1. Introduction
In the field of modern power generation equipment manufacturing, the quality of large steel castings directly determines the safety and service life of critical components. This study focuses on a heavy-duty gas turbine compressor support ring casting, which serves as a crucial component in thermal power plants. Compared with ordinary castings, this type of component exhibits larger dimensions, more complex structures, higher molten metal consumption, and the presence of thick wall sections. During the filling and solidification stages, establishing appropriate temperature gradients and controlling the solidification sequence becomes particularly challenging, making the casting highly susceptible to defects such as misruns, shrinkage porosity, shrinkage cavities, and hot tearing deformation. These casting defects not only reduce the mechanical properties and service performance of the final product but also significantly increase production costs and lead times.
The compressor support ring, with a maximum external dimension of 3070mm × 1175mm × 1462mm, weighs approximately 2.5 tons, with the thickest wall section reaching 368mm in thickness. The material selected for this component is ZG13Cr9Mo2Co1NiVNbNB (referred to as CB2), a new type of 9%–12% Cr ferritic heat-resistant steel developed during the European COST522 program. This material exhibits excellent high-temperature mechanical properties and creep resistance, serving under extreme conditions of 600–620°C and 30MPa. The primary purpose of the compressor support ring is to compress the air and gas entering the gas turbine, providing sufficient oxygen for the combustion chamber while enabling the entire system to achieve higher thermal efficiency through the Brayton cycle. Given the long-term exposure to high-temperature, high-pressure, and high-corrosion environments, the casting quality requirements for this component are exceptionally stringent.
Traditional trial-and-error methods for optimizing casting processes are time-consuming, costly, and inefficient. Numerical simulation technology, particularly through the application of commercial software such as ProCAST, offers an efficient alternative for predicting casting defects before actual production. By accurately modeling the filling and solidification processes, engineers can identify potential casting defects, optimize process parameters, and reduce the likelihood of defect formation. This study employs the finite element method (FEM) based ProCAST software to systematically investigate the casting process of the compressor support ring, aiming to identify, analyze, and eliminate casting defects while establishing an optimized casting process framework.
The overall research roadmap begins with the design of the casting process scheme, followed by the establishment of numerical simulation models, then optimization of the casting scheme based on simulation results, and finally validation through ortho-experimental designs and actual production trials. Figure 1 illustrates the automated pouring line used in the production verification stage.
2. Casting Process Design Principles
2.1 Overview of Sand Casting Process
Large steel casting production through sand casting involves a systematic procedure including pattern making, mold preparation, melting, pouring, shakeout, fettling, and inspection. The sand casting method offers several advantages including cost-effectiveness, flexible production capacity, and the ability to produce large and complex components. Within the foundry industry, sand casting accounts for approximately 60%–70% of all castings produced globally, with clay-bonded sand representing roughly 70% of that proportion. Additional factors that make sand casting suitable for compressor support ring production include its adaptability to various alloy systems, low tooling costs for single or small-batch production, and the capacity to produce extremely large components.
2.2 Gating System Design
The gating system, which is a critical component of the casting design, comprises the pouring cup, sprue, runner, and ingates. It serves to guide molten metal smoothly into the mold cavity while preventing slag, gas, and other impurities from entering the casting. In selecting the appropriate gating system for the compressor support ring, several factors were evaluated including alloy composition, casting geometry, pouring temperature, and solidification characteristics. A bottom gating system was selected because of its advantageous characteristics: smooth filling behavior, reduced splashing and air entrapment, and superior slag removal capability. However, this configuration inherently creates an unfavorable thermal gradient during solidification, requiring careful design of riser systems and sometimes chilling systems to compensate for the bottom-up temperature distribution.
The bottom gating system design incorporated three ingates connecting to the casting body, positioned at the lower section. The pouring cup was designed at a diameter of 265 mm with a sprue cross-sectional diameter of 80 mm. The system configuration includes a pouring cup, straight sprue, cross gate, and multiple ingates arranged carefully to distribute metal flow evenly. The gating ratio between sprue:runner:ingates can be expressed as:
$$ \frac{\sum F_{sprue}}{\sum F_{runner}:\sum F_{ingates}}=1:2:1.5 $$
To ensure complete filling of the mold before the metal loses excessive superheat, the filling time was calculated using the following relationship derived from fluid flow principles and mass conservation:
$$ t_f = \frac{V_m \cdot \rho_m}{MFR} $$
where $V_m$ is the mold cavity volume (m³), $\rho_m$ is the density of molten metal (kg/m³), and $MFR$ is the mass flow rate (kg/s). This calculation yielded a theoretical filling time of approximately 107 seconds.
