In our foundry, the axle box body for metro vehicles is a thin-walled box-type steel casting and an important load-bearing component in the bogie assembly. As operating speed and axle load increase, the demand for higher integrity in such castings has become severe. During early production, we repeatedly encountered casting defects such as shrinkage porosity, sand holes, cracks, core tilting during mold closing and pouring, black skin after machining, and insufficient machining allowance. These casting defects are closely related to the original casting process, so we systematically analyzed the process and implemented improvements. The following sections describe the component requirements, the original process, the root causes of the casting defects, the improved process design, and the production validation results.
The axle box body has a complex geometry with main wall thickness between 20 and 30 mm. The as-cast mass is 95 kg. The material is ZG25MnCrNiMo, a C-grade steel. We use CO2-hardened sodium silicate sand for molding and cores. Two castings are produced per mold. Melting is carried out in a 10 t electric arc furnace. Bottom pouring is applied, and the pouring temperature is controlled between 1560 and 1580 °C. The linear shrinkage is taken as 2%. Table 1 summarizes the key production parameters.
| Parameter | Value |
|---|---|
| Material | ZG25MnCrNiMo (C-grade steel) |
| Main wall thickness | 20–30 mm |
| As-cast mass | 95 kg |
| Linear shrinkage | 2% |
| Molding and core sand | CO2-hardened sodium silicate sand |
| Molding arrangement | Two castings per mold |
| Melting furnace | 10 t electric arc furnace |
| Pouring method | Bottom pouring |
| Pouring temperature | 1560–1580 °C |
| Surface inspection | Magnetic particle testing on the whole surface |
| Internal inspection | Ultrasonic testing by batch |
| Acceptance criteria | No shrinkage porosity or cracks on all machined surfaces and main load-bearing areas |
The pattern dimensions must compensate for the linear shrinkage of the steel. For any casting dimension \(L_{casting}\), the corresponding pattern dimension \(L_{pattern}\) is calculated as:
$$L_{pattern} = L_{casting} \left(1 + s\right)$$
where \(s = 0.02\) is the linear shrinkage. This 2% allowance is applied to all critical dimensions, especially the large and small axle housings, the spring plate, and the connecting ribs. Because the casting has thin walls and several thermal nodes, the shrinkage allowance alone cannot prevent casting defects. The thermal modulus of each region must also be controlled to achieve directional solidification.
The modulus \(M\) of a casting region is defined as the ratio of its volume \(V\) to its heat-dissipating surface area \(A\):
$$M = \frac{V}{A}$$
According to Chvorinov’s rule, the solidification time \(t_s\) of a region is proportional to the square of its modulus:
$$t_s = B M^2$$
where \(B\) is a constant depending on the mold material and the pouring temperature. To avoid shrinkage porosity, the modulus of the feeder or riser \(M_r\) must be larger than the modulus of the casting region \(M_c\):
$$M_r \ge 1.2 M_c$$
In the original process, we did not sufficiently account for the modulus differences between the large axle housing, the small axle housing, the ribs, and the spring plate. As a result, several casting defects appeared during batch production. The next section describes the original process and the observed problems.
Original casting process and observed defects. The original process used a parting line along the middle of the axle housing. A腰形随型补贴, which I will describe in English as a contoured padding, was placed inside the large axle housing to promote directional solidification and feeding. The small axle housing and the connecting rib formed a thermal node. Because its modulus and radius were relatively small, a semicircular blind riser was placed near the top connection to ensure internal quality. A 16 mm thick chill plate was installed on the lower mold of the large axle housing and on the spring plate face to enhance chilling. The ingate was located at the central rib. This original process generally ensured the quality of machined surfaces and main load-bearing areas, but in batch production, casting defects mainly occurred in the large axle housing, the ingate, and the spring plate. In addition, sand shedding and non-uniform machining allowance were observed.
