My research focuses on the formability and defect control of box-type steel castings, particularly those used in locomotive components. The railway industry demands high reliability, and castings such as axle box housings and brackets must be free from internal flaws. Throughout this work, I systematically analyzed sand foundry defects using numerical simulation and experimental validation. The goal was to predict, prevent, and remediate defects such as shrinkage porosity, hot tears, and gas porosity. By integrating casting process simulation with practical foundry trials, I developed optimized gating and risering systems that significantly improved casting soundness.
1. Foundry Process Fundamentals
The production of steel castings involves a sequence of operations, from raw material selection to final inspection. In my foundry, the electric arc furnace (EAF) is used for melting. The steelmaking process is divided into five stages: raw material collection, preparation, melting, oxidation, and reduction. Table 1 summarizes the key parameters and objectives for each stage.
| Stage | Key Operations | Objectives | Typical Parameters |
|---|---|---|---|
| Raw materials | Select clean scrap steel | Ensure low S, P, Cu, Sn, Pb | Scrap size ≤ 450 mm |
| Preparation | Calculate charge composition | Achieve target C, Mn, Si | C: specification +0.05–0.10% |
| Melting | Power input, slag formation | Rapid melting, early dephosphorization | Power: max after bore-in |
| Oxidation | Add iron ore, oxygen blowing | Remove P, reduce gases and inclusions | Decarburization rate ≥ 0.02%/h |
| Reduction | Slag deoxidation, alloy addition | Deoxidize, desulfurize, adjust composition | White slag time ≥ 20 min |
The oxidation stage is critical for reducing hydrogen and nitrogen, which otherwise cause sand foundry defects such as pinholes. The carbon-oxygen reaction drives melt stirring:
$$ [\text{C}] + [\text{O}] \rightarrow \text{CO(g)} \quad \Delta G^\circ = -22.4 – 21.0T \, \text{kJ/mol} $$
Complete deoxidation during tapping uses aluminum shot placed on the tap spout. After tapping, the steel is held in the ladle for at least 2 minutes to allow inclusions to float.
2. Moulding Materials and Ester-Cured Sodium Silicate Sand
The moulding sand system is the direct origin of many sand foundry defects. Traditional CO₂ sodium silicate sand has poor collapsibility, making casting cleaning difficult. I adopted an ester-cured sodium silicate sand process. This system offers high strength, good dimensional accuracy, and excellent collapsibility, significantly reducing the risk of hot tearing. The sand mixture used for cores is given in Table 2.
| Component | Specification | Proportion (%) |
|---|---|---|
| Silica sand (rounded) | AFS 50–60, moisture < 0.2% | 100 |
| Modified sodium silicate | M 2O·nSiO2, n=2.2–2.6 | 2.8–3.5 |
| Ester–based hardener | Glycerol triacetate | 0.4–0.7 |
The hardening reaction proceeds via ester hydrolysis under alkaline conditions:
$$ \text{RCOOR’} + \text{OH}^- \rightarrow \text{RCOO}^- + \text{R’OH} $$
The sodium silicate polymerizes, forming a silica gel network that bonds sand grains. I optimized the mix parameters to achieve a tensile strength of 1.8–2.5 MPa after 24 h, while maintaining a residual strength below 0.4 MPa to ensure easy shakeout. This approach reduces several sand foundry defects, including thermal cracking caused by poor mould collapsibility, and improves surface finish by reducing sand expansion defects.