2.3 Riser Design for Shrinkage Compensation
Riser design constitutes one of the essential elements in preventing shrinkage-related casting defects. In sand castings, risers function as reservoirs of molten metal that compensate for volumetric contraction during liquid and solid-state transformations. The modulus method, also known as Chvorinov’s rule, was selected for riser design, subsequently modified according to Caine’s and other researchers’ recommendations. The solidification time of a casting region can be expressed by Chvorinov’s rule:
$$ t_s = B\left(\frac{V}{SA}\right)^n $$
where $t_s$ represents the solidification time, $B$ is the mold constant, $V$ is the volume, $SA$ is the surface area, and $n$ is typically 2. The term $(V/SA)$ constitutes the casting modulus ($M$). For effective feeding, the riser modulus must satisfy the condition $M_r \geq 1.2 M_c$, where $M_r$ and $M_c$ represent the riser and casting modulus respectively. Three risers were positioned at the upper thick sections of the casting, with two waist-shaped risers measuring 750 mm × 250 mm positioned at the extremes and one circular riser of 450 mm diameter placed at the center. These risers were complemented by riser pads to maintain open feeding channels during solidification, thereby preventing premature channel closure and ensuring adequate metalavailable for shrinkage compensation.
Additionally, to enhance feeding efficiency and improve temperature gradient management, exothermic-insulating riser sleeves were applied in the optimized design phase. These sleeves reduce heat loss from the riser, maintain metal temperature longer, and substantially increase effective feeding distance. Empirical data indicates that exothermic-insulating riser sleeves can achieve feeding efficiencies of 30%–60%, significantly higher than conventional sand risers at 12%–15%. Table 1 summarizes the feeding efficiencies of various riser types.
| Riser Type | Feeding Efficiency (%) |
|---|---|
| Cylindrical/waist-shaped sand riser | 12–15 |
| Spherical riser | 15–20 |
| Refilled riser | 15–20 |
| Exothermic-insulating riser | 25–30 |
| Atmospheric pressure riser | 15–20 |
| Compressed-air riser | 35–40 |
| Gas explosion riser | 30–35 |
2.4 Chills for Thermal Gradient Control
Chills were strategically incorporated in the optimized process design to accelerate local cooling rates and modify the solidification sequence. External chills made of low-carbon steel were placed on both sides of the central boss region of the casting. Their function was threefold: first, they expand the effective feeding distance of the risers; second, they accelerate the cooling rate of the hot spot region, promoting directional solidification; and third, they refine the local microstructure, thereby improving mechanical properties. The chill dimensions were determined as 740mm × 210mm × 130mm and 217mm × 200mm × 156mm, positioned to directly address shrinkage porosity tendencies adjacent to the central riser. The thickness of the chills was calculated as 0.3–0.8 times the local hot spot thickness for steel castings, consistent with established industrial practice.
3. Numerical Simulation Methodology
3.1 ProCAST Software Overview
ProCAST was selected as the numerical simulation platform for this investigation. Developed initially by UES Inc. (USA) and subsequently acquired by ESI Group (France), ProCAST operates on the finite element method (FEM) to solve coupled fluid flow, heat transfer, and stress evolution equations. The software enables complete simulation of the casting process from mold filling through solidification to final cooling, incorporating a comprehensive suite of physical models including turbulent flow, free surface tracking, and microporosity prediction. The process flow in ProCAST involves several interlinked subroutines: Visual-Mesh for finite element mesh generation, Visual-Cast for parameter assignment, DataCast/ProCast solver for computation, and Visual-Viewer for post-processing and visualization. FEM was preferred over finite difference method (FDM) because of its superior capability in handling complex geometrical features and curved boundaries characteristic of large castings.
