In this work, I aimed at eliminating the severe shrinkage porosity defects that occur during the investment casting of a large supported frame component made of K4169 nickel-based superalloy. The as-cast product, which measures nearly 900 mm in outer diameter with a nominal wall thickness of only 3 mm, connects an inner ring and an outer ring through eight hollow struts. Both rings carry integral bosses for bolting. The large difference in section thickness between the thin walls and the thick bosses makes this component highly susceptible to solidification shrinkage. The sand foundry defect that most affected the yield was the presence of extensive micro-shrinkage and macro-shrinkage in the flanges and bosses. To systematically resolve this problem, I combined ProCAST numerical simulation with production-scale experiments. I first analyzed the original feeding system, then predicted the shrinkage porosity distribution using solidification modeling, redesigned the gating and risering system, and finally validated the improvements by casting two full-scale components followed by X-ray inspection, metallographic examination, room-temperature tensile testing, and high-temperature stress-rupture testing.
1. Numerical Prediction of Shrinkage Porosity in the Original Feeding System
The original feeding system was designed empirically. In that design, the outer ring was oriented with the large end flange at the bottom and the small end flange at the top, while the ingates were attached to the small end flange. The alloy K4169 has a liquidus temperature of about 1350°C and a solidus temperature of about 1200°C. The mold shell was preheated to 900°C, and the pouring temperature was 1500°C. Using ProCAST, I carried out a coupled flow-thermal-solidification analysis. For the prediction of shrinkage porosity, I employed the Niyama criterion, which is computed as
$$ \mathrm{Niyama} = \left| \frac{G}{\sqrt{\dot{T}}} \right| $$
where \(G\) is the local temperature gradient and \(\dot{T}\) is the local cooling rate. A low value of the Niyama parameter indicates a high probability of micro-shrinkage formation. The simulation results for the unsupplemented casting, i.e., without any gating system, revealed the following characteristic distribution of sand foundry defect intensity:
The struts connecting the inner ring and the outer ring showed a very low tendency for shrinkage porosity formation, which is consistent with their uniform 3 mm wall thickness and rapid cooling. The inner ring wall, the outer ring wall, and the junction regions between the struts and the rings also exhibited a low porosity risk. In contrast, the simulation predicted that the outer ring large-end flange would contain extensive shrinkage porosity distributed almost continuously along the annular direction. The outer ring small-end flange showed a much lower density of porosity than the large-end flange. The inner ring bosses were all predicted to be prone to shrinkage porosity, with the maximum probability of porosity exceeding 50% at some locations.
To quantify the severity of the predicted defects, I summarized the simulation results in Table 1.
| Region | Porosity probability | Relative defect volume | Spatial distribution |
|---|---|---|---|
| Struts (8 locations) | <1% | Very low | Isolated |
| Strut-ring junctions | <1% | Very low | Isolated |
| Outer ring small-end flange | 5–30% | Low | Sparse |
| Inner ring bosses | 50–60% | High | Localized in each boss |
| Outer ring large-end flange | >50% | Very high | Almost continuous along the annulus |
The simulation results agreed well with the actual X-ray inspection of production castings. The large-end flange showed the most severe shrinkage porosity, while the inner ring bosses exhibited visible shrinkage cavities. This agreement confirmed that the numerical model was capable of predicting the sand foundry defect distribution with acceptable engineering accuracy.
2. Influence of Mold Shell Preheat Temperature
After establishing the baseline defect distribution, I investigated the influence of cooling rate by varying the mold shell preheat temperature from 800°C to 1000°C. The cooling rate of the casting is directly related to the mold preheat temperature. A lower mold preheat temperature produces a higher cooling rate, which normally refines the dendrite structure and reduces segregation. However, in this particular casting, the effect of cooling rate on shrinkage porosity was not monotonic.
When the mold preheat temperature was reduced to 800°C, the outer ring large-end flange still exhibited shrinkage porosity that was almost continuous along the annulus. The porosity probability in some areas approached 60%, which was slightly higher than the value predicted at 900°C. The inner ring bosses also showed an increased tendency toward shrinkage, with the porosity probability rising to above 60%. The small-end flange porosity was not significantly changed. This behavior can be attributed to the fact that the higher cooling rate caused earlier solidification of the thin outer ring wall, thereby blocking the feeding path from the upper ingates to the lower large-end flange.