3. Classification and Mechanisms of Sand Foundry Defects
I systematically studied the main defects encountered in box-type steel castings. Table 3 lists the defect types, their typical features, and primary causes.
| Defect | Visual Features | Root Cause | Prevention |
|---|---|---|---|
| Gas porosity | Smooth-walled spherical holes | High gas pressure from mould or dissolved gas | Reduce moisture, enhance venting |
| Sand Inclusion | Irregular cavity with sand particles | Mould erosion, loose sand | Increase mould surface hardness |
| Slag inclusion | Irregular dross, often at top surface | Improper deoxidation, poor gating design | Use slag traps, bottom pouring |
| Shrinkage cavity/porosity | Rough, shaggy internal void | Insufficient feeding, hot spots | Optimize risers, chill, feeding aids |
| Hot tearing | Irregular oxidized crack | Solidification contraction restricted | Improve mould collapsibility, add fillets |
| Cold shut | Rounded-edge seam, incomplete fusion | Low pouring temperature, turbulence | Increase pouring temperature, improve gating |
For shrinkage prediction, I used the Niyama criterion and the f s (solid fraction) feeding criterion. The Niyama criterion is defined as:
$$ N = \frac{G}{\sqrt{\dot{T}}} $$
where G is the local temperature gradient (K/mm) and dot{T} is the cooling rate (K/s). When N is below a critical value (0.7–1.0 for steel castings), microporosity is likely. The pressure-difference criterion for macro‑shrinkage uses the pressure drop ΔP in the liquid–solid region:
$$ \Delta P = \frac{\mu f_L \beta V_{L}}{k} $$
where μ is the liquid viscosity, f L the liquid fraction, β the solidification shrinkage, V a volume parameter, and k the permeability.
4. Case Study 1: Axle Box Housing – Hot Tear and Shrinkage
The axle box housing is a half-open cylindrical steel casting (grade ZG230-450). It had stringent quality requirements because it experiences alternating loads during service. Table 4 shows the casting characteristics and original process.
| Parameter | Value |
|---|---|
| Material | ZG230-450 |
| Gross weight | ≈ 220 kg |
| Moulding process | Ester-cured sodium silicate sand |
| Gating system | Two pouring ladles, side injection at flange |
| Risers | Two open risers at flanges, four blind risers at bolt bosses |
Figure 1 shows a typical engine cylinder block, which has similar box-type geometry and feeding challenges.
The initial design placed ingates directly at the flange–bolt boss junction, creating an isolated hot spot. After machining, cracks appeared near the root of the flange in nearly half of the first production batch. Nondestructive inspection showed large shrinkage cavities beneath those cracks. The mechanism involved two issues: first, the ingate region became a thick section with slow solidification, which the adjacent risers could not feed because the feeding path was blocked; second, the two large end risers restrained contraction, generating high tensile stress at the fillet root. These sand foundry defects were caused by improper gating and riser placement, not by poor steel quality.
I used ProCAST finite element stress analysis to simulate the evolution of stress during solidification. The von Mises stress at the critical fillet was as high as:
$$ \sigma_{eq} = \sqrt{\frac{1}{2}\left[(\sigma_{1}-\sigma_{2})^2 + (\sigma_{2}-\sigma_{3})^2 + (\sigma_{3}-\sigma_{1})^2 \right]} \simeq 350 \, \text{MPa} $$
which exceeded the zero-ductility strength at that temperature. The improved process included three changes:
- Moved the ingates below the blind risers, eliminating the hot spot near the flange root.
- Added riser pads to the open end risers, enlarging the feeding distance and improving the feeding path.
- Added sacrificial stress-relief blocks at the flange root fillets. These blocks shifted the stress concentration away from the critical zone.
After the changes, the simulated residual stress at the fillet dropped below 150 MPa. Actual dissection of the modified castings showed no shrinkage porosity and no cracks. Table 5 compares the original and improved processes.
| Feature | Original | Improved |
|---|---|---|
| Ingate location | At flange–boss junction | Below blind risers |
| Riser geometry | Open riser without pad | Open riser with pad, extended neck |
| Stress relief | None | Transition block at root |
| Result | Cracks and shrinkage | Sound casting, no defects |
5. Case Study 2: Bracket – Shrinkage Prediction and Riser Redesign
The bracket is a key structural component of a high-power locomotive frame. It is a flat plate–like structure with a sudden thick circular boss (∅160 mm). The material is ZG25MnNi (equivalent to ASTM LCC). The casting weight is 180 kg. Because of the thick boss, internal shrinkage was likely. My initial process used open risers with dimensions ∅160 × 200 mm and ∅140 × 180 mm. I performed solidification simulations using ProCAST. Figure 2 shows the temperature field at 10 min after pouring.
