In this work, I investigated the manufacturing route for a large, complicated stainless steel impeller used in pharmaceutical machinery. The component had an outer diameter of 700 mm, a height of 130 mm, a net weight of 202 kg, and featured 54 complex flow-path blades with a wall thickness of only 5 mm. The technical requirements demanded a surface roughness of Ra 3.2–1.6 μm, dimensional accuracy of CT4–CT5, an ovality of the rim within 0–4 mm, and a maximum adjacent blade pitch variation of 2 mm. Most importantly, no cracks, shrinkage cavities, porosity, or inclusions were permitted on the blade surfaces, and the customer required complete freedom from magnetic response after heat treatment while avoiding any defects larger than 2 mm in the hub, rim, or spoke sections.
My goal was to develop an economical and practical method that combined investment casting and sand casting in one composite mould, thereby overcoming the limitations of either process when applied alone. The key concept was to split the impeller into two regions: the complex blade assembly and the simpler hub/spoke/rim structure. For the blade section, I designed a pressure die to produce individual wax patterns, welded them into an integral blade cluster using a specially built fixture, and then applied a hybrid shell system based on ethyl silicate and sodium silicate. For the hub and spoke sections, I used wooden patterns and core boxes, and fabricated water-glass sand moulds with a properly designed gating and riser system. The final mould was assembled by seating the investment shell within the sand mould using a sodium silicate sand sleeve as the locating reference. After melting ZG1Cr18Ni9Ti in a medium-frequency induction furnace and pouring at 1580 ± 10 °C, I obtained sound castings that satisfied all technical requirements.
Process Selection and Critical Design Considerations
I first compared three options: pure sand casting, pure investment casting, and the proposed hybrid. Pure sand casting failed because the 54 thin and twisted blades would require numerous cores with poor positioning accuracy, leading to excessive dimensional scatter and severe sand casting defects such as misruns, cold shuts, and sand inclusions. Pure investment casting failed because the large mass and thick sections would require very heavy shell sections and complex risering, which is impractical with conventional ceramic shell systems. The hybrid approach was therefore chosen because it exploits the excellent surface finish and geometrical precision of investment casting for the blade passages, while the sand mould provides robust feeding through large risers and simplifies the mould assembly for the heavy hub and rim.
I selected a two-part, horizontally parted mould with a top-pour riser system. The final layout included one central riser that also served as the sprue for the hub, and three open risers located on the rim for feeding and gas escape. This design enabled rapid filling of the thin blades, minimized oxidation of the stainless steel, and ensured directional solidification toward the risers. I verified the riser dimensions using modulus calculations. For the hub section, the modulus of the casting was calculated as:
$$ M_{\text{hub}} = \frac{a b}{2(a+b-c)} = \frac{55 \times 138}{2(55+138-30)} = 2.07 \, \text{cm} $$
With a safety factor of 1.2 for an open top riser, the required riser modulus was:
$$ M_{\text{hub, riser}} = 1.2 \times 2.07 = 2.48 \, \text{cm} $$
Similarly, for the rim section I obtained:
$$ M_{\text{rim}} = \frac{50 \times 138}{2(50+138-30)} = 1.89 \, \text{cm} $$
$$ M_{\text{rim, riser}} = 1.2 \times 1.89 = 2.27 \, \text{cm} $$
From standard riser charts, a waist-shaped open riser of 120 mm × 180 mm × 150 mm was selected for the hub, and three risers of 110 mm × 165 mm × 150 mm were placed on the rim. The total shrinkage compensation was checked against the volumetric contraction of ZG1Cr18Ni9Ti. The body shrinkage factor was estimated by summing the contributions of alloying elements, as shown in Table 1.
| Element | Content (wt%) | Contribution factor | Shrinkage contribution (%) |
|---|---|---|---|
| C | 0.10 | 3.2 | 3.200 |
| Mn | 1.20 | 0.0585 | 0.070 |
| Si | 1.20 | 1.03 | 1.236 |
| Cr | 18.00 | 0.10 | 1.800 |
| Ni | 10.00 | −0.0354 | −0.354 |
| Total | — | — | 5.952 |
Thus, the total volume contraction was taken as 6%. The central riser alone could feed a volume of 4.3 × 10³ cm³, which exceeded the required 3.94 × 10³ cm³ for the hub region. The three rim risers together could feed 3 × 3.0 × 10³ cm³, which was sufficient for the rim volume of 8.89 × 10³ cm³. The effective feeding distances were also checked, by considering the blade sections as chills that improved directional solidification.
