In my thesis work, I focused on a heat-resistant steel casting used in a gas turbine compressor support ring. This component is larger and more complex than ordinary castings, requiring a large amount of molten metal and containing thick-wall regions. During filling and solidification, it is difficult to establish a favorable temperature gradient and control the solidification sequence, which often leads to defects such as misruns, shrinkage porosity, shrinkage cavities, hot tears, and deformation. Therefore, I used the casting simulation software ProCAST to simulate the entire casting process, identify potential defects, optimize the casting process design, and improve the overall quality of the steel casting. The research was carried out within an industrial collaboration with a foundry, and the goal was to provide a reliable casting solution for the actual production of the compressor support ring.
In order to deeply understand the behavior of the steel casting, I first reviewed the theoretical foundations of sand casting for large steel casting parts. The conventional process for producing heavy steel castings in a foundry includes pattern making, molding, melting, pouring, shakeout, cleaning, and inspection. Each step must be carefully controlled to avoid defects. I chose sand casting with a gravity pouring system because of its low cost, high flexibility, and suitability for large steel casting components. The major challenge in sand casting of large steel casting parts is to ensure complete filling of the mold cavity and adequate feeding during solidification. To solve this problem, I designed a gating system, risers, and chills based on the geometry and alloy characteristics of the compressor support ring.
The first step in my design was to select the gating system type. For this large steel casting, I compared several types of gating systems. A top gating system provides fast filling and good feeding, but it can cause severe turbulence and oxidation. A bottom gating system offers smooth filling and better gas removal, but the temperature distribution tends to be higher at the bottom, which is not favorable for sequential solidification. A step gating system combines advantages of both, but its structure is complex. Considering the actual shape of the compressor support ring, I chose a bottom gating system with three ingates. The metal is introduced at the lower part of the ring, which ensures a stable rising liquid level and reduces the risk of gas entrapment. Table 1 summarizes the main gating system types and their characteristics.
| Type | Filling stability | Temperature gradient | Slag removal | Application |
|---|---|---|---|---|
| Top | Poor | Favorable (bottom to top) | Poor | Small castings |
| Bottom | Good | Unfavorable | Good | Large steel casting |
| Middle | Medium | Medium | Medium | Medium and small castings |
| Step | Good | Favorable | Good | Complex tall castings |
The gating system was designed with an open-to-closed ratio. The sprue, runner, and ingate cross-sectional areas were determined to control the flow velocity. For a bottom gating system used in sand casting of heavy steel casting parts, I adopted a non-pressurized system where the total cross-sectional area of the ingates is larger than that of the runner, and the runner is larger than the sprue. This helps to reduce the velocity of the molten metal after entering the mold cavity. The theoretical filling time was calculated using the mass flow rate and the volume of the cavity. The relevant formula is:
$$ t_{\text{fill}} = \frac{M}{\dot{m}} = \frac{\rho V}{\dot{m}} $$
where \( \rho \) is the density of the molten steel casting, \( V \) is the cavity volume, and \( \dot{m} \) is the mass flow rate. In the initial design, the pouring temperature was set to 1575 °C, and the pouring speed was set to 100 kg/s, which gave a filling time of approximately 107 seconds.
Riser design is another critical aspect for producing sound steel casting. The primary function of a riser is to provide molten metal to compensate for the volume shrinkage during solidification. I employed the modulus method to calculate the riser dimensions. The modulus of a casting region is defined as:
$$ M = \frac{V}{A} $$
where \( V \) is the volume of the region and \( A \) is its cooling surface area. According to the modulus principle, the riser modulus must be larger than the modulus of the casting section to ensure that the riser solidifies later than the casting. For steel casting, the following relationship is often used:
$$ M_r = (1.1 \sim 1.2) M_c $$
where \( M_r \) is the riser modulus and \( M_c \) is the casting modulus at the hot spot. I also checked the feeding capacity of the riser using the solidification shrinkage coefficient \( \beta \) of the alloy. The minimum riser volume \( V_r \) must satisfy:
$$ \beta (V_r + V_c) \le V_r \eta $$
where \( V_c \) is the volume of the casting section being fed, and \( \eta \) is the riser efficiency. For an ordinary sand riser, \( \eta \) is about 12-15%, while for an insulating riser, \( \eta \) can reach 25-30%, and for an exothermic riser, it can be as high as 30-60%. Since the compressor support ring has several thick sections, I placed three risers on the top of the casting: two lateral risers above the flanges and one central riser above the central boss. The initial riser design used ordinary sand risers, but later I replaced them with exothermic insulating risers to improve the feeding efficiency and establish a better temperature gradient.
