I studied a heavy-section nuclear main pump bearing support produced by sand casting. The component had an overall envelope of approximately 1000 mm × 1000 mm × 1103 mm, a wall thickness greater than 200 mm, and a mass close to 2.5 t. The material was ZG12MnMoV, a microalloyed cast steel equivalent to 12MDV6. Because the part operates under high pressure, high temperature, and radiation exposure, the allowable level of any casting defect was extremely low. The production route required ultrasonic inspection according to MC2000, and internal shrinkage cavity, porosity, and sand burn-on were unacceptable. My work therefore focused on understanding the casting defect formation mechanism, improving the foundry process, optimizing process parameters, and simulating the solidification microstructure.
1. Component and Material Characteristics
The bearing support is a thick and large casting with relatively uniform wall thickness but with a pronounced hot spot where the triangular ribs meet the vertical support face. This hot spot is critical because it remains liquid longer than the surrounding sections and is difficult to feed. The chemical composition and mechanical requirements of ZG12MnMoV are summarized below. The carbon content is low, and the manganese, molybdenum, and vanadium additions provide strength, hardenability, and grain refinement. However, the alloy also increases the tendency toward dendrite segregation and casting defect formation if the solidification sequence is not properly controlled.
| Element | C | Si | Mn | P | S | Mo | V |
|---|---|---|---|---|---|---|---|
| Content, wt.% | ≤0.15 | ≤0.60 | 1.20–1.70 | ≤0.025 | ≤0.020 | 0.20–0.40 | 0.05–0.10 |
| Property | Yield strength, MPa | Tensile strength, MPa | Elongation, % |
|---|---|---|---|
| ZG12MnMoV | ≥400 | ≥500 | ≥18 |
I noted that ZG12MnMoV passes through a peritectic region during solidification. The peritectic reaction can be incomplete, leading to microsegregation and a casting defect known as dendritic segregation. This type of casting defect reduces impact toughness and ductility and increases hot cracking sensitivity. Therefore, the casting process must not only avoid macro-shrinkage and sand burn-on but also produce a microstructure that can be homogenized by heat treatment. I used this requirement as a guiding condition throughout the process design and simulation work.
2. Original Process Design
I began with the original foundry process used in trial production. The original design selected a stepped gating system, self-hardening resin-bonded sand for the mold and cores, manual molding, and a water-based zircon coating applied by brushing. The casting was heat treated by quenching and tempering. The main process parameters are listed below.
| Process item | Original design |
|---|---|
| Pouring position | Large base plane facing downward |
| Solidification principle | Directional solidification |
| Gating system | Stepped gating system built with refractory tubes |
| Sprue diameter | 80 mm |
| Runner diameter | 80 mm |
| Ingate diameter | 50 mm |
| Pouring cup outlet diameter | 80 mm |
| Nozzle diameter | 45 mm |
| Riser system | One waist-shaped insulating blind riser on the arc top and two circular insulating blind risers on the triangular ribs |
| Chills | Two strip chills and eleven rectangular chills placed near thick sections |
| Coating | Water-based zircon coating, brushed once |
| Mold material | Self-hardening resin sand |
| Inspection | MC2000 ultrasonic inspection |
The original process intended to use the base plane facing downward so that the dense lower part of the mold would improve feeding and reduce inclusion entrapment. However, I found that this orientation made the hot spot at the triangular-rib-to-vertical-face junction difficult to feed. The circular blind risers on the triangular ribs were limited by the rib thickness, and their modulus was smaller than the local hot-spot modulus. As a result, the risers could not provide sufficient liquid metal to compensate for liquid and solidification shrinkage. This became the primary source of the casting defect observed in trial production.
3. Simulation Framework and Governing Equations
I used UG for three-dimensional modeling, exported the assemblies as STL files, and imported them into the casting simulation software. The simulations included pure solidification calculation, pure filling calculation, and coupled filling-and-thermal calculation. The mold, cores, chills, riser sleeves, and casting were assigned priority levels in the order casting, core, chill, riser sleeve, and mold. For pure solidification, I used a fine mesh of about 8 million cells to resolve the shrinkage cavity and porosity. For filling and coupled calculations, I reduced the mesh to about 600,000 cells to keep the computation efficient.
