I approached the production of a centrifugal pump volute as a casting defect control problem rather than a simple shape-forming task. The volute is a pressure-containing component, and any internal casting defect such as shrinkage porosity, gas entrapment, or oxide inclusion can reduce hydraulic efficiency, initiate fatigue cracking, and compromise operating stability. Because the component is produced by sand casting in small batches, I needed a process design that was robust, inexpensive, and repeatable. I therefore combined geometric analysis, gating system design, thermal reasoning, and FLOW-3D numerical simulation to identify and eliminate casting defect risks before production. The alloy was ASTM A744 Gr CF8M, a cast stainless steel with good corrosion resistance and acceptable strength at moderate temperature. The casting envelope was approximately 480.86 mm by 464.91 mm by 292.38 mm, with a maximum wall thickness of 57.51 mm, a minimum wall thickness of 12.51 mm, an average wall thickness near 16.00 mm, and a rough casting weight of 20.67 kg. Those dimensions created a strongly non-uniform thermal field. Thick flange and support regions cooled slowly, while thin spiral walls cooled quickly. This thermal contrast is one of the most common origins of a casting defect in pump housings.
The volute geometry consists of a spiral volute chamber, inlet and outlet flanges, reinforcing ribs, bosses, and a support base. The internal flow passage is a smooth curved surface, so the sand core must remain stable during filling. If the liquid metal enters with excessive velocity, it can erode the sand core, fold over itself, entrain air, and generate loose oxide skins. Those oxide skins are later trapped in the solidifying shell and appear as a casting defect after machining or radiographic inspection. I therefore treated velocity control, stream-line continuity, and directional solidification as the three main levers for casting defect reduction. The central cavity and curved passage also mean that a simple top-gating design would be risky. A top pour would direct liquid metal against the core, increasing sand erosion and causing turbulence. I selected a bottom-gated, open gating system for the initial production trial, with a pouring time of 7 s, a pouring temperature of 1560 degrees Celsius, and a minimum choke-area ratio of approximately 0.77 to 1 between the sprue and the ingates. The ingate was placed at the inlet flange, and a riser was positioned above the spiral volute to improve feeding.
Before modifying the process, I analyzed the thermal and flow behavior of that initial design. The first goal was to determine whether the filling pattern was smooth enough to avoid a casting defect caused by surface turbulence. The second goal was to determine whether the feeding path remained open long enough to prevent shrinkage porosity. The third goal was to see whether oxide inclusions could be floated into the riser or trapped in a controlled location. I used a transient three-dimensional volume-of-fluid model with a standard turbulence closure. The mass conservation equation was solved in the form
$$ \frac{\partial \rho}{\partial t} + \nabla \cdot (\rho \mathbf{u}) = 0 $$
where \(\rho\) is the mixture density and \(\mathbf{u}\) is the velocity vector. The momentum equation included gravity, pressure, viscous stress, and the surface tension source term:
$$ \frac{\partial (\rho \mathbf{u})}{\partial t} + \nabla \cdot (\rho \mathbf{u} \mathbf{u}) = -\nabla p + \nabla \cdot \boldsymbol{\tau} + \rho \mathbf{g} + \mathbf{F}_{s} $$
The energy transport equation was coupled to the solidification model through the enthalpy:
$$ \frac{\partial (\rho H)}{\partial t} + \nabla \cdot (\rho \mathbf{u} H) = \nabla \cdot (k \nabla T) + S_{h} $$
where \(H\) is the total enthalpy, \(k\) is thermal conductivity, \(T\) is temperature, and \(S_{h}\) is a source term for phase change and boundary conditions. The free surface was tracked with the volume fraction \(\alpha\):
$$ \frac{\partial \alpha}{\partial t} + \nabla \cdot (\alpha \mathbf{u}) = 0 $$
For turbulence, I used a two-equation model because a fully laminar assumption would underestimate the mixing and air entrainment that produce a casting defect in sand castings. The turbulent kinetic energy equation was written as
$$ \frac{\partial (\rho k)}{\partial t} + \nabla \cdot (\rho k \mathbf{u}) = \nabla \cdot \left( \frac{\mu_{t}}{\sigma_{k}} \nabla k \right) + G_{k} – \rho \varepsilon $$
and the dissipation equation was written as
$$ \frac{\partial (\rho \varepsilon)}{\partial t} + \nabla \cdot (\rho \varepsilon \mathbf{u}) = \nabla \cdot \left( \frac{\mu_{t}}{\sigma_{\varepsilon}} \nabla \varepsilon \right) + C_{1} \frac{\varepsilon}{k} G_{k} – C_{2} \rho \frac{\varepsilon^{2}}{k} $$
I also monitored the local Reynolds number, because the transition from ordered filling to chaotic filling is a practical indicator of casting defect formation:
$$ Re = \frac{\rho u D}{\mu} $$
The critical velocity for cast steel was taken as approximately 0.50 m/s. Below that value, the free surface can remain relatively calm; above it, the liquid metal can fold, splash, and entrain oxide films. I used this threshold as an engineering criterion, not as an absolute law, because local geometry and surface tension also matter. For a pump volute with internal cores, however, the threshold is a useful screening tool for casting defect risk.