When the mold preheat temperature was increased to 1000°C, the cooling rate decreased. In this condition, the outer ring large-end flange still contained substantial shrinkage porosity, with the porosity probability slightly higher than at 900°C but not dramatically different. The inner ring bosses still showed porosity at approximately 50% probability. Interestingly, the small-end flange showed an improvement at the higher mold temperature. However, the struts began to show a small tendency toward shrinkage porosity, which was not present at 900°C. This indicates that the slower cooling rate allowed the thin struts to remain liquid for a longer time, making them more susceptible to solidification shrinkage.
| Mold preheat (°C) | Outer large-end flange | Outer small-end flange | Inner ring bosses | Struts |
|---|---|---|---|---|
| 800 | ~60%, continuous | 5–30% | >60% | <1% |
| 900 | >50%, continuous | 5–30% | 50–60% | <1% |
| 1000 | ~60%, continuous | <10% | ~50% | 2–5% |
From these results, I concluded that simply adjusting the mold preheat temperature could not eliminate the shrinkage porosity. The fundamental problem was the inadequate feeding system design. The original orientation of the casting, with the large-end flange at the bottom and the small-end flange at the top, produced an unfavorable temperature gradient. The thin outer ring wall solidified first and cut off the feeding path to the heavy large-end flange at the bottom. Therefore, I decided to redesign the casting orientation and the gating/risering system.
3. Optimization of the Feeding System
3.1 Change of casting orientation
Based on the principle of directional solidification, the casting should be oriented so that the heaviest sections are at the top and the lightest sections are at the bottom. In this way, the temperature gradient is established from the bottom (cooler) to the top (hotter), and the liquid metal in the upper sections can feed the lower sections under the action of gravity. Accordingly, I inverted the casting orientation. The outer ring large-end flange was placed at the top, while the outer ring small-end flange was placed at the bottom. The ingates were then attached to the large-end flange, which now acted as a riser-like reservoir for feeding the entire outer ring.
This change in orientation allows the liquid metal to solidify first at the small-end flange, then progressively upward through the thin outer ring wall, and finally at the large-end flange. The ingate, being attached to the large-end flange, can supply liquid metal to compensate for the solidification shrinkage of the large-end flange. The temperature gradient becomes favorable for directional solidification, and the feeding channel remains open for a longer time.
3.2 Riser design for the inner ring bosses
The inner ring bosses are local thick sections that solidify last. Because the thin inner ring wall cannot provide a sufficient feeding path to these bosses, I designed cylindrical risers placed directly on top of each boss. The dimensions of the risers were determined using the modulus method. The modulus \(M\) of a casting section is defined as the ratio of its volume \(V\) to its cooling surface area \(A\):
$$ M = \frac{V}{A} $$
For the riser to be effective, its modulus \(M_r\) must satisfy the condition:
$$ M_r \geq f \cdot M_c $$
where \(M_c\) is the modulus of the boss (the section to be fed) and \(f\) is a safety factor typically taken as 1.2 to 1.3. For a cylindrical riser with diameter \(D_r\) and height \(H_r\), the modulus is:
$$ M_r = \frac{\pi D_r^2 H_r / 4}{\pi D_r H_r + 2 \pi D_r^2 / 4} = \frac{D_r H_r}{4 H_r + 2 D_r} $$
By equating this modulus to the required value, I determined the riser dimensions. The risers were designed to solidify after the bosses so that they could effectively feed the solidification shrinkage of the bosses. The improved feeding system configuration is summarized in Table 3.
| Design parameter | Original design | Improved design |
|---|---|---|
| Casting orientation | Large-end flange at bottom | Large-end flange at top |
| Ingate location | Small-end flange | Large-end flange |
| Riser on inner ring bosses | None | Cylindrical riser on each boss |
| Feeding path | Blocked by thin wall | Open from top to bottom |
| Mold preheat temperature | 900°C | 900°C (optimum) |
3.3 Simulation of the improved feeding system
After implementing the improved feeding system, I repeated the solidification simulation. The results showed a dramatic reduction in shrinkage porosity. The outer ring large-end flange showed almost no shrinkage porosity, with the porosity probability less than 1%. The outer ring small-end flange also exhibited a very low porosity probability, with only isolated regions exceeding 2%. The inner ring bosses, which previously had a porosity probability of over 50%, now showed almost no shrinkage porosity, with the maximum probability less than 1%. The shrinkage cavity was effectively transferred into the cylindrical risers, which could be removed after casting.
I also evaluated the sensitivity of the improved system to the cooling rate. When the mold preheat temperature was reduced to 800°C, the outer ring small-end flange showed a slightly increased porosity tendency. When the mold preheat temperature was increased to 1000°C, the outer ring large-end flange showed a marginal increase in porosity. However, in both cases, the porosity remained below the level of the original design. The optimum mold preheat temperature was confirmed to be approximately 900°C, where the overall shrinkage porosity tendency was the lowest.