The thermal field was solved using the transient heat conduction equation:
$$ \rho c_p \frac{\partial T}{\partial t} = \nabla \cdot \left( k \nabla T \right) + Q $$
where \(\rho\) is density, \(c_p\) is specific heat, \(T\) is temperature, \(t\) is time, \(k\) is thermal conductivity, and \(Q\) is the latent heat source term. The latent heat released during solidification was handled through an effective specific heat or enthalpy method. For the filling process, I used the continuity equation and the momentum equation for an incompressible Newtonian fluid:
$$ \nabla \cdot \mathbf{u} = 0 $$
$$ \rho \left( \frac{\partial \mathbf{u}}{\partial t} + \mathbf{u} \cdot \nabla \mathbf{u} \right) = -\nabla p + \mu \nabla^2 \mathbf{u} + \rho \mathbf{g} $$
Here \(\mathbf{u}\) is the velocity vector, \(p\) is pressure, \(\mu\) is dynamic viscosity, and \(\mathbf{g}\) is gravitational acceleration. The solid fraction \(f_s\) was estimated from the liquidus and solidus temperatures:
$$ f_s = \frac{T_L – T}{T_L – T_S} $$
where \(T_L\) and \(T_S\) are the liquidus and solidus temperatures, respectively. For shrinkage porosity prediction, I used the Niyama criterion:
$$ N = \frac{G}{\sqrt{\dot{T}}} $$
where \(G\) is the temperature gradient and \(\dot{T}\) is the cooling rate. A smaller Niyama value indicates a higher probability of shrinkage porosity. I also quantified the total shrinkage cavity volume and the total Niyama porosity volume from the post-processing module. These two quantities served as the main responses in the process optimization.
| Process scheme | Mesh size | Number of cells | Calculation type |
|---|---|---|---|
| Original process | 7.7 mm | 8,172,840 | Pure solidification |
| Improved process | 7.8 mm | 8,250,468 | Pure solidification |
| Filling and coupled cases | Variable | About 600,000 | Filling and coupled thermal-flow |
4. Original Process Simulation and Casting Defect Analysis
I first simulated the original process to identify the location and cause of the casting defect. The temperature field during filling showed that the base corners away from the ingates had the lowest temperature because of the chills. The central part of the base remained hotter because of its larger thickness. When filling reached the upper part, the lowest temperature appeared on the two sides of the triangular ribs, indicating that the chills were acting. However, the hot spot at the junction between the triangular ribs and the vertical support face remained liquid for a long time. The risers on the ribs were too small and too weakly insulated to feed this hot spot.
The velocity field showed that the metal velocity was relatively high near the ingate and near the rib-to-riser connection. High local velocity can cause mold erosion, gas entrapment, and inclusion entrapment, all of which can become a casting defect. The filling sequence also showed that the base filled slowly because of its large volume. After the base was filled, the metal rose into the vertical body and finally into the top risers. The upward flow helped float some inclusions into the risers, but the hot spot at the rib junction was not adequately fed.
The solidification simulation gave the most important result. The top risers fed the arc top region well, but the circular blind risers on the triangular ribs had insufficient modulus. The final solidification result showed a large amount of shrinkage cavity inside the triangular ribs. Ultrasonic inspection confirmed that the trial casting contained internal shrinkage cavity and porosity. This was a fatal casting defect for a nuclear pump component.
I concluded that the original process failed because the feeding system did not match the hot-spot modulus. The casting defect was not caused by a single parameter but by the combination of pouring position, riser size, riser type, and chill placement. The original orientation placed the largest hot spot in a position where a sufficiently large riser could not be placed. The blind risers could not be enlarged because of the rib geometry, and their insulating sleeves were too thin to maintain feeding for the required time. The chills helped cool the ribs but could not replace the missing liquid metal. Therefore, I decided to change the pouring position and redesign the riser system.
