As a graduate student specializing in materials engineering, I focused my research on the casting process optimization of a self-priming pump body. This component is a critical pressure-containing part of a canned pump, which has to withstand significant impact and pressure during operation. The casting is made of ZG07Cr19Ni10 stainless steel and features complex curved surfaces that make the casting process exceptionally challenging. In this paper, I present a systematic approach combining numerical simulation with experimental verification to eliminate casting defects and improve process yield.
1. Introduction to the Casting Process Design
The self-priming pump body is quite different from ordinary centrifugal pump casings. It integrates a suction chamber, a volute, a liquid separation chamber, and several flanges into one complex casting. The overall dimensions are approximately 610 mm × 360 mm × 380 mm, with an average wall thickness of 12 mm and a maximum thickness variation of 23 mm. The total weight of the finished casting is 90 kg. Because this part must handle corrosive fluids at high pressure, the material specification requires an austenitic stainless steel corresponding to ZG07Cr19Ni10 (similar to 304 grade).
From the casting process point of view, this alloy has poor fluidity, a strong tendency to oxidize, and a large volumetric shrinkage. Therefore, the initial casting process design must address three critical issues: (1) ensuring a smooth and complete mold filling without cold shuts, (2) avoiding oxide slag entrapment, and (3) providing adequate feeding to prevent shrinkage cavities and porosity.
Based on the existing production conditions in the foundry, I chose sand casting with manual molding. The mold was made of sodium silicate-CO₂ hardened sand, while the cores were produced with ester-cured alkaline phenolic resin sand. These materials offer high strength at elevated temperatures, good collapsibility, and they do not introduce sulfur or phosphorus into the steel. The pouring position was arranged with the inlet flange at the bottom and the large flange at the top, using a three-part molding process. The parting lines were placed at the outlet flange/bleed flange center plane and at the inlet flange end face.
2. Initial Casting Process Scheme
According to the casting process design principles, I first established the initial process parameters. The shrinkage allowance was set to 2.0% in the longitudinal direction and 3.0% in the radial direction. The dimensional tolerance was CT12, with a draft angle of 1° for the volute region and 1°30′ for other regions. A stepped gating system was initially designed, consisting of a sprue of Ø90 mm, two ingates, and no runner. The pouring time was calculated using the empirical formula:
$$
t = A G^{n}
$$
where \(A = 1.5\text{–}2.35\), \(n = 0.5\), and \(G = 127\) kg (total weight of casting plus gating and risers). This gave a pouring time range of 17–26.5 s, and I selected 20 s. The average rising velocity of the metal was verified as:
$$
v = \frac{h}{t} = \frac{500}{20} = 25 \ \text{mm/s}
$$
which satisfied the recommended 20–30 mm/s for thin-walled castings. The choke area was determined by the well-known formula:
$$
A_{\text{choke}} = \frac{G_L}{\rho_L \ \mu \ t \ \sqrt{2 g H_p}}
$$
with \(G_L = 127\) kg, \(\rho_L = 7000\) kg/m³, \(\mu = 0.5\), and \(H_p = 30.3\) cm, resulting in a minimum cross-sectional area of 6.87 cm². The dimensions of the initial gating system are listed below.
| Component | Length (mm) | Cross-section shape | Dimensions (mm) | Area (cm²) | Total area (cm²) |
|---|---|---|---|---|---|
| Pouring cup | 55 | Circular | Ø100 | 78.5 | 78.5 |
| Sprue | 440 | Circular | Ø90 | 63.6 | 63.6 |
| Ingate 1 | 220 | Rectangular | 175 × 38.5 | 67.4 | 75.4 |
| Ingate 2 | 200 | Trapezoidal | 44/40 × 20 | 8 |
For the feeding system, I adopted insulating sleeve risers on the thick sections and chills at the flanges. The riser modulus was calculated using the modulus method:
$$
M_r = 1.2 \ M_j
$$
where \(M_j\) is the modulus of the casting section. Three insulating sleeves with a modulus of 1.6 cm were placed above the large flange, one insulating blind riser with a modulus of 1.1 cm was set at the bottom end of the volute, and one large open riser was placed above the outlet flange. External chills were attached to the inlet and outlet flanges, while an internal chill was embedded in the exhaust boss. Figure 1 shows the initial casting layout including all these elements.
