Stainless steel pump bodies combine several conflicting requirements in one part: a pressure-tight internal passage, heavy bolted flanges, a wear ring that must remain dimensionally stable, and thin feet that are difficult to fill. In CF8 austenitic stainless steel these sections cool at different rates, and a gating system that fills the mould completely can still leave isolated liquid at the ring–foot junction or behind the inlet flange. Investment casting provides the dimensional accuracy and surface finish needed for the internal contour, but the process is economically sensitive to casting yield—the mass of the finished casting divided by the total poured mass, including gating and feeders. For complex pump bodies, conventional layouts often remain below about 35%.
A 24.75 kg CF8 pump body produced at Panjin Tiangong Precision Casting Co., Ltd. illustrates the problem. The casting envelope is 304 × 246 × 165 mm, with wall thicknesses from 11 to 30 mm. The inlet flange is 197 mm in diameter and 30 mm thick, the outlet flange is 160 mm in diameter and 25 mm thick, and the ring is 208 mm in diameter and 28 mm thick. These thick sections are separated by thinner walls, so the hot spots are not concentrated in a single riser-fed region. The alloy composition is given in Table 1.
| Element | C | Si | Mn | S | Cr | Ni | Mo | Fe |
|---|---|---|---|---|---|---|---|---|
| wt.% | 0.06 | 0.7 | 1.1 | 0.009 | 17.5 | 8.2 | 0.2 | balance |
Table 1. Chemical composition of the CF8 stainless steel pump body.
The solidus and liquidus temperatures used in simulation were 1370 °C and 1451 °C. Wax patterns were invested in a zircon-face, mullite-backup shell and dewaxed before firing at 1100 °C for 60 min. Initial filling and solidification simulations used a shell temperature of 1050 °C, a pouring temperature of 1600 °C, gravity pouring at 4 kg/s, a casting–shell heat-transfer coefficient of 500 W/m2·K, and an external air coefficient of 50 W/m2·K. The later parameter study examined a lower shell-temperature window of 700–800 °C, corresponding to the shell condition at pouring after transfer from the burnout furnace.
Pouring orientation and initial gating systems
Six pouring orientations were screened. With the inlet flange upward, the heaviest upper section can be fed directly and the ring remains below the main feed path. With the ring upward, multiple gates can feed the ring, but the inlet flange sits at the bottom and is remote from the riser. Side orientations either place the outlet flange and feet on different levels or require gates on curved ring surfaces, which complicates shell building and gate removal. Two top-poured schemes were carried forward.
Scheme 1 placed the inlet flange upward. Two 80 × 60 × 40 mm gates fed the inlet flange, a 90 × 55 × 60 mm gate with a 40 mm pad fed the outlet flange, and two 30 × 40 × 20 mm gates fed the feet. The gates were connected by 330 × 70 × 70 mm and 250 × 40 × 45 mm runners and a 60 × 60 × 80 mm sprue. Two open risers of Φ83 × Φ120 × 160 mm supplied liquid metal and a static head; the assembly height was 435 mm.
Scheme 2 placed the ring upward. Four 50 × 35 × 40 mm gates fed the ring, a 50 × 70 × 20 mm gate fed the outlet flange, and a 40 × 40 × 112 mm bottom gate fed the inlet flange. The runners measured 340 × 70 × 70 mm and 100 × 70 × 75 mm, and a single Φ83 × Φ120 × 160 mm open riser was used in a 410 mm assembly.
Filling was not the limiting condition
Both initial schemes filled completely. Scheme 1 filled in about 15 s and Scheme 2 in about 14 s, but their solidification patterns differed sharply. In Scheme 1, the shell cooled the foot and the lower outlet flange first. By about 330 s, the shell had also solidified across the passage between the foot and the ring. At about 540 s, the foot–ring junction became an isolated liquid region with no feed path. The casting solidified completely at about 1600 s, leaving a shrinkage volume of 0.6 cm3 at that junction. Production castings showed the same defect location after shell removal and inspection.
Scheme 2 performed worse. The bottom gate was too long to maintain a feed path to the inlet flange. At about 1008 s, the feed channel from the third gate was blocked, creating a large isolated liquid region at the inlet flange. Total solidification took about 1658 s and the shrinkage volume reached 10.3 cm3. Cutting off the third gate exposed the same cavity in production castings. In both cases, the defect was a feeding problem rather than a filling problem: the metal reached the mould cavity, but the liquid in a thick section lost contact with a live feeder before solidification ended.
Corrected designs and production verification
Scheme 1 was corrected by moving the third gate from above the foot to below the ring. The gate was enlarged to 160 × 60 × 40 mm, the sprue to 60 × 60 × 250 mm, and a 20 × 25 × 135 mm tie was added above the foot for venting and slag capture. The feeding sequence then became sprue → third gate → ring → foot, so the thin foot no longer had to feed the ring junction through a passage that closed early. Filling took about 16 s, total solidification was completed at about 1443 s, and shrinkage was confined to the gating system.
