Composite Casting of Large Complex Stainless Steel Impellers

In my research, I focused on the manufacturing challenges of large, complex, integral stainless steel impellers. These components are critical in shipbuilding, power generation, pharmaceutical equipment, pump and valve systems, and transportation. The performance of such equipment depends heavily on the impeller’s dimensional accuracy, surface finish, internal soundness, and mechanical properties. Conventional single-process approaches, such as standalone investment casting or standalone sand mold casting, often fail to meet all technical requirements simultaneously. Therefore, I developed a melting-module and sand mold composite casting process that combines the advantages of investment casting for complex blade geometries with the advantages of sand mold casting for large, simple hub and spoke sections.

Throughout this article, I use the term sand mold casting repeatedly because it forms the backbone of the simple-section molding strategy. The composite process integrates investment casting for the blade portion and sand mold casting for the hub and spoke portion. This hybrid route enables the production of impellers that are large, geometrically complex, and highly demanding in terms of surface roughness, dimensional tolerance, and internal quality. My experimental work involved pattern design, wax injection, shell building, gating and riser design, sand mold preparation, melting and pouring of ZG1Cr18Ni9Ti stainless steel, post-processing, and comprehensive inspection.

1. Structural Analysis and Technical Requirements

The target impeller had an outer diameter of 700 mm, a height of 130 mm, and a net weight of 202 kg. It contained 54 thin, streamlined blades with a wall thickness of only 5 mm. The blades were closely spaced and arranged in a complex overlapping pattern. The hub and spoke section formed a relatively large flat disc. The material was ZG1Cr18Ni9Ti austenitic stainless steel, which is prone to oxidation, has relatively low fluidity, and exhibits significant volume shrinkage. The technical requirements were stringent: chemical composition according to GB2100-80, no casting defects such as slag inclusion, porosity, cracks, shrinkage cavities, or porosity in the blades; surface roughness Ra3.2–1.6 μm; dimensional tolerance CT4–CT5; rim ovality 0–4 mm; adjacent blade distance variation not exceeding 2 mm; defect diameter in rim, hub, and spoke not exceeding Φ2 mm; non-magnetic after heat treatment; and mechanical properties meeting the casting standard.

Parameter Requirement
Material ZG1Cr18Ni9Ti
Outer diameter 700 mm
Height 130 mm
Net weight 202 kg
Number of blades 54
Blade thickness 5 mm
Surface roughness Ra3.2–1.6 μm
Dimensional tolerance CT4–CT5
Rim ovality 0–4 mm
Adjacent blade distance variation ≤2 mm
Maximum allowable defect diameter Φ2 mm
Magnetic permeability after heat treatment Non-magnetic

Based on these requirements, I divided the impeller into two major regions using the inner diameter of the rim, Φ416 mm, as the boundary. The outer region contains the complex, thin, closely spaced blades. The inner region contains the relatively simple hub and spoke. This division was the key to my composite strategy: investment casting for the complex blades and sand mold casting for the simple hub and spoke.

2. Process Selection: Why Composite Casting?

I evaluated three possible routes: standalone sand mold casting, standalone investment casting, and a composite melting-module/sand mold process.

Standalone sand mold casting: This method is economical and suitable for large simple parts. However, for the blade section, the complex geometry, thin walls, and tight spacing make molding and core assembly extremely difficult. The surface roughness and dimensional accuracy would not reach Ra3.2–1.6 μm and CT4–CT5. Sand mold casting alone would also struggle with the 5 mm blade walls because of poor fluidity and oxidation of ZG1Cr18Ni9Ti.

Standalone investment casting: This method can produce the complex blade geometry with excellent surface finish and dimensional accuracy. However, for a 202 kg impeller, the ceramic shell strength is insufficient. The large flat spoke section is prone to shell deformation. Gating and riser design for the thick hub and rim sections is difficult in a pure investment casting shell. Therefore, standalone investment casting also cannot guarantee the required internal quality and dimensional stability.