3. Numerical Simulation Setup
I used Flow-3D cast software to simulate both the mold filling and solidification processes. The three-dimensional model was created with a CAD system and saved in STL format. After checking and repairing the surfaces, the model was imported into Flow-3D. A virtual mold was defined because the sand mold and cores were made of the same material. A locally refined mesh was generated to ensure at least three elements across the thinnest wall. The total number of cells was controlled to balance accuracy and computational time.
The material properties of ZG07Cr19Ni10 were assigned as follows: solidus temperature 1399 °C, liquidus temperature 1454 °C, and densities of 7.6 g/cm³, 7.4 g/cm³, and 7.0 g/cm³ in the solid, mushy, and liquid regions respectively. The thermal conductivity values were interpolated from a temperature-dependent table. The heat transfer coefficients were set as: metal–mold 500 W/(m²·K), metal/ mold–air 300 W/(m²·K), metal–insulating riser 100 W/(m²·K), metal–chill 2000 W/(m²·K), and mold–chill 500 W/(m²·K). The initial process parameters were a pouring temperature of 1560 °C, pouring time of 20 s, and mold temperature of 20 °C.
4. Simulation Results of the Initial Scheme
4.1 Mold Filling Behavior
The filling pattern was extracted at different filling ratios. At the beginning, the liquid metal entered the sprue and immediately impinged on the bottom of the cavity with a high velocity. The velocity difference between the sprue and the ingate reached 1.6 m/s, causing severe turbulence and splashing. This behavior was observed in the velocity field plots, where the liquid metal hit the core surface and created local vortices. Such turbulent flow can entrain air and form oxide films, leading to casting defects.
The temperature field during filling showed that the metal temperature dropped only about 20 °C in the early stage, but the thin walls caused a more significant drop of about 40 °C. By the end of filling, the temperature ranged from 1450 °C to 1550 °C. All regions remained above the liquidus temperature, so cold shut and misrun defects were not expected from a thermal perspective. However, the upper ingate did not help to heat the inlet flange; instead, the cooler metal from the flange flowed backward into the sprue, lowering the sprue temperature. This could lead to a lack of feeding later.
The pressure field showed a maximum value of approximately 0.13 MPa at the bottom of the sprue. This was below the compressive strength of the sand (0.22 MPa for sodium silicate sand), so mold erosion was unlikely. Nevertheless, the turbulent filling was not acceptable for a stainless steel casting requiring high integrity.
4.2 Solidification and Shrinkage Prediction
The solid–liquid phase distribution was captured at various solidification fractions. I found that several isolated liquid pools were formed in the lower wall, under the partition plate, and inside the exhaust boss. These isolated regions could not be fed by the risers because they were cut off by earlier solidified areas. Consequently, shrinkage cavities were predicted in those regions, which matched the actual defects found later. The temperature field also showed that the thick sections near the flange remained hot for a long time, which caused coarse grains and potential hot spots.
Using the residual melt modulus (RMM) criterion in Flow-3D, I calculated the probability of shrinkage defects. The defects with a probability greater than 0.9 were filtered after removing the gating and riser systems. The total defective volume was 3.0×10⁵ mm³, which represented 2.6% of the pump body volume. The critical locations were the partition plate-to-body junction, the flange-neck junction, the suction pipe bottom, and the exhaust boss interior.