Manufacturing the Blade Cluster Pattern
I designed and machined a two-part steel pressure die for the individual blades. The die cavity accounted for the combined shrinkage of the wax pattern, shell expansion, and metal contraction. The total linear shrinkage was evaluated as:
$$ \varepsilon = \varepsilon_1 – \varepsilon_2 + \varepsilon_3 = 0.75\% – 0.50\% + 2.5\% = 2.75\% $$
However, because the actual contraction in complex thin sections is restrained, I adopted a practical value of 2.6% for the die cavity dimensions. The surface roughness of the die cavity was specified as Ra 0.2–0.8 μm, while the parting surfaces were finished to Ra 0.8–1.6 μm.
I formulated a modified rosin–wax pattern material, replacing conventional polyethylene with EVA and part of the natural waxes with synthetic substitutes. Table 2 lists the final composition and key properties.
| Polymerized rosin (%) | Modified rosin (%) | Paraffin wax (%) | EVA (%) | Melting point (°C) | Softening point (°C) | Tensile strength (MPa) | Shrinkage (%) |
|---|---|---|---|---|---|---|---|
| 17 | 40 | 30 | 3 | 74–78 | >40 | 5.4 | 0.4–0.7 |
Individual blades were injected using a manual wax press at a wax temperature of 70–85 °C, injection pressure of 0.3–1.5 MPa, and a holding time of 1–3 min. After cooling in water at 16–25 °C, the patterns were inspected. I rejected any pattern with flash, cold laps, cracks, sink marks, or bubbles. The remaining patterns were degreased in a neutral soap solution at 20–25 °C, rinsed in clean water, and stored on racks to avoid distortion.
The critical step was assembling 54 blades into an integral cluster. I designed a welding fixture consisting of an upper plate, a lower plate, a locating ring, two tie rods, and two base plates. The locating ring had an outer diameter of 425.15 ± 0.25 mm, which corresponded to the inner diameter of the rim after accounting for metal shrinkage. Its height was 141 mm, equal to the casting height plus machining allowances and shrinkage. The lower plate had a diameter of 1240 mm and included a stepped recess of 973 mm diameter to accommodate the sand sleeve that would later be formed around the shell.
The assembly procedure was as follows. First, I placed all 54 wax blades around the locating ring, ensuring that the curved surface of each blade matched the ring radius. After verifying the positions with a template, I welded the blades together at their upper ends using a heated soldering iron at about 80 °C. I then mounted the lower plate, clamped the fixture, and turned the whole assembly over so that I could weld the lower joints. After completing both sides, I covered the gap between the wax cluster and the fixture plates with molten wax to prevent any hardening liquid from penetrating during shell building. This fixture method guaranteed that the blade spacing and overall envelope remained within the required tolerance, and it also provided a robust support for coating operations.
Hybrid Shell System: Ethyl Silicate and Sodium Silicate
For the shell, I selected a composite system: the first two layers used an ethyl silicate binder to produce a smooth, refractory surface; the subsequent four layers used sodium silicate binder as the reinforcing back-up. This combination reduced cost while maintaining excellent surface finish. The raw materials and their quality specifications are summarized in Table 3.
| Material | Specification |
|---|---|
| Ethyl silicate | SiO₂ 30–34%, HCl ≤ 0.15%, density ≤ 1 at 20 °C |
| Ethanol | Purity > 95% |
| Distilled water | Primary or secondary distilled |
| Hydrochloric acid | Chemically pure, 36–38% |
| Sulfuric acid | Chemically pure, 95–98% |
| Acetic acid | Chemical grade |
| Ammonia | 20–30% solution |
| Ammonium chloride | Industrial grade |
| Quartz powder | SiO₂ ≥ 98.5%, refractory > 1680 °C, 0.055 mm |
| Quartz sand | 0.212–1.70 mm grades |
| Bauxite | Al₂O₃ ≥ 60%, refractory > 1770 °C, 0.106 mm |
| Sodium silicate | Modulus 2.9–3.4, density 1.40–1.45 g/cm³ |
I prepared two types of ethyl silicate coatings. The first was a compounded hydrolysis coating with a density of 1.69–1.71 g/cm³, using quartz powder at 3.9–4 times the mass of ethyl silicate. The second was a one-shot hydrolysis coating, with the composition given in Table 4, and was aged for at least 2 hours before use.
