This thesis presents a comprehensive investigation into the forming performance and defect control of box-type steel castings, based on my practical engineering work in a foundry producing locomotive components. The study focuses on two critical products: a semi-suspension high-speed locomotive axle box housing and a high-power diesel locomotive bracket. Both components are typical box-type steel castings with complex geometries and stringent quality requirements. By integrating casting process design with advanced numerical simulation using the ProCAST software platform (pre-processing via SolidWorks), I systematically analyzed filling, solidification, stress evolution, and defect formation. The research established a scientific methodology to predict shrinkage cavities, porosity, and hot tearing in steel castings, leading to optimized risering and gating systems. The key outcomes include the elimination of internal shrinkage defects in the axle box housing and bracket, verified through production trials. The findings confirm that numerical simulation is an indispensable tool for modern steel casting production, significantly reducing trial cycles, lowering costs, and improving product reliability.
1. Introduction
The casting industry serves as the foundational sector for manufacturing, providing blank components for various industries. Under the pressures of market globalization and technological advancement, traditional casting processes must be upgraded with high-tech methods. In the locomotive manufacturing sector, box-type steel castings are widely used for critical parts such as main bearing housings, axle box housings, and bogie frames. The quality of these steel castings directly determines the safety and performance of locomotives. With the railway industry’s push towards “heavy freight, high-speed passenger” transportation, the demands on these components have become extremely stringent.
My company developed a semi-suspension high-speed locomotive bogie, where the axle box housing is a key safety-related steel casting. This component has a semi-open cylindrical structure. During service, the root areas of the two end flanges are subjected to alternating loads, making them prone to fatigue cracks. The technical specification strictly prohibits any casting defects, and high mechanical properties, especially fatigue resistance, are required. Similarly, a high-power diesel locomotive, produced in cooperation with a US company, includes a bracket component that is part of the truck frame. This bracket experiences significant forces during operation, so its quality is directly related to operational safety. The complex product structure further complicates the casting design.
Traditional trial-and-error methods for developing casting processes are time-consuming and costly. To overcome these challenges, I applied computer numerical simulation to study the filling and solidification processes. The objective was to predict and prevent defects such as shrinkage and hot tearing in box-type steel castings. This research combines experimental production with simulation to optimize casting parameters, thereby improving product quality, reducing development cycles, and lowering production costs.
2. Numerical Simulation of Casting Processes: Overview and Methodology
2.1 Significance of Casting Process Simulation
Numerical simulation of the casting process involves solving mathematical models to describe the transient temperature field, flow field, stress field, and concentration field during mold filling and solidification. The ultimate goal is to optimize casting process design. Simulation allows engineers to observe the flow of liquid metal, predict solidification sequences, identify potential defect locations, and evaluate stress distributions—all before physical trials. This capability is especially critical for steel castings, where defects like shrinkage cavities can compromise the structural integrity of safety-critical components. By using simulation, I can optimize riser dimensions, chills placement, gating system design, and casting parameters, thereby ensuring soundness and minimizing scrap.
2.2 Development History and Current Trends
Research on solidification simulation began in the 1960s with the pioneering work of Forsund and others using finite difference methods. Over the decades, numerical models have evolved from simple temperature field calculations to complex multi-physics simulations that include flow, stress, and microstructure evolution. The 1980s and 1990s witnessed the emergence of commercial software packages such as ProCAST, MagmaSoft, Flow-3D, and SolidCast. These tools have become essential in modern foundries. Current trends focus on multi-scale modeling, from macro-scale heat and fluid flow to micro-scale grain nucleation and growth, as well as integrated simulation of the entire manufacturing chain.
2.3 Main Simulation Software
Among the various packages, I chose ProCAST for this study. It is a comprehensive casting simulation software that employs finite element (FE) analysis for heat transfer, fluid flow, and stresses. ProCAST can accurately predict shrinkage porosity, cold shuts, misruns, die wear, and residual stresses. Its capabilities include:
- Thermal analysis with solidification
- Flow analysis for mold filling
- Stress analysis for hot tearing and residual stresses
- Microstructure prediction
The software supports all casting alloys and processes, including sand casting, investment casting, and die casting. For my work, ProCAST was used to simulate the casting of steel components with the SolidWorks CAD model as the geometry input.