Composite melting-module and sand mold casting: By combining the two methods, I could use investment casting for the blade portion and sand mold casting for the hub and spoke portion. The blade shell provides the necessary surface finish and dimensional precision, while the sand mold provides strength, rigidity, and feeding capability for the thick sections. This hybrid approach overcomes the limitations of each individual process. I selected a two-box molding scheme with a top pouring system, three open risers on the rim, and one open riser with a pouring gate on the hub.

3. Technical Route

My experimental route consisted of the following steps: pattern design and manufacture; wax material preparation; blade wax injection; blade assembly welding; shell building using an ethyl silicate–water glass composite; sand set fixation; dewaxing; shell roasting; sand mold casting pattern and core box preparation; sand mixing and molding; composite mold assembly; melting and pouring of ZG1Cr18Ni9Ti; knockout; riser cutting; heat treatment; and final inspection. Each step was carefully controlled to ensure the final impeller met all technical requirements.

4. Blade Section: Investment Casting Shell Preparation

4.1 Pattern Design for a Single Blade

The blade pattern was machined from 45 steel, quenched and tempered to HB 280–350. The pattern cavity dimensions were calculated using the comprehensive average linear shrinkage. The total shrinkage was estimated as:

$$ \varepsilon = \varepsilon_1 – \varepsilon_2 + \varepsilon_3 $$

where \(\varepsilon_1\) is the wax pattern shrinkage, \(\varepsilon_2\) is the shell expansion, and \(\varepsilon_3\) is the metal shrinkage. For a rosin-based wax, I selected \(\varepsilon_1 = 0.75\%\). For the ethyl silicate–water glass shell, \(\varepsilon_2 = 0.50\%\). For ZG1Cr18Ni9Ti, \(\varepsilon_3 = 2.5\%\). Therefore:

$$ \varepsilon = 0.75\% – 0.50\% + 2.5\% = 2.75\% $$

Considering practical variations, I used an actual comprehensive shrinkage of \(\varepsilon_s = 2.6\%\). The pattern cavity nominal dimension was calculated as:

$$ L \pm \delta = L_p (1 + \varepsilon_s \%) \pm \delta $$

where \(L_p\) is the average casting dimension, and \(\delta\) is the manufacturing tolerance, typically 1/3 to 1/5 of the casting tolerance. To achieve CT4–CT5, I designed the pattern to CT3 and then adjusted it after trial injections. The surface roughness of the pattern was specified as Ra 0.2–0.8 μm for the cavity, Ra 0.8–3.2 μm for matching surfaces, Ra 0.8–1.6 μm for the parting plane, and Ra 6.3–12.5 μm for non-working surfaces.

Pattern Location Surface Roughness Ra (μm)
Cavity 0.2–0.8
Insert matching and locating surfaces 0.8–3.2
Parting plane 0.8–1.6
Non-working surfaces 6.3–12.5

4.2 Wax Material and Injection

I developed a modified rosin–wax based wax. It contained polymerized rosin, modified rosin, paraffin, and EVA. The composition and properties are given in the following table.

Component Content (%) Property Value
Polymerized rosin 17 Melting point 74–78 °C
Modified rosin 40 Softening point >40 °C
Paraffin 30 Tensile strength 54 MPa
EVA 3 Shrinkage 0.4–0.7%

The wax injection process used a manual or semi-automatic wax injection machine. The workshop temperature was 15–28 °C, the wax temperature was 70–85 °C, the pattern temperature was 20–30 °C, the injection pressure was 0.3–1.5 MPa, and the holding time was 1–3 min. The injected wax pattern was cooled in water at 16–25 °C for no more than 1 h, then stored at 15–28 °C on dedicated racks. Each blade wax pattern was inspected for flash, cold shuts, cracks, shrinkage, and bubbles. Dimensional inspection used calipers, templates, and special gauges.