4.3 Actual Casting Defects
To confirm the simulation predictions, I produced castings according to the initial process. Visual inspection revealed several defects: cold shots on the drain flange surface, surface wrinkles on the upper surface, cracks, and small cavities. Scanning electron microscopy was used to examine the cracked and cavity areas. The fracture surface exhibited a sponge-like appearance with coarse dendrites, which is typical of severe shrinkage. The small cavities also showed dendritic morphology rather than smooth gas pores, indicating that they formed due to oxide inclusions that interrupted local feeding. These observations corresponded well with the simulation results, confirming that the gating system was the primary source of oxide slag, while the feeding system was insufficient to avoid shrinkage.
5. Gating System Optimization
5.1 First Optimization Attempt
To reduce the turbulence and prevent oxide formation, I modified the gating system in two ways. First, I removed the upper ingate and kept only a lower ingate. Second, I changed the system from an open to a closed type by reducing the sprue diameter to Ø40 mm. A sprue well of Ø80 mm was added at the bottom of the sprue to dissipate kinetic energy and allow bubbles to escape. The ingate was placed tangentially with respect to the flange so that the metal would swirl and stabilize. The revised dimensions are shown below.
| Component | Length (mm) | Cross-section shape | Dimensions (mm) | Area (cm²) |
|---|---|---|---|---|
| Pouring cup | 50 | Circular | Ø50 | 19.6 |
| Sprue | 410 | Circular | Ø40 | 12.6 |
| Ingate | 220 | Rectangular | 35 × 25 | 7.5 |
| Sprue well | 80 | Circular | Ø80 | 50.3 |
The area ratio was \(\sum A_{\text{sprue}} : \sum A_{\text{ingate}} = 1.7:1\). Simulation of this scheme showed improved velocity stability during the middle and late filling stages. However, at the beginning, the metal jet still struck the core in the inlet flange and produced a local vortex. The velocity difference at that point was 0.8 m/s, which was not acceptable. Moreover, there was no component to trap slag particles entering the ingate.
5.2 Second Optimization and Final Gating Design
In the second iteration, I inserted a runner between the sprue and the ingate. The runner acts as a slag trap because the first turbulent metal that contains entrapped oxides can be held there and not enter the cavity. I also changed the gating system from closed to semi-closed, with the area ratio \(\sum A_{\text{sprue}} : \sum A_{\text{runner}} : \sum A_{\text{ingate}} = 1.2 : 0.8 : 1\). The ingate area was increased to 10 cm² to reduce the filling velocity. The final gating system dimensions are listed in the following table.
| Component | Length (mm) | Cross-section shape | Dimensions (mm) | Area (cm²) |
|---|---|---|---|---|
| Pouring cup | 50 | Circular | Ø50 | 19.6 |
| Sprue | 410 | Circular | Ø40 | 12.6 |
| Runner | 240 | Rectangular | 20 × 40 | 8.0 |
| Ingate | 150 | Rectangular | 40 × 25 | 10.0 |
| Sprue well | 80 | Circular | Ø80 | 50.3 |
The simulation of this final gating design showed that the velocity difference across the mold cavity was less than 0.5 m/s during most of the filling process. The temperature distribution remained uniform, with a maximum temperature difference of 110 °C. The maximum pressure was only 0.15 MPa, well below the sand strength. The runner successfully trapped the slag particles, and no casting defects were observed in the simulated filling results.
6. Feeding System Optimization
6.1 Reducing Riser Sizes and Adding Chills
After optimizing the gating system, I turned to the feeding system. The initial riser configuration was too large and created excessive metal consumption. In the first optimization, I reduced the three insulating risers above the large flange to a smaller size (70 mm × 36 mm × 60 mm) and removed the blind riser at the volute bottom, replacing it with chills. The large riser above the outlet flange was also reduced to Ø50 mm × 120 mm. External chills were added under the partition plate, around the exhaust boss, and between the risers to improve the temperature gradient and prevent isolated hot spots.
