In our investigation of advanced manufacturing techniques for complex pump components, we focused on the application of 3D printing casting technology to produce integral sand cores for multistage pumps. The multistage pump is a critical fluid transport device that integrates multiple impeller stages to generate high pressure. Its internal flow channels are narrow, curved, and demand excellent surface finish to avoid hydraulic losses and defects. Traditionally, these cores were made by manual sand molding, which involved numerous individual core boxes, extensive assembly, and significant storage requirements. This study systematically compares the manual process with 3D printing casting, highlighting improvements in quality, efficiency, and dimensional control.
We selected a typical eight-stage pump casting with overall dimensions of approximately 1177 mm × 714 mm × 315 mm and a weight of 590 kg. The main wall thickness was 16 mm, with the thinnest section at 6 mm and the narrowest flow channel width at 22 mm. The flow passage surfaces are required to be free from cold shuts, flash, protrusions, and sand inclusions. Because the channels are narrow and curved, post-casting grinding is extremely difficult. Therefore, the casting must achieve high first-pass yield without additional window opening or repair. Our goal was to validate whether 3D printing casting could deliver such quality while reducing manufacturing steps.
Traditional Manual Core Making Process
Traditional manual core making uses relatively coarse sand (40–70 mesh). The ramming process relies on human effort, leading to inconsistent density. Localized under-ramming can cause low-density zones that become sources of sand adhesion, cracks, and flash. After ramming, the cores are often patched manually, further degrading quality. During assembly, multiple cores (typically nine cores from five different core boxes for an eight-stage pump) must be combined into a single core assembly. The numerous joints increase the risk of misalignment, damage, and incomplete seal repair. Storage of core boxes also consumes significant floor space, and most cores are used only once per casting, leading to low utilization of the molds.
3D Printing Casting Approach
We utilized an ink-jet 3D printer (provided by a domestic supplier) to fabricate the entire flow channel core as a single piece. The printing process spreads a layer of pre-mixed sand (with a curing agent) of 70–140 mesh, and the print head selectively deposits furan resin binder according to the sliced 2D data. Each layer thickness is 0.3–0.5 mm. The entire core is built layer by layer, resulting in a dense and uniform structure. After printing, the unbound sand is removed, leaving the finished core. To facilitate handling and flipping during assembly, we designed sand reinforcing ribs between flow channels, which are cut off after assembly. Additionally, we reserved a cylindrical hole along the shaft axis, into which a steel pipe of matching diameter is inserted to provide extra strength during hoisting and inversion.
The following image illustrates the appearance of a 3D printed integral core before trimming.
Comparative Analysis of Core Properties
Surface Finish and Density
We examined the macro morphology of both manual and 3D printed sand cores. The manual core exhibited rough surfaces with visible ramming marks and uneven distribution of sand. In contrast, the 3D printed core showed a smooth, uniform surface without apparent defects. The finer sand particles (70–140 mesh) used in 3D printing casting contribute to a better as-printed surface. Table 1 summarizes the key differences.
$$
\begin{array}{|c|c|c|}
\hline
\text{Property} & \text{Manual Core} & \text{3D Printed Core} \\
\hline
\text{Sand mesh size} & 40–70 & 70–140 \\
\text{Surface roughness} & \text{Coarse, uneven} & \text{Smooth, consistent} \\
\text{Density uniformity} & \text{Poor, local voids} & \text{Uniform} \\
\text{Apparent defects} & \text{Ramming marks, patches} & \text{None} \\
\hline
\end{array}
$$
Strength and Handling
During casting, the core must withstand buoyancy forces and handling stresses. The manual cores can be reinforced by embedding steel wires or rods, but 3D printed cores cannot contain internal metal inserts because the printing process lays sand and binder layer by layer. To compensate, we integrated sand ribs and a steel pipe at the shaft hole. We then performed a flipping test: the 3D printed core (with ribs and pipe) was successfully turned over and assembled. The only failure occurred at the outlet core, which was a single-point cantilevered section with a small base. This part broke under its own weight after the ribs were removed. For future designs, we recommend that such delicate regions be produced separately using manual methods or with additional support structures that are later machined away. Table 2 compares the strength characteristics.
$$
\begin{array}{|c|c|c|}
\hline
\text{Parameter} & \text{Manual Core} & \text{3D Printed Core} \\
\hline
\text{Embedded reinforcement} & \text{Steel wire/rod possible} & \text{Sand ribs + steel pipe at shaft} \\
\text{Core strength (relative)} & \text{Varies, local low} & \text{Uniform, adequate with ribs} \\
\text{Handling success} & \text{Fragile joints} & \text{Good except outlet core} \\
\hline
\end{array}
$$
Dimensional Shrinkage Control
All cast alloys undergo shrinkage during solidification. For multistage pump castings, the critical dimensions are the center distances of flow channels, channel widths, and channel cross-section lengths. We applied empirical shrinkage allowances based on our experience with CA6NM stainless steel (pouring temperature 1580±10 °C). After casting, cleaning, and sand blasting, we measured the key dimensions of the 3D printed core casting. The shrinkage factors matched the design values within ±0.15% tolerance. The shrinkage rate can be defined as:
$$
\text{Shrinkage} = \frac{D_{\text{mold}} – D_{\text{casting}}}{D_{\text{mold}}} \times 100\%
$$
where \(D_{\text{mold}}\) is the dimension on the printed core and \(D_{\text{casting}}\) is the corresponding dimension on the cleaned casting. Our measurements confirmed that the actual shrinkage was within the expected range. Table 3 lists the comparison of nominal and measured shrinkage factors for three principal directions.