Process Step Parameter
Wax melting ≤200 °C
Workshop temperature 15–28 °C
Wax temperature 70–85 °C
Pattern temperature 20–30 °C
Injection pressure 0.3–1.5 MPa
Holding time 1–3 min
Cooling water temperature 16–25 °C
Cooling time ≤1 h
Storage temperature 15–28 °C

4.3 Assembly Tooling and Blade Welding

The assembly tooling consisted of an upper plate, a lower plate, a locating ring, a fastening shaft, and two backing plates. The locating ring was a critical component. Its outer diameter was calculated from the rim inner diameter of 416 mm with a shrinkage allowance of 2.2%:

$$ D_{ring} = 416 \times (1 + 0.022) = 425.15 \pm 0.25 \text{ mm} $$

The ring height accounted for the impeller height plus machining allowances:

$$ H = (130 + 3 + 5) \times (1 + 0.022) \approx 141 \text{ mm} $$

The upper plate outer diameter was 840 mm, and the lower plate outer diameter was 1240 mm. The lower plate included a Φ973 mm step to locate a water glass sand set. The fastening shaft had an effective length of about 318 mm and a diameter of Φ30 mm. Each fastening rod had an M30 nut welded to it.

Before welding, each wax blade was trimmed, cleaned in a neutral soap solution, rinsed, and inspected. The 54 blades were placed on the tooling, located by the locating ring and the rim arc. The upper surface joints were welded first at about 80 °C. The lower plate was then installed, the assembly was flipped, and the lower surface joints were welded. The tooling was then closed and fastened. Any gaps between the fastening nuts and the plates were filled with wax to prevent the hardening liquid from penetrating during shell building. The welding seams had to be smooth, free of wax beads, concave gaps, or cold welds.

5. Ethyl Silicate–Water Glass Composite Shell

5.1 Slurry Preparation

The face and second layers used an ethyl silicate slurry. The composite hydrolysis formula is shown below. I mixed ethyl silicate and alcohol, then added quartz powder, sulfuric acid, and acidified water under stirring. The temperature was controlled at 50 ± 5 °C. After stirring for 40–60 min, the density was measured at 1.69–1.71 g/cm³. The one-time hydrolysis formula is also given. The face coat density was 1.62–1.65 g/cm³, and the second coat density was 1.56–1.60 g/cm³. The slurry was allowed to rest for at least 40 min before use.

Component Amount for Composite Hydrolysis Amount for One-Time Hydrolysis
Ethyl silicate 1 kg 1 kg
Alcohol (93–96%) 654–680 mL 0.6 kg
Sulfuric acid 6.8 mL 3.5 mL
Hydrochloric acid 8.3 mL 3.0 mL
Distilled water 179–205 mL 0.18 kg
Acetic acid — 3.5 mL
Quartz powder (0.055 mm) 3.9–4 kg —

5.2 Shell Building

The shell consisted of two ethyl silicate layers followed by three to four water glass layers. The detailed shell building parameters are given in the table below. The face and second layers used quartz sand of 0.425/0.212 mm. The third layer used a water glass slurry with quartz sand of 0.850/0.425 mm. The fourth to sixth layers used a water glass slurry with quartz sand of 1.70/0.850 mm. The ethyl silicate layers were dried in air and then hardened with ammonia. The water glass layers were hardened with an ammonium chloride solution.

Layer Slurry Viscosity (s) Sand Size (mm) Drying (h) Hardening Agent Hardening Time (min)
Face Ethyl silicate + quartz 28–32 0.425/0.212 >2 Air + ammonia 30–40
Second Ethyl silicate + quartz 28–32 0.425/0.212 >2 Air + ammonia 30–40
Third Water glass + bauxite + quartz 30–45 0.850/0.425 — NH4Cl solution 5–10
Fourth–sixth Water glass + bauxite + quartz 30–35 1.70/0.850 — NH4Cl solution 5–10

I paid special attention to the interface between the second and third layers. To prevent delamination, I reduced the viscosity of the third water glass slurry and used a finer sand size than in a conventional water glass shell. I also extended the air drying time by 1–2 h before hardening. This ensured good bonding between the ethyl silicate and water glass layers.