The initial geometry and material data are summarized in Table 1. I used these values to calculate local modulus, cooling rate, and feeding distance. The modulus \(M\) of a section is
$$ M = \frac{V}{A} $$
where \(V\) is the volume and \(A\) is the heat-transfer surface area. A large modulus indicates a slow-cooling hot spot, which is a likely location for shrinkage porosity. The solidification time \(t_{s}\) can be estimated with Chvorinov’s rule:
$$ t_{s} = B \left( \frac{V}{A} \right)^{n} $$
where \(B\) is a mold constant and \(n\) is commonly close to 2. This relation explains why the thick inlet flange and support base are more prone to a shrinkage casting defect than the thin spiral wall. I also used the Niyama criterion to interpret feeding potential:
$$ N = \frac{G}{\sqrt{R}} $$
where \(G\) is the temperature gradient and \(R\) is the cooling rate. Low \(N\) values indicate a high probability of micro-shrinkage. The Niyama criterion does not replace simulation, but it helped me rank the hot spots in the volute.
| Parameter | Value | Unit | Effect on casting defect risk |
|---|---|---|---|
| Alloy | ASTM A744 Gr CF8M | – | Stainless steel, pressure-containing, oxide-forming |
| Envelope length | 480.86 | mm | Large sand core, multiple thermal regions |
| Envelope width | 464.91 | mm | Non-symmetric filling path |
| Envelope height | 292.38 | mm | Vertical feeding distance |
| Maximum wall thickness | 57.51 | mm | Slow cooling, hot spot |
| Minimum wall thickness | 12.51 | mm | Fast cooling, premature freezing |
| Average wall thickness | 16.00 | mm | Thin-wall dominated solidification |
| Rough casting weight | 20.67 | kg | Small steel casting, limited feeding reservoir |
| Pouring temperature | 1560 | degrees Celsius | High superheat, more oxide risk |
| Initial pouring time | 7 | s | Short fill, high velocity |
The initial gating design was a bottom-gated open system. I chose bottom gating to avoid direct core impingement. The sprue-to-ingate area ratio was set at approximately 0.77 to 1, which is an open ratio intended to reduce pressure and velocity in the mold cavity. In practice, however, the simulation showed that the liquid metal still reached velocities well above the critical value. The inlet flange received a high-speed jet, and the two streams that entered the flange region merged and formed a turbulent zone. That turbulent zone became the main source of a casting defect in the form of oxide inclusions and entrained gas. Table 2 lists the initial process parameters. Table 3 summarizes the initial gating ratios and the resulting risks.
| Simulation input | Value | Unit |
|---|---|---|
| Filling time | 7 | s |
| Pouring temperature | 1560 | degrees Celsius |
| Sprue-to-ingate ratio | 0.77 to 1 | – |
| Gating type | Bottom, open | – |
| Ingate location | Inlet flange | – |
| Riser location | Above spiral volute | – |
| Turbulence model | Two-equation eddy viscosity | – |
| Free-surface method | Volume of fluid | – |
| Critical velocity | 0.50 | m/s |
| Initial gating feature | Design value | Observed or predicted behavior | Casting defect risk |
|---|---|---|---|
| Sprue outlet | Direct drop | Liquid metal accelerated downward | High turbulence |
| Runner | No separate runner | Poor slag interception | Oxide inclusion |
| Ingate | Inlet flange | High-speed jet into flange | Gas entrapment |
| Riser | Above spiral volute | Partial feeding only | Shrinkage porosity |
| Choke ratio | 0.77 to 1 | Open system, low back pressure | Unstable filling |
The first simulation result I examined was the velocity field. At 3 percent filling, the liquid metal fell rapidly through the sprue and struck the bottom of the mold. A large part of the vertical kinetic energy was converted into horizontal velocity, and the maximum velocity reached about 1.014 m/s. The inflow was turbulent, and air was entrained almost immediately. At 5 percent filling, the metal entered the inlet flange with a maximum velocity near 1.139 m/s. This was more than twice the critical velocity of 0.50 m/s. At 10 percent filling, the stream split and rejoined inside the flange, creating a strong recirculation zone. At 30 percent filling, the metal entered the spiral volute at about 0.70 m/s. That value was still above the critical velocity, so the filling mode remained splash-like and turbulent. At 50 percent filling, the rising metal encountered increasing resistance, and the velocity dropped to about 0.30 m/s. At 70 percent filling, the free surface became smoother and no strong turbulence was observed. The late-stage calmness did not compensate for the early turbulence, because most of the oxide film and gas entrapment had already formed. Table 4 summarizes the velocity history and the associated casting defect risk.
