| Casting – Sand mold | 300–1000 (I used 500) |
| Casting – Air | 5–10 |
| Mold – Air | 5–10 |
| Casting – Chill | 1000–5000 (I used 2000) |
| Chill – Mold | 300–1000 |
| Riser sleeve – Mold | 50 |
For the interface mesh conditions, I used the “COINC” option for interfaces between dissimilar materials (e.g., casting and mold, casting and chill) to allow a temperature discontinuity, and the “EQUIV” option for continuous same-material regions such as the casting and ingates.
4. Modeling and Simulation Setup
4.1 Three-Dimensional Modeling and Mesh Generation
I built the three-dimensional solid models of the casting, gating system, risers, chills, and riser sleeves using NX11.0 software. The assembly was then exported in IGES format and imported into ProCAST’s Visual-Mesh module. A box-shaped sand mold with dimensions 4500 mm × 2500 mm × 2700 mm was created around the casting. Figure 1 shows a typical sand casting assembly used for simulation; the image represents the general concept of complex casting modeling.
After importing the assembly, I performed surface mesh generation with different element sizes for different regions. The sand mold was meshed with a nominal size of 100 mm, the casting body with 30 mm, and the ingates and the connection regions with 20 mm. For the optimized model, the riser sleeve and chill surfaces were refined to 10 mm. After generating the surface mesh, I checked and repaired the mesh quality, and then generated the volume mesh consisting of about 2.24 million elements for the original scheme and about 1.16 million elements for the optimized scheme. The mesh statistics for both schemes are summarized in Table 4.
| Scheme | Surface elements | Volume elements | Total elements |
|---|---|---|---|
| Original | ~70,000 | ~2,140,000 | ~2,230,000 |
| Optimized | ~53,000 | ~1,162,000 | ~1,220,000 |
4.2 Material Properties and Initial Conditions
Since the CB2 alloy is not present in the standard ProCAST database, I manually defined the chemical composition and utilized the built-in thermodynamic calculator to determine the temperature-dependent physical properties. The liquidus temperature was calculated as 1494 °C and the solidus temperature as 1186 °C. Figure 2 displays the key thermal-physical properties of the CB2 alloy, including thermal conductivity, density, enthalpy, and solid fraction as functions of temperature. In the Visual-Cast module, I assigned the casting, ingates, risers, and runner as “Alloy” type with the CB2 material; the sand mold was set as “Mold” with silica sand; the chills were assigned as low-carbon steel; and the insulating sleeve was defined as FT400 exothermic insulating material.
The initial temperature of the molten metal was set to 1575 °C, while the sand mold and chills had an initial temperature of 20 °C. The gravity direction was set as the +Y direction. The mold filling simulation was driven by a mass flow rate boundary condition at the pouring cup, with a flow rate of 100 kg/s and a fill limit of 100%. This resulted in a total filling time of about 107 s for the original scheme and 117 s for the optimized scheme, as the added chills and sleeves slightly affected the mesh and flow resistance.
4.3 Simulation Parameters
I selected the “Gravity Filling” template in ProCAST and set the simulation to stop at a temperature of 1176 °C, slightly below the solidus temperature. The number of time steps was set to 500,000 to ensure the simulation could complete. The temperature field was saved every 50 time steps. The flow was modeled as incompressible Newtonian fluid, and the k-epsilon turbulence model was used for the filling phase. The stress module was activated with a linear-elastic material model for the casting and a rigid model for the mold, allowing the calculation of thermal stresses and potential hot tearing.
5. Simulation Results and Analysis of the Original Scheme
5.1 Filling Process Results
The temperature field during mold filling is of primary importance because it affects the fluidity of the metal and the subsequent solidification pattern. I extracted the temperature field at several time instants during filling. At the beginning of pouring, the molten metal entered the sprue and quickly filled the runner system. After about 4 seconds, the metal reached the ingates and began to enter the mold cavity. Since the bottom gating system was used, the melt rose steadily and smoothly. At 48 seconds, about 45% of the cavity volume was filled, and the temperature distribution showed a relatively uniform hot region in the center, while cooler zones were observed near the mold walls due to heat extraction. At the end of filling (107 s), the entire cavity was filled with molten metal, and the overall temperature range across the casting was approximately 1575 °C to 1510 °C. No immediate cold shuts or misruns were observed, which indicated that the gating design was generally adequate for filling completeness.
