This paper presents a comprehensive investigation into the forming technology of an integral pump valve with complex internal geometry, emphasizing the integration of digital manufacturing, rapid prototyping, and a novel submerged-gate casting process. The study focuses on the feasibility of applying this innovative technique to complex valve castings traditionally manufactured through sand casting foundry operations. A series of water simulation experiments were conducted to analyze flow field stability, slag removal efficiency, and control of the submerged gate tube motion. The casting design was verified by numerical simulation using JSCAST, and the sand core was fabricated by the PIRP rapid prototyping method. Final pouring experiments were carried out with ductile iron, and the obtained casting was subjected to three-dimensional scanning and defect inspection. The results demonstrate that the combination of rapid prototyping and submerged-gate casting can produce high-quality complex pump valves with dimensional deviations mostly within -1.5 to -0.5 mm and without visible internal defects. This work provides a new practical route for producing complex pump valves within the framework of sand casting foundry technologies.
1 Introduction
Cast metal pump valves play an essential role in the foundry industry due to their wide application in fields such as machinery, automotive, aerospace, and marine engineering. Among them, complex integral pump valves are particularly challenging because of their intricate internal channels, thin wall sections, and demanding performance requirements. Traditional sand casting foundry methods often rely on manually produced cores that are segmented and then bonded together, which inevitably introduces dimensional errors and reduces core strength. Such shortcomings become especially critical when producing integral valves with complex curved passages.
To overcome these limitations, this research adopts a digitalized casting route that combines advanced rapid prototyping for the fabrication of monolithic sand cores, computer-aided simulation for process design, and a novel submerged-gate casting technique that enables controlled filling of the mold cavity. The main objectives are:
- To develop a systematic method for producing complex integral pump valve castings using sand casting foundry concepts.
- To explore the behavior of the flow field during submerged-gate casting by means of water modeling experiments.
- To optimize the motion trajectory and speed control of the submerged gate tube using an industrial robot.
- To validate the casting design through numerical simulation and actual pouring experiments.
- To evaluate the dimensional accuracy and internal integrity of the final casting using three-dimensional scanning and non-destructive testing.
The rest of this paper is organized as follows: Section 2 describes the water simulation experimental platform and the results obtained for different casting conditions. Section 3 presents the detailed casting design and numerical simulation. Section 4 explains the rapid manufacturing of the sand mold and core. Section 5 covers the pouring experiments and post-processing. Finally, conclusions are summarized.
2 Water Simulation Experiments
2.1 Theoretical Basis of Water Simulation
Water modeling is widely accepted as an efficient method to study mold filling in sand casting foundry processes, since the Reynolds number of molten metal flow in the gating system is generally in a turbulent regime similar to that of water. The fundamental similarity criteria are geometric, kinematic, and dynamic similarity. The submerged-gate casting process was treated as a special type of layered filling, where the gate tube outlet always remains below the free surface. The flow velocity at the tube outlet can be derived from Bernoulli’s equation:
$$ \frac{P_1}{\rho g} + \frac{v_1^2}{2g} + H_1 = \frac{P_2}{\rho g} + \frac{v_2^2}{2g} + H_2 + h_w \tag{1}$$
in which \(P_1\) and \(P_2\) are the pressures at the in-gate and out-gate conditions, \(v_1\) and \(v_2\) are the corresponding velocities, \(H_1\) and \(H_2\) are the heights of liquid in the ladle and the mold respectively, and \(h_w\) represents the total energy loss. For free outflow (\(H_2 = 0\), \(v_1 \ll v_2\), \(P_1 = P_2\)), the equation reduces to:
$$ v_2 = \mu_1 \sqrt{2gH_1} \tag{2}$$
where \(\mu_1\) is the flow coefficient for free outflow. For submerged (orificed) outflow, the pressure terms cancel and the differential head \(H = H_1 – H_2\) governs:
$$ v_2 = \mu_1 \sqrt{2gH} \tag{3}$$
The flow continuity equation is expressed as:
$$ v_1 A_1 = v_2 A_2 = v_3 A_3 \tag{4}$$
where \(A_1\) is the ladle cross-sectional area, \(A_2\) is the effective area of the tube outlet, and \(A_3\) is the instantaneous horizontal cross-sectional area of the mold cavity. From Eq. (4), the required rising speed of the submerged tube can be obtained once the outflow velocity is known:
$$ v_{rise} = \frac{A_2}{A_3} v_2 \tag{5}$$
Since the mold cross-sectional area varies with height, the rising speed must be adjusted continuously. In practice, the height range was divided into discrete sections, and each section was assigned an average speed based on its average cross-sectional area.