5.3 Sand Set Fixation

After completing the shell, I fixed the blade shell in a water glass sand set. A ring-shaped sand set mold was placed on the Φ973 mm step of the lower plate. A cast iron reinforcing ring with two lifting lugs was inserted. The same water glass sand formula used for molding was poured into the ring and hardened with CO₂ gas. The sand set thickness was about 100 mm. This sand set prevented deformation during dewaxing, roasting, and transport, and it provided a precise locating surface for later composite mold assembly.

5.4 Dewaxing and Roasting

Dewaxing was performed in a high-pressure steam dewaxing autoclave. The steam pressure was 0.6–1 MPa, the inflation time was less than 5 s, and the dewaxing time was 6–10 min. After dewaxing, the shell was stored for more than 12 h but not more than 2 days before roasting. Roasting was carried out in a 75 kW box-type electric furnace at 850 ± 30 °C for 1.5–2 h. The shell was placed horizontally in a stainless steel tray filled with 20/40 mesh quartz sand to prevent deformation. A properly roasted shell appeared white and did not emit black smoke. Repeated roasting was avoided because it could cause microcracks or delamination.

6. Sand Mold Casting for the Hub and Spoke

The hub and spoke section was produced by sand mold casting. This section is much simpler in shape than the blade section, but it contains thick sections that require adequate feeding. The sand mold casting process provided the necessary mold strength, thermal stability, and feeding capability. I used a wooden pattern and a wooden core box because the production was a single-piece batch with manual molding. The pattern working dimensions were calculated as:

$$ A_M = (A_C \pm A_t)(1 + \varepsilon) $$

where \(A_M\) is the pattern working dimension, \(A_C\) is the product dimension, \(A_t\) is the casting allowance, and \(\varepsilon\) is the linear shrinkage, taken as 2.2%. The plus sign was used for external dimensions, and the minus sign for internal dimensions.

For the molding sand, I selected a chromium铁矿 sand (chromite sand) facing sand and a quartz sand backing sand. Chromite sand has high refractoriness, good thermal stability, and resistance to metal penetration, which is important for stainless steel castings. The facing sand and core sand used a water glass CO₂ hardening system. The sand mixture is given below.

Component Amount Property Value
Chromite sand (40/70 mesh) 100% Wet permeability 4–5
Water glass 6–7% Wet compressive strength >200 kPa
Water 4–5% Hardened compressive strength >1.5 MPa

The sand was mixed in the order: original sand → powder additives → water glass → liquid additives. Dry mixing time was 2–3 min, and wet mixing time was 2–4 min. The sand was used promptly and kept covered to prevent drying and premature hardening. Molding and core making were performed manually. The sand was compacted, then hardened by inserting a perforated metal tube and blowing CO₂ gas. The blowing pressure was 0.10–0.15 MPa, the flow rate was 0.5 m³/h, and the blowing time was 20–30 s.

Parameter Value
Blowing pressure 0.10–0.15 MPa
Flow rate 0.5 m³/h
Blowing time 20–30 s

After hardening, the mold and core surfaces were coated with a zircon-based alcohol coating. The coating formula is given below. The coating was applied by brushing to a thickness of 0.3–1 mm. The alcohol-based coating self-hardened or could be ignited to harden, which reduced drying time and improved surface quality.

Component Content (%)
Zircon powder 70
Bauxite 30
Resin 1.2–2.0
Polyvinyl butyral 1–4
Na-bentonite 1–5
Clay 0.5–1.5
Urotropine 8
Ethanol Balance
Water 3.5–4.5

7. Gating and Riser Design

The gating and riser system was designed to ensure complete filling and adequate feeding of the thick hub and rim sections. I selected a top pouring system with one open riser on the hub and three open risers on the rim. The riser dimensions were calculated using the modulus method.