| Ethyl silicate (kg) | Ethanol (kg) | Distilled water (kg) | HCl (mL) | Acetic acid (mL) | H₂SO₄ (mL) |
|---|---|---|---|---|---|
| 1.0 | 0.6 | 0.18 | 3.0 | 3.5 | 3.5 |
The first-layer coating density was adjusted to 1.62–1.65 g/cm³, and the second-layer to 1.56–1.60 g/cm³. During application, I did not dip the entire large assembly; instead, I poured the coating over the slowly rotating wax cluster and allowed the excess to drain. This ensured that the thin blade gaps were completely wetted. After applying each ethyl silicate layer, I stuccoed with 0.425–0.212 mm quartz sand, air-dried for more than 2 h, and then hardened in an ammonia atmosphere for 30–40 min at 18–20 °C and 45–80% relative humidity.
For the reinforcement layers, I used a sodium silicate coating with a fine bauxite–quartz powder blend. The third layer had a lower viscosity to promote adhesion with the ethyl silicate surface and prevent delamination. The stucco sand size was 0.850–0.425 mm for the third layer and 1.70–0.850 mm for the fourth through sixth layers. Each sodium silicate layer was treated with an aqueous ammonium chloride solution for 5–10 min. The hardening depth was strongly influenced by the ammonium chloride temperature and concentration, as shown in the following empirical relations:
$$ \delta_h \propto t^{0.5} \exp\left(-\frac{E_a}{RT}\right) $$
where δ_h is the hardened layer thickness, t is the hardening time, and T is the temperature of the ammonium chloride solution. Higher temperature and concentration increased the hardening depth, hence I controlled the bath at 26–40 °C and replaced it regularly to avoid sodium chloride accumulation, which would depress the hardening depth as indicated in Figure 12 of the original study.
The entire shell coating sequence is summarized in Table 5.
| Layer | Binder | Coating viscosity (s) | Stucco size (mm) | Air drying (h) | Hardening | Hardening time (min) |
|---|---|---|---|---|---|---|
| 1 | Ethyl silicate | 28–32 | 0.425/0.212 | >2 | Ammonia | 30–40 |
| 2 | Ethyl silicate | 28–32 | 0.425/0.212 | >2 | Ammonia | 30–40 |
| 3 | Sodium silicate | 30–45 | 0.850/0.425 | 2 (before hardening) | NH₄Cl | 5–10 |
| 4–6 | Sodium silicate | 30–45 | 1.70/0.850 | — | NH₄Cl | 5–10 |
After finishing the sixth layer, the shell was still not strong enough to support its own weight during handling and subsequent dewaxing. I therefore formed a thick sodium silicate sand sleeve around the shell. A split cylindrical steel ring with an inner diameter of 973 mm was clamped onto the lower fixture plate, and a cast iron reinforcement ring incorporating two lifting lugs was placed around the shell. The annular space was filled with CO₂-hardened sodium silicate sand and compacted carefully. This sand sleeve served three purposes: it prevented deformation during dewaxing and firing, it provided lifting points for transportation, and it acted as a precise datum for locating the shell inside the sand mould.
Dewaxing and Shell Firing
Because the pattern material was rosin-based and the cluster was large, I used a pressure steam autoclave for dewaxing. The parameters were 0.6–1 MPa steam pressure, a charging time under 5 s, and a holding time of 6–10 min. The molten wax was drained from the bottom of the autoclave. After dewaxing, the shell was allowed to rest for 12–48 h to stabilize, but not longer than 2 days, to prevent surface spalling of the ethyl silicate layer.