3.2 Theoretical Fundamentals
For accurate simulation results, the calculation was performed using the fundamental governing equations of fluid dynamics and heat transfer. During filling, molten metal flow was assumed incompressible and Newtonian, with laminar flow modeled through the following equations:
Continuity equation:
$$ \nabla \cdot \mathbf{V} = \frac{\partial u}{\partial x} + \frac{\partial v}{\partial y} + \frac{\partial w}{\partial z} = 0 $$
Momentum conservation (Navier-Stokes equations):
$$ \rho\left(\frac{\partial u}{\partial t} + u\frac{\partial u}{\partial x} + v\frac{\partial u}{\partial y} + w\frac{\partial u}{\partial z}\right) = -\frac{\partial P}{\partial x} + \rho g_x + \mu\nabla^2 u $$
$$ \rho\left(\frac{\partial v}{\partial t} + u\frac{\partial v}{\partial x} + v\frac{\partial v}{\partial y} + w\frac{\partial v}{\partial z}\right) = -\frac{\partial P}{\partial y} + \rho g_y + \mu\nabla^2 v $$
$$ \rho\left(\frac{\partial w}{\partial t} + u\frac{\partial w}{\partial x} + v\frac{\partial w}{\partial y} + w\frac{\partial w}{\partial z}\right) = -\frac{\partial P}{\partial z} + \rho g_z + \mu\nabla^2 w $$
Energy conservation:
$$ \rho c_p\left(\frac{\partial T}{\partial t} + u\frac{\partial T}{\partial x} + v\frac{\partial T}{\partial y} + w\frac{\partial T}{\partial z}\right) = \frac{\partial}{\partial x}\left(k\frac{\partial T}{\partial x}\right) + \frac{\partial}{\partial y}\left(k\frac{\partial T}{\partial y}\right) + \frac{\partial}{\partial z}\left(k\frac{\partial T}{\partial z}\right) + S $$
For the solidification stage, conduction heat transfer through the casting and mold was computed using Fourier’s law with the latent heat of fusion handled through the enthalpy method:
$$ H = H_0 + \int_0^T C \, dT + (1 – f_s)L $$
where $H$ represents enthalpy, $C$ is specific heat capacity, $f_s$ is solid fraction, and $L$ is latent heat of fusion. Thermal radiation at exposed surfaces was included using the Stefan-Boltzmann law:
$$ q = \varepsilon \sigma_0 T_s^4 $$
3.3 Material and Boundary Conditions
The chemical composition of ZG13Cr9Mo2Co1NiVNbNB alloy used in this study follows the specification given in Table 2. This composition was verified to satisfy the CB2 steel standard.
| Element | C | Si | Mn | P | S | Cr | Ni | Mo | Co | V | Nb | B | N | Al |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Specification | 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 |
Because the ProCAST database does not include this specific alloy, thermal-physical and mechanical properties were computed internally using the Lever rule solidification model. The liquidus temperature was computed as 1494°C, and solidus temperature as 1186°C. From the energy balance calculations, the solidification shrinkage of the alloy was estimated at approximately 3-4% by volume.
For the mold material, silica sand (quartz sand) was selected. The interface heat transfer coefficients between different material pairs were assigned as follows: molten metal/sand: 500 W/(m²·K); chill/casting: 2000 W/(m²·K); chill/sand: 500 W/(m²·K); insulation sleeve/sand: 50 W/(m²·K). These values, listed in Table 3, follow established practices derived from previous experimental validation studies and industrial experience.
| Interface | Heat Transfer Coefficient [W/(m²·K)] |
|---|---|
| Casting-Mold | 500 |
| Casting-Chill | 2000 |
| Chill-Mold | 500 |
| Casting-Air | 10 |
| Mold-Air | 10 |
| Sleeve-Mold | 50 |
3.4 Mesh Generation
3D modeling of the casting, gating system, risers, chills, and insulation sleeves was performed in NX 11.0 software. The completed assembly was exported in IGES file format and imported into Visual-Mesh for finite element discretization. The sand mold box measured 4500mm × 2500mm × 2700mm. Mesh generation involved two sequential steps: first, surface mesh generation followed by volume mesh generation using tetrahedral elements. Different mesh densities were employed based on the required computational accuracy and component importance—the mold box was meshed at 100mm, the casting body at 30mm, and the gating system and chill components at 10-20mm resolution. The original scheme discretization produced approximately 70,000 surface elements and 2.14 million volume elements, totaling roughly 2.23 million elements. After optimization, the new scheme contained 53,033 surface elements and 1,162,161 volume elements. Mesh quality checks confirmed zero negative Jacobian determinants, ensuring acceptable element quality for numerical computation.