To further control the solidification sequence and reduce the risk of shrinkage defects at the central boss, I added external chills on both sides of the central riser. Chills accelerate cooling in specific regions, thereby increasing the temperature gradient toward the riser and enlarging the effective feeding distance of the riser. The chill thickness was determined as 0.3–0.8 times the hot spot thickness for steel casting. In my new design, I used two pairs of chills with dimensions 740 mm × 210 mm × 130 mm and 217 mm × 200 mm × 156 mm, respectively. These chills were positioned at the thick-walled sections adjacent to the central riser.
After designing the initial casting process, I built a three-dimensional model of the casting, including the gating system, risers, and chills. The geometric model was created in NX11.0 and then imported into the pre-processing module of ProCAST. I generated finite element meshes using the Visual-Mesh tool. The mesh size was set to 30 mm for the casting, 20 mm for the ingates and connections, 10 mm for the riser sleeves and chills, and 100 mm for the sand mold. The total number of tetrahedral elements was about 2.23 million. The mesh quality was checked and repaired to ensure accurate numerical results.
The material used for the compressor support ring is a high-temperature steel casting alloy designated as ZG13Cr9Mo2Co1NiVNbNB (also known as CB2). This alloy is a 9-12% Cr martensitic heat-resistant steel casting used for steam turbine components operating at temperatures up to 620 °C. Its chemical composition is listed in Table 2.
| C | Si | Mn | P | S | Cr | Ni | Mo | Co | V | Nb | B | N | Al |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 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 liquidus and solidus temperatures of this alloy were calculated using ProCAST’s thermodynamics engine. The liquidus temperature was found to be 1494 °C and the solidus temperature 1186 °C. The thermal and mechanical properties such as thermal conductivity, density, enthalpy, and solid fraction were automatically generated as functions of temperature. Since ProCAST does not include this material in its default database, I manually entered the composition and used the lever rule solidification model to obtain the required data.
In the pre-processing stage, I defined the boundary conditions. The interface heat transfer coefficient between the steel casting and the sand mold was set to 500 W/(m²·K), while the coefficient between the casting and the chills was set to 2000 W/(m²·K). The interface between the casting and the riser sleeves was set to 50 W/(m²·K). The mold material was silica sand with an initial temperature of 20 °C. The gravity direction was set along the positive Y direction. For the initial simulation, I set the pouring temperature to 1575 °C, the pouring speed to 100 kg/s, and the mold temperature to 20 °C. The filling limit was 100%, and the flow rate was prescribed at the inlet of the sprue. I also specified heat boundary conditions on the outer surfaces of the mold, riser tops, and sprue cup as air cooling boundaries.