2.2 Experimental Setup and Equipment
The water simulation platform was constructed using an industrial robot, a transparent acrylic intermediate ladle, a slag pocket, and a mold box. The robot arm held the intermediate ladle and controlled the vertical motion of the submerged gate tube, while an electromagnetic valve regulated the water flow. The entire system was controlled via the robot controller, enabling precise speed and position profiles.
Several types of submerged tubes were tested: three bottom-discharge tubes with inner diameters of 12 mm, 16 mm, and 20 mm, and three side-discharge tubes with an inner diameter of 20 mm and lateral orifices of 8 mm, 11 mm, and 14 mm. The mold box was made of transparent acrylic to allow direct observation of the flow field.
| Tube Type | Inner Diameter (mm) | Outlet Configuration | Outlet Diameter (mm) |
|---|---|---|---|
| Direct Flow 20 | 20 | Bottom | 20 |
| Direct Flow 16 | 16 | Bottom | 16 |
| Direct Flow 12 | 12 | Bottom | 12 |
| Lateral Flow 14 | 20 | Side (two orifices) | 14 |
| Lateral Flow 11 | 20 | Side (two orifices) | 11 |
| Lateral Flow 8 | 20 | Side (two orifices) | 8 |
2.3 Measurement of Flow Coefficients
To determine the actual outflow velocity for each tube, a calibration experiment was conducted by filling a cylindrical container of known volume and recording the filling time. The first attempt accounted for the entire time from the opening of the valve to the moment the container was full. However, it was discovered that the water required approximately 0.8 s to travel from the ladle valve to the tube outlet, irrespective of the tube geometry. After correcting for this transit time, the accurate filling times were obtained.
| Tube | Initial filling time (s) | Corrected filling time (s) | Outflow velocity (mm/s) | Mold rise speed (mm/s) |
|---|---|---|---|---|
| Direct 20 | 2.8 | 1.5 | 1332.5 | 7.0 |
| Direct 16 | 2.0 | 1.9 | 1644.5 | 5.5 |
| Direct 12 | 2.8 | 2.3 | 2415.5 | 4.6 |
| Lateral 14 | 2.3 | 2.0 | 1020.4 | 5.2 |
| Lateral 11 | 2.6 | 2.4 | 1376.0 | 4.4 |
| Lateral 8 | 3.9 | 3.0 | 1847.8 | 3.5 |
The corrected speeds were entered into the robot program. When the actual motion was visually monitored, the submerged tube remained at a relatively constant depth below the free surface, confirming that the corrected values were accurate and reliable.
2.4 Effect of Intermediate Ladle and Slag Pocket
The purpose of the intermediate ladle is to prevent slag and impurities from entering the mold cavity. The intermediate ladle used in this study had a vertical baffle dividing it into two chambers: an inlet side and an outlet side. Water entered from one side, flowed under the baffle, and exited through the bottom of the other side. Two operational modes were compared:
- Batch mode: The intermediate ladle was first filled with water, then the valve was opened without any further supply. In this mode, the baffle was effective while the water level remained above the baffle height. However, as the water level dropped near the bottom, the remaining water flowed through the gap and carried floating impurities into the mold. Moreover, air bubbles were entrained at the end of the discharge.
- Continuous mode: The intermediate ladle was kept completely full throughout the test by continuously supplying water through the inlet. In this case, all impurities were trapped on the inlet side of the baffle, and the outgoing stream remained clean and free of bubbles even at the end of the process.
Therefore, the continuous mode was selected as the preferred strategy for actual pouring in sand casting foundry operations. The slag pocket placed at the bottom of the mold also served to collect the first turbulent portion of the metal jet, which may contain gas bubbles and oxide inclusions. This pocket allowed the liquid to stabilize before entering the main cavity, significantly reducing the risk of gas porosity in the final casting.
