In this research, I systematically investigated the casting process design and optimization of an aluminum alloy turbine volute used in an aerospace turbo engine. The component is characterized by thin walls, complex internal hollow channels, and multiple functional bosses, which make the casting process extremely challenging. The primary objective was to predict and minimize casting defects such as shrinkage porosity, gas porosity, cold shuts, and misruns through numerical simulation using the ProCAST software, and then to verify the optimized process through actual production and heat treatment experiments.
The study began with a detailed geometric analysis of the turbine volute. The maximum overall dimensions were 167 mm × 150 mm × 160.35 mm, with a maximum diameter of 144 mm and a minimum wall thickness of only 3 mm. The internal structure consists of several intersecting hollow passages, forming a complex network that serves the turbocharging function. Because of this complexity, conventional trial-and-error casting process design would be time-consuming and costly. Therefore, I employed computer-aided engineering (CAE) to simulate both the mold filling and solidification stages, aiming to identify potential casting defects before committing to physical tooling.
The CAD model of the part was created using Unigraphics NX 12.0. The model was then imported into ProCAST for meshing and simulation. The alloy selected was ZL105A, a high-strength aluminum-silicon-copper-magnesium alloy. The nominal chemical composition is listed in Table 1.
| Element | Si | Cu | Mg | Ti | Al |
|---|---|---|---|---|---|
| Mass fraction (%) | 4.50–5.50 | 1.00–1.50 | 0.40–0.60 | 0.08–0.20 | Balance |
Table 1. Designed chemical composition of ZL105A aluminum alloy.
I used the lever rule solidification model available in ProCAST to calculate the temperature-dependent thermophysical properties of the alloy. The calculated liquidus temperature was 625.9°C and the solidus temperature was 551.6°C. The temperature–solid fraction curve indicated a wide freezing range, which is typical for this alloy and requires careful feeding design to avoid shrinkage porosity.
For the initial process design, I chose a bottom-gated runner system with a sprue and two horizontal runners feeding the casting from the bottom. This configuration is often preferred for complex thin-wall castings because it promotes smooth filling and directional solidification from the bottom upward. The pouring temperature was set to 800°C, and four symmetrical risers were placed on the top of the casting to feed the solidification shrinkage. The cross-sectional area of each riser was 1761.65 mm². The geometry of the initial gating system is shown schematically in the figure below.
The bottom-gated system was designed to allow the melt to rise uniformly, but the initial simulation revealed several issues. The filling simulation at 2.17 seconds showed that the melt reached the runner end and began entering the cavity. At 3.08 seconds, the melt velocity at the ingate was excessively high, causing direct impingement on the bottom of the casting and visible gas entrainment. This is a significant source of casting defects, as gas entrapment leads to porosity. By 4.53 seconds, the melt front became uneven and tilted, indicating non-uniform filling. Such unstable filling can cause cold shuts and oxide film entrapment, further increasing casting defects.
The solidification simulation of the initial design showed that the thin central wall solidified first, as expected, while the thicker protruding pipe ports solidified much later. This caused the central region to shrink away from the still-liquid thicker sections, leading to poor feeding and potential shrinkage defects. The lower three protrusions also solidified slower than the central thin wall, making them susceptible to shrinkage porosity. These observations correlated well with the Niyama criterion predictions.
I applied the Niyama criterion to predict micro-shrinkage porosity. The criterion is based on the local temperature gradient \(G\) and cooling rate \(R\) as
$$Ny = \frac{G}{\sqrt{R}}$$
where \(G\) is the temperature gradient and \(R\) is the cooling rate. Lower values of Ny indicate a higher risk of shrinkage porosity. The simulation results showed three major zones of high porosity probability. The risers themselves showed shrinkage, which is acceptable. However, the intersection of the internal hollow structures in the center of the casting exhibited a high porosity probability because this region is thicker than the surrounding thin walls and is remote from the risers and gates. Additionally, an annular shrinkage porosity distribution appeared at the bottom of the casting, which is the thickest section of the part.