5. Improved Process Design and Simulation
I inverted the casting so that the large base plane faced upward. This change allowed a large open insulating riser to be placed on the base directly above the hot spot. The new riser had a diameter of 560 mm, a height of 900 mm, and an insulating sleeve thickness of 112 mm. Because the riser was placed on a flat upper surface, it was easy to mold, easy to remove, and capable of supplying enough liquid metal. I also removed the small blind risers on the triangular ribs. The new riser acted as a feeder, an exhaust path, and a slag trap.
| Process item | Improved design |
|---|---|
| Pouring position | Casting inverted, large base plane facing upward |
| Riser type | Large open insulating riser |
| Riser dimensions | Φ560 mm × 900 mm |
| Riser sleeve thickness | 112 mm |
| Rib chills | 24 cylindrical chills, Φ30 mm × 70 mm, placed along the fillet at the rib-to-body junction |
| Base and arc chills | 10 strip chills, 180 mm × 120 mm × 90 mm, placed on four circular bosses and the top arc |
| Coating | Water-based zircon coating, brushed five times, local chromite facing sand |
| Facing sand | Chromite sand, 40–70 mesh, thickness 40–50 mm, at the hot spot |
| Backing sand | Furan resin sand |
The improved filling simulation showed a smoother and more progressive filling pattern. The metal first filled the lower body, then rose into the base, and finally entered the upper riser. The velocity remained moderate throughout most of the filling stage, which reduced the risk of mold erosion and inclusion entrapment. The temperature field showed that the chills on the base bosses and the top arc produced controlled cooling. The lowest temperature reached about 1493°C in the chill regions, while the bulk of the casting stayed above the liquidus long enough to avoid cold shuts. The top riser remained hot until the end of solidification.
The improved solidification simulation showed that the hot spot at the rib junction solidified before the riser. The large open riser continued to feed liquid metal into the hot spot, and the final shrinkage cavity was confined to the riser. The casting body showed no internal shrinkage cavity or porosity. The simulation result agreed with production trial results. After the improved process was used, ultrasonic inspection found a sound internal structure, and the casting met the customer requirements. This confirmed that the redesigned riser and chill system eliminated the casting defect.
6. Coating and Sand Burn-On Control
Even after the shrinkage cavity and porosity were eliminated, sand burn-on remained a serious problem at the hot spot. The burn-on was especially severe where the triangular ribs met the vertical support face. This region remained at a high temperature for a long time, and the original zircon coating was penetrated by the liquid steel. The mechanism was both mechanical and chemical. Liquid steel entered the pores of the sand through capillary action, and iron oxide reacted with silica to form low-melting silicates. The reaction products bonded strongly to the casting surface and produced a hard casting defect that was difficult to remove.
I analyzed the coating failure and found that the zircon coating decomposed at the hot spot because the local temperature stayed above 1540°C for an extended period. The decomposition produced silica and zirconia. The silica reacted with iron oxide to form fayalite, which is a low-melting compound. Once the refractory filler was consumed, the coating layer was broken through, and the steel directly contacted the molding sand. The result was severe penetration burn-on. In other regions, the casting cooled quickly enough that the zircon coating formed a dense sintered ceramic layer, which prevented burn-on.
| Coating item | Original process | Improved process |
|---|---|---|
| Carrier | Water | Water |
| Refractory filler | Zircon powder | Zircon powder |
| Binder | Molasses | Water-soluble phenolic resin and silica sol |
| Application method | Brushing | Brushing |
| Number of coats | 1 | 5 |
| Coating thickness | Not controlled | 1.2–1.5 mm |
| Hot-spot sand | Resin sand | Chromite sand facing, 40–50 mm thick |
| Result | Severe burn-on casting defect | Burn-on eliminated |
I proposed using chromite sand as a local facing sand at the hot spot. Chromite sand consists mainly of FeO·Cr₂O₃, has a refractoriness above 1900°C, and has high thermal conductivity. Under the action of liquid steel, chromite sand precipitates iron oxide, which softens plastically and changes the sand-to-sand contact from point contact to surface contact. The sand grains expand and press against each other, reducing the interstitial spaces and blocking steel penetration. The sintered chromite layer remains as a barrier between the liquid steel and the backing sand. Because chromite sand is expensive, I used it only at the hot spot, with a thickness of 40–50 mm. The backing sand remained furan resin sand. The entire mold cavity was then coated with zircon coating. I also changed the binder from molasses to water-soluble phenolic resin and silica sol because molasses loses its binding strength at high temperature and can ferment in summer. The brushing operation was repeated five times to obtain a coating thickness of 1.2–1.5 mm. Production trials confirmed that this coating and facing sand strategy eliminated the burn-on casting defect and greatly reduced cleaning labor.