The solidification simulation of this modified design showed that the shrinkage defects decreased by 46%, from 3.0×10⁵ mm³ to 1.6×10⁵ mm³. However, defects still appeared in the large flange center and inside the suction pipe. The smaller risers were not powerful enough to feed those regions completely. The temperature field indicated that the large flange solidified later than the risers, so the risers could not supply liquid metal to compensate for shrinkage.
6.2 Final Feeding System Design
In the second feeding optimization, I increased the riser size back to the original modulus of 1.6 cm for the large flange, but used only two insulating risers instead of three. I reinstated a blind riser at the volute bottom with a modulus of 1.1 cm. The chills at the suction pipe were moved from the outer wall to the bottom face to avoid cutting off the feeding channel before solidification was complete. I also added a chill at the partition plate and at the bottom of the cavity.
The final feeding system is illustrated in the process layout drawing. The solidification simulation results showed that the temperature difference between the risers and the casting remained above 90 °C during most of the solidification process, enabling excellent directional solidification. The isolated liquid regions gradually disappeared, and the only residual porosity was located in small non-critical areas. The defective volume was reduced to 1.0×10⁵ mm³, which corresponds to 0.88% of the pump body volume. Compared with the initial scheme, the shrinkage defects were reduced by 69%.
The process yield also improved significantly. The initial yield was 55.9%, while the optimized scheme achieved 77% due to the reduced riser volume and better placement of chills. This represents an 11% improvement in material utilization.
7. Orthogonal Experiment for Production Parameters
After the casting process geometry was optimized, I needed to determine the best combination of production parameters. I employed an orthogonal experimental design, \(L_9(3^4)\), with three factors: pouring temperature (A), mold preheating temperature (B), and pouring time (C). Each factor had three levels, as listed below.
| Level | A: Pouring temp (°C) | B: Mold preheat temp (°C) | C: Pouring time (s) |
|---|---|---|---|
| 1 | 1530 | 100 | 10 |
| 2 | 1560 | 200 | 15 |
| 3 | 1590 | 300 | 20 |
I created a total of nine schemes. For each simulation, I monitored the velocity, oxide concentration, and pressure at the ingate during filling, as well as the casting solidification time, ingate solidification time, and shrinkage defect volume during solidification. The results are summarized in the following two tables.
7.1 Filling Process Results
| Scheme | A | B | C | D (empty) | Ingate velocity (m/s) | Oxide concentration | Ingate pressure (Pa) |
|---|---|---|---|---|---|---|---|
| L1 | 1 | 1 | 1 | 1 | 2.01 | 22.33 | 152210 |
| L2 | 1 | 2 | 2 | 3 | 1.54 | 39.47 | 142046 |
| L3 | 1 | 3 | 3 | 2 | 1.19 | 47.19 | 138700 |
| L4 | 2 | 1 | 3 | 3 | 1.20 | 47.94 | 137530 |
| L5 | 2 | 2 | 1 | 2 | 2.19 | 24.03 | 150885 |
| L6 | 2 | 3 | 2 | 1 | 1.44 | 36.09 | 141331 |
| L7 | 3 | 1 | 2 | 2 | 1.57 | 38.12 | 142228 |
| L8 | 3 | 2 | 3 | 1 | 1.23 | 48.21 | 137407 |
| L9 | 3 | 3 | 1 | 3 | 2.19 | 23.58 | 151357 |
From the velocity field analysis, I observed that the schemes with a pouring time of 10 s (L1, L5, L9) exhibited the highest ingate velocities (above 2 m/s) and turbulent filling. Schemes with 15 s and 20 s pouring times produced lower velocities, but the 20 s schemes had higher oxide concentrations because the metal stayed in contact with air longer. The 15 s schemes gave a good balance. For example, scheme L6 had a stable velocity of about 0.8 m/s during later filling and a reasonable oxide concentration of 220 in the pump body.
The pressure results showed that a pouring time of 10 s caused a maximum pressure of 0.23 MPa, exceeding the sand strength limit. Schemes with 15 and 20 s had maximum pressures of 0.17 MPa and 0.15 MPa, respectively, which were safe for the mold.