4. Simulation Results and Analysis of Original Scheme
4.1 Filling Process Simulation
The filling process simulation of the original casting scheme revealed critical information regarding metal flow behavior and temperature distribution. Temperature field evolution during mold filling is shown in the temporal sequence of snapshots. At the initial stage (t=0s), molten metal began entering the sprue from the pouring cup. At t=4s, the gating system was completely filled and metal started entering the mold cavity. By t=24s, the filling level reached approximately 20%, with the molten metal exhibiting stable flow through the bottom gating system with minimal turbulence. The filling process proceeded uniformly, with the metal front advancing steadily upward.
At t=48s, the filling fraction reached approximately 45%. The temperature field showed higher temperatures in the central regions of the casting while lower temperatures were recorded near the mold walls, correctly reflecting the heat transfer process to the surrounding sand. As filling continued to 78s, the temperature distribution became more stratified, establishing a bottom-to-top thermal gradient. By the completion of filling at t=107s, the casting was completely filled with the temperature field showing a favorable vertical gradient from 1510°C at the bottom to 1575°C near the top, theoretically supporting upward directional solidification and riser feeding.
However, the velocity vector analysis revealed concerning patterns. During the early filling phase (up to approximately 30% fill), the molten metal velocity at the central ingate exhibited significant fluctuations. The velocity at the central ingate reached 2.4 m/s momentarily before dropping to 0.9 m/s, then rising again to stabilize at approximately 1.5 m/s. This initial velocity fluctuation pattern suggests possible turbulent flow conditions, which could promote gas entrainment and oxidation at the lower portion of the casting. The two side ingates showed lower velocities of 0.5 m/s initially, gradually increasing and stabilizing at 1.0 m/s. The velocity fluctuation at the central ingate was likely caused by initial temperature-induced solidification of a thin metal layer on the gate walls, followed by remelting and flushing of the solidified layer. This phenomenon can contribute to the formation of oxide inclusions and gas porosity, which are common casting defects.
4.2 Solidification Process Simulation
The solidification simulation of the original scheme provided significant insight into the casting’s freezing behavior. Solid fraction evolution tracking showed that by the completion of filling, the casting had already achieved 6.9% solidification, predominantly in thin-walled sections and regions in proximity to the mold walls. When the solid fraction reached 20%, the gating system had completely solidified, while the casting edges and riser peripheries had started freezing. At higher solid fractions, the solidification proceeded progressively from the external surfaces toward the interior.
At 50% solidification (9774s), the arc center and the regions beneath the side risers remained molten. The final solidification sequence, as revealed by the simulation, showed that several regions not associated with the risers were the last to freeze. The bottom of both side riser connections with the casting and the base platform beneath the central riser remained at higher temperatures compared to the rest of the casting. This created an undesirable thermal profile where the riser tops solidified before their bottoms, effectively closing off the feeding channels and preventing the risers from performing their intended compensation function.
Temperature field analysis during solidification further confirmed these concerns. At 5074s, the temperature across the casting ranged from 1494°C at the hot spots to 1186°C at the most rapidly cooled sections. The solidification time was calculated as 37424s, with the total solidification duration being about 37117s. The prolonged solidification time at the hot spots, combined with the ineffective riser feeding, created ideal conditions for shrinkage cavity formation at multiple casting locations.
To quantitatively analyze the solidification behavior at critical locations, five observation nodes were established on the casting. The cooling curves and solid fraction evolution at these nodes showed that nodes 4 and 5, representing thin-wall regions, initiated solidification much earlier (before 2000s) and completed solidification relatively quickly. In contrast, nodes 1, 2, and 3 (all associated with thick-walled areas or riser connections) remained molten for significantly longer periods. At 7000s, nodes 4 and 5 exceeded 50% solidification while nodes 1–3 had barely begun their solidification process. This divergent solidification pattern without proper riser feeding created the conditions for solidification shrinkage to manifest as volumetric contraction defects.