The simulation results showed that the open risers solidified faster than the thick boss. Hence no feeding occurred. The Niyama criterion predicted a large axial shrinkage region in the boss. The cause was clear: open risers had a large surface area, losing heat too quickly. Because the product structure limited the maximum riser diameter, I switched to insulating riser sleeves. Two designs were simulated:
- Design A: insulating risers ∅160 × 160 mm and ∅140 × 140 mm.
- Design B: insulating risers ∅180 × 180 mm and ∅160 × 160 mm.
Table 6 lists the key simulation parameters for the bracket.
| Parameter | Value |
|---|---|
| Pouring temperature | 1560 °C |
| Pouring time | 18 s |
| Mould material | Ester-cured sodium silicate sand |
| Riser sleeve type | Insulating (ceramic fiber based) |
| Interfacial heat transfer coefficient | 300 W/(m²·K) at metal–sand |
For Design A, the solidification time of the riser was still slightly shorter than that of the boss. The feeding distance was insufficient, leading to residual porosity in the boss as shown by the fs contour. I further increased the riser diameters. In Design B, the temperature of the riser remained significantly higher than that of the boss throughout solidification. The feeding channel stayed open, and the Niyama values at the boss centerline were all above 1.0. The predicted shrinkage cavity was confined entirely within the riser. Table 7 compares the simulation results.
| Design | Riser Dimension | Boss Temperature (relative) | Niyama (centerline) | Shrinkage in Boss |
|---|---|---|---|---|
| Open riser (original) | ∅160 × 200 | Higher than riser | < 0.5 | Large cavity |
| Insulating A | ∅160 × 160 | Riser slightly hotter | 0.6–0.8 | Minor porosity |
| Insulating B | ∅180 × 180 | Riser much hotter | > 1.0 | None |
To quantify the feeding efficiency, I calculated the riser modulus M using Chvorinov’s rule:
$$ t_s = C \left(\frac{V}{A}\right)^2 = C M^2 $$
where t s is the solidification time, V is the cooled volume, and A is the cooling surface area. For a cylindrical riser with diameter D and height H, the modulus is:
$$ M = \frac{V}{A} = \frac{\pi D^2 H /4}{\pi D^2 /4 + \pi D H} = \frac{D H}{D + 4H} $$
For Design B (D=180 mm, H=180 mm), M = 36 mm. The boss modulus was about 30 mm. Thus the riser had a 20% modular advantage, ensuring adequate feeding. Figure 3 is the production bracket section after final validation.
After implementing Design B in the foundry, two trial brackets were cast and sectioned. The machined section showed no internal defects. The measured dimensions and the fracture test results confirmed the simulation prediction. Thus, the optimization workflow for sand foundry defects in box-type castings was validated.
6. General Feeding System Design Equations
To systematically reduce sand foundry defects related to feeding, I used the geometric-feeding criterion. For a feed path from riser to hot spot, the critical feeding distance L f depends on the local temperature gradient G:
$$ L_f = C \cdot M_{hot} \cdot e^{\alpha (T_l – T_s)} $$
where C is a geometric constant, M hot the hot spot modulus, α the alloy shrinkage coefficient, and T l, T s the liquidus and solidus temperatures. In steel castings, the riser must have a modulus at least 1.2 times that of the hot spot:
$$ M_{riser} \geq 1.2 \, M_{hot} $$
The total volume feeding equation:
$$ V_{riser} \geq \frac{\beta V_{feeding}}{\eta} $$
with β = total volumetric shrinkage (liquid + solidification, about 5% for steel), V feeding the volume to be fed, and η the riser efficiency. For an open riser, η ≈ 0.14; for an insulating riser, η ≈ 0.35. This explains why Design B, even with a smaller volume than the original, worked better: the insulating sleeve increased the effective efficiency by more than twice, improving feeding efficacy and reducing sand foundry defects.