7. Uniform Design Optimization of Process Parameters
After the geometry and coating improvements, I optimized the pouring parameters to further reduce the casting defect tendency. I selected pouring temperature \(A\), pouring time \(B\), and mold initial temperature \(C\) as the three factors. Each factor was assigned five levels. The ranges were 1550–1570°C for pouring temperature, 80–120 s for pouring time, and 10–50°C for mold initial temperature. I used a uniform design table \(U_{10}^*(8^{10})\) because it provides good uniformity with only ten experiments. The selected columns were 1, 5, and 6 according to the usage table. The factor levels and design matrix are given below.
| Level | A: Pouring temperature, °C | B: Pouring time, s | C: Mold initial temperature, °C |
|---|---|---|---|
| 1 | 1550 | 80 | 10 |
| 2 | 1555 | 90 | 20 |
| 3 | 1560 | 100 | 30 |
| 4 | 1565 | 110 | 40 |
| 5 | 1570 | 120 | 50 |
| Experiment | A, °C | B, s | C, °C |
|---|---|---|---|
| 1 | 1550 | 120 | 20 |
| 2 | 1555 | 120 | 30 |
| 3 | 1560 | 110 | 50 |
| 4 | 1565 | 110 | 10 |
| 5 | 1570 | 100 | 20 |
| 6 | 1550 | 100 | 40 |
| 7 | 1555 | 90 | 50 |
| 8 | 1560 | 90 | 10 |
| 9 | 1565 | 80 | 30 |
| 10 | 1570 | 80 | 40 |
I used the total shrinkage cavity volume and the total Niyama porosity volume as the responses. The shrinkage cavity volume results are listed below.
| Experiment | A, °C | B, s | C, °C | Shrinkage cavity volume, cc |
|---|---|---|---|---|
| 1 | 1550 | 120 | 20 | 25,680 |
| 2 | 1555 | 120 | 30 | 25,940 |
| 3 | 1560 | 110 | 50 | 26,370 |
| 4 | 1565 | 110 | 10 | 26,640 |
| 5 | 1570 | 100 | 20 | 26,980 |
| 6 | 1550 | 100 | 40 | 25,960 |
| 7 | 1555 | 90 | 50 | 26,430 |
| 8 | 1560 | 90 | 10 | 26,620 |
| 9 | 1565 | 80 | 30 | 27,140 |
| 10 | 1570 | 80 | 40 | 27,420 |
The Niyama porosity volume results are listed below. These values were used to check whether the same factors also controlled microporosity.
| Experiment | A, °C | B, s | C, °C | Niyama porosity volume, cc |
|---|---|---|---|---|
| 1 | 1550 | 120 | 20 | 4,450 |
| 2 | 1555 | 120 | 30 | 4,523 |
| 3 | 1560 | 110 | 50 | 4,432 |
| 4 | 1565 | 110 | 10 | 5,482 |
| 5 | 1570 | 100 | 20 | 5,931 |
| 6 | 1550 | 100 | 40 | 7,157 |
| 7 | 1555 | 90 | 50 | 6,956 |
| 8 | 1560 | 90 | 10 | 6,742 |
| 9 | 1565 | 80 | 30 | 5,920 |
| 10 | 1570 | 80 | 40 | 6,482 |
I analyzed the results with a statistical software package. First, I checked the interactive effects using a factorial design. The main effects and two-factor interactions were evaluated. The results showed that pouring temperature and pouring time had significant effects on the shrinkage cavity volume, while mold initial temperature had a very small effect. The two-factor interactions were not significant. The Niyama porosity response was mainly controlled by pouring time, while pouring temperature and mold initial temperature had little significant effect. Therefore, I built a linear regression model for the shrinkage cavity volume.