4.3 Defect Prediction for the Original Scheme
The Niyama criterion was selected as the porosity prediction parameter because of its well-established performance in predicting shrinkage porosity in steel castings. The dimensionless Niyama criterion was evaluated at each node using the formula:
$$ \frac{G}{\sqrt{R}} < C_{Niyama} $$
where $G$ is the local temperature gradient, $R$ is the cooling rate, and $C_{Niyama}$ is the critical value, typically set between 0.8 and 1.1. Values falling below the critical threshold indicate the formation of shrinkage porosity or shrinkage cavities. The defect prediction results of the original scheme identified three significant shrinkage regions (macroscopic cavities) beneath the three risers, extending downward into the casting body. Additionally, two prominent shrinkage porosity zones were identified on both sides of the central boss area. The total porosity volume reached approximately 16% of the casting volume. These regions coincide directly with the locations where improper temperature gradients — the absence of true directional solidification — produced conditions for shrinkage formation. The identified casting defects were consistent with the observed premature solidification of the riser tops and the resulting feeding channel obstruction.
5. Optimization of Casting Process and Simulation Verification
5.1 Process Modifications
The comprehensive analysis of the original casting scheme enabled the formulation of targeted corrective measures. The primary casting defects — shrinkage cavities and porosity — resulted from two principal shortcomings: improper thermal gradient distribution in the upper zones and flow velocity fluctuations during early filling. To address these issues comprehensively, the following modifications were implemented in the optimized casting process plan:
1) Replacement of conventional sand risers with exothermic-insulating risers: All three risers were surrounded by a 50 mm thick FT400 exothermic-insulating material, with the central riser sleeve thickness set at 25mm. This modification slows the cooling rate of the riser metal, extends the liquid feeding duration, and establishes a favorable temperature gradient that promotes upward directional solidification of the riser itself while maintaining the feeding channel open for a longer period.
2) Addition of external chills: Two pairs of chills were placed on opposite sides of the central boss. Their primary function is to accelerate local solidification by absorbing heat, lowering the local metal temperature, and consequently reducing metal fluidity and flow velocity at this critical location. This reduces the likelihood of gas entrapment and oxidation from turbulent flow during filling while simultaneously adjusting the local solidification rate.
3) Chill material: Low-carbon steel was selected as the chill material because of its high thermal diffusivity and compatibility with the casting alloy system.
5.2 Filling Simulation of Optimized Scheme
With the modified process parameters, the filling simulation was repeated. The filling time increased slightly to 117s due to the additional thermal mass of the chills. During filling, the metal flowed more smoothly through the ingates. The velocity profile at the critical locations was significantly improved, with reduced initial fluctuations and more stable filling behavior overall. The temperature field distribution at various filling times demonstrated more uniform temperature gradients and reduced temperature variation across the casting cross-section. The absence of gas entrapment regions suggested improved metal quality, potentially containing fewer oxidized inclusions. The improved velocity control was particularly evident in the region between the chills, where the combination of heat absorption and slowed flow significantly reduced turbulence-associated risks.
5.3 Solidification Simulation of Optimized Scheme
In the optimized scheme solidification simulation, a vertical cross-section was added through the entire assembly to allow direct visualization of the temperature and solid fraction evolution within the risers. At the completion of filling, the casting had achieved 12% solidification, with thin sections and the chill-adjacent regions beginning to freeze. This higher initial solidification fraction, compared with the original scheme, indicates that the chills were effectively extracting heat during the filling stage.
The solid fraction evolution at 60% solidification demonstrated a well-established progressive solidification pattern: thin sections solidified first, followed progressively by the thick sections, with the remaining liquid metal consistently contained within the risers. The final solidification regions — the last pockets of liquid to solidify — were entirely located within the riser cavities. This is the essential condition for effective feeding: a continuous liquid feeding channel from the riser to the solidification front. The chills successfully created a lower temperature zone between the central riser and the boss, contributing to proper solidification directionality.
The cross-sectional temperature profiles of the risers during the intermediate solidification stages confirmed ideal temperature distributions. The temperature at the riser bottom exceeded that at the riser top, establishing precisely the temperature gradient required for upward directional solidification and ensuring complete utilization of the available liquid metal for feed purposes.
5.4 Defect Prediction of the Optimized Scheme
The Niyama criterion-based defect prediction for the optimized process showed a dramatic improvement. The shrinkage cavities that existed in the original casting were eliminated entirely. The residual porosity indicators were confined to the internal riser cavities and portions of the gating system — regions removed during subsequent finishing operations. The casting body itself was completely free of shrinkage defects, confirming that the redesigned riser system effectively achieved its functional purpose. However, the simulation still identified a few localized zones of minor shrinkage porosity at the bottom pedestal edges of the casting and at the connection between the central ingate and the casting. These residual casting defects, though significantly less severe, provided the motivation for further process parameter optimization through the orthogonal experiment design methodology.