2.5 Influence of Discharge Type and Outlet Area
For each submerged tube, the flow field in the mold was observed during the experiment. Figure-like observations (not reproduced here) showed that for bottom-discharge tubes, a turbulent region formed directly beneath the tube outlet. The size and intensity of this region depended on the outlet diameter and velocity. Larger diameters produced a wider but gentler turbulent zone, while smaller diameters produced a narrow high-velocity jet with almost negligible disturbance to the surrounding flow. Side-discharge tubes created turbulent zones on both sides of the tube, whereas the region below the tube remained relatively calm. This behavior is important when deciding which tube configuration to use for a particular casting geometry.
| Configuration | Disturbance location | Intensity | Suitability |
|---|---|---|---|
| Direct 20 | Below tube, broad | Low | Large flat cavities |
| Direct 16 | Below tube, moderate | Moderate | Mixed geometry |
| Direct 12 | Below tube, narrow | High | Thin vertical sections |
| Lateral 14 | Sides of tube | Low | Complex internal cores |
| Lateral 11 | Sides of tube | Moderate | General use |
| Lateral 8 | Sides of tube, narrow | High | Very narrow pockets |
2.6 Optimization of the Submerged Tube Trajectory
Two major issues were observed during the initial trajectory design. First, when a bottom-discharge tube was lifted out of the slag pocket into the main cavity, the turbulent zone below the tube could stir the impurities accumulated in the slag pocket, causing them to enter the main cavity. Second, when the tube passed through the transition zone between the slag pocket and the mold bottom, the sudden reduction in cross-sectional area caused the liquid level to rise faster than the tube movement, leading to splashing at the slag pocket opening.
To solve the first issue, the tube was placed at the bottom of the slag pocket before the water valve was opened. The tube then rose together with the free surface, so that the impurities remained under calm conditions and were eventually retained by the upper baffle. Side-discharge tubes were even more effective for this purpose, since their flow disturbance did not act in the region below the tube, completely isolating the slag pocket once the tube had risen above it.
For the second issue, a modified trajectory was developed. The tube started at the slag pocket bottom, waited until the water level covered the outlet, then rose at a speed corresponding to the slag pocket cross-sectional area. As the liquid surface approached the slag pocket opening, the tube was moved upward rapidly to the mold bottom level and then waited for the liquid to cover its outlet again before resuming the normal speed schedule. This two-stage waiting strategy eliminated the splashing and kept the flow calm throughout the transition.
2.7 Simulation of Variable Cross-Section Filling
To emulate real castings with varying cross-sections, two different core models were inserted into the mold: one with a gradually changing cross-section and one with abrupt changes. The submerged tube speed was updated stepwise based on the horizontally sliced area. For small changes, the area was averaged over the height interval. Experiments confirmed that the robot could precisely follow the required velocity changes and that the tube maintained proper immersion depth in both cases. This validated the control methodology for complex geometries.
2.8 Water Simulation for the Multi-Way Valve
As a final validation, a real multi-way valve core was placed inside the transparent mold. The core had been fabricated by the PIRP rapid prototyping process. The total height of the core was divided into twelve zones. The horizontal cross-sectional area for each zone was calculated by intersecting the core with a plane of 0.3 mm thickness in the Magics software. The results are shown in Table 4.
| Zone | Height position (mm) | Core area (mm²) | Mold cavity area (mm²) | Rising speed (mm/s) |
|---|---|---|---|---|
| 1-1 | 10 | 28453 | 58453 | 8.9 |
| 1-2 | 15 | 28797 | 58797 | 8.3 |
| 1-3 | 20 | 30682 | 60682 | 7.2 |
| 2 | 25 | 33540 | 63540 | 6.6 |
| 3 | 41 | 40580 | 70580 | 5.4 |
| 4 | 51 | 38470 | 68470 | 5.7 |
| 5 | 63 | 41350 | 71350 | 5.3 |
| 6 | 68 | 41390 | 71390 | 5.3 |
| 7 | 83 | 36620 | 66620 | 6.0 |
| 8 | 87 | 42210 | 72210 | 5.2 |
| 9 | 92 | 45780 | 75780 | 4.8 |
| 10 | 97 | 43000 | 73000 | 5.1 |
| 11 | 115.5 | 45820 | 75820 | 4.8 |
| 12 | 130.5 | 41300 | 71300 | 5.3 |
The velocity curve derived from these areas was successfully implemented in the robot. During the entire water simulation, the submerged tube remained continuously below the free surface and followed the liquid level as intended. This confirmed the practical feasibility of applying submerged-gate casting to the multi-way valve production within a sand casting foundry environment.
