As a foundry engineer engaged in the production of critical railway components, I have long been confronted with the challenges posed by housing-type steel castings. These components, which include axle box housings and brackets for high-power diesel locomotives, demand exceptional internal soundness and mechanical performance because they operate under severe alternating loads and directly affect running safety. In this paper, I present my research on the forming behavior and defect mechanisms of steel castings, with particular emphasis on the application of numerical simulation for process optimization. Throughout this work, the central theme has been to predict and eliminate solidification defects such as shrinkage porosity, hot tearing, and cold cracking in steel casting production.
The foundry industry has entered a new era where traditional trial-and-error methods are being replaced by computer-aided engineering. My study focuses on two representative housing-type steel castings: the axle box housing used in semi-suspended high-speed bogies and the bracket used in a high-power diesel locomotive frame. Both components are classic examples of box-like structures with thick flanges, abrupt section changes, and limited accessibility for feeding. The quality requirements are stringent: no porosity, no cracks, and high fatigue resistance. To achieve these goals, I combined practical foundry knowledge with advanced simulation tools, namely the ProCAST software for pre-processing and the ProCAST solver for thermal, flow, and stress analysis. The results have enabled me to redesign gating and risering systems, eliminate casting defects, and establish a reliable process window for the production of high-integrity steel castings.
Steel Casting Process Fundamentals
The production of steel castings in my workshop begins with electric arc furnace (EAF) melting. The entire melting cycle comprises raw material preparation, melting period, oxidation period, and reduction period. Each stage must be carefully controlled to achieve the desired chemistry and cleanliness. Table 1 summarizes the key parameters for EAF steelmaking that I have utilized in this work.
| Stage | Purpose | Typical Control |
|---|---|---|
| Charging | Provide proper carbon and alloy content | Carbon 0.20–0.40% above lower limit; phosphorus <0.05% |
| Melting | Melt down scrap and form slag | Use maximum power after electrode bore-in; slag basicity 2.0–2.5 |
| Oxidation | Remove phosphorus, gases, inclusions | Temperature >1600°C; decarburization rate ≥0.3%/h; oxygen blowing or ore additions |
| Reduction | Deoxidize and desulfurize | White slag; carbon powder additions; final aluminum deoxidation (0.1–0.15%) |
For the housing-type steel castings discussed here, I used low-alloy steel grades equivalent to ASTM A148 or similar. The pouring temperature was set between 1560°C and 1580°C, and the mold material was ester-cured sodium silicate sand, which offers better collapsibility and less environmental impact than conventional CO₂-process sand. The molding sand composition is given in Table 2.
| Component | Proportion (by weight) |
|---|---|
| Silica sand (AFS 45–55) | 100 parts |
| Modified sodium silicate (modulus 2.2–2.6) | 3.0–4.5 parts |
| Ester hardener (glycerol acetate) | 0.15–0.30 parts |
| Water (if needed) | ≤0.5 parts |
The use of ester-cured sodium silicate sand significantly improved dimensional accuracy and surface finish of steel castings compared with traditional CO₂ sand. I observed fewer sand defects and better collapsibility, which reduced hot tearing tendency because the mold yields more readily during solidification contraction.
Numerical Simulation Methodology
Numerical simulation has become an indispensable tool for analyzing the filling, solidification, and stress evolution in steel casting. In this study, I used ProCAST as the simulation platform. The workflow consists of three main steps: pre-processing, solving, and post-processing. In pre-processing, I created three-dimensional solid models of the casting, gating system, risers, and chills using Unigraphics NX and then exported them in a format compatible with ProCAST. Each component—casting, riser, gate, chill—was assigned to a separate volume and its material properties defined. The finite difference mesh was generated with adaptive refinement to capture thin sections and high thermal gradients. Table 3 lists the simulation settings used for the steel castings in this investigation.