7.1 Riser Modulus Calculation

For the hub section, treated as a rod–plate intersection with a wall thickness factor \(k = 1.1\):

$$ a = 1.1 \times 50 = 55 \text{ mm}, \quad b = 138 \text{ mm}, \quad c = 30 \text{ mm} $$

$$ M_{hub} = \frac{a b}{2(a + b – c)} = \frac{55 \times 138}{2(55 + 138 – 30)} = 20.7 \text{ mm} = 2.07 \text{ cm} $$

The hub riser modulus was:

$$ M_{hub\ riser} = k \times M_{hub} = 1.2 \times 2.07 = 2.48 \text{ cm} $$

For the rim section, considering the blade cooling effect, I used:

$$ a = 50 \text{ mm}, \quad b = 138 \text{ mm}, \quad c = 30 \text{ mm} $$

$$ M_{rim} = \frac{a b}{2(a + b – c)} = \frac{50 \times 138}{2(50 + 138 – 30)} = 18.9 \text{ mm} = 1.89 \text{ cm} $$

$$ M_{rim\ riser} = 1.2 \times 1.89 = 2.27 \text{ cm} $$

7.2 Volume Shrinkage of ZG1Cr18Ni9Ti

The volume shrinkage of the alloy was estimated from its composition and pouring temperature. The total shrinkage was calculated as 5.952%, rounded to 6%. The contribution of each element is given below.

Element Content (%) Effect on Shrinkage (%) Contribution (%)
C 0.10 — 3.2
Mn 1.20 0.0585 0.0702
Si 1.20 1.03 1.236
Cr 18.00 0.10 1.80
Ni 10.00 −0.0354 −0.354
Total 5.952

7.3 Riser Size and Number

For the hub, I selected a standard waist-circle open riser with \(a = 120\) mm, \(b = 180\) mm, and \(h = 150\) mm. The riser weight was about 25 kg, and its maximum feeding volume was \(4.3 \times 10^3\) cm³. The required feeding volume for the hub section was calculated as:

$$ V_{hub\ required} = (14.65^2 – 8.5^2) \times 3.14 \times 3 + (8.5^2 – 3.5^2) \times 3.14 \times 13.8 = 3.94 \times 10^3 \text{ cm}^3 $$

Since \(3.94 \times 10^3 < 4.3 \times 10^3\), one riser was sufficient for the hub.

For the rim, I selected a standard waist-circle open riser with \(a = 110\) mm, \(b = 165\) mm, and \(h = 150\) mm. The riser weight was about 21 kg, and its maximum feeding volume was \(3.0 \times 10^3\) cm³. The required feeding volume for the rim was:

$$ V_{rim\ required} = (24.3^2 – 20.8^2) \times 3.14 \times 13.8 + (20.8^2 – 14.65^2) \times 3.14 \times 3 = 8.89 \times 10^3 \text{ cm}^3 $$

The number of rim risers was:

$$ \frac{8.89 \times 10^3}{3.0 \times 10^3} = 2.96 \approx 3 $$

Therefore, three rim risers were required.

7.4 Verification of Riser Number and Pouring Time

For the hub, the effective feeding distance of one riser was:

$$ L = 4.5T = 4.5 \times 50 = 225 \text{ mm} $$

The riser length was 180 mm, so the total feeding length was \(225 + 180 = 405\) mm. The hub circumference was:

$$ \pi D = 3.14 \times 125 = 392.5 \text{ mm} < 405 \text{ mm} $$

Thus, one hub riser was sufficient.

For the rim, the circumference was:

$$ \pi D = 3.14 \times 456 = 1431.84 \text{ mm} $$

With three risers, each having an effective feeding distance of \(4T = 4 \times 60 = 240\) mm and a riser length of 165 mm, the total feeding length was \((240 + 165) \times 3 = 1215\) mm. Although this is slightly less than the rim circumference, the cooling effect of the blades (cold ribs) provided additional feeding, so three risers were adequate.

The minimum liquid rising speed was taken as 0.01 m/s. The casting height was 0.138 m, so the required pouring time was 10–12 s. The actual rising speed was:

$$ v = \frac{0.138}{10 \text{ to } 12} = 0.0115 \text{ to } 0.0138 \text{ m/s} > 0.01 \text{ m/s} $$

Therefore, a pouring time of 10–12 s ensured no sand inclusion defects.

The casting yield was calculated as:

$$ \text{Yield} = \frac{202}{202 + 88} \times 100\% = 69.6\% $$

This is within the normal range for such castings.