3 Casting Design and Numerical Simulation
3.1 Casting and Core Model
The researched component is a multi-way valve used in hydraulic systems. It possesses multiple intersecting ports, thin sections, and a complex internal channel network. A monolithic sand core was designed to form all internal passages. Because of the complexity, traditional core-making methods could not guarantee dimensional consistency and strength. Instead, the core was manufactured by the PIRP rapid prototyping method as described in Section 4.
3.2 Gating System Design
A bottom-gating system was adopted to ensure calm filling and minimize the impact of metal flow on the brittle sand core. The gating system comprised a pouring cup, a sprue, a cross runner, and two ingates. The ratio of cross-sectional areas for the sprue, runner, and ingates was chosen as 1.5:2:1, with the ingate as the choke. The choke area was calculated using the following empirical formulas:
$$ A_{choke} = \frac{G}{\rho t \mu \sqrt{2g H_p}} \tag{6}$$
where \(G\) is the total weight of metal in the mold (20 kg), \(\rho\) is the density of liquid iron (7.3 g/cm³), \(t\) is the filling time (7.6 s), \(\mu\) is the flow coefficient (0.6), and \(H_p\) is the average static head (20 cm). The filling time was obtained from:
$$ t = S \sqrt[3]{G} \tag{7}$$
with \(S = 1.7\) for this type of casting. Substituting the values yields \(A_{choke} = 8.6\,\text{cm}^2\), and consequently \(A_{runner} = 12.9\,\text{cm}^2\), \(A_{sprue} = 17.2\,\text{cm}^2\).
Since the casting material was ductile iron, no external risers were used. Instead, the graphitic expansion during solidification was relied upon to compensate for shrinkage. The pouring temperature was controlled between 1360 °C and 1420 °C.
3.3 Numerical Simulation with JSCAST
The designed casting and gating system were converted into STL files and imported into the JSCAST simulation package. Direct finite difference method was used for solving the flow and thermal fields. The mold was defined as sand, and the material data corresponded to gray cast iron. The simulation computed the filling pattern, temperature distribution, and solidification sequence.
The filling simulation showed complete mold filling without any misrun or cold shut defects. The solidification simulation revealed that the last solidifying zones were located in the thick central regions of the valve body. Due to the self-feeding ability of ductile iron, no shrinkage porosity was predicted. The simulation thus confirmed the adequacy of the initial design.
4 Rapid Manufacturing of Sand Mold and Core
4.1 Mold Design and Machining
The sand mold was obtained by subtracting the casting and gating system geometries from a rectangular sand block. A minimum molding allowance of 40 mm was maintained around all surfaces. The mold was split into three parts: an upper cope, a middle drag, and a lower base. The upper cope contained the pouring cup and vent holes. The middle drag accommodated the main cavity and the core prints. The lower base allowed the machining of the slag pocket and the integrated choke from the bottom. This split was specifically designed for the CNC milling process used to cut the sand blocks.
The sand material was a resin-bonded silica sand block that had been pre-hardened. CNC machining was performed using a three-axis engraving machine. The upper and lower parts were machined from one side only, whereas the middle part required machining from both the top and the bottom. Care was taken to avoid thin unsupported sections and to maintain sharp edges at the parting lines.
4.2 PIRP Rapid Prototyping of the Core
The core was manufactured using the PIRP (Profile Invalidation Rapid Prototyping) method, a technique originally developed in our research group. The principle involves heating a layer of coated sand with a hot plate to cure it, then using a laser to scan the outer contour of the desired profile. The high-energy laser overheats and deactivates the resin along the contour, creating a weakened boundary line. Successive layers are built up until the entire block is formed. After post-processing, the waste sand is removed, leaving the solid core.
For the multi-way valve core, a CAD model was created and additional auxiliary cutting bodies were added to facilitate the removal of waste material after curing. The model was sliced into layers and converted into CLI format for the PIRP machine. The machine used was a PIRP-1500 with a maximum build volume sufficient for this core.
| Parameter | Value / Description |
|---|---|
| Material | Phenolic resin coated sand |
| Layer thickness | 0.3 mm |
| Laser power | 100 W (CO₂ laser) |
| Heating plate temperature | 200 °C |
| Build speed | Approx. 10 mm/h in height |
| Post-curing temperature | 180 °C for 2 h |
After the PIRP process, the core block was removed from the machine. Loose sand on the surface was carefully brushed off. The block was then placed in an oven and cured at 180 °C for 2 hours to increase its strength. After cooling, the waste sand was stripped away manually. The resulting monolithic core had a smooth surface and accurate dimensions. Additional support structures were not required because the core was fully surrounded by sand until the final stripping.