$$
\begin{array}{|c|c|c|c|}
\hline
\text{Direction} & \text{Nominal Shrinkage (\%)} & \text{Measured Shrinkage (\%)} & \text{Deviation (\%)} \\
\hline
\text{Center distance (horizontal)} & 2.0 & 2.04 & +0.04 \\
\text{Channel width} & 1.8 & 1.78 & -0.02 \\
\text{Channel length (vertical)} & 2.2 & 2.18 & -0.02 \\
\hline
\end{array}
$$
Cast Surface Quality
We poured the casting using the same CA6NM alloy and standard pouring conditions. After cooling, the casting was cleaned and sandblasted. Visual inspection of the flow channel surfaces revealed a striking contrast between areas produced by manual cores and those from 3D printing casting. The manual core surfaces showed defects such as flash, flow marks, sand holes, scabs, and missing metal (especially on the half-shaft core). In contrast, the 3D printed core surfaces, after being coated with a refractory wash, exhibited a clean, smooth finish with no visible defects. The improvement is attributed to the finer sand, uniform density, and elimination of core joints. Table 4 summarizes the observed defects.
$$
\begin{array}{|c|c|c|}
\hline
\text{Defect Type} & \text{Manual Core Area} & \text{3D Printed Core Area} \\
\hline
\text{Flash / split seam} & \text{Present} & \text{Absent} \\
\text{Flow marks} & \text{Present} & \text{Absent} \\
\text{Sand hole} & \text{Occasional} & \text{Absent} \\
\text{Scab / burn-on} & \text{Visible} & \text{Absent} \\
\text{Missing metal} & \text{On half-shaft core} & \text{Absent} \\
\hline
\end{array}
$$
The 3D printing casting process eliminated the need for extensive core assembly and repair, directly yielding high-quality internal channels. The only problematic area was the outlet core, as noted earlier, which did not have adequate support. This confirms that 3D printing casting is particularly beneficial for complex monolithic cores, but delicate cantilevered sections may still require design optimization or hybrid manufacturing.
Discussion and Benefits of 3D Printing Casting
The adoption of 3D printing casting for multistage pump cores brings several quantifiable advantages. First, the number of core pieces is reduced from nine separate cores to a single integral core, drastically simplifying the assembly process. Second, the storage of core boxes is eliminated; instead, digital models are stored, and cores are printed on demand. Third, the dimensional accuracy is improved, reducing rework and scrap. Fourth, the surface quality of the casting is superior, which is critical for hydraulic performance. Finally, the lead time for prototype and small-batch production is shortened because no tooling is required.
We can express the overall efficiency improvement using a simple metric. Let \(T_{\text{manual}}\) be the total time for manual core making, assembly, and mold storage, and \(T_{\text{3D}}\) be the time for 3D printing, post-processing, and assembly. For our eight-stage pump, we estimated:
$$
\Delta T = T_{\text{manual}} – T_{\text{3D}} \approx 40\% \text{ reduction}
$$
This does not include the savings from eliminated mold storage and reduced defect repair. In terms of material utilization, 3D printing casting uses only the sand necessary for the core, while manual core boxes often require oversize cores to accommodate assembly tolerances.
Limitations and Future Work
Despite its many benefits, 3D printing casting has limitations. The current binder system (furan resin) produces cores with lower collapsibility than phenolic resin, but still acceptable for CA6NM steel. The inability to embed metal reinforcements internally may restrict its use for very large or heavily loaded cores. In our test, the outlet core failed, indicating that design rules need to be developed for 3D printed cores – for instance, increasing local thickness or adding temporary supports that are later machined away. The cost of 3D printing casting is still higher than manual molding for very high volumes, but for medium-to-low volume production (typical for multistage pumps), the overall cost becomes competitive when considering tooling amortization and scrap reduction.
Future work will focus on optimizing the reinforcement strategy, exploring alternative binder materials with better high-temperature performance, and developing hybrid processes that combine 3D printing casting for complex cores with manual methods for simple parts. We also plan to conduct flow simulations to correlate surface quality improvements with hydraulic efficiency gains.
Conclusion
Our study demonstrates that 3D printing casting is a viable and advantageous method for producing integral sand cores for multistage pumps. The printed cores exhibit superior surface finish, uniform density, and adequate strength when auxiliary supports are used. Dimensional shrinkage is consistent with conventional allowances, and the resulting castings have significantly fewer surface defects than those made by manual core assembly. The outlet core remains a challenge that can be addressed by design modifications or separate manual fabrication. Overall, 3D printing casting reduces manufacturing steps, saves time and materials, improves dimensional accuracy and surface quality, and lowers the difficulty of producing complex structural products. As the technology matures and costs decrease, we expect wider adoption in the pump and valve industry.