However, the velocity field analysis revealed a concern in the early stage of filling. Figure 3 shows the velocity vectors at different filling fractions. During the first 30 seconds, the flow in the ingates was turbulent, with the central ingate exhibiting a peak velocity of about 2.4 m/s, while the side ingates had lower velocities. The central ingate experienced a rapid velocity fluctuation: from 2.4 m/s down to 0.9 m/s and then up to about 1.5 m/s. This fluctuation was attributed to the large temperature difference between the mold and the molten metal, which caused partial solidification of the metal layer on the runner walls. The solidified layer reduced the effective cross-section of the ingate, leading to a temporary decrease in flow rate. Later, the hot metal remelted the solidified layer and the flow resumed its normal rate. This turbulent behavior can entrain air and produce oxide inclusions, which are precursors of casting defects. The velocity distribution of the three ingates is presented in Table 5, listing the characteristic velocities in different phases.
| Ingate location | Initial peak velocity (m/s) | Steady-state velocity (m/s) | Remarks |
|---|---|---|---|
| Central ingate | 2.4 | 1.5 | High turbulence and velocity fluctuation |
| Side ingate 2 | 0.9 | 0.5 | Slower response, gradual increase |
| Side ingate 3 | 1.0 | 0.5 | Similar to side ingate 2 |
5.2 Solidification Process Results
The solidification behavior was examined by monitoring the solid fraction evolution. At the end of filling, about 7% of the casting volume had already solidified, mainly at the thin walls and edges. Figure 4 shows the solid fraction distribution at various stages. When the solid fraction reached 20%, the ingates and runner were almost completely solidified. The casting continued to solidify from the outer surfaces toward the inner thick sections. At a solid fraction of 80%, the last remaining liquid regions were located at the arc top center and beneath the side risers. However, at the final stages, the last liquid did not reside entirely inside the risers. This indicated an improper solidification sequence, i.e., the riser necks solidified before the casting sections that they were supposed to feed. Consequently, the feeding path became blocked, and shrinkage cavities were inevitable in the thick-walled regions.
I also tracked the temperature curves at five representative nodes located in the thick-wall areas and the thinner sections. The cooling curves showed that the thin-wall nodes reached 50% solidification much earlier (around 7000 s) while the thick-wall nodes remained liquid for a much longer time. The lack of a well-aligned temperature gradient from the casting to the riser caused the riser to lose its effectiveness. The overall solidification time was about 37,117 seconds.
5.3 Prediction of Casting Defects in the Original Scheme
Using the Niyama criterion, which is a dimensionless parameter defined as
$$ \text{Niyama} = \frac{G}{\sqrt{R}} $$
where \(G\) is the local temperature gradient and \(R\) is the cooling rate, I identified the regions with a high risk of shrinkage porosity. For large steel castings, a critical value of 1.1 is usually adopted; regions with a Niyama value lower than this threshold are prone to micro-porosity. The simulation predicted three large shrinkage cavities beneath the three risers, all extending into the casting body, and two large shrinkage porosity regions on both sides of the central riser. The predicted volume fraction of shrinkage defects was as high as 16% of the casting volume. The location of these casting defects corresponded well with the blocked feeding paths and the turbulent flow patterns observed in the filling analysis.