Scheme 2 was corrected by moving the third gate to the side of the inlet flange and extending a 50 × 70 × 50 mm flange pad over a 60 mm length. The sprue became 70 × 70 × 205 mm and the connecting runner 175 × 70 × 70 mm. The revised sequence was third runner → sprue → third gate → inlet flange. Filling took about 15 s and solidification was completed at about 2752 s, again with no shrinkage in the casting itself.
Scheme 1 was selected for production trials because its layout was easier to handle and offered more room for gating reduction. The trial castings passed pressure testing and X-ray inspection at the ring, with no porosity in the casting. The two corrected schemes had casting yields of 33% and 34%, respectively, so the next step was to reduce the gating system itself.
Modulus matching to raise casting yield
The modulus of a casting section—its volume divided by its heat-transfer surface area—is a practical proxy for local solidification time. Three hot spots controlled the design, and ProCAST was used to calculate their moduli. Their positions and values are listed in Table 2.
| Hot spot | Location | Hot-spot circle diameter, mm | Modulus, cm |
|---|---|---|---|
| 1 | Outlet flange / central runner junction | 25 | 0.65 |
| 2 | Inlet flange | 30 | 0.76 |
| 3 | Bottom thick section opposite inlet flange | 25 | 0.60 |
Table 2. Hot-spot locations and calculated moduli.
The minimum defect-free modulus of each gate was then determined by simulation, with the gate size reduced until shrinkage appeared at the corresponding hot spot. The results are summarised in Table 3. The ratio k is the gate modulus divided by the hot-spot modulus.
| Gate | Fed hot spot | Minimum gate modulus, cm | Hot-spot modulus, cm | k | Defect condition |
|---|---|---|---|---|---|
| 1 | 1 | 0.76 | 0.65 | 1.17 | Defect at k = 1.10 |
| 2 | 2 | 0.60 | 0.76 | 0.78 | Defect at k = 0.63 |
| 3 | 3 | 0.90 | 0.60 | 1.50 | Defect at k = 1.40 |
Table 3. Minimum defect-free modulus ratios for the three gates.
Gate 2 is the exception to a simple “gate modulus greater than hot-spot modulus” rule: its minimum defect-free modulus was below that of the inlet flange. In that position the gate sits above the flange and acts as a short feed channel, so proximity and thermal gradient allowed a smaller modulus than the flange itself. Gate 3, which feeds the ring and foot from below, behaved more like a blind riser and required a modulus 1.5 times that of hot spot 3.
The runners and sprue were then reduced under the same criterion. A runner modulus of at least 1.6 cm was required, giving ratios of 2.1 relative to gate 1 and 2.6 relative to gate 2; at a runner modulus of 1.5 cm, shrinkage returned at the inlet flange. The sprue required a modulus of at least 1.15 cm, or 1.28 times gate 3; at 1.04 cm, defects approached the gate–casting junction.
Once these feed paths were established, removing the open riser did not produce shrinkage in the casting. The riser had served mainly as a pouring cup and liquid head rather than as the primary feeder. It was replaced by a Φ45 × 120 mm pouring cup. The final gating dimensions are given in Table 4.
| Feature | Dimension, mm |
|---|---|
| Pouring cup | Φ45 × 120 |
| Runner | 70 × 70 × 390 |
| Sprue | 50 × 50 × 290 |
| Gate 1 | 62 × 50 × 100 |
| Gate 2 | 35 × 35 × 40 |
| Gate 3 | 40 × 60 × 75 |
Table 4. Final gating dimensions after modulus-based downsizing.
The result was a rise in casting yield from 34% to 47%, a 13-percentage-point increase. The largest contribution came from reducing the runner and sprue cross-sections and eliminating the open risers, while keeping the feed paths above their minimum defect-free moduli.
Thermal parameter optimisation by design of experiments
The modulus-controlled layout fixed the feeding structure but not the thermal window. Shell temperature, shell thickness, and pouring temperature change local cooling rates and therefore the shrinkage volume in each hot spot. A three-factor, three-level L9(33) design was used, with the factors and levels listed in Table 5. The responses were the shrinkage volumes at three positions: position 1 at the outlet flange, position 2 at the ring, and position 3 at the ring–foot junction.
| Level | A: shell temperature, °C | B: shell thickness, mm | C: pouring temperature, °C |
|---|---|---|---|
| 1 | 700 | 7.5 | 1560 |
| 2 | 750 | 8.0 | 1580 |
| 3 | 800 | 8.5 | 1600 |
Table 5. Factor levels for the L9 orthogonal design.