| Parameter | Value |
|---|---|
| Initial pouring temperature | 1570°C |
| Filling time | 15–25 seconds |
| Mold material | Ester-cured sodium silicate sand |
| Interface heat transfer coefficient (casting-mold) | 500–1000 W/m²K |
| Ambient temperature | 25°C |
| Critical liquid fraction | 0.6 |
| Feeding efficiency | 0.7 (open riser), 0.9 (insulated riser) |
The governing equation for heat conduction during solidification is the Fourier equation:
$$ \rho c_p \frac{\partial T}{\partial t} = \nabla \cdot (k \nabla T) + \dot{Q} $$
where ρ is density, cp is specific heat, k is thermal conductivity, and Q̇ represents the latent heat evolution. In ProCAST, the latent heat is handled by the enthalpy method:
$$ \frac{\partial H}{\partial t} = \nabla \cdot (k \nabla T) $$
with enthalpy H defined as:
$$ H(T) = \int_{T_0}^{T} \rho c_p \, dT + \rho L f_s(T) $$
where L is the latent heat of fusion and fs is the solid fraction. This formulation allows accurate tracking of the solidification front and the prediction of shrinkage porosity.
For the prediction of shrinkage defects, I employed the Niyama criterion, which is expressed as:
$$ N_y = \frac{G}{\sqrt{\dot{T}}} $$
where G is the temperature gradient and Ṫ is the cooling rate. When the local Niyama value falls below a critical threshold, porosity is expected. For steel castings, I typically use a critical value of 0.7–1.0 depending on section size. Additionally, the ProCAST porosity module calculates the actual percent porosity as a function of feeding resistance. A combined use of temperature field, temperature gradient, and porosity criteria provides reliable predictions for housing-type steel castings.
Case Study 1: Axle Box Housing – Cracking and Shrinkage Defects
The first product I investigated is the axle box housing for a semi-suspended high-speed bogie. The material is a cast steel with a nominal composition of ZG25MnNi (similar to AISI 1025 with nickel and manganese additions). The gross weight of the casting is 92 kg, and it has a semi-open cylindrical box structure with two end flanges connected by a central body. During service, the root regions between the flanges and the body experience alternating bending loads, making them highly susceptible to fatigue cracking. Therefore, the casting must be free of any defects that could act as stress raisers.
The initial casting process design used a two-part mold with a mid-plane parting line and a horizontal gating system. Two open risers were placed on the end flanges, while four insulated side risers were positioned above the bolt bosses on the central body. The ingates were located at the intersection of the parting plane and the bolt bosses, as shown in the original layout. This design produced a thick hot spot at the location where the ingate merged with the flange–boss junction. Because the hot spot was outside the feeding range of the adjacent risers, severe shrinkage porosity formed in that region. In addition, the large open risers on the flanges exerted high thermal contraction forces against the sand mold, generating concentrated tensile stresses at the flange roots and causing hot cracks.
After the first production batch, machining revealed fine cracks near the flange roots in nearly half of the castings. Further sectioning showed large internal shrinkage cavities beneath those cracks. It was clear that the problem was process-induced and would affect every casting produced with that gating/risering scheme. To resolve this issue, I carried out a systematic analysis using ProCAST stress solver. The finite element stress simulation allowed me to visualize the residual stress distribution and the evolution of thermally induced strains during cooling. Figure 1 illustrates the residual stress field predicted by the modified process. The simulation indicated that stress concentration at the flange root was excessive under the original design.
Based on the simulation, I implemented three key modifications to the casting process:
1. Relocation of ingates. The ingates were moved from the flange–boss junction to the lower part of the insulated side risers. This eliminated the additional hot spot created by the ingate thickness and allowed the riser to feed both the boss and the adjacent flange region. The new ingate configuration promoted a more uniform temperature distribution and reduced the local thermal stress.
2. Improvement of feeding channels. I increased the riser contact area by adding a built-up seat (choke) to the open risers on the end flanges. This enlarged the effective feeding distance and improved the thermal gradient from the casting to the riser. The modified feeding geometry allowed the riser to supply liquid steel to the lower flange sections that were previously underfed.