6. Process Optimization Based on Simulation
6.1 Improvements in Riser Design
To eliminate the shrinkage cavities, I redesigned the feeding system. The conventional sand risers were replaced with exothermic insulating risers. A 50-mm-thick FT400 exothermic insulating sleeve was placed around each riser, and for the central riser the sleeve thickness was 25 mm. The exothermic insulating material can significantly extend the solidification time of the riser and increase the feeding efficiency from 12–15% for conventional sand risers to 30–60% for exothermic insulating risers. This ensures that the riser remains liquid for a longer time, maintaining an open feeding channel and allowing the molten metal to feed the casting during the later stages of solidification. The modulus of an insulating riser can be evaluated using the modified modulus formula:
$$ M’ = \frac{V}{a S_{\text{side}} + b S_{\text{top}}} $$
where \(a\) and \(b\) are the cooling surface fractions that are active. For a fully insulated (dark) riser with \(a = b = 0.7\), the effective modulus becomes
$$ M’ = \frac{M}{0.7} = 1.43 M $$
which implies a 43% increase in the effective modulus compared with a sand riser of the same shape.
6.2 Addition of Chills
To solve the shrinkage porosity problem on both sides of the central riser, I added two pairs of external steel chills. The chills accelerate the cooling rate at those locations, thus shifting the solidification sequence so that those regions solidify earlier. This reduces the demand on the central riser and prevents the formation of casting defects. The chill dimensions were selected according to the rule of thumb for steel castings, where the chill thickness is about 0.3 to 0.8 times the local thermal modulus. The final chill dimensions were 740 mm × 210 mm × 130 mm for the larger pair and 217 mm × 200 mm × 156 mm for the smaller pair. The chills were placed on the surface of the mold cavity in contact with the casting boss on both sides of the middle riser.
7. Simulation of the Optimized Scheme
7.1 Filling Behavior
I repeated the simulation using the optimized geometry. Although the gating system was unchanged, the presence of the riser sleeves and chills altered the thermal boundary conditions. The filling time increased slightly to about 117 s due to the additional mass of the sleeves and the modified flow paths. The filling temperature field was similar to that of the original scheme, but the velocity near the chills was noticeably reduced. The metal flow remained smooth, and no visible turbulence or air entrainment was observed in the lower part of the casting. The improvement was primarily attributed to the chilling effect, which increased the viscosity of the metal locally and thereby damped the turbulent fluctuations.
7.2 Solidification Sequence
The solid fraction distribution in the optimized scheme exhibited a much more favorable pattern. The casting solidified progressively from the thin sections and from the chilled areas toward the risers. At the end of filling, the solid fraction was about 12%. As solidification proceeded, the last liquid regions were always located inside the risers, which is exactly the desired condition for effective feeding. The temperature field confirmed a clear bottom-to-top directional solidification in the riser regions. The chills successfully reduced the local temperature in the boss areas, creating a steeper temperature gradient that promoted the transport of liquid metal toward the solidification front.
7.3 Defect Prediction in the Optimized Scheme
The Niyama-based defect prediction for the optimized scheme showed that the large shrinkage cavities disappeared from the casting body. The only remaining casting defects were located inside the riser cavities and in the runner system, which are normally cropped away after solidification. The small shrinkage porosity regions near the bottom boss and near the central ingate connection were still present, but their size and extent were drastically reduced compared to the original scheme. This suggested that further improvement could be achieved by optimizing the pouring process parameters.
8. Orthogonal Experiment for Parameter Optimization
8.1 Experimental Design
To further minimize the remaining casting defects, I employed an orthogonal experimental design to optimize the three most influential casting parameters: pouring temperature, pouring rate, and mold initial temperature. Each factor was studied at three levels, as shown in Table 6.
| Factor | Level 1 | Level 2 | Level 3 |
|---|---|---|---|
| A: Pouring temperature (°C) | 1585 | 1575 | 1565 |
| B: Pouring rate (kg/s) | 100 | 90 | 105 |
| C: Sand mold temperature (°C) | 20 | 25 | 30 |
I selected the \(L_9(3^4)\) orthogonal array, which is the smallest valid table for three factors at three levels. The nine experimental schemes are listed in Table 7.