| Filling stage | Maximum velocity | Unit | Flow condition | Casting defect risk |
|---|---|---|---|---|
| 3 percent | 1.014 | m/s | Sprue jet and splash | Air entrainment |
| 5 percent | 1.139 | m/s | High-speed flange entry | Oxide film entrapment |
| 10 percent | 1.000 to 1.139 | m/s | Stream splitting and merging | Turbulent recirculation |
| 30 percent | 0.700 | m/s | Spiral volute splash filling | Gas and slag entrapment |
| 50 percent | 0.300 | m/s | Slowing rise | Reduced but not eliminated |
| 70 percent | 0.300 or less | m/s | Smooth free surface | Late-stage calm |
The oxide slag simulation confirmed what the velocity field suggested. At 7 percent filling, the liquid metal passed directly from the sprue into the ingate, and the slag density reached about 100.00 kg/m3. At 9 percent and 15 percent filling, the two streams collided at the inlet flange, producing a turbulent zone with an oxide density near 280.00 kg/m3. At 30 percent and 50 percent filling, the metal entered the spiral volute in a splashing mode, and the maximum oxide density reached about 450.00 kg/m3. A large amount of oxide slag entered the outlet flange and the spiral chamber. At 100 percent filling, some of the slag floated into the riser, but a fraction remained trapped near the outlet flange. That residual slag is a classical casting defect in a pressure-containing pump component. Table 5 lists the oxide density history. Table 6 gives the corresponding qualitative assessment.
| Filling stage | Maximum oxide density | Unit | Main location | Likely consequence |
|---|---|---|---|---|
| 7 percent | 100.00 | kg/m3 | Sprue and ingate | Initial oxide entrainment |
| 9 percent | 280.00 | kg/m3 | Inlet flange | Turbulent collision |
| 15 percent | 280.00 | kg/m3 | Inlet flange | Gas and slag mixing |
| 30 percent | 450.00 | kg/m3 | Spiral volute and outlet flange | Trapped inclusion |
| 50 percent | 450.00 | kg/m3 | Spiral volute | Slag retention |
| 100 percent | Variable | kg/m3 | Riser and outlet flange | Partial flotation, residual casting defect |
| Defect type | Formation mechanism | Initial risk | Detection method |
|---|---|---|---|
| Gas porosity | Air entrainment during turbulent filling | High | Radiography |
| Oxide inclusion | Surface folding and slag entrapment | High | Radiography and machining |
| Shrinkage porosity | Isolated liquid zone with insufficient feeding | Medium to high | Radiography and pressure test |
| Sand erosion | High-velocity impingement on core | Medium | Visual and dimensional inspection |
| Cold shut | Low local temperature and poor fusion | Low to medium | Visual and penetrant inspection |
The solidification sequence showed a second class of casting defect risk. At 5 percent solidification, the thin outer contour, edges, and bottom regions began to freeze first. The average metal temperature was about 1394.50 degrees Celsius. This is consistent with the expected outside-in cooling pattern and provides a basis for directional solidification. At 20 percent and 40 percent solidification, the thin walls were nearly frozen, while a small amount of liquid remained in the central region and the support frame. At that stage, the central region and the gating path had already disconnected, forming an isolated liquid zone. An isolated liquid zone cannot be fed effectively, so it is a direct precursor to shrinkage porosity. At 60 percent solidification, most of the cavity had frozen, but an isolated liquid zone remained at the inlet flange. At 70 percent and 90 percent solidification, the riser was still able to supply some liquid, but the inlet flange hot spot remained a concern. The initial riser system was therefore only partially successful. It could feed the upper spiral region, but it could not reliably feed the lower inlet flange. Table 7 summarizes the solidification sequence and the feeding condition.
| Solidification stage | Average temperature | Unit | Frozen region | Remaining liquid | Casting defect risk |
|---|---|---|---|---|---|
| 5 percent | 1394.50 | degrees Celsius | Thin outer contour, edges, bottom | Most of the casting | Low at this stage |
| 20 percent | Below solidus locally | degrees Celsius | Thin walls | Center and support frame | Isolated liquid zone begins |
| 40 percent | Below solidus locally | degrees Celsius | Thin walls and outer shell | Center and support frame | Shrinkage porosity risk |
| 60 percent | Below solidus locally | degrees Celsius | Most of the cavity | Inlet flange | Hot spot and shrinkage |
| 70 percent | Decreasing | degrees Celsius | Riser and upper regions | Inlet flange | Partial feeding |
| 90 percent | Decreasing | degrees Celsius | Riser neck | Small isolated zones | Possible micro-shrinkage |
Because the initial process produced a high risk of a casting defect from both filling and feeding, I optimized the gating system in two stages. The first optimized scheme, which I call Scheme 1, kept the same pouring time and temperature but added a runner and an extension segment. The runner height was 40.00 mm, approximately twice the ingate height. The gating ratio was changed to a semi-closed system with a minimum area ratio of approximately 0.9 to 0.8 to 1 for sprue, runner, and ingate. I also added a pouring basin and a sprue well below the sprue. The sprue well had the same diameter as the sprue and was intended to absorb part of the vertical momentum and turn the flow smoothly into the runner. The ingate was introduced tangentially to the inlet flange so that the metal would swirl along the flange instead of striking it as a jet. Table 8 lists the Scheme 1 design parameters. Table 9 compares the initial design and Scheme 1.
