Investment casting of large superalloy structural components is a critical manufacturing route for aero-engine applications. However, the complexity of thin-walled geometries with local thick sections often leads to severe solidification-related defects, particularly shrinkage porosity and macro-shrinkage. The present study addresses the persistent occurrence of sand foundry defects in a large bearing frame casting made of K4169 nickel-based superalloy. The work combines computational solidification modeling with experimental validation to optimize the feeding system, aiming at eliminating shrinkage defects and improving the overall integrity of the component. This research is carried out from the perspective of a manufacturing engineer who needs to solve real production problems, and the entire development process is described in first person to highlight the decision-making and practical outcomes.
In the initial phase of my study, I investigated the existing production route of the bearing frame casting. The component has an outer diameter close to 900 mm, with a nominal wall thickness of only 3 mm. Eight hollow struts connect the inner ring and the outer ring, and both rings carry integrally cast bosses for bolting. The huge difference in section thickness between the thin webs and the thick flanges or bosses creates a severe thermal center condition. As a result, the original casting process produced numerous sand foundry defects, primarily concentrated at the outer large end flange, the outer small end flange, and the inner ring bosses. X-ray inspection of production castings showed extensive shrinkage porosity in these regions, and some inner bosses exhibited macro-shrinkage cavities. These defects not only reduce the load-bearing cross-section but also act as stress raisers, significantly degrading fatigue and creep performance. It became evident that the original feeding system, which oriented the casting with the large end flange downward and the small end flange upward, could not establish a favorable temperature gradient for directional solidification. The liquid metal had to travel through the thin outer wall to reach the massive large end flange, and once the thin wall solidified, the feeding path was blocked. Consequently, solidification shrinkage at the large flange could not be compensated, leading to extensive dispersed porosity.
To systematically tackle the problem, I began by constructing a detailed numerical model of the casting without any feeding system. The objective was to predict the natural solidification behavior and identify the regions most susceptible to shrinkage porosity. I used the ProCAST simulation software, which is widely applied for predicting casting defects such as those associated with sand foundry defects. The geometry of the bearing frame was meshed with an unstructured tetrahedral mesh, refined locally at the thin struts and the boss regions. The K4169 alloy was defined with a liquidus temperature of 1350°C and a solidus temperature of 1200°C. The mold was a ceramic shell produced by the investment casting route, and the heat transfer coefficient at the metal-shell interface was set to 500 W/(m²·K). Three different shell pre-heat temperatures were considered: 800°C, 900°C, and 1000°C. The initial simulation was performed without any gating or feeding system, using a shell temperature of 900°C and a pouring temperature of 1500°C. This allowed me to observe the intrinsic solidification sequence and the natural locations of hot spots.
The simulation results revealed that the connecting struts between the inner and outer rings have a low probability of shrinkage porosity, because they are uniformly thin and cool quickly. The junctions between the struts and the rings also showed little susceptibility, thanks to the small geometric thermal centers. However, the outer large end flange displayed a high-volume fraction of shrinkage porosity, almost forming a continuous ring-shaped defect. The outer small end flange showed some porosity, but the defect intensity was much lower than that in the large end flange. Notably, every inner boss was predicted to have a significant probability of shrinkage, with the maximum local probability exceeding 50%. These predictions matched the X-ray observations from actual castings, confirming that my numerical model could capture the essential solidification behavior and reproduce the characteristic sand foundry defects.
Following the baseline prediction, I examined the effect of shell preheat temperature, which directly controls the cooling rate of the casting. The simulation was repeated with shell temperatures of 800°C and 1000°C while keeping all other parameters unchanged. The results are summarized in Table 1, in terms of the qualitative shrinkage tendency at the critical locations. A lower shell temperature (800°C) increases the cooling rate and was expected to refine the solidification structure. However, the simulations showed that increasing the cooling rate actually aggravated the shrinkage porosity at the inner bosses, with the probability rising above 60%. The outer large end flange also exhibited a slightly higher defect intensity at 800°C compared with the 900°C case. In contrast, a higher shell temperature (1000°C) decreased the cooling rate and improved the soundness of the outer small end flange. Nevertheless, the overall shrinkage tendency at the large flange increased, and the strut regions began to show some porosity, although at a low level. These findings indicate that adjusting the shell temperature alone cannot eliminate the severe shrinkage defects. The fundamental problem lies in the improper feeding geometry and the lack of a directed solidification path.
