In this work, we investigate the casting process of a large double-suction impeller made of C95820 aluminum bronze. The component features a maximum outer diameter of 1015 mm, a height of 630 mm, a central through-hole, and uneven wall thickness with a minimum of 5 mm. The inner cavity contains two asymmetric layers of blades connected by a plate only 15 mm wide, and the upper and lower covers are joined to the blade layers by 19 mm wide connectors. Traditional sand casting methods face significant challenges due to the complex geometry, the tendency of aluminum bronze to form oxide inclusions and gas porosity, and the difficulty of achieving directional solidification. To overcome these issues, we combine sand casting with 3D printing technology, which offers unparalleled freedom in mold design and core fabrication. This paper details our casting process design, including gating and riser optimization using ProCAST simulation, the integrated 3D printing of critical sand cores with conformal gas vents, and the successful production of sound castings.
Our objective is to demonstrate how sand casting and 3D printing can synergistically produce high-quality aluminum bronze double-suction impellers. We emphasize the importance of smooth filling and directional solidification to avoid defects common in this alloy. By leveraging 3D printing, we eliminate the need for complex core boxes, reduce lead time, and enable intricate internal features that would be impossible with conventional pattern-making. The results show that the combination of sand casting and 3D printing is a reliable route for manufacturing large, thin-walled, and geometrically intricate impeller components.
1. Material Properties and Challenges
Aluminum bronze C95820 has a narrow crystallization temperature range, good fluidity, and is resistant to composition segregation and dispersed shrinkage. However, it is prone to oxidation during melting and pouring, leading to oxide inclusions and gas porosity. The alloy also exhibits significant volumetric contraction during solidification, which can cause concentrated shrinkage cavities and slag entrapment. The chemical composition of C95820 is given in the table below.
| Cu | Al | Ni | Fe | Mn |
|---|---|---|---|---|
| ≥77.5 | 9.0–10.0 | 4.5–5.8 | 4.0–5.0 | ≤1.5 |
The double-suction impeller geometry presents additional difficulties. The two blade layers are asymmetric, and the connecting plates are thin. In conventional sand casting, the core for the internal cavities would be fragile and prone to breakage during mold stripping or core removal. Moreover, the large diameter and complex curvature make it challenging to design a risering system that ensures feeding without causing hot spots. With sand casting and 3D printing, we can design integrated cores that include gas vents exactly where needed, and we can optimize the gating system through simulation.
2. Casting Process Design
2.1 Gating System
We chose a bottom-gating system to ensure smooth, progressive filling of the mold cavity. This approach minimizes turbulence, which is critical for preventing oxide entrapment in aluminum bronze. After several design iterations using ProCAST simulation, we adopted a configuration with four bottom ingates directly connected to four blind risers. This design not only controls metal flow but also enhances the feeding capability of the risers. The gating system parameters are summarized below.
| Parameter | Value |
|---|---|
| Type | Bottom-gated |
| Number of ingates | 4 |
| Ingate connection | Direct to blind risers |
| Pouring time | 120 s |
| Pouring temperature | 1150 °C |
2.2 Risering Design
We performed an initial simulation without risers to identify defect-prone regions. The results showed concentrated shrinkage porosity at three thick sections: the blade tips, the blade root junctions, and the central plate. Based on this analysis, we designed a set of risers tailored to the geometry:
- Six open (top) risers at the upper ring cover to feed the thick upper section.
- One cylindrical open riser at the central through-hole to feed the hub area.
- Small cylindrical risers at the tip of each blade to promote directional solidification and to act as vents for gas and inclusions.
- Four blind risers at the bottom connected to the lower ring cover via the ingates.
This combination ensures that all thick sections are fed until the end of solidification, and that any oxides or gas bubbles are carried into the risers rather than remaining in the casting. The riser geometry and placement are summarized in the following table.
| Location | Type | Number | Purpose |
|---|---|---|---|
| Upper ring | Open (top) | 6 | Feed upper thick section |
| Central hole | Open (top) | 1 | Feed hub |
| Blade tips | Open (top) | 12 (one per blade) | Feed thin blade ends and vent |
| Lower ring (bottom) | Blind | 4 | Feed lower thick section, connected to ingates |
3. Simulation Results
We used ProCAST to model the filling and solidification. The mold material was resin-bonded sand with an initial temperature of 80 °C. The heat transfer coefficient between casting and mold was set to 500 W/m²·K. The simulation parameters are listed below.
| Parameter | Value |
|---|---|
| Mold material | Resin sand |
| Pouring temperature | 1150 °C |
| Mold initial temperature | 80 °C |
| Heat transfer coefficient (casting–mold) | 500 W/m²·K |
| Pouring time | 120 s |
| Other parameters | Software defaults |
The filling simulation showed that the liquid metal entered through the blind risers and then filled the cavity from bottom to top without any impingement or splashing. The velocity distribution during filling remained stable, as indicated by the calculated maximum velocities. The kinematic equation for the filling front can be approximated by
$$ \frac{\partial F}{\partial t} + \mathbf{u} \cdot \nabla F = 0 $$
where \(F\) is the volume fraction of liquid and \(\mathbf{u}\) is the velocity field. In our simulation, \(F\) progressed uniformly. At 25% fill, the velocity magnitude was below 5 m/s, and at 98% fill it remained under 3 m/s, confirming a tranquil filling pattern.