7. Role of Simulation in Defect Prediction
I used ProCAST with the finite element method to simulate, at each time step, the temperature, velocity, and solid fraction. One of the most useful quantities for predicting macro-shrinkage is the criterion proposed by Niyama:
$$ N = \frac{G}{\sqrt{\dot{T}}} = \frac{|\nabla T|}{\sqrt{\frac{\partial T}{\partial t}}} \times 10^3 \, (\mathrm{K^{1/2} s^{1/2} m^{-1}}) $$
In all critical sections of the bracket, the Niyama value at centerline decreased below 0.7 in the original design, indicating high shrinkage risk. After insulating risers, all values remained above 1.0. The temperature gradient direction at the boss–riser junction also reversed. Table 8 shows the Niyama values measured along the centerline of the ∅160 mm boss for three designs.
| Distance from boss bottom (mm) | Open Riser | Insulating A | Insulating B |
|---|---|---|---|
| 0 | 0.35 | 0.58 | 1.12 |
| 20 | 0.42 | 0.66 | 1.25 |
| 40 | 0.37 | 0.71 | 1.14 |
| 60 | 0.45 | 0.69 | 1.18 |
| 80 | 0.33 | 0.62 | 1.08 |
During the filling simulation, the molten metal velocity at the ingate was kept under 0.5 m/s to avoid mould erosion and sand entrainment, thereby eliminating sand-related sand foundry defects. I also used tracer particles to visualize the flow paths. The optimized gating system used multiple ingates, which balanced the flow and reduced local overheating.
8. Practical Recommendations for Box-Type Castings
Based on my research and production trials, I compiled a set of guidelines for producing sound box-type steel castings:
- Position ingates away from isolated hot spots and directly beneath risers whenever possible.
- Use insulating or exothermic riser sleeves when the riser geometry is constrained by the casting envelope.
- Add stress-relief blocks or cast-on fins in areas with stress concentrations to prevent hot tears.
- Increase mould collapsibility by using ester-cured sodium silicate sand with appropriate hardener control.
- Apply the Niyama criterion together with the feeding resistance criterion to identify shrink-prone regions.
- Ensure proper venting of cores. Core gas can be a major source of sand foundry defects such as blowholes.
- Control pouring temperature slightly above the alloy liquidus (1560°C for carbon steel) for smooth filling and adequate feeding.
- Use computer simulation iteratively to compare multiple designs before adding any physical trials.
9. Conclusion
I studied the forming behavior and defect formation in two representative box-type steel castings: the axle box housing and the bracket. Through a combination of foundry practice, finite element simulation, and dissection validation, I successfully eliminated shrinkage cavities, hot tears, and microporosity. The key points are:
- Original gating and riser designs caused isolated hot spots and feeding blockage, leading to severe sand foundry defects.
- Moving ingates under risers and adding stress-relief blocks solved the cracking and shrinkage problems in the axle box housing.
- For the bracket, simulation showed that open risers cooled too fast. Replacing them with insulated risers of adequate modulus established efficient feeding and eliminated all internal porosity.
- Numerical simulation calibrated by Niyama and feeding criteria proved to be a powerful tool for predicting and preventing sand foundry defects in complex steel castings.
- The validated process has been used for batch production, significantly reducing rejection rates and shortening development cycles.
This research confirms that applying physics-based simulation to sand castings of box-type geometry is an essential step in modern foundry engineering. It improves the understanding of thermal and stress fields and enables defect-free production. Continuous improvement using simulation technology will be my ongoing approach to minimizing sand foundry defects and advancing foundry quality control.