Analysis of variance (ANOVA) was performed on the filling indicators. The F-values and P-values showed that the pouring time had a highly significant effect on all three filling indicators, while the pouring temperature and mold preheat temperature had no significant effect. The significance order was C > A > B.
7.2 Solidification Process Results
| Scheme | A | B | C | D (empty) | Casting solidification time (s) | Ingate solidification time (s) | Shrinkage defect volume (mm³) |
|---|---|---|---|---|---|---|---|
| L1 | 1 | 1 | 1 | 1 | 1912 | 183 | 103809 |
| L2 | 1 | 2 | 2 | 3 | 2022 | 209 | 102014 |
| L3 | 1 | 3 | 3 | 2 | 1770 | 198 | 99536 |
| L4 | 2 | 1 | 3 | 3 | 1973 | 234 | 100737 |
| L5 | 2 | 2 | 1 | 2 | 2169 | 229 | 100289 |
| L6 | 2 | 3 | 2 | 1 | 2378 | 262 | 98373 |
| L7 | 3 | 1 | 2 | 2 | 2089 | 244 | 95415 |
| L8 | 3 | 2 | 3 | 1 | 1805 | 277 | 96073 |
| L9 | 3 | 3 | 1 | 3 | 2581 | 277 | 96169 |
Scheme L7, which had a pouring temperature of 1590 °C, a mold preheat temperature of 100 °C, and a pouring time of 15 s, produced the smallest shrinkage defect volume of 95,415 mm³. This represents 0.84% of the pump body volume. Scheme L6, with 1560 °C/300 °C/15 s, gave 98,373 mm³ (0.87%), while scheme L1, with 1530 °C/100 °C/10 s, gave 103,809 mm³ (0.91%).
I also analyzed the temperature curves at the ingate and the insulating riser. Higher pouring temperatures resulted in steeper cooling rates above the liquidus, which is beneficial for directional solidification. The ingate solidification time increased with increasing pouring temperature and mold preheat temperature. A longer ingate solidification time allows the riser to feed the casting for a longer period. However, an excessively long casting solidification time lowers productivity. Therefore, a compromise was needed.
The ANOVA results for the solidification indicators showed that the pouring temperature had the most significant effect on casting solidification time and shrinkage defect volume, followed by the mold preheat temperature. The pouring time had a weaker effect. The significance order was A > B > C, which was opposite to the filling stage.
8. Determination of the Optimal Production Parameters
Combining the filling and solidification results, I concluded that the best comprehensive performance was achieved with a pouring temperature of 1590 °C, a mold preheat temperature of 300 °C, and a pouring time of 15 s. This combination corresponds to scheme A3B3C2, which was not among the original nine trials. I performed an additional simulation using these parameters. The results are shown in the table below.
| Indicator | Maximum value |
|---|---|
| Ingate velocity (m/s) | 1.43 |
| Oxide concentration | 757 |
| Metal pressure (MPa) | 0.17 |
| Casting solidification time (s) | 2150 |
| Shrinkage defect volume (mm³) | 95931 |
This optimized parameter set yielded a shrinkage defect volume of 95,931 mm³ (0.83% of the pump body), which was the smallest among all tested schemes. The maximum pressure of 0.17 MPa was safely below the sand strength, and the oxide concentration was acceptable because the slag was confined to the runner and risers. Therefore, I selected the final production parameters as 1580–1600 °C pouring temperature, 280–300 °C mold preheat, and 14–17 s pouring time, allowing practical variations during manual pouring.
9. Production Verification and Testing
To validate the optimized process, I conducted a full-scale production trial using the final casting design and the optimal production parameters. After casting and solid solution treatment, the pump body was inspected according to the acceptance criteria.