Table 2 lists the main casting defects and their locations in the original process.
| Casting defect | Typical location | Consequence |
|---|---|---|
| Sand shedding | Connecting rib between large and small axle housings; ear area of the hung core | Sand inclusions, scrap in small batches |
| Sand holes | Inside the large axle housing | Repair welding after rough machining, increased cost |
| Shrinkage porosity | Inside the large axle housing; ingate area | Reduced load-bearing capacity, repair welding |
| Cracks | Ingate ribs | Gouging and repair welding |
| Core tilting | Spring plate core during mold closing and pouring | Uneven machining allowance |
| Black skin | Machined surfaces | Insufficient machining allowance, additional cleaning |
| Insufficient machining allowance | Spring plate face | Repair welding or scrap |
The presence of these casting defects indicated that the original gating and feeding system, core fixing method, and mold compaction control were not robust enough. We therefore carried out a detailed root cause analysis.
Root cause analysis of casting defects. We analyzed each casting defect category separately. For sand shedding, the possible causes included low sand strength, poor bonding, non-uniform mixing, insufficient mold bars or box bars, lack of hooks or nails in the upper mold or hung sand areas, insufficient ramming, uneven compaction, excessive mold cavity depth, rough handling during mold closing, and vibration during alignment or clamping. In our case, the ear area of the axle housing used a hung core that was glued and hung during molding. During transportation and lifting, collisions occurred, and the core paste did not always reach the required strength. This led to sand shedding. In the connecting rib cavity between the large and small axle housings, the height was 50 mm and the depth was relatively large. After compaction on the jolt-squeeze machine, the sand strength in this area was insufficient. During pouring, the molten steel scoured the sand and caused sand shedding. These sand particles were trapped in the steel and could not be removed, producing sand inclusion casting defects.
For the large axle housing, the main casting defects were sand holes and shrinkage porosity. Sand holes were caused by careless cleaning of the mold cavity. Because the inner wall of the cylindrical housing is a machined surface, most sand holes could be removed by machining if the machining allowance was sufficient. We therefore decided to increase the machining allowance in this region. Shrinkage porosity, however, is a more serious internal casting defect. It occurs when the casting does not receive sufficient liquid metal feeding during solidification, so directional solidification cannot be achieved. The large cylindrical housing had a wall thickness of 25 mm before machining. The end zone of feeding was about 65 mm, and the riser zone was about 50 mm. The vertical feeding distance was therefore about 115 mm. The axle housing height was 245 mm. To achieve full directional solidification and feeding, the required padding depth was about 130 mm. The distance between the two insulating risers was 120 mm, and the feeding distance between the two risers was about 100 mm. Therefore, the feeding distance between the risers was insufficient. Because the pouring temperature was difficult to control precisely, shrinkage porosity occurred. The original two oval insulating risers could not provide enough feeding. We considered replacing them with three circular insulating exothermic risers, reducing the riser spacing to 90 mm. Theoretically, this would provide a better feeding effect without reducing the process yield.
For the spring plate, the insufficient machining allowance was caused by the core design. The spring plate core was a circular disc core with a lower core print. There was an error between the core box print position and the core seat draft. The wooden pattern deformed during service, so the gap between the core print and the core seat became too large. During mold closing, the core tilted forward. After pouring, the machining allowance on the spring plate face was uneven, and the upper half had insufficient allowance, requiring repair welding. The lower core print did not provide good positioning. We decided to change the lower core print to two side core prints. The core print size was reduced, and the fit between the core print and the core seat was improved, giving better positioning than the original design.
For the ingate, shrinkage porosity and cracks occurred. The original ingate size was 40/35 mm × H, where H was the section height of the ingate. The height transitioned from the runner to the casting rib, and the height at the rib was 15 mm. The rib thickness at the ingate was 7.5 mm. This thickness difference caused the rib to solidify before the ingate. The shrinkage stress produced a tensile stress at the contact, leading to cracks during solidification. To solve the shrinkage porosity, the ingate and the casting rib thickness should be made more uniform. An open gating system with a flat ingate would be required, but steel has a high melting point and poor fluidity. If the ingate were changed to a flat shape, the pouring temperature might not be controlled well, and misruns could occur. We therefore considered relocating the ingate to the left slot. The slot is a machined surface with a thickness of 25 mm. Locating the ingate there is more reasonable. Even if cracks or other casting defects occur, they can be removed by machining.