| Experiment | A | B | C | Y1, cm3 | Y2, cm3 | Y3, cm3 |
|---|---|---|---|---|---|---|
| 1 | 1 | 1 | 1 | 1.10 | 0.36 | 1.90 |
| 2 | 1 | 2 | 2 | 1.21 | 0.76 | 1.90 |
| 3 | 1 | 3 | 3 | 0.80 | 0.83 | 1.90 |
| 4 | 2 | 1 | 2 | 1.10 | 0.734 | 2.23 |
| 5 | 2 | 2 | 3 | 0.50 | 0.66 | 1.90 |
| 6 | 2 | 3 | 1 | 0.03 | 0.16 | 0 |
| 7 | 3 | 1 | 3 | 1.40 | 0.94 | 0.40 |
| 8 | 3 | 2 | 1 | 0.02 | 0.21 | 0 |
| 9 | 3 | 3 | 2 | 0.024 | 0.032 | 0 |
Table 6. L9 simulation results: shrinkage volume at the outlet flange (Y1), ring (Y2), and ring–foot junction (Y3).
| Factor and levels | Position 1 mean, cm3 | R1 | Position 2 mean, cm3 | R2 | Position 3 mean, cm3 | R3 |
|---|---|---|---|---|---|---|
| A: 700 / 750 / 800 °C | 1.04 / 0.54 / 0.48 | 0.56 | 0.65 / 0.52 / 0.39 | 0.26 | 1.90 / 1.38 / 0.13 | 1.77 |
| B: 7.5 / 8.0 / 8.5 mm | 1.20 / 0.58 / 0.28 | 0.92 | 0.68 / 0.54 / 0.34 | 0.34 | 1.51 / 1.27 / 0.63 | 0.88 |
| C: 1560 / 1580 / 1600 °C | 0.38 / 0.78 / 0.90 | 0.52 | 0.24 / 0.51 / 0.81 | 0.57 | 0.63 / 1.38 / 1.40 | 0.77 |
Table 7. Mean shrinkage volume for levels 1/2/3 and range R for each factor and position.
The three positions did not respond to the same factor. At position 1, shell thickness had the largest range (0.92 cm3), followed by shell temperature (0.56 cm3) and pouring temperature (0.52 cm3). A thicker shell slows heat extraction from the outlet flange and keeps the feed path open, so B3 was preferred. At position 2, pouring temperature dominated (range 0.57 cm3), followed by shell thickness (0.34 cm3) and shell temperature (0.26 cm3). A lower pouring temperature reduced superheat and shrinkage demand in the ring, so C1 was preferred. At position 3, shell temperature had the largest effect (range 1.77 cm3), followed by shell thickness (0.88 cm3) and pouring temperature (0.77 cm3). A cold shell accelerated solidification of the thin foot and blocked feed to the ring–foot junction, so A3 was preferred.
All three positions pointed to the same main-effect combination: A3B3C1, that is, a shell temperature of 800 °C, a shell thickness of 8.5 mm, and a pouring temperature of 1560 °C. This combination was not one of the nine L9 runs; it is an additive main-effect prediction. The result is therefore a process window for confirmation in production, not a directly tested factorial optimum. The mechanism is consistent across the three locations: slower cooling at the outlet flange and ring–foot junction preserves feeding, while a lower pouring temperature limits shrinkage demand in the ring.
Transfer to a second pump body
The modulus ratios were applied to a second, larger pump body weighing 33.15 kg, with an envelope of 335 × 270 × 180 mm, wall thicknesses of 11–33 mm, an inlet flange of 217 mm diameter and 33 mm thickness, and an outlet flange of 176 mm diameter and 27 mm thickness. Its hot-spot moduli were 0.70 cm at the outlet flange, 0.84 cm at the inlet flange, and 0.63 cm at the bottom thick section. Applying the same minimum ratios gave gate moduli of 0.82, 0.66, and 0.95 cm, a runner modulus of 1.72 cm, and a sprue modulus of 1.22 cm. ProCAST predicted no shrinkage in the casting, with defects confined to the gating system.
This transfer is evidence that the modulus approach scales to a geometrically similar pump body, not proof that the same ratios apply to arbitrary castings. The minimum defect-free modulus depends on local geometry, alloy, shell conductivity, and the direction of the thermal gradient.
Process design implications
The case shows a useful order of operations for investment-cast pump bodies. First, eliminate isolated liquid regions by choosing a pouring orientation and local feed paths that match the hot-spot layout; filling simulation alone will not reveal the problem. Second, reduce gates, runners, and sprues to their minimum defect-free moduli, because gating mass is the largest removable contribution to low casting yield. Third, tune shell and pouring parameters within the window that preserves those feed paths. Parameter optimisation cannot compensate for a feed channel that closes before the hot spot solidifies.
For this CF8 pump body, the minimum defect-free modulus ratios were k1 = 1.17, k2 = 0.78, and k3 = 1.50; the runner-to-gate 2 ratio was 2.6 and the sprue-to-gate 3 ratio was 1.28. The open risers were unnecessary once these paths were established. The final layout raised casting yield from 34% to 47%, while production castings passed pressure testing and X-ray inspection. The main-effect thermal window was 800 °C shell temperature, 8.5 mm shell thickness, and 1560 °C pouring temperature, with the caveat that this combination was predicted from the L9 main effects and should be verified in a production trial.