5 Pouring Experiments and Post-Processing
5.1 Coating Application
To improve the surface finish and prevent sand-metal reaction, a refractory coating was applied to all surfaces of the mold and core that would come into contact with molten metal. A zircon-based alcohol coating was chosen due to its rapid drying nature. The coating viscosity was controlled by measuring the Baume degree. For the upper and lower mold parts, brushing was used with a heavier coating (55–60 °Bé). For the middle mold part and the core, flow coating was applied at a lower viscosity (45–50 °Bé). The core was flow-coated thoroughly to ensure that the internal passage surfaces were completely covered. After coating, the solvent was ignited and burned off, leaving a dense refractory layer.
5.2 Venting and Assembly
Vent holes were drilled through the upper cope to allow the escape of gases generated during pouring. The mold parts were dried in an oven at 80 °C overnight to remove any residual moisture. After drying, the core was placed in the lower drag using locating prints. The upper cope was then lowered carefully, and the mold was assembled. To prevent run-out at the parting lines, a layer of clay was applied around the mold periphery, and weights were placed on top of the mold.
5.3 Pouring
The pouring experiment was conducted in a steel foundry to ensure high-quality molten iron. The material was HT250 gray iron, melted in an induction furnace. The melt was tapped into a preheated ladle, and the temperature was adjusted to approximately 1400 °C before pouring. The pouring was executed continuously and steadily until the metal rose to the top of the pouring cup. No bubbling or overflow was observed, indicating a well-vented mold.
5.4 Inspection and Results
After cooling to room temperature, the mold was broken out, and the casting was separated from the sand. The gating system was cut off, and the casting was sand-blasted to remove all adhering sand particles. The resulting casting showed a smooth external surface and fully defined ports.
Dimensional accuracy was assessed using a three-dimensional optical scanner. The scanned point cloud was aligned with the original CAD model using best-fit algorithms. The deviation map indicated that the majority of surface points were within a range from -2.0 mm to +0.5 mm, with the most concentrated deviations between -1.5 mm and -0.5 mm. The standard deviation of the point cloud was less than 0.7 mm, which is satisfactory for a sand-cast valve of this complexity. Table 6 summarizes the dimensional results.
| Statistical measure | Value (mm) |
|---|---|
| Minimum deviation | -2.3 |
| Maximum deviation | +0.8 |
| Average deviation | -0.9 |
| Standard deviation | 0.62 |
| Percentage within -1.5 to -0.5 | 68% |
| Percentage within -2.0 to +0.5 | 93% |
To evaluate the internal integrity, the casting was cut along a plane passing through its main channels. The cross-section revealed fully formed internal passages without any sand blockage or core shift. Subsequently, the casting was subjected to dye penetrant inspection. No cracks, cold shuts, or visible porosity were found. The absence of gas holes was attributed to the combination of the ventilated mold, the controlled pouring, and the presence of the slag pocket that absorbed the initial turbulent metal.
Conclusions
The following conclusions can be drawn from this work:
- The integration of industrial robot with submerged-gate casting was successfully demonstrated in water simulation experiments. The robot could precisely follow the required velocity profiles and maintain the submerged tube depth at the desired level.
- Water simulation experiments revealed that side-discharge tubes produced less disturbance in the bottom region and performed better in retaining impurities inside the slag pocket. The intermediate ladle, when operated in continuous mode, effectively removed slag and prevented gas entrainment.
- The trajectory optimization of the submerged tube, including the two-stage waiting near the slag pocket opening, suppressed splashing and ensured a calm filling process even when the cross-section changed abruptly.
- The casting design for the multi-way valve was verified by numerical simulation using JSCAST. Complete filling and controlled solidification were predicted and later confirmed by actual pouring.
- Rapid prototyping by PIRP enabled the production of a monolithic sand core for the multi-way valve, avoiding the inaccuracies and strength issues of segmented cores. The resulting casting had smooth internal channels and met the dimensional requirements.
- The final multi-way valve casting exhibited no observable defects, and its dimensions deviated mostly within -1.5 mm to -0.5 mm. The proposed process is therefore a viable alternative for producing complex integral pump valves in a modern sand casting foundry.
Acknowledgments
The author expresses sincere gratitude to all the team members and laboratory colleagues who contributed to this research through valuable advice, experimental assistance, and continuous support during the entire project.