The governing equations for the filling process are the continuity equation, the momentum (Navier-Stokes) equation, and the energy equation. For an incompressible Newtonian fluid, the continuity equation is:
$$ \frac{\partial u}{\partial x} + \frac{\partial v}{\partial y} + \frac{\partial w}{\partial z} = 0 $$
The momentum equation in the x-direction is:
$$ \rho \left( \frac{\partial u}{\partial t} + u \frac{\partial u}{\partial x} + v \frac{\partial u}{\partial y} + w \frac{\partial u}{\partial z} \right) = -\frac{\partial P}{\partial x} + \rho g_x + \mu \nabla^2 u $$
Similar equations hold for the y and z directions. The energy equation is given by:
$$ \rho c \left( \frac{\partial T}{\partial t} + u \frac{\partial T}{\partial x} + v \frac{\partial T}{\partial y} + w \frac{\partial T}{\partial z} \right) = \frac{\partial}{\partial x} \left(k \frac{\partial T}{\partial x}\right) + \frac{\partial}{\partial y} \left(k \frac{\partial T}{\partial y}\right) + \frac{\partial}{\partial z} \left(k \frac{\partial T}{\partial z}\right) + S $$
where \( \rho \) is the density, \( u, v, w \) are the velocity components, \( P \) is the pressure, \( \mu \) is the dynamic viscosity, \( c \) is the specific heat, \( k \) is the thermal conductivity, \( T \) is the temperature, and \( S \) is the internal heat source due to latent heat release. The latent heat is handled by the enthalpy method:
$$ H = H_0 + \int_{0}^{T} C \, dT + (1 – f_s) L $$
where \( H \) is the enthalpy, \( C \) is the specific heat, \( f_s \) is the solid fraction, and \( L \) is the latent heat.
For the prediction of shrinkage porosity and cavities, I used the Niyama criterion. The Niyama criterion is defined as:
$$ \frac{G}{\sqrt{R}} < C_{\text{Niyama}} $$
where \( G \) is the local temperature gradient, \( R \) is the cooling rate, and \( C_{\text{Niyama}} \) is a critical value. For large steel casting parts, a critical value of 1.1 is commonly used. If the Niyama parameter is below this threshold, shrinkage porosity is likely to form. The temperature gradient \( G \) can be calculated from the temperature field, and the cooling rate \( R \) is the time derivative of the temperature. This criterion was utilized in ProCAST to display the porosity distribution in the casting.
I first ran the simulation for the original casting process design. The filling process was analyzed using the temperature field and velocity field results. At t=0 s, the molten metal began to enter the sprue. After 4 s, the metal filled the gating system and started to enter the mold cavity. The bottom gating system provided a smooth filling pattern with no severe splashing. At t=48 s, the filling percentage reached about 45%. The temperature distribution was relatively uniform, but the center of the casting was hotter than the surfaces in contact with the mold. At t=107 s, the filling was complete. The overall temperature ranged from 1575 °C to 1510 °C. No misrun or cold shut defects were observed in the temperature field.
The velocity field analysis revealed that the metal velocity in the sprue was about 2.5 m/s, and the velocity in the central ingate was about 1.3 m/s initially, while the lateral ingates had about 0.5 m/s. During 30% to 90% filling, the velocity in the cavity was stable at about 0.5 m/s. However, the velocity-time curves at the three ingates showed that during the first 30 s, the flow was turbulent. The central ingate velocity rapidly increased to 2.4 m/s, then dropped to 0.9 m/s, and later rose again to 1.5 m/s. This behavior was caused by the initial chilling of the molten metal on the cold runner walls, which reduced the effective cross-section, followed by remelting of the solidified layer as hot metal continued to flow. This unsteady flow could cause gas entrapment and oxidation at the lower part of the casting.
The solidification process was then analyzed. The simulation showed that when the filling was completed, 6.9% solid fraction already existed in the casting. At a solid fraction of 20%, the gating system was fully solidified, and the edges of the casting and risers began to solidify. The solidification continued from the outer surfaces toward the interior. At a solid fraction of 80% and 90%, the last solidification regions were observed around the bottom of the risers and the central boss. The total solidification time was about 37117 seconds. The temperature field during solidification indicated that the temperature gradient was not ideal. The riser tops solidified earlier than the riser bottoms, which closed the feeding channels and prevented adequate liquid metal supply to the thick sections. This inevitably led to shrinkage cavities and porosity.