| Shell preheat (°C) | Cooling rate effect | Outer large flange | Outer small flange | Inner bosses | Struts |
|---|---|---|---|---|---|
| 800 | High | Severe porosity, nearly continuous | Moderate, dispersed | High probability >60% | No defects |
| 900 | Medium | Severe porosity, continuous | Moderate, sparse | High probability >50% | No defects |
| 1000 | Low | Severe porosity, slightly increased | Improved, fewer defects | High probability ≈50% | Some micro-porosity |
The influence of cooling rate on the solidification structure can be described by the relationship between the secondary dendrite arm spacing (SDAS) and the local cooling rate:
$$ \lambda_2 = A \left( \frac{\partial T}{\partial t} \right)^{-n} $$
where \(A\) and \(n\) are alloy-dependent constants. A higher cooling rate reduces the SDAS and normally reduces micro-shrinkage. However, the problem in this casting is not simply micro-shrinkage, but rather macro-shrinkage caused by inadequate feeding. The thin wall sections solidify rapidly and cut off the liquid supply to the heavier flanges and bosses, regardless of the SDAS. Therefore, the key is to redesign the feeding system so that the solidification progresses directionally from the smaller sections toward the larger sections, and finally toward the risers or ingates.
Based on the numerical predictions and the analysis of the actual sand foundry defects, I decided to reverse the casting orientation. The original orientation placed the outer large end flange at the bottom and the small end flange at the top. In the new design, the outer large end flange is located at the top, and the outer small end flange is placed at the bottom. The ingates are attached to the large end flange, which is now the highest thermal mass region. This arrangement creates a natural vertical temperature gradient after filling: the lower and thinner small end flange cools first, while the upper and heavier large end flange remains hot for a longer period. The ingates act as effective risers, feeding the large flange through the liquid path that remains open during the later stages of solidification. In addition to the orientation change, I introduced cylindrical risers on top of each inner boss. The inner bosses are local hot spots that cannot be fed through the thin inner ring walls. According to the modulus principle, the riser modulus \(M_r\) must be greater than the modulus of the boss \(M_b\). The modulus is defined as:
$$ M = \frac{V}{A} $$
where \(V\) is the volume and \(A\) is the cooling surface area. For a cylindrical riser with height \(h\) and diameter \(d\), the modulus is:
$$ M_r = \frac{\pi d^2 h / 4}{\pi d h + 2 \pi d^2 / 4} = \frac{d h}{4h + 2d} $$
By selecting appropriate dimensions, the riser solidification time exceeds that of the boss, ensuring that liquid metal is available to compensate for solidification shrinkage. The new feeding system is illustrated in Figure 1, which shows the arrangement with the large end flange at the top and the risers on the bosses.
After redesigning the feeding system, I performed a new series of solidification simulations to verify the effect of the modifications. The simulations used the same boundary conditions and material properties as before. The shell preheat temperature was initially kept at 900°C. The results showed a dramatic reduction in shrinkage defects at all critical locations. The outer large end flange no longer exhibited the continuous ring-shaped porosity; the probability of shrinkage at any given location dropped to less than 1%. Similarly, the outer small end flange became almost sound, with only isolated spots showing a probability slightly above 2%. The inner bosses, which previously had a defect probability exceeding 50%, now showed almost no shrinkage, because the added risers intercepted the solidification shrinkage and moved the porosity into the riser bodies. This clearly demonstrated that the orientation change and the riser placement effectively eliminated the primary sand foundry defects in the casting.
I subsequently investigated the sensitivity of the improved feeding system to shell preheat temperature. The simulations were repeated at 800°C and 1000°C. At 800°C, the outer large end flange remained sound, with a shrinkage probability below 1%. However, the outer small end flange showed a noticeable increase in porosity compared to the 900°C case. The reason is that the faster cooling of the thin outer wall blocks the feeding path from the large end flange to the small end flange before the small flange has fully solidified. This indicates that an excessively high cooling rate can reintroduce defects in the lower regions. At 1000°C, the outer small end flange exhibited some porosity as well, but the defect intensity was lower than at 800°C. The inner bosses remained sound at both 800°C and 1000°C, because the risers effectively feed those hot spots regardless of the cooling rate. The overall optimum condition was found at a shell temperature of around 900°C, where the combined tendency of shrinkage porosity across all regions was minimized. These results are summarized in Table 2.