| Scheme 1 parameter | Value | Unit | Purpose |
|---|---|---|---|
| Pouring time | 7 | s | Maintain productivity |
| Pouring temperature | 1560 | degrees Celsius | Maintain fluidity |
| Runner height | 40.00 | mm | Increase slag residence time |
| Runner-to-ingate height ratio | 2 to 1 | – | Reduce ingate velocity |
| Gating ratio | 0.9 to 0.8 to 1 | – | Semi-closed system |
| Sprue well | Same diameter as sprue | mm | Absorb impact |
| Ingate direction | Tangential to inlet flange | – | Prevent jet impingement |
| Runner extension | Included | – | Collect slag |
| Feature | Initial design | Scheme 1 | Expected effect on casting defect |
|---|---|---|---|
| Runner | None | Added | Improved slag interception |
| Sprue well | None | Added | Reduced splashing |
| Gating ratio | 0.77 to 1 | 0.9 to 0.8 to 1 | More stable pressure |
| Ingate entry | Direct | Tangential | Less turbulence |
| Flow pattern | Splitting and collision | Guided swirl | Lower oxide entrainment |
The Scheme 1 velocity field improved markedly. At the beginning of filling, the sprue well absorbed part of the kinetic energy, and the velocity in the runner dropped to about 0.90 m/s. The flow was stable, with no pronounced splash or turbulence. When the metal entered the ingate, the velocity was about 0.88 m/s. This was still slightly above the critical value, so some gas entrainment remained possible. The metal entered the inlet flange in a stable condition and did not split into two colliding streams. In the spiral volute, the splash-type turbulence disappeared, and the filling front advanced more evenly. At the end of filling, the velocity decreased to about 0.15 m/s, and the free surface rose smoothly. The remaining problem was the ingate velocity, which was close to but still above the critical threshold. That residual velocity was enough to create a local casting defect risk, especially if the metal carried oxide films from the runner.
The oxide slag simulation for Scheme 1 showed a clear improvement but not a complete solution. At the beginning of filling, the semi-closed gating system and sprue well reduced the oxide density in the runner to about 30.00 kg/m3. At later stages, the runner extension collected some of the slag, and the local density in the extension reached about 100.00 kg/m3. However, the runner still contained regions with about 50.00 kg/m3 of oxide. Because the ingate velocity remained above the critical value, the metal could still entrain gas and generate additional oxide. The oxide density in the ingate increased to about 200.00 kg/m3. In the spiral volute, the flow was smoother, and the additional oxide generation was limited. Most of the existing slag was carried toward the inlet flange. At 55 percent filling, the oxide density in the casting reached about 400.00 kg/m3. At the end of filling, a large amount of slag was discharged into the riser, but about 100.00 kg/m3 remained in the support region. That residual level was lower than in the initial design, but I did not consider it acceptable for a pressure-containing casting. Table 10 summarizes the Scheme 1 velocity and oxide results. Table 11 compares the initial design and Scheme 1 in terms of casting defect risk.
| Scheme 1 stage | Velocity | Unit | Oxide density | Unit | Observation |
|---|---|---|---|---|---|
| Early runner fill | 0.90 | m/s | 30.00 | kg/m3 | Stable, low slag |
| Runner extension | 0.90 | m/s | 100.00 | kg/m3 | Slag collected |
| Ingate entry | 0.88 | m/s | 200.00 | kg/m3 | Still above critical velocity |
| Spiral volute | 0.70 or less | m/s | 400.00 | kg/m3 | No splash, but slag present |
| Late filling | 0.15 | m/s | 100.00 | kg/m3 | Support region residual |
| Casting defect indicator | Initial design | Scheme 1 | Improvement |
|---|---|---|---|
| Maximum fill velocity | 1.139 m/s | 0.90 m/s | Moderate |
| Ingate velocity | Above 1.00 m/s | 0.88 m/s | Slight |
| Maximum oxide density | 450.00 kg/m3 | 400.00 kg/m3 | Slight |
| Residual oxide density | High at outlet flange | 100.00 kg/m3 at support | Moderate |
| Flow stability | Poor | Improved | Good |
| Overall casting defect risk | High | Medium | Partial |
I therefore developed Scheme 2. The main changes were an increase in pouring time from 7 s to 8 s and the installation of a ceramic filter in the runner. The filter rating was 40 PPI. The longer pouring time reduced the average filling velocity, and the filter provided both a physical barrier for oxide slag and a flow-straightening effect. The filter also increased the local pressure drop, which helped suppress the formation of a strong vortex. The pressure drop across the filter can be approximated as
$$ \Delta p = C_{1} \rho u^{2} + C_{2} \mu u $$
where \(C_{1}\) and \(C_{2}\) are empirical coefficients that depend on the filter structure and porosity. The first term represents inertial resistance, and the second term represents viscous resistance. Because the filter creates a pressure drop, the downstream velocity is lower and the flow is more uniform. In my simulation, the velocity before the filter was about 0.90 m/s, and the velocity after the filter was about 0.55 m/s. That reduction was significant. The ingate velocity decreased to about 0.75 m/s. Although this was still above the 0.50 m/s threshold, the flow direction was more coherent, and the oxide density at the ingate and inlet flange dropped to about 30.00 kg/m3. The filter therefore acted as a combined flow conditioner and slag trap. The resulting casting defect risk was much lower than in Scheme 1.