The initial regression equation was:
$$ Y = -56022 + 53.9A – 15.9B + 1.21C $$
The analysis of variance gave \(R^2 = 99.6\%\), and the model probability was zero, indicating that the regression was highly significant. The coefficients and their probabilities are summarized below.
| Predictor | Coefficient | Standard error | T value | P value |
|---|---|---|---|---|
| Constant | -56022 | 3773 | -14.85 | 0.000 |
| A | 53.905 | 2.368 | 22.77 | 0.000 |
| B | -15.881 | 1.184 | -13.42 | 0.000 |
| C | 1.214 | 1.046 | 1.16 | 0.290 |
| Source | Degrees of freedom | Sum of squares | Mean square | F value | P value |
|---|---|---|---|---|---|
| Regression | 3 | 2,805,524 | 935,175 | 508.42 | 0.000 |
| Residual error | 6 | 11,036 | 1,839 | ||
| Total | 9 | 2,816,560 |
Because the coefficient for \(C\) was not significant, I removed it from the model. The refined regression equation was:
$$ Y = -54421 + 52.9A – 16.4B $$
The refined model had \(R^2 = 99.5\%\). The F values for \(A\) and \(B\) were 1244.17 and 208.19, respectively, both much larger than the critical value \(F_{0.05}(1,7)=5.59\). This confirmed that pouring temperature had the strongest effect, followed by pouring time. To minimize the shrinkage cavity volume, \(A\) should be as low as possible and \(B\) should be as long as possible within the tested range. The mold initial temperature should also be as low as possible according to the first regression equation. Therefore, the optimal combination was:
$$ A = 1550^\circ\text{C}, \quad B = 120\text{ s}, \quad C = 10^\circ\text{C} $$
I performed an additional verification experiment with this combination. The predicted shrinkage cavity volume was lower than all ten original experiments. The simulated total shrinkage cavity volume was 25.66 cc, which confirmed the optimum. I also noted that the Niyama porosity response did not provide a suitable regression model because only pouring time was significant; therefore, I did not use it to determine the final parameter set. The final optimized parameters reduced the casting defect risk and provided a reliable basis for production.
8. Microstructure Simulation of ZG12MnMoV
After the macro-scale casting defect control, I studied the microstructure of ZG12MnMoV. The alloy solidifies through a peritectic reaction, and the peritectic transformation is often incomplete. This leads to dendritic segregation, which is a micro-scale casting defect. Dendritic segregation reduces impact toughness and ductility and increases hot cracking tendency. The secondary dendrite arm spacing is a key parameter because it controls the diffusion distance for homogenization. A smaller secondary dendrite arm spacing means less segregation, a finer structure, and easier heat treatment. I used the microstructure module of Procast to simulate the secondary dendrite arm spacing as a function of pouring temperature.
| Property | Value |
|---|---|
| Density | 7.9 g/cm³ |
| Liquidus temperature | 1511°C |
| Solidus temperature | 1469°C |
| Pouring velocity | 1.2 m/s |
| Calculation steps | 5000 |
| Stop temperature | 300°C |
| Micro module | Enabled |
The secondary dendrite arm spacing was calculated using the classic coarsening relationship:
$$ \lambda_2 = A \dot{T}^{-n} $$
where \(\lambda_2\) is the secondary dendrite arm spacing, \(\dot{T}\) is the local cooling rate, \(A\) is a material constant, and \(n\) is an exponent typically between 0.30 and 0.50 for cast steels. The initial grain radius can be related to the nucleation density and cooling rate. I compared three pouring temperatures: 1550°C, 1565°C, and 1580°C. The mesh had 445,088 elements and 89,219 nodes. The thermal properties, density, enthalpy, viscosity, and solid fraction were temperature dependent.