7.5 Other Process Parameters

The sand allowance (distance between pattern and flask wall) was 50–100 mm. The facing sand thickness was 20–25 mm for normal sections and 30–35 mm for thick sections. The lifting force on the upper mold was calculated as:

$$ F_{lift} = K \rho g S h = 1.2 \times 7800 \times 10 \times \pi \times 0.33^2 \times 0.335 = 10.56 \times 10^3 \text{ N} $$

Therefore, the mold clamping force was \(10.56 \times 10^3\) N, achieved by placing weights on the flask.

8. Composite Mold Assembly

After the blade shell and the hub/spoke sand mold casting molds were prepared, I assembled the composite mold. The core for the hub bore was placed on the core seat of the lower mold. The blade shell, with its water glass sand set, was located in the Φ973 mm locating circle of the sand mold cavity. The upper mold was then placed on the lower mold using locating pins. I checked for damage, venting, and cleanliness before closing. The parting line and riser bases were sealed with clay strips to prevent metal leakage. A pouring cup and riser collars were placed, and the mold was clamped with weights. The composite mold combined the precision of the investment casting shell for the blades with the strength and feeding capability of the sand mold casting for the hub and spoke.

9. Melting and Pouring of ZG1Cr18Ni9Ti

9.1 Charge Calculation

I used a 300 kg medium-frequency induction furnace with a basic lining. The charge was calculated for 300 kg, based on a 100 kg balance multiplied by 3. The target composition and raw material compositions are given below. The recovery rates for alloying elements were considered, and the charge was adjusted accordingly.

Element Standard (%) Target (%) Recovery (%) Required (kg per 100 kg)
C ≤0.12 0.07 100 0.07
Si ≤1.5 0.80 90 0.89
Mn 0.8–2.0 1.5 95 1.58
S ≤0.03 0.020 100 0.020
P ≤0.045 0.030 100 0.025
Cr 17–20 18 98 18.36
Ni 8.0–11.0 10 99.9 10.0
Ti 0.5–0.7 0.50 80 0.625
Charge Material Amount (kg per 100 kg) C Si Mn Cr Ni Ti
Return scrap 60.00 0.048 0.480 0.840 10.5 5.10 0.30
Electrolytic nickel 5.010 0.0015 — — — 5.00 —
Ferrochrome 12.68 0.0063 0.049 0.038 7.867 — —
Ferrosilicon 0.324 0.0006 0.2441 0.004 — — —
Electrolytic manganese 0.601 0.0002 — 0.5943 — — —
Low-phosphorus scrap 20.302 0.0217 0.0651 0.1086 — — —
Ferrotitanium 1.083 0.0007 0.054 — — — 0.325
Aluminum 0.10 — — — — — —
Total 100.18 0.079 0.887 1.585 18.37 10.10 0.625

For a 300 kg charge, all amounts were multiplied by 3. The final charge consisted of 3.25 kg ferrotitanium, 15.03 kg electrolytic nickel, 38.04 kg ferrochrome, 0.97 kg ferrosilicon, 1.80 kg electrolytic manganese, 60.9 kg low-phosphorus scrap, 0.24 kg aluminum powder, and 0.3 kg aluminum block, totaling 300.54 kg.

9.2 Furnace Lining and Sintering

The furnace lining was rammed from metallurgical magnesia, boric acid, and water glass. The magnesia had MgO ≥ 90%, SiO₂ ≤ 4%, CaO < 1%, and Fe₂O₃ < 1%. The particle size distribution was 15% 4–6 mesh, 30% 10–20 mesh, 20% 20–40 mesh, and 35% finer than 120 mesh. Boric acid was added at 0.8–1.0%, and a water glass solution with a density of 1.15 g/cm³ was added at 15 L. The mixture was rammed in layers of 25–35 mm, each layer rammed for 25 min. After air drying for 1–2 days, the lining was sintered at 1500–1550 °C for 20–30 min. The furnace was then checked for cracks and repaired if necessary.