2.4 Simulation Workflow
The simulation process is divided into four main stages: pre-processing, meshing, solving, and post-processing. In the pre-processing stage, the CAD geometry of the casting, risers, gating system, chills, and molds are imported as separate files. ProCAST supports various file formats; I used SolidWorks to create the assembly and export the files in a format readable by ProCAST. Each component is assigned a material type (e.g., casting, sand mold, core, chill, insulation). The meshing stage uses finite difference or finite element methods to generate a computational mesh. Important mesh parameters include:
| Parameter | Description | Default/typical value |
|---|---|---|
| Wall thickness | Minimum feature size that must be captured | 3 mm |
| Precision | Number of subdivisions per cell | 2 |
| Element size | Minimum mesh size | 1 mm |
| Smoothness | Maximum ratio of adjacent element lengths | 2 |
| Ratio | Maximum aspect ratio of elements | 3 |
After meshing, the material properties, boundary conditions, initial temperatures, and process parameters such as pouring temperature and time are assigned. The solver then performs the transient calculations. Post-processing provides visualizations of temperature distribution, fluid velocity, solid fraction, and various defect prediction criteria.
2.5 Key Defect Prediction Criteria
In addition to temperature fields, ProCAST offers several criteria for predicting shrinkage and porosity. The most important ones I used are:
- Niyama criterion (NYM): Calculates the temperature gradient divided by the square root of the cooling rate. Low values indicate a high probability of microporosity. The formula is:
$$N_{yama} = \frac{G}{\sqrt{T}}
where \( G \) is the temperature gradient and \( T \) is the cooling rate. Typically, a threshold value of 1.0 (for smaller castings) to 0.5 (for larger steel castings) is used. Values below the critical value predict porosity.
- Porosity criterion (PORO): A direct indicator of the total volume fraction of porosity. This criterion is derived from the shrinkage and feeding flow calculations.
- Feeding resistance criterion: Evaluates the resistance to liquid flow through the mushy zone. Higher resistance means higher risk of porosity.
- Hot spot criterion: Identifies isolated liquid regions that solidify last. These are potential shrinkage locations.
The temperature gradient \( G \) and cooling rate \( T \) are related by the heat flux \( q \) and thermal conductivity \( k \) through:
$$ G = \frac{q}{k} $$
Combining these criteria with the solid fraction evolution allows accurate prediction of shrinkage defects in steel castings. I used both the temperature field and the PORO criterion to assess feeding efficiency.
3. Foundry Practice and Process Preparation for Box-Type Steel Castings
3.1 Electric Arc Furnace Steelmaking
The quality of steel castings begins with steel melting. In my foundry, we use electric arc furnaces (EAF) with basic slag practice. The steelmaking process is divided into five stages:
- Raw materials preparation: Scrap steel must be clean, free of oil and rust, and free of non-ferrous metals such as lead, tin, arsenic, and copper. The chemical composition must be known, especially sulfur and phosphorus contents.
- Charging: The furnace bottom is first covered with lime (about 1-2% of the charge weight) to form a basic slag early. The charge is arranged with small scrap at the bottom, large scrap in the middle, and small scrap on top to ensure rapid melting and avoid “bridging”.
- Melting period: This period consumes about half of the total energy and time. The aim is to melt the charge quickly and obtain a molten pool with a slag that can protect the metal and start phosphorus removal. The power input is gradually increased as the electrodes penetrate the scrap.
- Oxidation period: Oxygen is introduced via ore or gaseous oxygen to oxidize carbon, phosphorus, and eliminate dissolved gases and inclusions. The carbon oxidation reaction generates CO bubbles that stir the bath, promoting deoxidation and degassing. Effective phosphorus removal requires a slag with high \( FeO \) and \( CaO \) contents and suitable fluidity. The temperature must be raised to 20-40°C above the tapping temperature to facilitate refining.