Table 3 summarizes the root causes and the corresponding improvement measures for each casting defect.
| Casting defect | Root cause | Improvement measure |
|---|---|---|
| Sand shedding at rib | Insufficient sand strength in deep cavity; scouring by molten steel | Use welding rod folded into a core bone to reinforce the core; improve compaction |
| Sand shedding at ear | Weak glued hung core; collision during handling | Reserve a hole in the core; insert nails through the hole after hanging |
| Sand holes in large axle housing | Incomplete mold cavity cleaning | Increase machining allowance; strengthen cavity cleaning |
| Shrinkage porosity in large axle housing | Insufficient feeding distance between risers | Replace two oval risers with three circular insulating exothermic risers; reduce spacing to 90 mm |
| Insufficient machining allowance on spring plate | Poor lower core print positioning; pattern deformation | Change to two side core prints; reduce core print size; improve fit |
| Shrinkage porosity and cracks at ingate | Thickness difference between ingate and rib; premature rib solidification | Relocate ingate to left slot (25 mm thick machined surface) |
These improvements were incorporated into a new casting process. Before production, we used numerical simulation to verify the filling and solidification behavior. The simulation results agreed with our analysis. The following sections describe the new process design and the production control measures in detail.
Improved process design and feeding system. The new process retained the overall parting line along the middle of the axle housing, but we changed the feeding system significantly. For the large axle housing, we replaced the two oval insulating risers with three circular insulating exothermic risers. The new riser spacing was 90 mm. The riser modulus was selected according to the casting modulus. The required riser modulus is:
$$M_r = 1.2 M_c$$
For the large axle housing, the casting modulus \(M_c\) was calculated from the volume and surface area of the cylindrical section. If the outer diameter is \(D_o\), the inner diameter is \(D_i\), and the height is \(H\), then the volume is:
$$V_c = \frac{\pi}{4} \left(D_o^2 – D_i^2\right) H$$
The heat-dissipating surface area includes the outer surface, the inner surface, and the end faces. For a cylindrical shell, the area is:
$$A_c = \pi D_o H + \pi D_i H + \frac{\pi}{2} \left(D_o^2 – D_i^2\right)$$
Thus, the modulus is:
$$M_c = \frac{V_c}{A_c} = \frac{\frac{\pi}{4} \left(D_o^2 – D_i^2\right) H}{\pi D_o H + \pi D_i H + \frac{\pi}{2} \left(D_o^2 – D_i^2\right)}$$
Using the actual dimensions, we obtained the modulus and then selected a riser with \(M_r \ge 1.2 M_c\). The new three-riser arrangement reduced the maximum feeding distance. The feeding distance \(L_f\) between risers can be estimated as:
$$L_f = L_r + L_e$$
where \(L_r\) is the riser feeding zone and \(L_e\) is the end zone. For steel castings, the end zone effect depends on the modulus and the temperature gradient. With three risers, the distance between adjacent risers was 90 mm, which was within the effective feeding distance. This helped eliminate shrinkage porosity casting defects in the large axle housing.
For the small axle housing and the connecting rib, we kept the semicircular blind riser near the top connection, but we adjusted its size and position to ensure that the thermal node was properly fed. The blind riser modulus was also checked against the local casting modulus. For the spring plate, we changed the core print design. The original lower core print was replaced by two side core prints. The new core print dimensions were smaller, and the fit between the core print and the core seat was improved. This reduced the gap and prevented core tilting during mold closing and pouring. The spring plate machining allowance became uniform, and the black skin casting defect was eliminated.
For the ingate, we relocated it to the left slot. The slot is a machined surface with a thickness of 25 mm. The ingate cross-section was adjusted to match the local modulus. The new ingate area \(A_g\) was calculated from the pouring time and the allowable pouring rate. The pouring time \(t_p\) is related to the casting mass \(m\) and the pouring rate \(Q\):
$$t_p = \frac{m}{Q}$$
The ingate area is then:
$$A_g = \frac{m}{\rho v t_p}$$
where \(\rho\) is the density of liquid steel and \(v\) is the allowable flow velocity. For steel castings, the flow velocity in the ingate is usually limited to avoid mold erosion and turbulence. We also checked the Reynolds number to ensure smooth filling:
$$Re = \frac{\rho v D_h}{\mu}$$
where \(D_h\) is the hydraulic diameter and \(\mu\) is the dynamic viscosity. The new ingate location and size reduced the thickness difference between the ingate and the casting rib, lowering the risk of shrinkage porosity and cracks. Because the ingate is now on a machined surface, any minor casting defects can be removed during machining.