4. Experimental Validation
4.1 Casting of full-scale components
To validate the numerical predictions, I cast two full-scale supported frame components (designated L1 and L2) using the improved feeding system. The mold shell preheat temperature was maintained at 900°C, and the pouring temperature was 1500°C. After casting and removal of the ceramic shell, the components were subjected to X-ray inspection. The results showed a significant reduction in the number of shrinkage defects from more than sixty in the original process to about ten in the improved process. X-ray inspection of the struts, the large-end flange, the small-end flange, and the inner ring bosses showed no visible shrinkage cavities.
4.2 Microstructure analysis
I examined the microstructure of samples taken from the outer ring bosses, inner ring bosses, and the strut-to-ring transition regions. The samples were examined in two heat-treatment conditions:
(1) Homogenization + solution + aging
(2) Homogenization + hot isostatic pressing (HIP) + solution + aging
The heat-treatment parameters were as follows:
- Homogenization: 1090°C for 3 h
- HIP: 1100–1150°C at 120–130 MPa for 3 h
- Solution: 960°C for 1 h
- Aging: 700°C for 8 h
Metallographic observation was performed using an optical microscope and a scanning electron microscope. The etchant used was 50 ml HCl + 50 ml H₂O + 5 g CuSO₄.
In the outer ring boss samples, the microstructure consisted of primary γ dendrites with γ′ and γ″ precipitates in the dendrite cores. The interdendritic regions contained MC carbides, Laves phase, and δ phase. The grain size was approximately 3 mm in both the as-heat-treated and the HIP-treated conditions. HIP treatment did not cause any significant grain growth. In the normal heat-treated condition, large amounts of Laves phase and acicular δ phase were present in the interdendritic regions. After HIP treatment, the Laves phase and δ phase were largely dissolved back into the matrix. Only MC-type carbides and a small amount of δ phase remained at the grain boundaries.
The inner ring boss samples showed a similar behavior. The grain size was approximately 2–3 mm, and no significant grain growth occurred during HIP. The HIP treatment effectively eliminated the Laves phase and reduced the segregation. The strut-to-ring transition regions, which were thinner sections, exhibited a finer grain size of approximately 1–2 mm. The HIP treatment reduced the microsegregation and dissolved the Laves phase and δ phase in these regions as well.
| Region | Grain size (mm) | Laves phase (as-heat-treated) | Laves phase (after HIP) | δ phase (after HIP) |
|---|---|---|---|---|
| Outer ring boss | ~3 | Abundant | Dissolved | Minor |
| Inner ring boss | 2–3 | Abundant | Dissolved | Minor |
| Strut-to-ring transition | 1–2 | Abundant | Dissolved | Minor |
The metallographic results demonstrated that the improved feeding system produced a sound casting with a homogeneous microstructure. The two castings L1 and L2 showed nearly identical microstructural features, confirming the reproducibility of the process.
4.3 Mechanical properties
4.3.1 Room-temperature tensile properties
The room-temperature tensile tests were conducted on specimens cut from various locations of the castings, including the outer ring large flange, outer ring small flange, inner ring large flange, inner ring small flange, and the struts. The specimens were tested in both heat-treatment conditions. The test results are summarized in Table 5.
| Specimen condition | Ultimate tensile strength \(\sigma_b\) (MPa) | Yield strength \(\sigma_{0.2}\) (MPa) | Elongation \(\delta\) (%) | Reduction of area \(\Psi\) (%) |
|---|---|---|---|---|
| L-series, as-heat-treated | 977 ± 40 | 830 ± 20 | 15 ± 3 | 30 ± 3 |
| L-series, after HIP | 920 ± 30 | 788 ± 25 | 16 ± 2 | 32 ± 2 |
| D-series, as-heat-treated | 960 ± 35 | 823 ± 15 | 14 ± 2 | 31 ± 2 |
| D-series, after HIP | 982 ± 28 | 815 ± 18 | 14 ± 3 | 37 ± 3 |
| Technical requirement | ≥825 | ≥640 | ≥5 | ≥8 |
All tested specimens exceeded the technical requirements. The tensile properties were remarkably consistent across different locations and between the two castings. The HIP treatment did not significantly alter the room-temperature tensile properties, which is consistent with the fact that the improved feeding system already produced a sound casting with minimal micro-shrinkage porosity.