The defect prediction using the Niyama criterion for the original design is shown in the results. Three large shrinkage cavities were found under the three risers, extending into the casting interior. In addition, two large shrinkage porosity regions were observed on both sides of the central riser. The total porosity volume was about 16% of the casting volume. These defects were unacceptable for the compressor support ring, which must withstand high pressure and temperature.
To eliminate these defects, I modified the casting process. First, I replaced the ordinary sand risers with exothermic insulating risers. I added a 50 mm thick FT400 exothermic insulating sleeve around each riser, and the central riser sleeve was 25 mm thick. The exothermic material reacts upon contact with the molten metal and releases heat, which significantly prolongs the liquid state of the riser and increases the feeding efficiency. Second, I added two pairs of chills on both sides of the central riser to accelerate cooling at the adjacent thick sections. The chills create a steeper temperature gradient toward the central riser and promote sequential solidification. The modified design is shown in the model layout.
After modifying the process, I re-meshed the new geometry and repeated the simulation. The filling time slightly increased to 117 s due to the increased thermal mass of the insulating sleeves. The temperature field during filling was more uniform, and the metal flow in the central region was slower and calmer because the chills lowered the local temperature and increased the viscosity. No turbulence or gas entrapment was observed.
The solidification sequence of the new design was greatly improved. At t=117 s, the solid fraction was 12%. The edges and thin walls solidified first, and the last regions to solidify were located inside the risers. The temperature profiles across the riser cross-sections showed a clear bottom-to-top gradient, which allowed the risers to feed the casting continuously until the end of solidification. The chills effectively accelerated cooling in the central boss area, eliminating the hot spot that had caused the porosity.
The Niyama defect prediction for the optimized design showed that the shrinkage cavities and porosity were almost completely eliminated from the casting body. Some porosity was predicted in the riser interiors and in the gating system, but these are removed after cutting off the risers. However, small shrinkage porosity regions still appeared at the bottom edge of the central boss and near the connection of the central ingate to the casting. To further reduce these defects, I decided to optimize the pouring process parameters using an orthogonal experiment.
Three key process parameters were selected: pouring temperature (A), pouring speed (B), and mold temperature (C). Each parameter was varied at three levels. The levels were chosen based on practical foundry experience. Table 3 lists the factor levels.
| Level | A: Pouring temperature (°C) | B: Pouring speed (kg/s) | C: Mold temperature (°C) |
|---|---|---|---|
| 1 | 1585 | 100 | 20 |
| 2 | 1575 | 90 | 25 |
| 3 | 1565 | 105 | 30 |
I selected the \( L_9(3^4) \) orthogonal array, which is appropriate for three factors with three levels. The empty column was used to estimate the experimental error. Table 4 shows the nine experimental schemes.
| Experiment No. | A | Empty | B | C |
|---|---|---|---|---|
| L1 | 1585 | 1 | 100 | 20 |
| L2 | 1585 | 2 | 90 | 25 |
| L3 | 1585 | 3 | 105 | 30 |
| L4 | 1575 | 1 | 90 | 30 |
| L5 | 1575 | 2 | 105 | 20 |
| L6 | 1575 | 3 | 100 | 25 |
| L7 | 1565 | 1 | 105 | 25 |
| L8 | 1565 | 2 | 100 | 30 |
| L9 | 1565 | 3 | 90 | 20 |
For each experiment, I performed a full filling and solidification simulation using ProCAST. After each simulation, I evaluated the shrinkage porosity volume in the casting body using the post-processing module. The porosity percentage, calculated as the volume of shrinkage defects divided by the total casting volume, was used as the response variable. The results are listed in Table 5.
| Experiment No. | Porosity (%) |
|---|---|
| L1 | 15.34 |
| L2 | 14.36 |
| L3 | 16.63 |
| L4 | 13.37 |
| L5 | 10.53 |
| L6 | 12.97 |
| L7 | 17.63 |
| L8 | 14.67 |
| L9 | 18.52 |
I performed a range analysis (also called the intuitive analysis) to determine the significance of each factor. For each factor level, I calculated the sum of the porosity percentages \( K_{ij} \) and the average \( k_{ij} \). The range \( R \) is the difference between the maximum and minimum average values. Table 6 summarizes the range analysis.