The total shrinkage porosity simulation, based on the Chvorinov modulus method, was used to predict macro-shrinkage cavities. The equation for the local modulus is
$$M = \frac{V}{A} = K \left( \frac{T_{p} – T_{mold}}{\Delta H_{sol}} \right)^{1/2} \sqrt{t_{sol}}$$
where \(V\) is the volume, \(A\) is the surface area, \(K\) is a constant, \(T_{p}\) is the pouring temperature, \(T_{mold}\) is the mold temperature, \(\Delta H_{sol}\) is the latent heat of solidification, and \(t_{sol}\) is the solidification time. The simulation predicted a large shrinkage cavity at the center of the casting and an annular cavity at the protruding pipe port. These regions would be extremely difficult to feed because they solidify late and are far from the risers.
To address the casting defects identified in the initial simulation, I devised several process modifications. First, I added two auxiliary side runners to the horizontal runners. These side runners act as additional ingates to distribute the melt more evenly and reduce the velocity of the aluminum liquid entering the cavity. I used a control variable method to determine the optimal thickness of the side runners. I simulated thicknesses of 2, 3, 4, 5, and 6 mm. The results at 2.5 seconds showed that 2 mm and 3 mm had insufficient flow to stabilize the filling, while 5 mm and 6 mm caused the side gates to fill faster than the main gate, disturbing the intended flow pattern. A thickness of 4 mm provided the most balanced flow, with the main gate and auxiliary gates filling simultaneously and smoothly. Therefore, the side runners were designed as oblique trapezoids with a contact surface of 20 mm × 4 mm.
Second, I optimized the pouring temperature. The initial 800°C was too high, leading to excessive shrinkage and energy consumption. I tested temperatures from 640°C to 800°C in increments of 20°C. Below 720°C, the melt displayed poor fluidity and signs of premature solidification, especially at the thin wall sections. At temperatures above 720°C, filling was satisfactory, but higher temperatures increased the total shrinkage. Considering both quality and energy efficiency, I selected 740°C as the optimal pouring temperature. The mass flow rate was recalculated to 0.6097 kg/s.
Third, I added an additional riser specifically for the large protruding pipe port. This riser had a cross-sectional area of 2310.31 mm², which was larger than the original risers, to ensure adequate feeding of the thick section. The gating system dimensions were also refined using analytical calculations. The sprue entrance diameter was set to 64.5 mm, the vertical sprue diameter to 30 mm, and the horizontal runner cross-section to 31.45 mm × 31.45 mm. These values were derived to minimize excessive metal consumption while maintaining sufficient filling rates.
Fourth, I placed nine chillers at the hot spots identified by the simulation. The chillers were made of a high-thermal-conductivity material and were embedded in the sand mold in direct contact with the casting. Their purpose was to accelerate local solidification and promote directional solidification. The locations included the lower protruding structures and the external card holes, where the thermal center was likely to form. The interface heat transfer coefficients were set to 2000 W/(m²·K) for the casting-chiller interface and 800 W/(m²·K) for the mold-chiller interface.
After implementing these modifications, I re-ran the ProCAST simulation. The improved filling process showed a completely different behavior. At 1.32 seconds, the melt filled the runner system and began entering the cavity from both the main and auxiliary ingates. At 2.74 seconds, the four streams merged gently without any sloshing or gas entrainment. The melt front remained horizontal and rose steadily. At 4.14 seconds, the front reached the maximum diameter of the part, and the speed decreased appropriately but the front remained stable. By 7.93 seconds, the main body was filled, and the risers began to fill. The filling completed at 9.57 seconds with a uniform, laminated flow pattern, which is characteristic of ideal filling.
The solidification simulation of the improved process showed that the chillers effectively accelerated the cooling of the previously problematic lower protrusions. By 435.5 seconds, these regions had solidified almost simultaneously with the central thin walls. The new riser successfully fed the large protruding pipe port, preventing the formation of a shrinkage cavity there. The overall solidification pattern became truly directional, with the casting solidifying from the thinner sections toward the risers. This is the desired condition for minimizing casting defects.