4.3.2 High-temperature stress-rupture performance
Stress-rupture tests were conducted at 650°C under a stress of 620 MPa. The results are presented in Table 6, comparing the improved feeding system with the original one.
| Sampling location | Condition | Original feeding system (min) | Improved feeding system (min) |
|---|---|---|---|
| Struts | Heat-treated | 37:42 | 36:10 |
| Struts | HIP | — | 36:15 |
| Outer ring large flange | Heat-treated | 7:26 | 43:24 |
| Outer ring large flange | HIP | — | 40:50 |
| Outer ring small flange | Heat-treated | 24:32 | 41:43 |
| Outer ring small flange | HIP | — | 40:46 |
| Inner ring large flange | Heat-treated | 29:26 | 38:43 |
| Inner ring large flange | HIP | — | 36:21 |
| Inner ring small flange | Heat-treated | 40:23 | 39:27 |
| Inner ring small flange | HIP | — | 41:36 |
The most striking improvement occurred at the outer ring large flange, where the stress-rupture life increased from 7 minutes in the original design to over 40 minutes in the improved design. This dramatic increase is a direct consequence of eliminating the severe shrinkage porosity in that region. The outer ring small flange also showed a significant improvement, from 24 minutes to over 40 minutes. All specimens from the improved castings exceeded the technical requirement of 30 minutes.
Fracture surface analysis of the improved specimens revealed a ductile fracture mode with fine dimples. No significant micro-shrinkage was observed on the fracture surfaces. In contrast, specimens from the original castings showed clear evidence of shrinkage porosity and even cold shut defects on the fracture surfaces, which were responsible for the premature failure.
5. Discussion
The results of this study clearly demonstrate that the sand foundry defect in the form of shrinkage porosity in the supported frame casting was primarily caused by an inadequate feeding system design. The original design, which placed the heavy large-end flange at the bottom and the ingates at the small-end flange, could not establish a favorable temperature gradient for directional solidification. The thin outer ring wall solidified early and blocked the feeding path to the heavy flange below.
By inverting the casting orientation so that the large-end flange was at the top and the ingates were attached to it, I established a temperature gradient that promoted directional solidification from the bottom to the top. The ingate, attached to the large-end flange, acted as a riser and effectively fed the solidification shrinkage of the large-end flange. This simple change eliminated the most severe shrinkage porosity in the casting.
The inner ring bosses required a separate solution. Since the thin inner ring wall could not provide a feeding path to the bosses, I placed cylindrical risers directly on top of each boss. This is a classic application of the modulus method for riser design. The risers effectively transferred the shrinkage cavity out of the boss and into the riser, which was subsequently removed. This eliminated the porosity in the bosses.
The sand foundry defect analysis also revealed that the mold preheat temperature had a significant but non-monotonic effect on porosity. At a low mold preheat temperature of 800°C, the high cooling rate caused premature solidification of the thin sections, which blocked the feeding paths. At a high mold preheat temperature of 1000°C, the slow cooling rate increased the tendency for shrinkage in the thin struts. The optimum mold preheat temperature was approximately 900°C, which balanced the need for directional solidification and the need to avoid excessive thermal gradients.
The mechanical property tests confirmed the effectiveness of the improved feeding system. All room-temperature tensile specimens exceeded the technical requirements. The stress-rupture lives at 650°C and 620 MPa were consistently above 30 minutes, with the most dramatic improvement observed at the outer ring large flange, where the life increased from 7 minutes to over 40 minutes. This improvement is directly attributable to the elimination of the sand foundry defect in the form of shrinkage porosity.
6. Conclusions
In this work, I successfully optimized the feeding system of a large supported frame investment casting of K4169 superalloy by combining numerical simulation with experimental validation. The main conclusions are as follows:
(1) The sand foundry defect in the original casting was predominantly located in the outer ring large-end flange, the outer ring small-end flange, and the inner ring bosses. The shrinkage porosity in the large-end flange was almost continuous along the annular direction and was far more severe than in the small-end flange. Every inner ring boss showed a porosity probability exceeding 50%.
(2) Adjusting the mold preheat temperature alone could not eliminate the shrinkage porosity. The optimum mold preheat temperature was approximately 900°C, where the overall shrinkage porosity tendency was minimal.
(3) The improved feeding system, with the outer ring large-end flange placed at the top, the ingates attached to the large-end flange, and cylindrical risers placed on the inner ring bosses, significantly reduced the sand foundry defect tendency. The shrinkage porosity probability in the large-end flange and the bosses was reduced to less than 1%.
(4) Full-scale casting validation confirmed the simulation results. The number of shrinkage defects was reduced from more than sixty in the original process to about ten in the improved process. X-ray inspection showed no visible shrinkage cavities in the critical regions.
(5) The microstructure of the castings produced by the improved process was uniform and reproducible. Hot isostatic pressing effectively dissolved the Laves phase and reduced microsegregation without causing grain growth.
(6) The room-temperature tensile properties of the improved castings met all technical requirements. The stress-rupture lives at 650°C and 620 MPa exceeded the 30-minute requirement, with a dramatic improvement at the outer ring large flange from 7 minutes to over 40 minutes. The elimination of the sand foundry defect in the form of shrinkage porosity was the key factor responsible for this improvement.