3. Addition of stress-relief transition blocks. To reduce stress concentration at the sharp flange root, I added temporary transition blocks (so-called “stress-relief pads”) at the root radius on the outside of the casting. These pads shifted the stress concentration away from the critical load-bearing section. After solidification and cooling, the pads were trimmed off during fettling, leaving the final casting with a sound root geometry.
The improved process was verified by sectioning actual production castings. No shrinkage cavities were found in the previously defective regions. The transition blocks successfully eliminated cracking. This confirmed that the stress-field simulation provided accurate guidance for process improvement. The lessons learned from this axle box housing case underline the importance of considering both thermal feeding and mechanical restraint when designing risering systems for steel castings.
Case Study 2: Bracket for Locomotive Frame – Shrinkage Porosity Elimination
The second product is a bracket for the frame of a high-power diesel locomotive built in cooperation with an American partner. The bracket weighs 78 kg and is made of low-alloy cast steel corresponding to ASTM A148 grade 105-85 (commonly designated as ZG25MnNiV in some standards). The component has a complex, thin-walled plate-like geometry with a sudden thickening at a circular boss of diameter Φ120 mm. The boss region is highly susceptible to shrinkage porosity because it acts as a local hot spot that is difficult to feed from the surrounding thin sections.
My initial casting process design used a two-part mold with a horizontal parting line and a middle-level gating system. The mold material was sodium silicate sand hardened by CO₂. For risering, I initially preferred open risers because insulated risers are known to cause surface sand adhesion at the riser neck. Using the modulus method, I calculated two open risers of dimensions Φ140 mm × 180 mm and Φ120 mm × 180 mm. The simulation, however, revealed a fundamental problem: at 50% solidification, the temperature in the open risers was lower than that in the underlying boss, meaning the risers solidified before the boss. This reversed thermal gradient prevented the risers from feeding the solidification shrinkage of the boss. As a result, the boss would develop significant centerline shrinkage porosity. The Niyama criterion and the porosity module both predicted a substantial defective zone at the center of the Φ120 mm boss. The open risers, despite their large size, could not compensate for the lack of directional solidification.
Because the product structure did not permit enlarging the open risers (the available space was limited), I changed to insulated risers. Insulated risers reduce the cooling rate of the riser metal, thereby increasing the thermal gradient from casting to riser. The new risers were Φ120 mm × 160 mm and Φ100 mm × 150 mm. The simulation of this revised design showed a dramatic improvement: at 50% solidification, the riser temperature was slightly higher than that of the boss, and the feeding channel remained open until the boss completely solidified. However, a small local shrinkage pore still appeared near the centerline of the right-hand boss. This was deemed acceptable by the design department, but I aimed for further improvement.
To achieve a completely sound boss, I increased the insulated riser diameters to Φ140 mm × 180 mm and Φ120 mm × 160 mm. This third simulation showed an even larger temperature difference between riser and boss, with the riser remaining significantly hotter throughout solidification. The resulting feeding action was sufficient to eliminate all shrinkage porosity in the boss. The porosity predictions are summarized in Table 4.
| Design | Riser type | Riser dimensions (mm) | Temperature gradient at 50% solidification (riser–boss) | Predicted porosity in boss |
|---|---|---|---|---|
| A (original) | Open | Φ140×180, Φ120×180 | −8°C (riser colder) | Severe centerline shrinkage |
| B (revised) | Insulated | Φ120×160, Φ100×150 | +5°C (riser hotter) | Small localized porosity |
| C (final) | Insulated | Φ140×180, Φ120×160 | +12°C (riser hotter) | No porosity |
The production trial with the final design confirmed the simulation. Two brackets were cast and sectioned. No shrinkage defects were observed in the boss or at the riser neck. The defect locations and sizes matched the ProCAST predictions with high accuracy. This demonstrates that numerical simulation can reliably guide the risering of complex housing-type steel castings, reducing costly trial runs.