9.3 Melting and Pouring

The melting process followed a non-oxidation route. The charge was loaded with small pieces at the bottom, ferroalloys in the middle, and larger pieces at the top. Power was initially 60% for 6–8 min, then increased to maximum. Slag was formed using a mixture of lime and fluorspar in a 2:1 ratio. After 95% of the charge was melted, a sample was taken for analysis, and the remaining charge was added. The slag was removed, and a new slag was formed. Deoxidation was performed by adding a mixture of lime and aluminum powder (1:2). The composition was adjusted, and ferrotitanium was added at 1630–1650 °C. Final deoxidation was done with 0.1% aluminum. The tapping temperature was 1620 ± 10 °C, and the pouring temperature was 1580 ± 10 °C. The pouring time was 10–12 s. Argon was used to purge the melt and protect the stream. After pouring, the risers were covered with insulating material.

Step Operation Parameter
1 Charging Small pieces at bottom, ferroalloys in middle, large pieces at top
2 Melting 60% power for 6–8 min, then maximum
3 Slagging Lime:fluorspar = 2:1, 1–1.5% of charge
4 Sampling At 95% melted
5 Deoxidation Lime:aluminum = 1:2
6 Composition adjustment Si adjusted 10 min before tapping
7 Titanium addition At 1630–1650 °C
8 Final deoxidation 0.1% Al
9 Tapping 1620 ± 10 °C
10 Pouring 1580 ± 10 °C, 10–12 s

10. Post-Processing

10.1 Knockout and Cleaning

The composite mold was knocked out 40–60 min after pouring to accelerate cooling and obtain a dense structure in the hub and rim. The water glass sand mold and sand set were removed mechanically after the casting cooled to 50–60 °C. The investment shell between the blades was removed by alkaline boiling. The alkaline solution contained 15–20% NaOH and was boiled for 4–8 h. After boiling, the casting was neutralized in a solution of 90 g chromic oxide, 30 g sulfuric acid, 1.2 g sodium oxide, and 1 kg water at 18–28 °C for 2–3 min, then rinsed with clean water. This method effectively removed the shell without damaging the blade surfaces.

10.2 Riser Cutting

The risers were cut using vibration gas cutting. This method uses a standard cutting torch that is vibrated to break through the refractory oxide film on stainless steel. The cutting parameters are given below. The residual riser height was 10 mm or less. The cutting was fast and economical.

Parameter Value
Amplitude 10–15 mm
Oscillation frequency About 80 cycles/min
Preheat Surface melting
Flame Neutral
Cutting oxygen pressure 0.8–1 MPa
Acetylene pressure 0.05–0.1 MPa

10.3 Heat Treatment

The impeller was solution treated and then stabilized. Solution treatment was performed at 1050–1100 °C for 3–4 h, followed by water quenching. The heating rate was 200–300 °C/h. After solution treatment, stabilization was performed at 850–950 °C for 2–4 h, followed by air cooling. The heat treatment improved mechanical properties, reduced segregation, and increased corrosion resistance. The impeller was non-magnetic after heat treatment.

11. Results and Analysis

11.1 Chemical Composition

The chemical composition was analyzed by direct-reading optical emission spectrometry. The results are shown below. All elements were within the specification, and the carbon content was well controlled.

Element C Si Mn Cr Ni Ti S P
Analysis 1 0.080 1.0 1.5 18.0 9.0 0.62 0.026 0.032
Analysis 2 0.083 1.1 1.3 18.5 10.0 0.58 0.025 0.030
Calculated 0.079 0.887 1.585 18.37 10.10 0.625 0.0175 0.0279
Standard ≤0.12 ≤1.5 0.8–2.0 17–20 8.0–11.0 0.5–0.7 ≤0.030 ≤0.045

11.2 Visual and Dimensional Inspection

Visual inspection using a 5× magnifier revealed no run-out, sand adhesion, swelling, sand inclusion, cold shuts, or misruns. The impeller contour was clear, the blades were fully formed, and the surfaces were smooth. The residual riser height after cutting was 5–8 mm. Flash was present on the upper and lower blade faces due to the mold parting line, but the machining allowances of 5 mm and 3 mm eliminated this issue. The surface roughness was Ra1.6 for the blades and Ra3.2–1.6 for the other surfaces, meeting the requirement. Dimensional inspection showed that the rim roundness error was within 0.5 mm, the adjacent blade distance variation was less than 0.8 mm, and the flatness error was within 1 mm. The casting dimensions met CT4–CT5, and the blade position dimensions reached CT3.