The improved process also included a revised chill design. The 16 mm thick chill plates on the lower mold of the large axle housing and the spring plate face were retained, but their placement and thickness were checked against the local modulus. Chills increase the local cooling rate and reduce the modulus, promoting directional solidification toward the risers. The chill thickness \(t_c\) should be sufficient to absorb heat without melting completely. A common rule is:
$$t_c \ge 0.5 M_c$$
We verified this rule for the large axle housing and the spring plate. The chill plates were placed flat and without sand backing, as required by the process card.
Table 4 compares the original and improved feeding and gating parameters.
| Parameter | Original process | Improved process |
|---|---|---|
| Large axle housing risers | 2 oval insulating risers | 3 circular insulating exothermic risers |
| Riser spacing | 120 mm | 90 mm |
| Feeding distance between risers | 100 mm | Designed to be within effective feeding distance |
| Small axle housing riser | Semicircular blind riser | Semicircular blind riser, size and position adjusted |
| Spring plate core print | Lower core print | Two side core prints |
| Spring plate machining allowance | Uneven, sometimes insufficient | Uniform, 7 mm |
| Ingate location | Central rib | Left slot (machined surface, 25 mm thick) |
| Ingate shape | 40/35 mm × H, H = 15 mm at rib | Adjusted cross-section, more uniform thickness |
| Chill plates | 16 mm thick on lower mold and spring plate | Retained, placement and thickness verified |
The new process was simulated before actual production. The simulation predicted the filling pattern, temperature distribution, and solidification sequence. The results showed that the new riser arrangement provided sufficient feeding to the large axle housing, and the relocated ingate reduced the risk of shrinkage porosity and cracks. The spring plate core print change improved the positioning accuracy and reduced the tendency for core tilting. The simulation also indicated that the new process would reduce sand-related casting defects if the mold and core handling were properly controlled.
Production control for the improved process. After the simulation, we made the new pattern and core boxes and conducted trial production. We strengthened control over the molding and core-making operations. The following measures were implemented:
1. The process card specified the placement of chills, pouring cups, insulating riser sleeves, and other consumables. The chills were not allowed to be tilted or backed with sand. The sand thickness between the core bone and the pattern wall was controlled between 15 and 50 mm, or beyond this range if specified by the process card.
2. The mold and core were stripped smoothly. The compaction of all mold and core parts was uniform. The compactability was controlled at 45 units or higher unless otherwise specified by the process card.
3. The mold and core surfaces were coated with refractory coating. The coating thickness was appropriate, without accumulation, flow marks, or unevenness.
4. Before hanging the core, the mold cavity was cleaned thoroughly. The loose sand around the chills and the vent holes was scraped flat and blown clean. The vent holes were kept open. The oxygen injection holes in the cavity were plugged with molding sand and smoothed.
5. The core surface was cleaned. Fins, burrs, and sharp edges were removed, and sharp corners were rounded.
6. During mold closing, the core was positioned carefully. The side core prints of the spring plate core were checked for proper fit. The mold was closed without vibration or impact that could dislodge sand.
7. The pouring temperature was maintained between 1560 and 1580 °C. The pouring rate was controlled to avoid turbulence and mold erosion. The pouring time was recorded for each mold.
8. After pouring, the castings were allowed to cool in the mold. The shakeout time was controlled to avoid hot cracking. The castings were then cleaned, shot-blasted, and inspected.
Table 5 lists the key production control parameters for the improved process.
| Control item | Specification |
|---|---|
| Sand thickness between core bone and pattern wall | 15–50 mm |
| Compactability | ≥ 45 units |
| Coating thickness | Uniform, no accumulation or flow marks |
| Mold cavity cleaning | No loose sand; vent holes open; oxygen holes plugged |
| Core surface cleaning | No fins, burrs, or sharp edges |
| Pouring temperature | 1560–1580 °C |
| Pouring rate | Controlled to avoid turbulence |
| Chill placement | Flat, no sand backing, per process card |
| Riser sleeve placement | Per process card, no tilt |
| Mold closing | No vibration or impact |
We also paid attention to the sand core reinforcement. For the connecting rib between the large and small axle housings, we used a welding rod folded into a core bone. The core bone was placed in the deep cavity before ramming. The sand around the core bone was compacted uniformly. This increased the core strength and prevented sand shedding during pouring. The core bone also helped to anchor the sand in the deep cavity, reducing the risk of sand inclusion casting defects.