Defect Mechanisms in Housing-Type Steel Castings
Through these two case studies, I have identified several common defect mechanisms that are particularly relevant to housing-type steel castings:
1. Hot spot formation at ingate–boss junctions. The ingate area often becomes a hot spot because the local mold is heated by the flowing steel and the ingate thickness adds extra metal. If the hot spot is not within the feeding zone of a riser, shrinkage porosity will form. This is exactly what happened in the axle box housing. The solution is to locate ingates so that they do not create an isolated hot spot, preferably by feeding through the riser itself.
2. Inadequate riser temperature gradient. In many industrial designs, the riser is placed on a thick section, but the cooling of the riser is faster than that of the casting because of the large surface-to-volume ratio of the riser. This leads to a reverse temperature gradient and poor feeding. For steel castings, the riser must have a sufficiently large modulus and/or be insulated to stay liquid longer than the casting. The simulation was able to quantify this effect and show the critical temperature difference.
3. Stress concentration at flange roots. Housing-type castings often have abrupt section changes at the junction of flanges and bodies. During cooling, the heavy flanges contract more than the thin webs, creating tensile stresses at the root radius. If the stress exceeds the hot strength of the steel, hot cracks form. The addition of transition blocks is a practical way to shift the stress concentration away from the critical section. Stress simulation allows the designer to evaluate the effectiveness of such pads before physical trials.
4. Centerline shrinkage in axi-symmetric bosses. When a circular boss is fed from a central riser, the last liquid to solidify is located along the axis. If the riser cannot maintain a positive temperature gradient toward its open end, the axis region experiences an “axial shrinkage line” where porosity forms. This is similar to what we observed in the bracket before the final riser modification. The Niyama criterion is particularly useful for detecting this type of defect.
Mathematical Models for Shrinkage Prediction
In addition to the Niyama criterion, I have used the concept of feeding efficiency and the temperature gradient criterion. For a cylindrical boss, the feeding distance can be expressed as:
$$ L_f = 2.5 \sqrt{\frac{A}{P}} + 0.15 M $$
where Lf is the feeding distance, A is the cross-sectional area, P is the perimeter, and M is the local thermal modulus. For steel castings, the critical modulus ratio of riser to casting is typically:
$$ \frac{M_r}{M_c} \geq 1.2 $$
where Mr is the riser modulus and Mc is the casting modulus. The modulus is defined as the volume-to-surface area ratio:
$$ M = \frac{V}{A_s} $$
For an insulated riser, the effective modulus is increased by a factor f based on the insulation material:
$$ M_{eff} = f \cdot M_r $$
Typical values of f are 1.3–1.5 for exothermic sleeves and 1.1–1.3 for ordinary insulating sleeves. In my bracket design, I achieved the required modulus increase by using insulated risers.
The solidification time is given by Chvorinov’s rule:
$$ t_s = \frac{\pi}{4} \left( \frac{k_m \rho_m c_m}{k_c \rho_c c_c} \right) \left( \frac{M}{T_m – T_0} \right)^2 $$
where Tm is the liquidus temperature and T0 is the mold initial temperature. This formula shows that increasing the riser modulus has a quadratic effect on solidification time, which is why insulated risers are so effective for steel castings.
Process Optimization Strategy
Based on my research, I propose a systematic approach for the design of casting processes for housing-type steel castings:
Step 1: Preliminary design. Calculate the modulus of each hot spot, determine the riser modulus and size using the modulus ratio, and design the gating system to avoid turbulent filling and air aspiration.
Step 2: Numerical simulation. Use a reliable software package to simulate filling, solidification, and stress. Analyze the temperature field at various solidification times, the temperature gradient distribution, and the porosity criteria. Identify any areas where the thermal gradient is reversed or where the Niyama value is below the critical threshold.