| Factors | A | B | C |
|---|---|---|---|
| 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 |
| R | 13.96 | 8.56 | 3.27 |
From the range values, I found that \( R_A > R_B > R_C \). Therefore, the pouring temperature had the greatest influence on the shrinkage porosity of the steel casting, followed by the pouring speed, and the mold temperature had the least influence. To minimize the porosity, I selected the level that gave the smallest \( k \) value for each factor. For factor A, the minimum \( k \) was 12.29 at level 2 (1575 °C). For factor B, the minimum \( k \) was 13.19 at level 1 (100 kg/s). For factor C, the minimum \( k \) was 14.33 at level 1 (20 °C). Thus, the optimal combination was \( A_2B_1C_1 \), i.e., pouring temperature 1575 °C, pouring speed 100 kg/s, and mold temperature 20 °C.
I then ran a final simulation using the optimal parameters. The defect prediction showed that almost no shrinkage porosity or cavities existed in the casting body. The porosity percentage was reduced to 8.4%, and the remaining porosity was confined to the risers above the cut-off line. The small defects near the central ingate connection and the bottom edge were also eliminated. The simulation confirmed that the optimized steel casting process produces high-quality compressor support rings.
To validate the simulation results, I carried out actual production runs at the foundry. The pattern, molding, melting, and pouring procedures strictly followed the optimized process parameters. The molten metal composition was checked by a direct-reading spectrometer and confirmed to be within the specification of ZG13Cr9Mo2Co1NiVNbNB. The pouring temperature was controlled at 1575 °C, the pouring speed at 100 kg/s, and the mold temperature at 20 °C. After pouring, the casting was allowed to cool slowly in the sand mold. The shakeout was then performed, and the risers were cut off. Figure 2 shows the riser appearance after cutting. The exothermic insulating risers had a lower height than ordinary risers, indicating less metal consumption. The cross-section of the riser revealed a sound internal structure with no shrinkage cavity, which proves that the riser feeding was effective.
The casting was then subjected to performance heat treatment in accordance with the specification for steel casting components. The heat treatment consisted of normalizing at 1110 °C ± 10 °C for a soaking time of 2 h plus 26.5 h, followed by tempering at 740 °C ± 10 °C for 12 h plus 28 h. After heat treatment, metallographic analysis was performed on test samples at magnifications of 200× and 400×. The microstructure consisted of tempered martensite, and the grain size was rated between 3 and 5. The metallographic results satisfied the requirements of the purchasing specification.
Finally, the casting was cleaned and machined for non-destructive testing (NDT). The inspection included visual testing (VT), magnetic particle testing (MT), ultrasonic testing (UT), and liquid penetrant testing (PT). The NDT results showed no macroscopic defects in the casting, which was consistent with the ProCAST simulation predictions. The optimized steel casting process successfully eliminated the shrinkage defects that had previously plagued the production of this compressor support ring.
In conclusion, my research demonstrated that the combination of sand casting simulation and orthogonal experimentation is an effective approach for optimizing the casting process of large and complex steel casting components. The use of ProCAST allowed me to identify the root causes of defects and to evaluate the benefits of using exothermic insulating risers and chills. The orthogonal experiment provided a systematic method to determine the best pouring parameters. The validated optimum parameters—pouring temperature 1575 °C, pouring speed 100 kg/s, and mold temperature 20 °C—ensure the production of sound steel casting parts with minimal shrinkage porosity. This methodology not only improves the quality of the compressor support ring but also reduces the cost and lead time associated with trial-and-error development. Future work may involve extending the material database for other heat-resistant steel casting alloys, adding microstructure simulation using the CAFE module, and studying the influence of solidification parameters on grain structure. Through continuous improvement, the steel casting industry can achieve higher efficiency and more reliable products for advanced power generation equipment.