| Shell preheat (°C) | Outer large flange | Outer small flange | Inner bosses | Overall evaluation |
|---|---|---|---|---|
| 800 | Sound (p<1%) | Moderate porosity | Sound (p<1%) | Not optimal |
| 900 | Sound (p<1%) | Sound (p<2%) | Sound (p<1%) | Optimal |
| 1000 | Very slight porosity | Slight porosity | Sound (p<1%) | Acceptable |
The numerical optimization strongly suggested that the improved feeding system would produce a sound casting. To confirm this, I manufactured two bearing frame castings (designated L1 and L2) using the new feeding system. The shell preheat temperature was maintained at 900°C, the pouring temperature at 1500°C, and the other process parameters were identical to those used in the original production. After casting, the components were subjected to X-ray inspection at the critical regions: the struts, the large end flange, the small end flange, and the inner bosses. The X-ray images showed no visible shrinkage porosity in any of these regions. In the original production, the large end flange always exhibited extensive dispersed porosity, and the inner bosses often had shrinkage cavities. After the improvement, the X-ray inspection indicated that the castings were essentially free from such sand foundry defects. Quantitatively, the number of detectable defects was reduced from more than sixty to less than fifteen across the entire casting. This substantial improvement validates the effectiveness of the simulation-driven redesign.
In addition to the internal quality assessment, I performed a comprehensive microstructural analysis of the castings. Samples were taken from the outer ring bosses, inner ring bosses, and the transitions between the struts and the rings. The sampling plan is summarized in Table 3. For each location, samples were examined in two conditions: (1) standard heat treatment (homogenization + solution + aging), and (2) hot isostatic pressing (HIP) followed by the same heat treatment. The heat treatment parameters were: homogenization at 1090°C for 3 hours, HIP at 1125°C and 125 MPa for 3 hours, solution treatment at 960°C for 1 hour, and aging at 700°C for 8 hours. The metallographic samples were polished and etched with a solution of 50 ml HCl, 50 ml H₂O, and 5 g CuSO₄. The grain size was evaluated using standard comparison methods, and the high-magnification microstructure was examined by optical microscopy.
| Sampling location | Specimen type | Mechanical test | Quantity |
|---|---|---|---|
| Outer large flange | Round bar Φ5 | Room-temperature tensile | 4 |
| Outer small flange | Round bar Φ5 | Room-temperature tensile | 4 |
| Inner large flange | Round bar Φ5 | Room-temperature tensile | 4 |
| Inner small flange | Round bar Φ5 | Room-temperature tensile | 4 |
| Strut sides | Flat 120×30×2 | Room-temperature tensile | 8 |
| Strut sides | Flat 120×35×2 | High-temperature creep | 8 |
| Outer large flange | Round bar Φ5 | High-temperature creep | 4 |
| Outer small flange | Round bar Φ5 | High-temperature creep | 4 |
| Inner large flange | Round bar Φ5 | High-temperature creep | 4 |
| Inner small flange | Round bar Φ5 | High-temperature creep | 4 |
| Outer boss | Metallographic | Microstructure | 8 |
| Inner boss | Metallographic | Microstructure | 8 |
| Strut-inner ring transition | Metallographic | Microstructure | 8 |
| Strut-outer ring transition | Metallographic | Microstructure | 8 |
The low-magnification metallographic examination revealed that the grain structure of the outer ring bosses was similar in both castings, with an average grain size of about 3 mm. The HIP treatment did not cause any noticeable grain growth. The high-magnification micrographs of the as-heat-treated samples showed the typical dendritic structure of K4169 alloy, with γ matrix, γ′ and γ″ precipitates, and a network of MC carbides, δ phase, and Laves phase in the interdendritic regions. The Laves phase, which is generally detrimental to mechanical properties, was present in significant amounts in the as-cast condition. After HIP, the Laves phase was largely dissolved, and the δ phase was significantly reduced. The microstructure became more homogeneous, which is beneficial for the stability of mechanical properties. The inner boss samples exhibited a similar behavior. The grain size in the inner boss region was approximately 2–3 mm, and no abnormal grain growth occurred during HIP. The strut-to-ring transition regions, which are thinner and cool faster, displayed a finer grain size of about 1–2 mm. In these regions, the initial microstructure also contained Laves phase and δ phase. After HIP, these phases were almost completely dissolved, leaving a clean microstructure with fine carbides and stable grain boundaries. The consistency between the two castings indicates that the improved feeding system provides reproducible microstructural quality.
The room-temperature tensile properties were evaluated on specimens taken from the designated locations. The tests were performed at a strain rate of 0.5 mm/min using an Instron 5569 universal testing machine. The results are presented in Table 4, which lists the ultimate tensile strength (UTS), yield strength (YS), elongation, and reduction of area for both heat-treated and HIPed samples. All measured values exceeded the technical requirements, which are a minimum UTS of 825 MPa, a minimum YS of 640 MPa, a minimum elongation of 5%, and a minimum reduction of area of 8%. The average UTS and YS of the L1 and L2 castings were in the range of 920–980 MPa and 780–830 MPa, respectively. The HIP treatment slightly decreased the UTS and YS but increased the elongation, which is typical for nickel-based superalloys because HIP eliminates micro-shrinkage and homogenizes the structure. The variation between the two castings was small, demonstrating good process repeatability.