Step 3: Defect interpretation. If shrinkage porosity is predicted, verify whether the riser volume and modulus are adequate. If cracking is predicted, examine the stress distribution and consider modifying the casting geometry (adding transition radii or pads) or improving mold collapsibility.
Step 4: Iterative modification. Change the gating/risering design, repeat the simulation, and compare results. Typically, two to three iterations are sufficient to achieve a sound casting design.
Step 5: Production verification. Cast test pieces and section them to confirm the absence of defects. If the predicted and actual defect locations match, the simulation is validated and the final process can be released.
This approach has been applied successfully to the axle box housing and the bracket, saving significant time and cost compared with traditional empirical methods.
Benefits of Numerical Simulation for Steel Casting
The application of ProCAST has transformed my approach to steel casting process design. Among the major benefits, I can list:
- Accurate prediction of shrinkage porosity and hot tears using the porosity module and the Niyama criterion, with results that correlate well with actual castings.
- Visualization of temperature and stress fields during solidification, allowing me to understand the root causes of defects rather than merely treating symptoms.
- Virtual experimentation that reduces the number of physical production trials, thereby lowering material, energy, and labor costs.
- Shortened development cycles for new products, as seen in the bracket case where the final process was developed in three simulation iterations without multiple shop-floor trials.
- Improved product quality, ensuring that critical components for railway applications are free of defects and meet the stringent requirements for fatigue resistance.
Table 5 presents a comparison of the trial-and-error method versus simulation-based design for the two steel castings studied.
| Metric | Traditional Trial-and-Error | Simulation-Based Design |
|---|---|---|
| Number of iterations | 4–6 | 2–3 |
| Material waste | High (multiple trial castings) | Low (only final verification) |
| Development time | 6–8 weeks | 2–3 weeks |
| Cost | High | Reduced by ~50% |
| Defect predictability | Poor | Excellent |
Conclusion and Outlook
In this research, I have investigated the formability and defect formation mechanisms of housing-type steel castings through a combination of practical foundry practice, rigorous numerical simulation, and experimental verification. The main conclusions are as follows:
(1) The formation of shrinkage porosity and hot cracks in housing-type steel castings is primarily governed by the thermal field, the feeding capability of risers, and the mechanical restraint of the mold. Defects tend to occur at ingate–boss junctions, at flange roots, and at the centerline of thickened bosses.
(2) For the axle box housing, relocating the ingates from the flange–boss junction to the side risers, increasing the riser seating area, and adding stress-relief transition blocks effectively eliminated both internal shrinkage cavities and surface cracks. The stress simulation using ProCAST provided detailed insight into the residual stress distribution and guided the design of the transition blocks.
(3) For the bracket, changing from open risers to insulated risers with optimized dimensions created the necessary positive thermal gradient for directional solidification. The final riser sizes of Φ140 mm × 180 mm and Φ120 mm × 160 mm completely eliminated shrinkage porosity in the Φ120 mm boss. Production sectioning confirmed the simulation results.
(4) Numerical simulation is an accurate and practical tool for optimizing steel casting processes. The Niyama criterion and the porosity module in ProCAST correctly predicted defect locations and severities. The use of simulation reduced the number of development iterations, saved cost, and improved the reliability of the final products.
Looking forward, I believe that the continued application of simulation technology will further enhance the quality and competitiveness of steel casting production. Microstructure modeling, which predicts grain size and phase fractions, will allow even better control of mechanical properties. The integration of casting simulation with finite element analysis of service loads will enable the design of safer and lighter railway components. For housing-type steel castings, the combination of advanced simulation, rigorous process control, and continuous learning from production feedback remains the most effective path to excellence.
In conclusion, my work demonstrates that through the systematic use of simulation and the understanding of steel casting solidification behavior, complex housing-type steel castings can be manufactured with high integrity and reliability. These methods are now a standard part of my foundry practice and will continue to contribute to the safe operation of high-speed and high-power locomotives.