| Run | Factor A | Factor B | Factor C | Combination |
|---|---|---|---|---|
| 1 | 1585 | 100 | 20 | A1B1C1 |
| 2 | 1585 | 90 | 25 | A1B2C2 |
| 3 | 1585 | 105 | 30 | A1B3C3 |
| 4 | 1575 | 90 | 30 | A2B2C3 |
| 5 | 1575 | 105 | 20 | A2B3C1 |
| 6 | 1575 | 100 | 25 | A2B1C2 |
| 7 | 1565 | 105 | 25 | A3B3C2 |
| 8 | 1565 | 100 | 30 | A3B1C3 |
| 9 | 1565 | 90 | 20 | A3B2C1 |
8.2 Simulation Results and Range Analysis
I performed nine full-cycle ProCAST simulations corresponding to the nine experimental runs. For each run, I extracted the total volume of the shrinkage cavities and porosity, and computed the shrinkage porosity percentage as the ratio of defect volume to the total casting volume. The results are listed in Table 8, along with the calculated range (R) values.
| Run | A (°C) | B (kg/s) | C (°C) | Shrinkage porosity (%) |
|---|---|---|---|---|
| 1 | 1585 | 100 | 20 | 15.34 |
| 2 | 1585 | 90 | 25 | 14.36 |
| 3 | 1585 | 105 | 30 | 16.63 |
| 4 | 1575 | 90 | 30 | 13.37 |
| 5 | 1575 | 105 | 20 | 10.53 |
| 6 | 1575 | 100 | 25 | 12.97 |
| 7 | 1565 | 105 | 25 | 17.63 |
| 8 | 1565 | 100 | 30 | 14.67 |
| 9 | 1565 | 90 | 20 | 18.52 |
Using the range analysis method, I calculated the average shrinkage porosity for each factor at each level. The results are summarized in Table 9.
| Statistical value | A (pouring temperature) | B (pouring rate) | C (mold temperature) |
|---|---|---|---|
| K1 | 46.33 | 46.34 | 42.98 |
| K2 | 36.86 | 39.56 | 46.25 |
| K3 | 50.82 | 48.12 | 44.79 |
| k1 | 15.44 | 15.45 | 14.33 |
| k2 | 12.29 | 13.19 | 15.42 |
| k3 | 16.94 | 16.04 | 14.93 |
| Range R | 13.96 | 8.56 | 3.27 |
The range values indicate the relative importance of each factor on the formation of casting defects. The largest range corresponds to pouring temperature (R=13.96), followed by pouring rate (R=8.56), and the smallest is the mold temperature (R=3.27). Therefore, the pouring temperature has the highest influence on the shrinkage porosity of the compressor support ring casting, while the sand mold temperature has the least influence.
Since the goal is to minimize the shrinkage porosity, the optimal level for each factor is the one with the smallest average defect percentage. From Table 9, the lowest average for factor A is at level 2 (1575 °C), for factor B it is at level 1 (100 kg/s), and for factor C it is at level 1 (20 °C). Thus, the optimal combination is \(A_2B_1C_1\), i.e., pouring temperature = 1575 °C, pouring rate = 100 kg/s, and sand mold temperature = 20 °C.
8.3 Verification of the Optimal Parameters
I then performed a final simulation using the optimal parameter combination. The predicted shrinkage porosity was only about 8.4% of the casting volume, and all the remaining porosity was confined to the top portions of the risers, which would be subsequently cut off from the casting. The small shrinkage areas near the bottom boss and the ingate connection were essentially eliminated. This confirms that the orthogonal experiment successfully identified a set of process parameters that minimizes the occurrence of casting defects while maintaining good filling behavior.
9. Industrial Verification
To validate the simulation results, I cooperated with the foundry to produce the compressor support ring under the optimized process conditions. The pattern, molding sand, and core were prepared according to the optimized casting design. The molten CB2 steel was melted in an induction furnace followed by VOD refining, and the chemical composition was verified by direct-reading spectrometry to be within the specification range. The pouring was executed at a melt temperature of 1575 °C, a mass flow rate of 100 kg/s, and a sand mold temperature of 20 °C. After pouring, the casting was allowed to cool slowly in the mold. After shakeout, the risers were cut off. The exothermic insulating risers showed a very good feeding effect: the riser cavities displayed sound metal, and the cross-section of the riser necks revealed no internal shrinkage porosity. The pouring metal consumption was reduced compared with conventional sand risers, because the exothermic sleeves allowed a smaller riser volume while maintaining the same feeding capacity.