The Scheme 2 oxide simulation showed that the large oxide concentrations in the inlet flange and spiral volute were largely eliminated. At 3 percent filling, the oxide density was low because the filter had already removed part of the initial slag. At 8 percent filling, the flow through the filter was stable. At 15 percent filling, the ingate and inlet flange had an oxide density of about 30.00 kg/m3, which was far below the values observed in the initial design and Scheme 1. At 100 percent filling, the remaining oxide was mainly in the riser and in a few low-velocity pockets. The support region no longer showed the high residual density that appeared in Scheme 1. I concluded that Scheme 2 was the best gating system among the three designs. Table 12 lists the Scheme 2 process parameters. Table 13 compares the filter performance. Table 14 summarizes the casting defect risk after Scheme 2.
| Scheme 2 parameter | Value | Unit | Effect |
|---|---|---|---|
| Pouring time | 8 | s | Lower average velocity |
| Filter rating | 40 | PPI | Slag interception and flow straightening |
| Filter location | Runner | – | Downstream velocity reduction |
| Velocity before filter | 0.90 | m/s | Incoming flow |
| Velocity after filter | 0.55 | m/s | Strong deceleration |
| Ingate velocity | 0.75 | m/s | Still above critical but stable |
| Oxide density at ingate | 30.00 | kg/m3 | Major reduction |
| Flow state | Stable and coherent | – | Less gas entrainment |
| Metric | Before filter | After filter | Reduction |
|---|---|---|---|
| Velocity | 0.90 m/s | 0.55 m/s | 38.9 percent |
| Ingate velocity | 0.88 m/s in Scheme 1 | 0.75 m/s in Scheme 2 | 14.8 percent |
| Oxide density at ingate | 200.00 kg/m3 in Scheme 1 | 30.00 kg/m3 in Scheme 2 | 85.0 percent |
| Flow pattern | Local turbulence | Stable | Qualitative improvement |
| Casting defect | Initial design | Scheme 1 | Scheme 2 |
|---|---|---|---|
| Gas porosity | High | Medium | Low |
| Oxide inclusion | High | Medium | Low |
| Sand erosion | Medium | Low | Low |
| Shrinkage porosity | Medium to high | Medium | Medium until chill added |
| Cold shut | Low to medium | Low | Low |
The illustration above shows a representative complex cast component with internal passages and thick-to-thin transitions. I include it here because the thermal and filling challenges are conceptually similar to those of the volute. In both cases, a casting defect is most likely where a thick boss or flange meets a thin wall, where the liquid metal must change direction rapidly, and where the local modulus is high. For the volute, the inlet flange was the critical thick region. Even with Scheme 2, the solidification simulation still showed an isolated liquid zone at the inlet flange. That zone was not caused by filling turbulence; it was caused by the thermal mass of the flange. I therefore added external chills to the bottom of the inlet flange. I placed four identical chills, each with an outline size of 40.00 mm by 20.00 mm by 14.00 mm. The chills increased the local heat extraction rate, reduced the local solidification time, and shifted the last liquid to a region that could be fed by the riser. The comparison at 60 percent solidification showed that the isolated liquid zone at the inlet flange was eliminated when the chills were used. This confirmed that the chill design addressed the remaining shrinkage casting defect risk.
The thermal effect of a chill can be understood through the local energy balance. The heat extracted by the chill increases the local cooling rate \(\dot{T}\):
$$ \dot{T} = \frac{\partial T}{\partial t} $$
A higher cooling rate reduces the local solidification time and suppresses the formation of a large isolated liquid pocket. The feeding distance also depends on the thermal gradient. A positive gradient toward the riser is desirable, because it promotes directional solidification. The chill creates a positive gradient from the flange toward the riser, which allows the riser to feed the last liquid. In addition, the chill reduces the local modulus \(M\), so the flange no longer behaves as a large hot spot. The modulus reduction can be expressed qualitatively as
$$ M_{eff} = \frac{V}{A + A_{chill}} $$
where \(A_{chill}\) is the additional heat-transfer area provided by the chill. The larger the effective surface area, the smaller the effective modulus and the shorter the solidification time. Table 15 lists the chill design parameters. Table 16 compares the solidification behavior with and without chills. Table 17 summarizes the feeding-related casting defect risk before and after the chill addition.