| Pouring temperature | Initial grain radius | Secondary dendrite arm spacing | Microsegregation tendency | Overall assessment |
|---|---|---|---|---|
| 1550°C | Smallest | Smallest | Lowest | Best refined structure |
| 1565°C | Intermediate | Intermediate | Moderate | Acceptable but coarser |
| 1580°C | Largest | Largest | Highest | Coarse structure and higher casting defect risk |
The simulation showed that increasing the pouring temperature increased both the initial grain radius and the secondary dendrite arm spacing. At 1550°C, the secondary dendrite arm spacing was the smallest over the entire section, including the bottom arc and the central region near the riser-to-base junction. The distribution was also more uniform. At 1580°C, the coarse dendrite regions expanded significantly, and the microsegregation tendency increased. This means that a high pouring temperature increases the risk of a micro-scale casting defect and makes subsequent heat treatment less effective. I therefore selected 1550°C as the optimal pouring temperature for ZG12MnMoV. This temperature agrees with the macro-scale optimization result, where the lowest pouring temperature minimized the shrinkage cavity volume.
9. Discussion of Casting Defect Control Strategy
The results from the macro-scale and micro-scale simulations can be combined into a single strategy for controlling the casting defect. First, the pouring position must place the largest hot spot where a sufficiently large riser can be placed. Second, the riser modulus must exceed the hot-spot modulus, and the riser sleeve must be thick enough to keep the riser liquid until the casting has solidified. Third, chills should be placed to promote directional solidification without causing excessive local cooling that could lead to cold shuts. Fourth, the coating and facing sand must resist penetration at the hot spot, because a coating failure can produce a burn-on casting defect even when the feeding system is correct. Fifth, the pouring parameters should be selected to minimize shrinkage cavity volume while keeping the filling process smooth. Sixth, the pouring temperature should be kept as low as practical to reduce dendritic segregation and refine the microstructure.
I found that the casting defect formation was not controlled by a single factor. The original process failed because the riser modulus, pouring position, and coating system were not matched to the thick-section geometry. When I inverted the casting and placed a large open riser on the base, the macro-shrinkage casting defect was eliminated. When I used local chromite facing sand and a thicker zircon coating, the burn-on casting defect was eliminated. When I used uniform design and regression analysis, the pouring parameters were optimized to further reduce the shrinkage cavity volume. When I used the microstructure module, the pouring temperature was confirmed to be optimal at 1550°C for minimizing microsegregation. The combination of these measures produced a sound casting that met ultrasonic inspection requirements.
10. Summary of Findings
I summarize the main findings in the following table.
| Area | Main finding | Result |
|---|---|---|
| Original process | Small blind risers on triangular ribs had insufficient modulus | Shrinkage cavity and porosity casting defect inside ribs |
| Improved process | Inverted casting with large open insulating riser and redesigned chills | Shrinkage cavity moved to riser; sound casting body |
| Coating | Zircon coating broke down at hot spot; chromite facing sand formed a barrier | Burn-on casting defect eliminated |
| Uniform design | Pouring temperature and pouring time significantly affected shrinkage cavity volume | Optimal parameters: 1550°C, 120 s, 10°C |
| Microstructure | Lower pouring temperature reduced secondary dendrite arm spacing | 1550°C gave the finest and most uniform structure |
The optimized process was verified by simulation and production trial. The casting showed no unacceptable internal shrinkage cavity or porosity, and the surface burn-on casting defect was eliminated. The microstructure simulation provided a theoretical basis for selecting the pouring temperature to minimize microsegregation. The work demonstrated that a combination of numerical simulation, uniform design, regression analysis, and coating optimization can effectively control casting defect formation in heavy-section microalloyed steel castings.
For future work, I would recommend studying the temperature-dependent interfacial heat transfer coefficient more accurately, because the simulation used constant values in some stages. I would also recommend investigating the effect of small changes in coating composition on burn-on resistance. Finally, more microstructural simulation tools should be applied to heavy-section cast steel so that dendritic segregation and other micro-scale casting defects can be predicted and controlled more precisely.