11.3 Nondestructive Testing

Penetrant testing was performed on the blade surfaces, blade roots, and the junctions between the spoke, hub, and rim. No linear or clustered indications exceeded the acceptance level. Ultrasonic testing was performed on the hub, rim, and spoke using a 2.5P20Z straight probe. The sensitivity was set to Φ2 mm flat-bottom hole. The scanning speed was adjusted using a 1:4 ratio for the hub and rim and a 1:1 ratio for the spoke. The calculations for sensitivity adjustment are shown below.

$$ \Delta = 20 \log \frac{P_f}{P_b} = 20 \log \frac{\lambda x^2}{\pi D_f^2} = 20 \log \frac{2.5 \times 3.14 \times 2^2}{5.9 \times 138} = 34.3 \text{ dB} \quad \text{(hub and rim)} $$

$$ \Delta = 20 \log \frac{2.5 \times 3.14 \times 2^2}{5.9 \times 30} = 21 \text{ dB} \quad \text{(spoke)} $$

No defects larger than Φ2 mm were detected in the hub, rim, or spoke. The internal quality met the technical requirements.

11.4 Microstructure

Metallographic samples were prepared from the riser material and etched with aqua regia. The as-cast microstructure consisted of an austenite matrix with dendritic ferrite and granular carbides. After solution treatment, the microstructure was austenite with a small amount of ferrite. After solution treatment plus stabilization, the microstructure was similar, with austenite and a small amount of ferrite. The impeller was non-magnetic after heat treatment.

11.5 Mechanical Properties

Hardness was measured on the blade and rim using a Brinell hardness tester with a 5 mm ball and 750 kg load. Tensile tests were performed on standard specimens machined from梅花试棒 (plum-blossom test bars) that were cast and heat treated together with the impeller. The results are shown below. All properties met the standard.

Property Test Value Standard
Hardness (HB) 170–180 ≤187
Tensile strength σb (MPa) 495–510 ≥450
Yield strength σs (MPa) 240–260 ≥196
Elongation δ5 (%) 35–39 ≥25
Reduction of area ψ (%) 45–50 ≥32
Impact toughness Ak (×10⁵ J/m²) 13–17 ≥10

12. Conclusion

My research demonstrated that the melting-module and sand mold casting composite process is an effective route for manufacturing large, complex stainless steel impellers. By dividing the impeller into a complex blade section and a simple hub/spoke section, I was able to apply investment casting for the blades and sand mold casting for the hub and spoke. The following conclusions can be drawn:

1. The blade assembly tooling solved the problems of positioning, welding, and slurry application for the complex blade section. It can be adapted for other similar castings.

2. The ethyl silicate–water glass composite shell provided a good surface finish and sufficient strength. It is a simple, practical, and low-cost shell system.

3. The composite mold combined the advantages of investment casting and sand mold casting, overcoming the limitations of each individual process. This made the production of large, complex integral impellers feasible.

4. For ZG1Cr18Ni9Ti, strict charge control, rapid melting in a medium-frequency basic induction furnace, a melting time within 60 min for a 300 kg heat, and a pouring temperature of about 1580 °C were essential.

5. The produced impellers, each weighing 202 kg, achieved dimensional accuracy CT4–CT5 and surface roughness Ra3.2–1.6. The blades were free of defects, and the hub, rim, and spoke showed no defects larger than Φ2 mm. The mechanical properties and chemical composition met the specification. The impellers performed at the level of imported components and were successfully used as replacements.

6. Ten impellers were produced in one batch, generating significant economic benefit. The process was stable and cost-effective.

This work provides a new approach for manufacturing similar large, complex castings. Further research can extend the composite process to other impeller types and other alloys.

Scroll to Top