| Chill parameter | Value | Unit |
|---|---|---|
| Number of chills | 4 | – |
| Length | 40.00 | mm |
| Width | 20.00 | mm |
| Thickness | 14.00 | mm |
| Location | Bottom of inlet flange | – |
| Function | Local chilling and directional solidification | – |
| Solidification condition | Without chill | With chill | Effect on casting defect |
|---|---|---|---|
| 60 percent solidification at inlet flange | Isolated liquid zone present | Isolated liquid zone eliminated | Shrinkage risk removed |
| Local cooling rate | Lower | Higher | Finer structure |
| Thermal gradient toward riser | Weak | Stronger | Better feeding |
| Effective modulus | Higher | Lower | Shorter solidification time |
| Riser efficiency | Partial | Improved | Less micro-shrinkage |
| Feeding indicator | Before chill | After chill |
|---|---|---|
| Isolated liquid zone at inlet flange | Yes | No |
| Shrinkage porosity probability | Medium to high | Low |
| Riser feeding distance | Limited | Sufficient |
| Directional solidification | Interrupted | Restored |
| Final casting defect risk | Medium | Low |
After selecting Scheme 2 for the gating system and the four external chills for the feeding system, I performed a production validation. The castings were poured, cleaned, and inspected. The external appearance met the shop requirements, and the dimensional check showed no unacceptable distortion. I then tested the mechanical properties. The yield strength was 336 MPa, and the tensile strength was 557 MPa. These values passed the strength acceptance criteria. For internal quality, I used industrial X-ray inspection with a single-wall technique. The inspected wall thickness was 37 mm. The X-ray voltage was 330 kV, the current was 5 mA, and the exposure time was 4 min. The source-side image quality indicator had a recognizable wire diameter of 0.81 mm, corresponding to an image quality index of 11. The radiographic results showed no unacceptable indications. In other words, the optimized process eliminated the casting defect that had been predicted by the initial simulation. Table 18 summarizes the validation results. Table 19 compares the final process with the initial process. Table 20 gives the final process control window that I would use for repeat production.
| Validation item | Result | Acceptance |
|---|---|---|
| Yield strength | 336 MPa | Pass |
| Tensile strength | 557 MPa | Pass |
| X-ray wall thickness | 37 mm | Single-wall technique |
| X-ray voltage | 330 kV | Procedure qualified |
| X-ray current | 5 mA | Procedure qualified |
| Exposure time | 4 min | Procedure qualified |
| Wire diameter | 0.81 mm | Image quality index 11 |
| Radiographic indications | None unacceptable | Pass |
| Visual and dimensional inspection | Acceptable | Pass |
| Aspect | Initial process | Final optimized process |
|---|---|---|
| Pouring time | 7 s | 8 s |
| Gating type | Bottom, open | Semi-closed with runner and filter |
| Runner | None | Added with extension |
| Filter | None | 40 PPI ceramic filter |
| Ingate entry | Direct | Tangential |
| Feeding | Riser only | Riser plus four external chills |
| Maximum fill velocity | 1.139 m/s | 0.90 m/s before filter, 0.55 m/s after filter |
| Oxide density at critical region | 450.00 kg/m3 | 30.00 kg/m3 |
| Isolated liquid zone | Present at inlet flange | Eliminated |
| Final casting defect risk | High | Low |
| Process variable | Recommended window | Reason |
|---|---|---|
| Pouring temperature | 1550 to 1570 degrees Celsius | Maintain fluidity without excessive oxidation |
| Pouring time | 7.5 to 8.5 s | Balance fill speed and turbulence |
| Filter rating | 40 PPI | Slag interception and flow straightening |
| Runner height | 40.00 mm | Slag residence and pressure control |
| Gating ratio | 0.9 to 0.8 to 1 | Semi-closed stable filling |
| Chill size | 40.00 mm by 20.00 mm by 14.00 mm | Local hot spot control |
| Chill quantity | 4 | Uniform flange cooling |
| Chill location | Bottom of inlet flange | Direct thermal control |
I also evaluated the simulation model itself to ensure that the conclusions were not artifacts. The model used a structured finite-volume mesh with local refinement near the ingate, the filter, the inlet flange, and the spiral volute. The time step was controlled by the Courant condition and by the volume-fraction transport. The solidification model used a temperature-dependent enthalpy and a mushy-zone permeability. The permeability was modeled with a Darcy-type relation:
$$ \mathbf{u} = -\frac{K}{\mu} (\nabla p – \rho \mathbf{g}) $$
where \(K\) is the permeability of the mushy zone. This term is important because it controls the interdendritic flow that feeds shrinkage. If the permeability is too low, the model predicts more shrinkage than the real casting; if it is too high, the model underestimates the casting defect. I calibrated the permeability against the measured thermal history and the final radiographic results. The calibration increased my confidence in the predicted hot spots and in the effect of the chills. I also checked the mesh sensitivity by comparing coarse, medium, and fine meshes. The medium mesh gave essentially the same flow pattern as the fine mesh, but with lower computational cost. The final simulations used the medium mesh with local refinement. Table 21 lists the model constants and numerical settings. Table 22 gives the mesh sensitivity results.