9.1 Chemical Composition
The chemical composition of the test bar was analyzed and compared with the ZG07Cr19Ni10 specification.
| Element | Si | Mn | Cr | Ni | Fe |
|---|---|---|---|---|---|
| Specification (wt%) | ≤1.5 | ≤1.5 | 18–20 | 8–11 | Balance |
| Measured (wt%) | 0.79 | 0.80 | 18.55 | 8.02 | 71.05 |
All elements met the requirements, confirming that the melting and deoxidation procedures were effective.
9.2 Metallographic Examination
The metallographic structure was examined at 20× and 100× magnifications. The ferrite content was measured as approximately 7.5%, and the austenitic grain size was determined to be 4.5. These values are within the acceptable range for this grade, indicating a sound microstructure without excessive delta ferrite or carbide precipitation.
9.3 Mechanical Properties
Tensile, impact, and hardness tests were performed on test bars that were heat-treated together with the casting. The results are compared with the required values in the following table.
| Property | Yield strength (MPa) | Tensile strength (MPa) | Elongation (%) | Impact energy (J) | Brinell hardness (HB) |
|---|---|---|---|---|---|
| Requirement | ≥175 | ≥440 | ≥30 | ≥60 | 178–230 |
| Measured | 336 | 557 | 56.5 | 247 | 221 |
All mechanical properties significantly exceeded the acceptance limits, demonstrating that the optimized casting process did not introduce internal defects that would degrade strength or ductility.
9.4 Radiographic Inspection
The critical thick sections of the casting, including the large flange, drain flange, inlet flange, and outlet flange, were inspected using X-ray radiography according to ASTM E446. The inspection results showed no indications of porosity, shrinkage, or inclusions exceeding the acceptance limits. The casting was rated as Grade I, which is the highest quality level.
9.5 Pressure Tests
Finally, the pump body was subjected to hydrostatic and pneumatic pressure tests to verify its leak-tightness and pressure-bearing capacity. The hydrostatic test was performed at 3 MPa for 30 minutes, and the pneumatic test was conducted with nitrogen gas at 2.2 MPa for 10 minutes. The results are summarized below.
| Test | Pressure | Medium | Result |
|---|---|---|---|
| Hydrostatic test | 3 MPa | Water | Pass |
| Pneumatic test | 2.2 MPa | N₂ | Pass |
No leaks or deformations were observed. This confirmed that the casting satisfied all functional requirements for use in a self-priming canned pump.
10. Conclusion
Through this research, I have successfully optimized the casting process for a complex ZG07Cr19Ni10 stainless steel self-priming pump body. The key findings and conclusions are summarized as follows:
1. The initial casting process design had a gating system that caused severe turbulence during mold filling, leading to oxide slag entrapment and subsequent shrinkage defects. The solidification simulation correctly predicted the locations of the defects, which were confirmed by actual casting inspection and scanning electron microscopy.
2. By modifying the gating system to a semi-closed design with a runner for slag trapping, the filling process became smooth with a maximum velocity difference of less than 0.5 m/s. The maximum pressure was reduced to 0.15 MPa, well below the sand strength.
3. The feeding system was optimized through two iterations of riser resizing and chill placement. The final design achieved directional solidification, reducing the shrinkage defect volume by 69% compared to the initial scheme. The process yield increased by 11%, reaching 77%.
4. The orthogonal experiment revealed that pouring time has the most significant effect on the filling process, while pouring temperature has the greatest influence on the solidification process. The optimal combination was determined as a pouring temperature of 1590 °C, a mold preheat temperature of 300 °C, and a pouring time of 15 s. This combination produced the smallest shrinkage defect volume of 0.83% of the pump body volume.
5. Production verification confirmed that the optimized casting process produces pump bodies that meet all acceptance criteria for chemical composition, metallographic structure, mechanical properties, radiographic inspection, and pressure resistance. The successful trial demonstrates that the combination of numerical simulation and orthogonal experimentation is a reliable and cost-effective method for optimizing complex sand castings and eliminating casting defects.