Firing was carried out in a 75 kW box furnace at 850 ± 30 °C for 1.5–2 h. During firing, the shell was placed horizontally on a bed of 20/40 mesh quartz sand in a stainless steel tray to prevent sagging. I rejected any shell that showed black smoke on removal, indicating incomplete burnout of residual wax. Only one firing was permitted, as repeated thermal cycles could cause micro-cracks or delamination between the ethyl silicate and sodium silicate layers. The fired shell displayed a clean white surface with very low roughness, meeting the requirements for thin-blade casting.
Sand Mould Design and Fabrication for Hub and Spoke
I designed a wooden pattern for the sand mould that formed the hub, spoke, and rim outer geometry. The pattern dimensions were calculated using the standard relation:
$$ A_M = (A_C \pm A_t)(1 + \varepsilon) $$
where A_M is the pattern dimension, A_C is the finished part dimension, A_t is the machining allowance (5 mm on top, 3 mm on bottom, and 5 mm on the bore), and ε is the metal shrinkage of 2.2%. The mould was parted horizontally. The upper and lower sand moulds were made from CO₂-hardened sodium silicate sand. For the facing sand, I used chromite sand with a 40/70 mesh size, which has excellent refractory properties and high thermal conductivity. The backing sand was ordinary quartz sand. The facing sand composition and properties are shown in Table 6.
| New chromite sand (mesh) | Content (%) | Sodium silicate (%) | Water (%) | Green permeability | Green compressive strength (kPa) | Hardened compressive strength (MPa) |
|---|---|---|---|---|---|---|
| 40/70 | 100 | 6–7 | 4–5 | >200 | 17–23 | >1.5 |
I used wooden core boxes to produce the sand core for the central bore. The core was made from the same chromite facing sand to avoid metal penetration at high temperatures. Exhaust passages were built into the core to vent gases safely during pouring.
The sand mixing sequence was dry blending first, followed by the addition of sodium silicate and water. The total mixing time was controlled to 2–4 min after wetting to avoid over-mixing, which would reduce strength. I then manually compacted the sand around the pattern, stabbed vents through the upper mould, and hardened the mould by blowing CO₂ through inserted tubes. The blowing pressure was 0.10–0.15 MPa, flow rate 0.5 m³/h, and time 20–30 s. The moulds were stripped carefully after a small lateral tap on the pattern to avoid damaging the mould cavity.
After stripping, I coated the mould surfaces with a zircon-based alcohol coating. The coating formulation is shown in Table 7. It was applied by brushing to a thickness of 0.3–1.0 mm, then ignited to remove the alcohol and harden the coating. This coating reduced the risk of metal penetration and also protected the sodium silicate-bonded mould from moisture absorption while waiting for assembly.
| Zircon flour | Bauxite | Resin | PVB | Sodium bentonite | Clay | Urotropine | Ethanol | Water |
|---|---|---|---|---|---|---|---|---|
| 70 | 30 | 1.2–2.0 | 1–4 | 1–5 | 0.5–1.5 | 8 | Balance | 3.5–4.5 |
Assembly of the Composite Mould
The fired investment shell, with its surrounding sodium silicate sand sleeve, was lowered into the lower sand mould. The 973 mm diameter outer surface of the sand sleeve fitted into the corresponding circular recess in the sand mould; this provided precise centring of the blade shell relative to the sand mould cavity shown in Figure 1.
The central core was placed in the lower mould core print. Then the upper mould was carefully turned over, checked for loose sand or debris, and closed over the lower mould using guide pins to ensure exact alignment. The parting line was sealed with asbestos rope or clay to prevent flash and run-out. The three rim risers and the central hub riser communicated directly with the corresponding openings in the shell and the blade cavities. Finally, the mould was clamped with weights to resist the ferrostatic lift. The lifting force was calculated as:
$$ F_{\text{lift}} = K \rho g S h $$
Using K = 1.2, ρ = 7800 kg/m³, g = 10 m/s², S = π(0.33 m)², and h = 0.335 m, I obtained:
$$ F_{\text{lift}} = 1.2 \times 7800 \times 10 \times \pi \times (0.33)^2 \times 0.335 \approx 10.56 \times 10^3 \, \text{N} $$
The assembled composite mould is shown schematically in the original work as a two-part system with the shell, sand sleeve, sand mould, core, risers, and pouring cup. The resulting mould was strong enough to withstand the thermal and mechanical loads during pouring.