The solidification sequence was examined through the fraction solid distribution. The thin blade sections solidified first, followed by the thicker plate regions. The risers solidified last, especially the central open riser. This directional solidification behavior is described by the thermal gradient condition:
$$ G = \frac{\partial T}{\partial x} > 0 $$
where \(G\) is the temperature gradient in the solidification direction. Our simulation showed positive gradients from the casting body toward the risers, confirming that feeding channels remained open until the end. The overall solidification time for the casting was approximately 400 s, while the risers solidified 50–80 s later.
The defect prediction indicated that all shrinkage porosity was confined to the risers and the gating system. No porosity or inclusions were observed in the casting itself. The table below quantifies the simulated defect volume.
| Region | Porosity Volume Fraction (%) |
|---|---|
| Open risers (upper) | 12.3 |
| Blind risers (bottom) | 8.7 |
| Blade tip risers | 5.1 |
| Gating system | 3.4 |
| Casting body | 0.0 |
4. 3D Printing of Sand Molds and Cores
4.1 Core Design and Integration
A key advantage of sand casting and 3D printing is the ability to print complex sand cores as a single piece. For the double-suction impeller, we divided the mold assembly into several parts: the lower outer mold, the upper outer mold, and three cores. The central core (forming the hub through-hole) is simple and has straight gas vents. However, the two blade cores are highly complex, with curved surfaces and thin protrusions. In traditional sand casting, these cores would be assembled from multiple pieces, leading to dimensional inaccuracies and potential core shifts. Using sand casting and 3D printing, we printed each blade core as a monolithic component, ensuring perfect alignment and dimensional stability.
Furthermore, we incorporated conformal gas vents directly into the 3D printed cores. Instead of drilling straight holes after printing, we designed the vents to follow the blade curvature, providing a direct path for gas escape. This is particularly important for aluminum bronze, which has a high tendency to absorb hydrogen and form gas pores. The conformal vents allow evolved gases (from resin decomposition and metal–mold reaction) to be expelled efficiently. The vent design is based on the principle that the maximum allowable gas pressure inside the mold can be expressed as
$$ P_{\text{max}} = \frac{2\sigma \cos\theta}{r} $$
where \(\sigma\) is the surface tension of the molten metal, \(\theta\) is the contact angle, and \(r\) is the effective radius of the vent channel. By keeping the vent radius above 1 mm, we ensured that the gas pressure never exceeded the threshold that could cause blowholes.
4.2 Printing Parameters and Post-Processing
The sand molds and cores were printed using our in-house 3D printing system. The printing material was coated ceramic sand (resin-coated). The key printing parameters are given in the table below.
| Parameter | Value |
|---|---|
| Laser power | 900 W |
| Spot diameter | 1.1 mm |
| Laser scanning speed | 2900 mm/s |
| Layer thickness | 0.3 mm |
| Printing material | Resin-coated ceramic sand |
After printing, we cleaned the loose sand from the cavities and then flame-torched the surface to create a hard crust. The cores and molds were then placed in a furnace, buried in dry sand to support overhangs, and heated to 180 °C for curing. They were subsequently cooled to room temperature inside the furnace. This post-processing improved the surface strength and prevented sand erosion during pouring.
5. Assembly and Pouring
The cured sand components were assembled in sequence. We first placed the lower outer mold, then the blade cores and hub core, followed by the upper outer mold. All joints were sealed with refractory paste. Gas vent tubes (ceramic fiber ropes) were inserted into the conformal vents and extended outside the mold. The assembly was then embedded in a steel flask using water-glass sand for support. Before pouring, the mold cavity was preheated to 80 °C using a hot air blower to reduce thermal shock.
Pouring was performed at 1150 °C with a pouring time of 120 s, as designed. During pouring, we ignited the vent tubes to burn off any combustible gases and to confirm that gases were flowing freely. The process was stable, and no signs of mold cracking or metal eruption were observed.
After solidification and cooling, the casting was shaken out. The risers and gating system were cut off. A visual inspection and subsequent non-destructive testing (ultrasonic and X-ray) confirmed that the casting was free from shrinkage porosity, gas holes, and oxide inclusions. The dimensional accuracy met the specifications, thanks to the integrated core printing. The final product is shown in the figure below.
6. Conclusion
In this work, we successfully demonstrated the application of sand casting and 3D printing to produce a large, complex C95820 aluminum bronze double-suction impeller. The key findings are:
- Bottom-gating with ingates connected to blind risers ensures smooth filling and avoids oxide entrapment, as confirmed by ProCAST simulation.
- Directional solidification was achieved by placing open and blind risers at all thick sections; shrinkage porosity was confined to the risers.
- 3D printing allowed us to fabricate the intricate blade cores as one piece, eliminating assembly errors and improving dimensional accuracy.
- Conformal gas vents integrated into the cores effectively evacuated mold gases, preventing gas porosity.
- The combination of sand casting and 3D printing significantly shortened the production lead time compared to traditional pattern-making, while maintaining high casting quality.
We believe that this approach can be extended to other large, complex castings in copper alloys and beyond. The use of sand casting and 3D printing not only reduces the need for expensive patterns but also enables designs that were previously impractical. Future work will focus on further optimizing the vent geometry and exploring the fatigue performance of the 3D printed sand molds under repeated use.