The casting was then subjected to the required heat treatment: normalizing at 1110 °C ± 10 °C with a holding time of 2 h + 26.5 h, followed by tempering at 740 °C ± 10 °C with a holding time of 12 h + 28 h. After heat treatment, metallographic specimens were prepared and examined at 200× and 400× magnifications. The microstructure consisted of tempered martensite with a grain size grade of 3–5 and 3.5–5, which satisfied the material specification. The final step was nondestructive testing (NDT) including visual testing (VT), magnetic particle testing (MT), ultrasonic testing (UT), and penetrant testing (PT). All NDT inspections confirmed that the casting was free from macroscopic defects such as cracks, shrinkage cavities, and harmful porosity. The quality of the optimized casting met the acceptance criteria of the purchase specification. This successful industrial trial strongly supports the validity and effectiveness of the simulation-based process optimization in reducing casting defects and improving casting integrity.
10. Conclusion
In this work, I systematically investigated the formation and prevention of casting defects in a large compressor support ring steel casting using ProCAST numerical simulation and process optimization. The main conclusions can be summarized as follows:
(1) The combination of a bottom gating system with three ingates, exothermic insulating risers, and external chills provides a robust feeding configuration that promotes directional solidification and minimizes casting defects in a large CB2 steel casting.
(2) The simulation of the original process scheme revealed that the main causes of casting defects were the turbulent flow in the early filling stage, which led to gas entrapment and oxide inclusions, and the improper solidification sequence, which resulted in blocked feeding paths and the formation of shrinkage cavities and porosity. The original design produced three large shrinkage cavities and two extensive shrinkage porosity regions in the casting body, with a defect volume fraction of about 16%.
(3) By replacing the conventional sand risers with exothermic insulating risers and adding two pairs of external chills, the solidification sequence was significantly improved. The chills accelerated the cooling rate in the problem areas, while the insulating risers maintained liquid metal for a longer period and provided effective feeding. The optimized scheme showed that the casting defects were confined to the risers and the runner system, and the casting body was essentially defect-free.
(4) The orthogonal experiment with three factors and three levels demonstrated that the pouring temperature has the most significant effect on the formation of shrinkage porosity in this casting, followed by the pouring rate, and finally the sand mold temperature. The optimal process parameters were determined to be a pouring temperature of 1575 °C, a pouring rate of 100 kg/s, and a sand mold temperature of 20 °C. The simulation using this parameter combination produced a casting with a very low shrinkage porosity of about 8.4%, all located in the risers, which are subsequently removed.
(5) The industrial production trial fully confirmed the simulation predictions. The casting produced with the optimized process parameters passed all required nondestructive tests, exhibited the expected tempered martensite microstructure, and showed no casting defects. This work demonstrates that the integration of ProCAST simulation with orthogonal experimental design is an effective approach to reduce casting defects, improve casting quality, and reduce production costs for large steel castings.
Future Outlook
Although the present study successfully optimized the casting process for the compressor support ring, further improvements can be made in several directions. First, the material database in ProCAST should be expanded with more precise temperature-dependent properties for CB2 and other advanced heat-resistant steels, as the current values are calculated rather than measured. Second, the simulation can be extended to the microscale using the CAFE module to predict grain structure and porosity distribution more accurately. Third, the influence of other parameters such as superheat, mold coating, and pouring time can be included in future investigations. Finally, the same methodology can be applied to other large steel castings used in ultra-supercritical power plants to systematically minimize casting defects and improve production efficiency. Through continued research and development, computer simulation will play an increasingly important role in the intelligence and greening of the foundry industry.