The Niyama criterion results after improvement showed a dramatic reduction in shrinkage porosity probability. The large annular defect at the bottom disappeared, and the central internal junction area showed only minimal residual risk. In the casting body, the area of shrinkage porosity probability greater than 15% decreased by more than 70% compared with the initial design. The total shrinkage porosity simulation showed that all macro-shrinkage cavities were relocated to the risers and the gating system, which are removed after casting. The shrinkage cavity area in the casting was reduced by more than 90%.
A comparison of the initial and optimized process parameters is summarized in Table 2.
| Parameter | Initial process | Optimized process |
|---|---|---|
| Pouring temperature (°C) | 800 | 740 |
| Number of risers | 4 | 5 |
| Riser cross-section area (mm²) | 1761.65 | 1761.65 and 2310.31 |
| Number of side runners | 0 | 2 |
| Side runner thickness (mm) | – | 4 |
| Number of chillers | 0 | 9 |
| Sprue entrance diameter (mm) | – | 64.5 |
| Vertical sprue diameter (mm) | – | 30 |
| Horizontal runner cross-section (mm×mm) | – | 31.45×31.45 |
Table 2. Comparison between initial and optimized casting process parameters.
To validate the simulation results, I produced the sand molds using a 3D printing technique. The molds were printed with alkali phenolic resin sand, which offers excellent dimensional accuracy and is suitable for complex cores. After assembly, the molds were placed in a gravity pouring station. The aluminum melt was prepared at 740°C and poured within 10 seconds. The pouring was controlled to ensure a steady flow. After pouring, the mold was left undisturbed for about 20 hours to allow complete cooling and solidification, preventing thermal shock cracking.
After cooling, the sand mold was removed. For the internal hollow channels, the sand cores were burned out in a muffle furnace to avoid damaging the thin walls. The cast part was then cleaned and cut off from the gating system. The resulting turbine volute blank is shown in a physical section view. The section revealed no macroscopic internal defects, and the material appeared uniformly dense.
I subjected the cast parts to X-ray inspection using a flat-panel rotating cabinet X-ray machine. The initial process samples showed clear indications of misrun in the thin-wall areas and gas porosity in the complex junction areas. These casting defects were eliminated in the optimized samples. The X-ray images of the optimized parts from the front, top, and left views showed no discernible porosities, cracks, or inclusions. This confirmed that the optimized process successfully eliminated the casting defects predicted by the simulation.
The alloy composition of the cast parts was verified by optical emission spectroscopy. The measured composition is given in Table 3.
| Element | Si | Cu | Mg | Ti | Fe | Zn | Mn | Al |
|---|---|---|---|---|---|---|---|---|
| Measured (wt%) | 5.032 | 1.135 | 0.506 | 0.114 | 0.063 | 0.004 | 0.001 | Balance |
Table 3. Measured chemical composition of the cast ZL105A alloy.
The composition met the requirements of GB/T 16865-2013. The low iron content was intentional to improve ductility, and the titanium was added for grain refinement.
To enhance the mechanical properties of the casting, I applied a T5 heat treatment, which consists of solution treatment followed by artificial aging. The heat treatment parameters were optimized through a series of experiments. I prepared three groups of tensile specimens, labeled S1, S2, and S3, which were solution treated at 535°C for 2, 4, and 6 hours, respectively. The heating rate was controlled at ≤3°C/min to avoid local overheating. After solution treatment, the specimens were quenched in water at 25°C within approximately 25 seconds. The aging treatment was performed at 170°C for 8 hours, followed by air cooling to room temperature.
Tensile tests were carried out using a SHT4305 servo-hydraulic universal testing machine with a stress rate of 7 MPa/s. The results are summarized in Table 4.
| Batch | Tensile strength before heat treatment (MPa) | Tensile strength after heat treatment (MPa) | Yield strength after heat treatment (MPa) | Elongation after fracture (%) |
|---|---|---|---|---|
| S1 (2h solution) | 183.3 | 239.6 | 179.3 | 3.7 |
| S2 (4h solution) | 181.9 | 279.8 | 195.7 | 2.2 |
| S3 (6h solution) | 184.1 | 333.6 | 274.1 | 1.7 |
Table 4. Tensile properties of specimens before and after T5 heat treatment.