Melting and Pouring of ZG1Cr18Ni9Ti
I melted the alloy in a 300 kg medium-frequency coreless induction furnace with a basic magnesia lining. The lining was rammed from fused magnesia (4–6 mesh, 15%; 10–20 mesh, 30%; 20–40 mesh, 20%; and −120 mesh powder, 35%) with 0.8–1% boric acid and a sodium silicate solution as temporary binder. After ramming in 25–35 mm layers, the lining was sintered by heating slowly to 1500–1550 °C and holding for 20–30 min. This basic lining was essential to avoid contamination of the stainless steel from an acid lining.
I calculated the charge carefully. The furnace capacity was 300 kg, and the target composition in the molten steel is given in Table 8.
| Element | C | Si | Mn | S | P | Cr | Ni | Ti |
|---|---|---|---|---|---|---|---|---|
| Target (wt%) | 0.07 | 0.80 | 1.50 | 0.020 | 0.030 | 18.0 | 10.0 | 0.50 |
Element recoveries were assumed as follows: C 100%, Si 90%, Mn 95%, S 100%, P 100%, Cr 98%, Ni 99.9%, Ti 80%. The required elemental amounts per 100 kg of charge were calculated as:
$$ m = \frac{100 A}{\rho} $$
where A is the target composition and ρ is the recovery fraction. The resulting charge calculation for 300 kg is listed in Table 9.
| Raw material | Weight (kg) | Purpose |
|---|---|---|
| Return scrap (same alloy) | 180.0 | Main charge |
| Low-phosphorus steel scrap | 60.9 | Balance iron |
| Ferrochromium (62% Cr) | 38.0 | Cr addition |
| Electrolytic nickel | 15.0 | Ni addition |
| Ferrotitanium (30% Ti) | 3.25 | Ti addition |
| Electrolytic manganese | 1.80 | Mn addition |
| Ferrosilicon (75% Si) | 0.97 | Si adjustment |
| Aluminum (block/powder) | 0.30/0.24 | Deoxidation |
| Total | 300.54 | — |
The melting procedure was carried out using the non-oxidizing (direct) method. I charged the furnace with small pieces at the bottom, ferroalloys in the middle, and bulky scrap on top. Power was initially set to 60% for the first 6–8 min, then increased to full power once the current stabilized. I watched for bridging and pushed the charge down as melting progressed. When about 95% of the charge was molten, I took a sample for chemical analysis and added a slag cover composed of lime and fluorspar (2:1). After complete melting, I removed the primary slag and formed a new reducing slag. Deoxidation was performed by adding a mixture of lime and aluminum powder to the slag, followed by a final aluminum deoxidation of 0.1% after adding ferrotitanium. The melt was protected by argon purging during final temperature adjustment to 1620 ± 10 °C for tapping.
Pouring was performed at 1580 ± 10 °C. I used a 300 kg bottom-pour ladle with a well-baked new lining. The pouring process was fast and steady, with a filling time of 10–12 s. I started slowly to avoid splashing, then increased the flow, and finally reduced the stream as the risers filled. After topping up each riser, I covered them with insulating material to improve feeding. The mould was knocked out after 40–60 min to accelerate cooling of the thick sections, which helped refine the microstructure and reduce macrosegregation.
Post-Casting Operations
After the mould had cooled to 50–60 °C, I removed the sand mould and the sand sleeve mechanically. The impeller was then subjected to oxyacetylene vibration cutting to remove the risers. Standard gas cutting of stainless steel is difficult because of the refractory chromium oxide layer. I therefore used a vibration technique in which the cutting torch oscillated with an amplitude of 10–15 mm and a frequency of about 80 cycles per minute. The cutting oxygen pressure was 0.8–1 MPa, and the flame was neutral. This method broke through the high-melting-point oxide scale and permitted controlled separation of the riser stubs. The residual riser height was only 5–8 mm, which was well within the machining allowance. This approach proved to be economical and safe, and it avoided thermal damage to the blades.
The remaining ceramic shell inside the blade passages was removed by alkaline boiling. I immersed the impeller in a boiling 15–20% NaOH solution for 4–8 h. After cleaning, I neutralized the surface in a solution containing 90 g/L chromic acid, 30 g/L sulfuric acid, and 1.2 g/L sodium chloride at 18–28 °C for 2–3 min, then rinsed thoroughly with clean water. This procedure eliminated all residual shell material without mechanically damaging the thin blade edges.