For the ear area of the hung core, we reserved a hole in the core during core making. After the core was hung, we inserted nails through the reserved hole to reinforce the core. This simple measure prevented the core from shifting or falling during handling and pouring. It also reduced the sand shedding casting defect that had previously occurred in this area.
The formula for the compressive strength of the core can be expressed as:
$$\sigma_c = \frac{F}{A}$$
where \(F\) is the applied force and \(A\) is the cross-sectional area. By adding the core bone and nails, we increased the effective cross-sectional area and the load-bearing capacity of the core, thereby improving its resistance to sand shedding.
For the large axle housing, we increased the machining allowance on the inner wall. The additional allowance allowed any small sand holes to be removed during machining. We also strengthened the cleaning of the mold cavity. The mold was cleaned with compressed air and a vacuum before closing. Any loose sand or debris was removed. This reduced the sand hole casting defect. The shrinkage porosity casting defect was addressed by the new three-riser arrangement. The feeding distance was reduced, and the riser modulus was sufficient to feed the large axle housing. The simulation showed that the new riser design achieved directional solidification from the far end toward the risers.
The Niyama criterion was used to predict shrinkage porosity in the simulation. The criterion is:
$$G / \sqrt{R} \le C$$
where \(G\) is the temperature gradient, \(R\) is the cooling rate, and \(C\) is a critical value. Regions with \(G / \sqrt{R}\) below the critical value are prone to shrinkage porosity. The simulation results showed that the new riser design kept the large axle housing above the critical value, indicating a low risk of shrinkage porosity casting defects.
For the spring plate, the new side core prints improved the positioning accuracy. The core print dimensions were reduced, and the fit between the core print and the core seat was tighter. The gap between the core print and the core seat was controlled to less than 0.2 mm. This prevented the core from tilting during mold closing. The spring plate machining allowance was measured after casting and was found to be 7 mm, uniform around the entire face. The black skin casting defect was eliminated because the machining allowance was sufficient to remove the surface layer.
For the ingate, the relocation to the left slot eliminated the thickness difference between the ingate and the rib. The new ingate was located on a 25 mm thick machined surface. The ingate cross-section was designed to provide a smooth flow of steel without excessive turbulence. The pouring time was calculated as:
$$t_p = \frac{m}{Q}$$
where \(m = 95\) kg and \(Q\) is the pouring rate. The ingate area was then determined from the flow rate:
$$A_g = \frac{Q}{v}$$
where \(v\) is the allowable flow velocity. The new ingate design reduced the risk of shrinkage porosity and cracks. After casting, the ingate area was inspected by magnetic particle testing. No cracks or shrinkage porosity casting defects were found.
Production validation and results. Using the improved process, we produced 10 sample castings. After cleaning and grinding, the castings were visually inspected. No sand shedding casting defects were found. The dimensions were checked by layout inspection. All main inspection dimensions were within tolerance. The spring plate machining allowance was 7 mm and uniform in thickness. The castings were then cleaned and subjected to ultrasonic testing on the axle housings. No shrinkage porosity casting defects were found in any of the 10 castings. External inclusions and sand holes were significantly reduced compared with the original process. The ingate area was inspected by magnetic particle testing. No cracks or surface casting defects were found. The magnetic particle testing passed. All 10 castings passed the first inspection and testing. This validated the root cause analysis and the feasibility of the improvement measures.