Only the S3 batch, which received a 6-hour solution treatment, met the national standard requirement of a minimum tensile strength of 280 MPa. The tensile strength increased by 81.2% compared to the as-cast condition, reaching 333.6 MPa. The yield strength also increased substantially to 274.1 MPa. However, the elongation decreased, indicating the heat treatment increased brittleness, which is typical for this precipitation-hardening alloy.
Brinell hardness tests were performed using an HB-3000C digital hardness tester with a 5 mm hard steel ball and a load of 1226 N. The results are shown in Table 5.
| Batch | Hardness before heat treatment (HB) | Hardness after heat treatment (HB) | Improvement (%) |
|---|---|---|---|
| S1 | 58.31 | 92.65 | 58.9 |
| S2 | 59.02 | 103.48 | 75.3 |
| S3 | 58.63 | 107.24 | 82.9 |
Table 5. Brinell hardness of specimens before and after heat treatment.
All heat-treated batches exceeded the national standard minimum hardness of 80 HB. The hardness increased significantly with increasing solution treatment time, but the difference between 4 hours and 6 hours was relatively small. Therefore, I selected 6 hours as the final solution treatment time to ensure complete dissolution of the eutectic phases and maximum hardening response.
Fracture surface analysis was performed using a scanning electron microscope (SEM). The macroscopic fracture surfaces of all specimens showed no necking, and the fracture plane was flat and perpendicular to the tensile axis, indicating brittle fracture. The microscale observation revealed the presence of cleavage facets and tear ridges. Some tiny dimples were found on the tear ridges, but they were shallow. In the as-cast specimen, the fracture surface contained smooth regions that were likely associated with pre-existing shrinkage porosity. The heat-treated S1, S2, and S3 specimens exhibited more distinct cleavage facets and fewer dimples, confirming that the T5 treatment increased the strength by sacrificing some ductility.
Metallographic observations were conducted using an optical microscope (OM). The as-cast microstructure consisted of α-Al dendrites and a large amount of coarse acicular and blocky eutectic silicon phase distributed heterogeneously in the α-Al matrix. In addition, a small amount of Al₂Cu phase was observed. The coarse eutectic silicon severely disrupted the continuity of the α solid solution, which is the primary cause of the low strength in the as-cast condition. After the T5 treatment, the eutectic silicon phase was fragmented and spheroidized. The sharp edges were blunted, and the silicon particles became fine and evenly distributed in a network-like pattern within the α-Al matrix. This morphological change improves the cohesion of the α matrix and substantially enhances the tensile strength and hardness. The Al₂Cu phase was largely dissolved into the α matrix during solution treatment, leaving only a minimal residue.
Energy dispersive X-ray spectroscopy (EDS) analysis confirmed that the α matrix is aluminum-rich, and the silicon particles are concentrated in the eutectic regions. The negligible amount of Cu detected in the matrix suggests that Cu was in solid solution, which is beneficial for precipitation strengthening during aging.
The relationship between process parameters and casting defects can be summarized by the modified Niyama criterion. For the ZL105A alloy, the critical Niyama value was found to decrease with better feeding design. The use of side runners and chillers effectively eliminated the high-risk zones. The total shrinkage porosity prediction correlated well with X-ray inspection results, confirming the reliability of the simulation approach.
In conclusion, this work demonstrates that numerical simulation is a powerful tool for designing casting processes for complex aluminum alloy components. Through iterative simulation, I was able to identify potential casting defects in the initial design and systematically optimize the gating system, riser configuration, pouring temperature, and cooling strategy. The optimized process produced defect-free turbine volute castings with significantly improved mechanical properties after T5 heat treatment. The final heat treatment cycle, consisting of solution treatment at 535°C for 6 hours, water quenching at 25°C, and aging at 170°C for 8 hours, yielded a tensile strength of 333.6 MPa and a hardness of 107.24 HB, exceeding the national standard requirements. This integrated approach, combining simulation, production verification, and heat treatment optimization, provides a robust framework for the manufacturing of high-quality aerospace aluminum castings while minimizing the occurrence of casting defects.