Heat Treatment
To obtain the required austenitic microstructure and mechanical properties, I applied a solution treatment followed by stabilization treatment. The solution treatment consisted of heating the impeller at 1050–1100 °C for 3–4 h, then water quenching from above 950 °C. The stabilization treatment was carried out at 850–950 °C for 2–4 h followed by air cooling. These treatments dissolved any chromium carbides and precipitated the stable titanium carbides instead, thus preventing intergranular corrosion and ensuring no magnetic response.
Results and Verification
Chemical Composition
I took samples from a separately cast test bar and analyzed them using optical emission spectroscopy. The results are presented in Table 10 together with the charge calculation and the standard specification.
| Element | Analyzed | Calculated charge | Standard GB2100-80 |
|---|---|---|---|
| C | 0.080–0.083 | 0.079 | ≤0.12 |
| Si | 1.0–1.1 | 0.887 | ≤1.5 |
| Mn | 1.3–1.5 | 1.585 | 0.8–2.0 |
| Cr | 18.0–18.5 | 18.37 | 17.0–20.0 |
| Ni | 9.0–10.0 | 10.10 | 8.0–11.0 |
| Ti | 0.58–0.62 | 0.625 | 0.5–0.7 |
| S | 0.025–0.026 | 0.0175 | ≤0.030 |
| P | 0.030–0.032 | 0.0279 | ≤0.045 |
All elements were within the specification, and the impurities were lower than the maximum allowable values.
Surface and Dimensional Inspection
Visual inspection after cleaning showed no flash, sand adhesion, veining, cold shuts, or misruns. The blade surfaces were smooth, and the transitions between blades and rims were well rounded. The residual riser stubs were within the machining allowance. I used dye penetrant testing on the blade roots, hub-to-spoke junctions, and rim-to-spoke junctions. The inspection followed the standard GB/T9443-1988. No linear indications longer than 1.5 mm were found, and the total indication length in any area was less than 3.5 mm, which corresponds to grade 01. This satisfied the user’s requirement for absence of cracks and surface-breaking porosity.
Surface roughness was evaluated by comparing the casting with standard roughness comparators. The blade surfaces achieved Ra 1.6 μm, while the hub and rim surfaces were in the Ra 3.2–1.6 μm range. After machining, the end faces were even smoother.
Dimensional inspection was performed with callipers, templates, and a coordinate measuring approach. The rim ovality measured by a rotary-roundness instrument was less than 0.5 mm, which is well within the specified 0–4 mm. The adjacent blade distance errors did not exceed 0.8 mm, which is below the allowed 2 mm. The flatness of the spoke surfaces was within 1 mm, easily covered by the 3–5 mm machining allowances. The overall dimensions met CT4–CT5 grade according to GB/T6414-1999, and the blade positions were even within CT3.
Ultrasonic Testing
I performed ultrasonic testing of the hub, rim, and spoke regions using a CTS-26 instrument with a 2.5 MHz, 20 mm diameter normal probe. The sensitivity was set to detect a 2 mm flat-bottom hole equivalent. For the hub and rim, using a 1:4 time-base calibration, the signal-to-noise ratio allowed evaluation at a gain computed from:
$$ \Delta = 20 \log \frac{\lambda x_f}{\pi D_f^2} = 20 \log \frac{5.9 \times 138}{2.5 \times \pi \times 2^2} = 34.3 \, \text{dB} $$
For the spokes, the corresponding gain was:
$$ \Delta = 20 \log \frac{5.9 \times 30}{2.5 \times \pi \times 2^2} = 21 \, \text{dB} $$
In both cases, no indications above the 2 mm reference level were found. This confirmed the absence of internal shrinkage cavities, gas porosity, and non-metallic inclusions, aligning with the desired avoidance of sand casting defects such as sand inclusions and internal blows.
Microstructure
Metallographic samples taken from the riser area were prepared by grinding, polishing, and etching with aqua regia. The as-cast microstructure consisted of austenite with a dendritic ferrite network and intergranular chromium carbides. After solution treatment, the carbides were largely dissolved, and the structure was austenite with a small amount of ferrite. After the final stabilization treatment, the structure remained austenitic with fine titanium carbides, and there was no significant change compared with the solution-treated state. The magnetic permeability test confirmed that the impeller was completely non-magnetic after heat treatment.