6. Orthogonal Experiment Design for Parameter Optimization
6.1 Design of Experiments
The Taguchi method, implemented as an orthogonal experiment design, was adopted to systematically investigate the influence of key casting process parameters on porosity formation and to identify the optimal parameter set. The selection was based on three factors identified as most influential in controlling the filling and solidification process of this casting: pouring temperature (A), pouring velocity (B), and sand mold initial temperature (C). Each factor was assigned three levels selected based on industrial experience and the physical constraints of the alloy system:
| Level | (A) Pouring Temperature (°C) | (B) Pouring Velocity (kg/s) | (C) Sand Mold Temperature (°C) |
|---|---|---|---|
| 1 | 1585 | 100 | 20 |
| 2 | 1575 | 90 | 25 |
| 3 | 1565 | 105 | 30 |
The L9(3³) orthogonal array was selected as the smallest orthogonal table capable of evaluating three factors at three levels. The experiment array is shown in Table 5, where each row represents one unique combination of process parameters.
| Experiment No. | A | Empty | B | C | Process Parameter Combination |
|---|---|---|---|---|---|
| L1 | 1585 | 1 | 100 | 20 | A1B1C1 |
| L2 | 1585 | 2 | 90 | 25 | A1B2C2 |
| L3 | 1585 | 3 | 105 | 30 | A1B3C3 |
| L4 | 1575 | 1 | 90 | 30 | A2B2C3 |
| L5 | 1575 | 2 | 105 | 20 | A2B3C1 |
| L6 | 1575 | 3 | 100 | 25 | A2B1C2 |
| L7 | 1565 | 1 | 105 | 25 | A3B3C2 |
| L8 | 1565 | 2 | 100 | 30 | A3B1C3 |
| L9 | 1565 | 3 | 90 | 20 | A3B2C1 |
6.2 Simulation Results and Analysis
All nine experimental combinations were simulated using ProCAST with the optimized casting scheme geometry. After each simulation, the shrinkage porosity volume fraction was computed from the Niyama criterion results and served as the sole quantitative response for determining the optimal parameter combination. The resulting porosity fractions for each test combination are summarized in Table 6.
| Experiment No. | A | B | C | Porosity Fraction (%) |
|---|---|---|---|---|
| L1 | 1 | 1 | 1 | 15.34 |
| L2 | 1 | 2 | 2 | 14.36 |
| L3 | 1 | 3 | 3 | 16.63 |
| L4 | 2 | 1 | 3 | 13.37 |
| L5 | 2 | 2 | 1 | 10.53 |
| L6 | 2 | 3 | 2 | 12.97 |
| L7 | 3 | 1 | 2 | 17.63 |
| L8 | 3 | 2 | 3 | 14.67 |
| L9 | 3 | 3 | 1 | 18.52 |
6.3 Range (Extreme Difference) Analysis
Range analysis, also referred to as extreme difference analysis or the intuitive analysis method, was applied to interpret the orthogonal experiment results. For each parameter level, the average porosity fraction was computed, and the range (R) for each factor was determined as the difference between the maximum and minimum average values. A larger R value indicates a stronger influence of that factor on the response variable. The calculations are presented in Table 7.
| Statistics | 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 |
Where $K_i$ represents the sum of porosity fractions for each factor at level $i$, and $k_i$ denotes the average value ($k_i = K_i/3$). Based on the range analysis, the order of significance of the three process parameters on porosity formation was determined as: pouring temperature (A) > pouring velocity (B) > sand mold temperature (C). Since the objective was porosity minimization, the optimal parameter combination corresponds to the lowest average porosity values for each factor, which yields the combination A2B1C1.
6.4 Verification of Optimized Process Parameters
The final optimized parameters were confirmed as: pouring temperature 1575°C, pouring velocity 100 kg/s, and sand mold temperature 20°C. A verification simulation was then conducted using these parameter values. The resulting defect prediction confirmed a substantial reduction in porosity — the porosity fraction decreased to approximately 8.4%. The residual shrinkage defects were entirely confined to the upper regions of the riser cavities, well above the critical sections of the casting body. The small shrinkage porosity regions identified earlier in the base platform and at the central ingate connection were completely eliminated. These results demonstrate that the selected combination of process parameters achieves superior casting quality, with complete elimination of casting defects from the returned product while maintaining all defects safely within the removed riser and gating materials.