| Model or numerical parameter | Setting |
|---|---|
| Free-surface method | Volume of fluid |
| Turbulence model | Two-equation eddy viscosity |
| Energy equation | Enthalpy based |
| Solidification model | Mushy-zone permeability |
| Mushy-zone flow | Darcy-type resistance |
| Mesh type | Structured finite volume |
| Local refinement | Ingate, filter, inlet flange, spiral volute |
| Time step control | Courant and volume-fraction limits |
| Convergence criteria | Residual and mass-balance checks |
| Mesh level | Cell count | Maximum velocity difference | Oxide density difference | Decision |
|---|---|---|---|---|
| Coarse | Low | Higher than medium | Underpredicted local peaks | Not used |
| Medium | Moderate | Reference | Reference | Used for final design |
| Fine | High | Small difference from medium | Small difference from medium | Used for verification |
From a practical standpoint, the most important finding was that a casting defect in this volute could not be eliminated by a single change. The initial design had a filling problem and a feeding problem. Scheme 1 improved the filling stability but left a residual oxide casting defect and did not solve the isolated liquid zone. Scheme 2 solved the filling and oxide problem, but the inlet flange still needed local chilling. Only the combination of a longer pouring time, a semi-closed gating system, a ceramic filter, a tangential ingate, and four external chills produced a robust process. This layered approach is consistent with the physics of casting defect formation. Turbulence creates inclusions and gas porosity. Slow cooling creates shrinkage porosity. A chill changes the local thermal field but cannot correct a globally turbulent filling pattern. A filter reduces inclusions but cannot feed a hot spot. Therefore, the process design must address each mechanism separately and then verify the combined effect.
I also examined the relationship between local velocity and oxide entrainment more quantitatively. The critical velocity criterion can be written as
$$ u_{cr} = 0.50 \ \text{m/s} $$
I used this value as a threshold for the onset of severe surface folding. However, the simulation showed that the transition is not perfectly sharp. In Scheme 1, the ingate velocity was 0.88 m/s, and the oxide density was still about 200.00 kg/m3. In Scheme 2, the ingate velocity was 0.75 m/s, and the oxide density dropped to about 30.00 kg/m3. The reduction in oxide density was much larger than the reduction in velocity. This indicates that flow direction and residence time are also important. The filter did not simply reduce the velocity; it also removed existing oxide and straightened the streamlines. Therefore, a velocity criterion alone is not sufficient. A better design rule is to keep the velocity below the critical value where possible, to use a filter to remove the oxide that does form, and to avoid sudden expansions or contractions that create recirculation. I summarized this rule in Table 23.
| Design rule | Quantitative target | Reason |
|---|---|---|
| Limit peak fill velocity | Below 0.50 m/s where possible | Suppress surface folding |
| Limit ingate velocity | Below 0.80 m/s, preferably below 0.60 m/s | Reduce local turbulence |
| Use a filter | 40 PPI or equivalent | Remove oxide and straighten flow |
| Avoid sudden area change | Smooth transitions | Prevent recirculation |
| Use a sprue well | Same diameter as sprue | Absorb impact energy |
| Use a runner extension | At least one runner height | Collect slag |
| Use tangential ingates | Tangent to flange | Prevent jet impingement |
| Control hot spots | Effective modulus below critical value | Promote directional solidification |
| Use chills where needed | Increase local cooling rate | Eliminate isolated liquid zones |
The thermal design can also be expressed with a simple hot-spot criterion. If the local modulus is larger than the feeding modulus that the riser can support, a shrinkage casting defect is likely. The critical modulus depends on the alloy, the mold, and the riser size. I estimated the hot-spot modulus of the inlet flange and compared it with the effective modulus after chilling. The chill reduced the local solidification time and shifted the last liquid toward the riser. The Niyama criterion was also improved. In the initial design, the inlet flange had a low Niyama value, indicating a high probability of micro-shrinkage. After adding the chills, the Niyama value increased above the critical level for the alloy. I did not rely on a single threshold, because the Niyama criterion is sensitive to mesh size and cooling rate, but the trend was consistent with the radiographic results. Table 24 shows the thermal criteria and the observed outcome. Table 25 gives a final process capability summary.
| Thermal criterion | Initial design | Final design | Outcome |
|---|---|---|---|
| Local modulus at inlet flange | High | Reduced by chill | Less hot spot |
| Solidification time at inlet flange | Long | Shorter | Less isolated liquid |
| Niyama value | Low | Higher | Lower shrinkage risk |
| Thermal gradient toward riser | Weak | Stronger | Better feeding |
| Isolated liquid zone | Present | Eliminated | No shrinkage casting defect |
| Capability item | Final result | Comment |
|---|---|---|
| Filling stability | Stable | Filter and tangential ingate |
| Peak velocity | 0.90 m/s before filter | Above critical but filtered |
| Downstream velocity | 0.55 m/s | Near critical |
| Oxide density | 30.00 kg/m3 | Low |
| Shrinkage control | Chill-assisted | Isolated zone removed |
| Mechanical properties | 336 MPa yield, 557 MPa tensile | Pass |
| Radiography | No unacceptable indications | Pass |
| Production repeatability | Controlled window established | Ready for batch production |
In summary, I used FLOW-3D simulation to diagnose a casting defect problem in a sand-cast stainless steel volute. The initial bottom-gated open system produced high velocities, turbulent filling, oxide entrainment, and an isolated liquid zone at the inlet flange. Scheme 1 added a runner, a sprue well, a runner extension, and a tangential ingate. It improved the flow but left a residual oxide casting defect because the ingate velocity remained high. Scheme 2 added a longer pouring time and a 40 PPI ceramic filter. It reduced the velocity after the filter to about 0.55 m/s and lowered the oxide density at the critical region to about 30.00 kg/m3. The filter also stabilized the flow direction and prevented the formation of a large turbulent vortex. The remaining feeding problem was solved by placing four external chills at the bottom of the inlet flange. The chills increased the local cooling rate, reduced the effective modulus, and eliminated the isolated liquid zone at 60 percent solidification. Production validation confirmed the design: the yield strength was 336 MPa, the tensile strength was 557 MPa, and X-ray inspection showed no unacceptable indications. The optimized process therefore provided a reliable route for producing a pressure-containing volute with a low casting defect risk.