| Condition | Casting | UTS (MPa) | YS (MPa) | Elongation (%) | Reduction of area (%) |
|---|---|---|---|---|---|
| Heat treated | L1 | 977.2 | 829.9 | 14.8 | 30.5 |
| L2 | 960.2 | 823.4 | 30.1 | 56.0 | |
| HIP + Heat treated | L1 | 920.5 | 787.8 | 16.3 | 32.0 |
| L2 | 981.8 | 814.8 | 14.2 | 37.3 | |
| Technical requirement | ≥825 | ≥640 | ≥5 | ≥8 | |
The high-temperature creep rupture tests were conducted at 650°C with an applied stress of 620 MPa. The results are summarized in Table 5. In the original production process, some locations, such as the outer large flange, showed a very short creep life of only 7 minutes, which was far below the required 30 minutes. The poor creep life was attributed to the severe shrinkage porosity that acted as crack initiation sites. After the feeding system improvement, all tested specimens exhibited creep life values exceeding 36 minutes, comfortably meeting the 30-minute requirement. The HIP treatment did not significantly change the creep life, which indicates that the castings were already sound and contained very few micro-porosities. The strut specimens displayed a creep life of about 36–37 minutes, which is consistent with the thin-wall geometry and the absence of macroscopic defects. The flange specimens showed improved life values between 36 and 43 minutes, with no large scatter. These results confirm that the elimination of sand foundry defects directly translates into a substantial improvement in high-temperature load-bearing capability.
| Location | Condition | Feeding system | Creep life (min) |
|---|---|---|---|
| Strut side | Heat treated | Original | 37:42 |
| Strut side | Heat treated | Improved | 36:10 |
| Strut side | HIP + Heat treated | Improved | 36:15 |
| Outer large flange | Heat treated | Original | 7:26 |
| Outer large flange | Heat treated | Improved | 43:24 |
| Outer large flange | HIP + Heat treated | Improved | 40:50 |
| Outer small flange | Heat treated | Original | 24:32 |
| Outer small flange | Heat treated | Improved | 41:43 |
| Outer small flange | HIP + Heat treated | Improved | 40:46 |
| Inner large flange | Heat treated | Original | 29:26 |
| Inner large flange | Heat treated | Improved | 38:43 |
| Inner large flange | HIP + Heat treated | Improved | 36:21 |
| Inner small flange | Heat treated | Original | 40:23 |
| Inner small flange | Heat treated | Improved | 39:27 |
| Inner small flange | HIP + Heat treated | Improved | 41:36 |
Fractographic examination of the creep rupture specimens provided further insight into the relationship between sand foundry defects and mechanical performance. The improved-casting specimens exhibited a transgranular fracture path along the dendrite arms, with limited necking and some fine dimples. No significant shrinkage porosity was observed on the fracture surface. In contrast, the original-casting specimens with short creep life showed pronounced necking and a fracture surface covered with shrinkage cavities and even cold shuts. These defects initiated cracks and accelerated the final rupture. The absence of such defects in the improved castings explains why the creep life exceeded the technical requirement even without HIP. Evidently, the numerical optimization of the feeding system was effective in eliminating the harmful macroscopic defects that governed the failure mechanism.
In summary, this study successfully solved the long-standing problem of shrinkage porosity in a large investment-cast K4169 bearing frame. The key conclusions can be drawn as follows. First, the original feeding system, with the large end flange downward, could not provide a favorable solidification sequence, leading to extensive sand foundry defects at the outer flanges and inner bosses. Second, varying the shell preheat temperature between 800°C and 1000°C was insufficient to eliminate these defects, as the fundamental issue was the feeding path and the lack of risers for the local hot spots. Third, by inverting the casting orientation (large end flange upward), attaching the ingates to the large flange, and placing cylindrical risers on the inner bosses, the numerical simulation predicted a dramatic reduction in shrinkage probability. Fourth, the actual castings produced with the improved feeding system exhibited a reduction in detectable defects from more than sixty to fewer than fifteen, confirming the simulation accuracy. Fifth, the microstructural analysis showed that the improved feeding system yielded consistent and stable grain structures, and HIP effectively dissolved the Laves phase and improved microstructural homogeneity. Finally, the room-temperature tensile properties and high-temperature creep life both satisfied the technical requirements, with creep life always exceeding 30 minutes. This study demonstrates that computer-aided casting simulation is a powerful tool for diagnosing and eliminating sand foundry defects in complex superalloy components. The methodology can be extended to other thin-walled investment castings with thick bosses and flanges, providing a robust route for defect-free production and reliable mechanical performance.