Table 6 summarizes the validation results for the 10 sample castings.
| Inspection item | Requirement | Result for 10 samples |
|---|---|---|
| Visual inspection for sand shedding | No sand shedding | No sand shedding casting defects found |
| Layout inspection | Main dimensions within tolerance | All main dimensions good |
| Spring plate machining allowance | Uniform, sufficient | 7 mm, uniform |
| Ultrasonic testing of axle housings | No shrinkage porosity | No shrinkage porosity casting defects found |
| External inclusions and sand holes | Minimized | Significantly reduced |
| Magnetic particle testing of ingate area | No cracks or surface defects | Passed, no cracks |
| Overall acceptance | All castings qualified | 10 out of 10 qualified |
The improved process not only reduced casting defects but also improved the process yield and reduced repair welding costs. The new riser arrangement with three circular insulating exothermic risers provided better feeding without reducing the process yield. The side core prints on the spring plate core improved the dimensional accuracy and reduced the machining allowance variation. The relocated ingate eliminated the crack casting defect at the ingate. The core reinforcement measures eliminated sand shedding casting defects. The increased machining allowance in the large axle housing allowed any minor sand holes to be removed during machining. The overall result was a robust casting process that consistently produced qualified axle box steel castings.
We also analyzed the economic benefits of the improved process. The reduction in repair welding and scrap directly lowered production costs. The improved process yield reduced the amount of molten steel required per casting. The reduction in casting defects also reduced the time spent on inspection and rework. Table 7 shows a qualitative comparison of the original and improved processes in terms of casting defects and production performance.
| Performance indicator | Original process | Improved process |
|---|---|---|
| Sand shedding casting defects | Frequent | Eliminated |
| Sand hole casting defects | Occasional | Significantly reduced |
| Shrinkage porosity casting defects | Occasional in large axle housing | Eliminated in validation samples |
| Crack casting defects at ingate | Occasional | Eliminated |
| Core tilting casting defects | Occasional | Eliminated |
| Machining allowance uniformity | Poor | Good |
| Repair welding cost | High | Low |
| Process yield | Acceptable | Improved |
| First-pass inspection pass rate | Variable | 100% for 10 samples |
In conclusion, the casting defects that occurred in the axle box steel castings were caused by a combination of insufficient feeding distance, weak core strength, poor core positioning, and ingate design issues. By systematically analyzing the root causes and implementing targeted improvements, we eliminated the sand shedding, shrinkage porosity, sand hole, crack, and core tilting casting defects. The improved process used three circular insulating exothermic risers, side core prints for the spring plate, a relocated ingate on the left slot, and reinforced cores with core bones and nails. The production validation with 10 sample castings confirmed that the improved process is capable of producing qualified axle box steel castings with no casting defects on machined surfaces and main load-bearing areas. The lessons learned can be applied to other thin-walled steel castings with similar thermal nodes and core handling challenges.
For future work, we plan to continue monitoring the production process and collect statistical data on casting defects. We will also refine the simulation model to further optimize the riser and chill design. The goal is to achieve zero casting defects and to reduce the need for repair welding. We will also explore the use of alternative molding materials and coatings to further improve sand strength and surface quality. The experience gained from this project will be valuable for improving the casting process of other bogie components.
Table 8 provides a summary of the key formulas used in the analysis and design of the improved process.
| Purpose | Formula | Description |
|---|---|---|
| Pattern dimension | \(L_{pattern} = L_{casting} (1 + s)\) | Linear shrinkage compensation |
| Modulus | \(M = V / A\) | Volume-to-surface-area ratio |
| Chvorinov’s rule | \(t_s = B M^2\) | Solidification time |
| Riser modulus | \(M_r \ge 1.2 M_c\) | Feeding requirement |
| Feeding distance | \(L_f = L_r + L_e\) | Riser feeding zone plus end zone |
| Pouring time | \(t_p = m / Q\) | Casting mass divided by pouring rate |
| Ingate area | \(A_g = Q / v\) | Flow rate divided by allowable velocity |
| Reynolds number | \(Re = \rho v D_h / \mu\) | Flow regime check |
| Niyama criterion | \(G / \sqrt{R} \le C\) | Shrinkage porosity prediction |
| Compressive strength | \(\sigma_c = F / A\) | Core strength |
| Chill thickness | \(t_c \ge 0.5 M_c\) | Chill design rule |
The improved casting process has been implemented in production. We continue to inspect each batch by magnetic particle testing and ultrasonic testing. The casting defects that previously occurred have been eliminated or significantly reduced. The axle box steel castings now meet the stringent requirements for metro vehicle bogie components. This project demonstrates that a systematic approach to casting defect analysis and process improvement can achieve reliable, high-quality steel castings even for thin-walled, complex components.