7. Industrial Production Validation
7.1 Production Trials
To validate the simulation findings and verify the effectiveness of the optimized casting scheme, an actual production run was carried out. The sand mold was prepared to the optimized casting design using silica sand bonded with resin. The CB2 alloy was melted in a medium-frequency induction furnace followed by VOD refining. Direct-reading spectroscopy of the molten metal before pouring confirmed that the chemical composition satisfied the specification requirements for ZG13Cr9Mo2Co1NiVNbNB within the allowable tolerance range. The pour was executed under the exact optimal process parameters obtained from the orthogonal experiment design: pouring temperature 1575°C, pouring flow rate 100 kg/s, and sand mold temperature 20°C. Following pouring, the casting was left to cool slowly in the mold to minimize residual stresses and thermal distortion.
7.2 Inspection and Testing
After shakeout and initial cleaning, the risers were removed. The removed riser sections displayed characteristic features of effective feeding: the exothermic-insulating riser had a lower final height than an equivalent sand riser, reflecting both metal conservation and adequate feeding capacity, and the interior surfaces of the risers showed uniform solidification without shrinkage defects. The riser profile confirmed the proper functioning of the feeding system.
After performance heat treatment (normalizing at 1110°C ± 10°C followed by tempering at 740°C ± 10°C), metallographic examination was performed on test coupons. The microstructure observed under both 200× and 400× magnifications showed tempered martensite with grain sizes of 3–5 grades, matching specification requirements. Following secondary fettling and rough machining, the casting underwent comprehensive non-destructive testing per the purchaser’s specification. Visual testing (VT) showed a clean surface without visible casting defects. Magnetic particle testing (MT) confirmed the absence of surface and near-surface discontinuities, while ultrasonic testing (UT) verified the internal integrity of the casting, revealing no internal shrinkage, gas porosity, or other volumetric casting defects. Finally, penetrant testing (PT) further confirmed that all examined surfaces were free of superficial cracks and pinholes. Collectively, the non-destructive test results matched the ProCAST simulations remarkably well, confirming that the optimized process successfully eliminated the casting defects that were present in the original scheme and achieving full conformity with procurement specifications.
8. Conclusions
The systematic investigation combining numerical simulation, design of experiments, and industrial production validation has yielded the following conclusions:
(1) The original bottom-gated casting process for the compressor support ring exhibited significant casting defects due to premature solidification of riser tops, closure of the feeding channels during the critical final stages of solidification, and turbulent metal flow during the initial filling period. The Niyama criterion analysis of the original scheme predicted three shrinkage cavities and two substantial shrinkage porosity regions within the casting body.
(2) The optimized casting process, featuring exothermic-insulating risers and chilled regions at the critical boss location, transformed the solidification behavior. The chills accelerated local solidification, achieving early closure of thick sections and enabling true directional solidification from thin sections toward the risers. The exothermic risers maintained metal temperature, preserved the feeding channel open, and allowed complete metal compensation for volumetric contraction. Simulation of the optimized scheme confirmed complete elimination of shrinkage cavities within the casting, with residual defects confined to the removable riser and gating materials.
(3) Three-parameter, three-level orthogonal experiments established the influence hierarchy: pouring temperature had the greatest effect, followed by pouring velocity, with sand mold temperature having the least effect. The optimal process parameter combination was determined as pouring temperature 1575°C, pouring velocity 100 kg/s, and sand mold temperature 20°C. The verification simulation with these parameters achieved the lowest porosity fraction of 8.4%, with defects confined to the riser area.
(4) Actual industrial production run with the optimized parameters validated all simulation predictions. The as-cast compressor support ring passed all relevant non-destructive inspection procedures, including visual test, magnetic particle inspection, ultrasonic inspection, and dye penetrant examination. The manufactured casting was fully compliant with procurement specifications, confirming the efficiency and accuracy of ProCAST simulation for process design and defect prevention in large heat-resistant steel castings. More broadly, this study establishes a methodological framework applicable to the casting design of other large components for power generation equipment.