Mechanical Properties
I machined standard tensile specimens from the separately cast and heat-treated test bar. Brinell hardness readings were within the specification. Table 11 presents the mechanical test results.
| Property | Test values | Required (GB228-87) |
|---|---|---|
| Ultimate tensile strength σ_b (MPa) | 495, 510, 502 | ≥450 |
| Yield strength σ_s (MPa) | 240, 260, 250 | ≥196 |
| Elongation δ₅ (%) | 35, 37, 39 | ≥25 |
| Reduction of area ψ (%) | 50, 50, 45 | ≥32 |
| Impact toughness Α_k (×10⁵ J/m²) | 15, 17, 13 | ≥10 |
| Brinell hardness HB | 170, 175, 180 | ≤187 |
All mechanical properties comfortably exceeded the standard. The data were consistent, and the values probably underestimate the properties of the thin blade sections because the test bar had a thicker section with slightly coarser dendrites.
Discussion of Avoiding Sand Casting Defects
The success of this hybrid process relied on systematically eliminating the most common sand casting defects. By using an investment shell for the blade passages, I avoided the irregular surfaces and dimensional mismatch typically associated with sand cores. The shell gave a smooth, accurate blade surface, eliminating sand erosion and metal penetration that cause roughness and inclusions. At the same time, the sand mould portion allowed the use of large risers that would have been impossible in a pure investment shell. This ensured adequate feeding of the thick hub and rim, preventing shrinkage-related sand casting defects. The top-pour design minimized the pouring time, which is critical for austenitic stainless steels because of their susceptibility to oxidation and cold shut formation. I also used chromite sand with a zircon coating to prevent burn-on and further reduce the risk of surface sand casting defects. The strict control of pouring temperature and the use of argon protection minimized oxide inclusions that can cause pinhole and slag-related sand casting defects. Finally, the controlled knock-out time and heat treatment cycle prevented thermal stress cracking, which is one of the most dangerous sand casting defects in large stainless steel castings.
The production trial resulted in ten impellers, all of which passed the rigorous quality inspection. The castings met the dimensional CT4–CT5 and surface Ra 3.2–1.6 requirements, contained no harmful internal discontinuities, and the mechanical properties were fully compliant. This process proved to be stable, repeatable, and much more economical than attempting to produce the same part by purely investment casting or purely sand casting. The use of a sodium silicate sand sleeve as both a structural support and a mould location feature was particularly valuable, as it solved the critical problem of integrating a large ceramic shell into a sand mould while preserving the shell’s shape during dewaxing and handling.
Conclusions
I demonstrated that a large, complex stainless steel impeller with 54 thin blades could be successfully manufactured by combining investment casting for the blade section and sand casting for the hub, spoke, and rim sections. The key findings are as follows:
- The specially designed welding fixture enabled precise assembly of 54 individual wax blades into a cluster, guaranteeing blade spacing within 0.8 mm and enabling uniform coating of the complex assembly.
- The ethyl silicate–sodium silicate composite shell provided excellent surface finish and adequate strength while reducing cost compared with a full ethyl silicate system.
- The sodium silicate sand sleeve around the shell prevented deformation during dewaxing and firing, and facilitated precise positioning of the shell in the sand mould.
- The water-glass sand mould with chromite facing and zircon coating produced a defect-free hub and rim, avoiding sand burning and preventing common sand casting defects such as erosion, penetration, and inclusions.
- Medium-frequency induction melting with a basic lining, strict charge calculation, argon protection, and a controlled pouring temperature of 1580 °C yielded clean steel with fully compliant composition and mechanical properties.
- The final impellers met all dimensional, surface, ultrasonic, and mechanical requirements. The process was stable, economical, and suitable for producing even larger components of similar complexity.
This work confirmed that the composite mould concept is a viable solution for large, intricate castings where neither investment casting nor sand casting alone can achieve the required quality and cost targets. Future studies could extend the method to other alloy systems and even more complex geometries, but the present results already demonstrate a significant advance in the practical manufacture of high-value impeller castings.