I would use the following final process for repeat production. The pouring temperature should be held between 1550 and 1570 degrees Celsius. The pouring time should be 8 s with a tolerance of about 0.5 s. The gating system should be semi-closed with a sprue-to-runner-to-ingate area ratio of approximately 0.9 to 0.8 to 1. A 40 PPI ceramic filter should be placed in the runner. The runner height should be 40.00 mm, and the runner extension should be at least one runner height. The ingate should enter the inlet flange tangentially. A sprue well with the same diameter as the sprue should be used. Four external chills, each 40.00 mm by 20.00 mm by 14.00 mm, should be placed at the bottom of the inlet flange. The riser should remain above the spiral volute. With this configuration, the peak velocity before the filter is about 0.90 m/s, the velocity after the filter is about 0.55 m/s, the ingate velocity is about 0.75 m/s, and the oxide density at the ingate is about 30.00 kg/m3. The isolated liquid zone at the inlet flange is eliminated, and the final casting defect risk is low. Table 26 presents the final recommended process card. Table 27 lists the key lessons that I would carry to similar pump castings. Table 28 gives the final acceptance criteria that I used.
| Process step | Recommended setting | Control reason |
|---|---|---|
| Melt preparation | ASTM A744 Gr CF8M, clean charge | Reduce initial oxide |
| Pouring temperature | 1550 to 1570 degrees Celsius | Fluidity and oxidation balance |
| Pouring time | 8 s plus or minus 0.5 s | Velocity control |
| Gating ratio | 0.9 to 0.8 to 1 | Semi-closed stable filling |
| Sprue well | Same diameter as sprue | Impact absorption |
| Runner height | 40.00 mm | Slag residence |
| Runner extension | At least one runner height | Slag collection |
| Filter | 40 PPI ceramic | Slag interception and flow straightening |
| Ingate entry | Tangential to inlet flange | Prevent jet impingement |
| Riser | Above spiral volute | Primary feeding |
| Chill | Four pieces, 40.00 mm by 20.00 mm by 14.00 mm | Local hot spot control |
| Chill location | Bottom of inlet flange | Directional solidification |
| Lesson | Application to similar volutes |
|---|---|
| Filling turbulence is a primary source of casting defect | Use a filter and avoid direct jet ingates |
| Velocity alone does not predict oxide entrainment | Control flow direction and residence time |
| Open gating systems can be too aggressive | Use a semi-closed ratio for small steel castings |
| Thick flanges require local chilling | Place chills at the highest modulus region |
| A riser cannot feed every hot spot | Combine riser design with chill design |
| Simulation must be validated by production | Use radiography and mechanical testing |
| Process windows matter | Define pouring time and temperature tolerances |
| Acceptance item | Criterion | Final result |
|---|---|---|
| Yield strength | At least 300 MPa | 336 MPa |
| Tensile strength | At least 500 MPa | 557 MPa |
| Radiographic indications | No unacceptable shrinkage, gas, or inclusion | No unacceptable indications |
| Surface quality | No visible sand erosion or cold shut | Acceptable |
| Dimensional accuracy | Within drawing tolerance | Acceptable |
| Casting defect rate | Below production target | Low risk after optimization |
The final conclusion is that the casting defect problem in the centrifugal pump volute was solved by a combination of numerical simulation, gating system redesign, filtration, and local chilling. The initial design had high turbulence and a high oxide density, and it also formed an isolated liquid zone at the inlet flange. Scheme 1 improved the filling pattern but did not eliminate the oxide casting defect. Scheme 2 introduced a 40 PPI filter and a longer pouring time, which reduced the oxide density to about 30.00 kg/m3 and stabilized the flow. The four external chills then eliminated the isolated liquid zone and restored directional solidification. The production castings passed mechanical and radiographic inspection. I therefore consider the optimized process suitable for manufacturing the volute with a low casting defect risk. The same approach can be applied to other pump housings, valve bodies, and pressure-containing castings where thin walls meet thick flanges and where internal quality is critical.