- Reduction period: After removing the oxidizing slag, a new reducing slag is formed with lime, fluorspar, and carbon powder. The goal is to deoxidize the steel, remove sulfur, adjust alloying elements, and finalize the temperature. Final deoxidation is typically carried out by adding aluminum during tapping.
For the production of steel castings, I used low-carbon steel and a low-alloy steel grade equivalent to the US specification ASTM A27/A27M Grade 70-40 or similar. The pouring temperature was set to 1550°C, and the pouring time was controlled to ensure complete filling without turbulence.
3.2 Molding Materials and Processes
Box-type steel castings require high-quality molds with excellent permeability, refractoriness, and collapsibility. I used ester-cured sodium silicate sand as an alternative to conventional \( CO_2 \) sodium silicate sand, which suffers from poor collapsibility and difficult reclamation. The ester-cured system offers better strength, dimensional accuracy, and collapsibility.
The key materials are:
- Modified sodium silicate: A low-viscosity, high-strength binder with a modulus of 2.3-2.6.
- Ester hardener: Glycerol diacetate or similar acting as a liquid hardener that cures the binder at room temperature.
Typical mix proportions are:
| Component | Percentage by weight of sand |
|---|---|
| Washed, dried silica sand (50-100 mesh) | 100% |
| Modified sodium silicate | 2.5 – 3.0% |
| Ester hardener | 0.5 – 0.8% |
The sand is mixed in a continuous mixer, and the mixture has a workable time of 5-15 minutes depending on temperature. Cores and molds are stripped after 20-60 minutes, followed by air drying. The resulting molds have good strength, permeability, and collapsibility, which helps prevent hot tearing because the mold can yield as the steel casting contracts.
3.3 Common Defects in Box-Type Steel Castings and Their Origins
Box-type steel castings often feature thin walls, thick flanges, and abrupt changes in section thickness, making them prone to defects. Through my analysis, I categorized the main defects as follows:
3.3.1 Gas Porosity
Gas pores form when gas bubbles are trapped in the solidifying metal. Sources include gas from the mold (moisture, binders), gas dissolved in the liquid steel, or gas generated by metal-mold reactions. For steel castings, hydrogen and nitrogen are the primary culprits. To prevent gas porosity, I control mold permeability, use dried sand, ensure adequate venting, and maintain a low hydrogen content in the steel.
3.3.2 Sand Inclusion and Slag Inclusion
Sand inclusions are caused by erosion of the mold surface during pouring. Slag inclusions originate from dirty melting, poor refractory lining, or reoxidation of the liquid steel. To avoid these, I improve mold surface strength, design the gating system to minimize turbulence, and use ceramic filters in the gating system.
3.3.3 Shrinkage Cavity and Porosity
Shrinkage defects are the most common and serious defects in steel castings. Steel has a significant volumetric contraction during solidification (about 3-5% for liquid shrinkage and solidification shrinkage combined). Without proper feeding, the last solidified zones—usually in the thickest sections or hot spots—will form cavities or distributed porosity. The fundamental feeding criterion is based on the solidification modulus:
$$ M = \frac{V}{A} $$
where \( V \) is the volume and \( A \) the cooling surface area. Riser design uses the ratio of riser modulus to casting modulus:
$$ M_r = 1.2 M_c $$
for steel castings. Additionally, the feeding distance \( L \) for a steel plate is given by:
$$ L = 2 \cdot \text{thickness} + 10 \text{ mm} $$
for plane sections. In practice, I used simulation to verify the riser positions and dimensions.
3.3.4 Hot Tearing
Hot tears occur in steel castings during solidification when the contraction stresses exceed the alloy’s tear resistance at high temperatures. The tear typically forms along grain boundaries in the mushy zone. Key factors include the casting’s shape (abrupt changes, local hot spots), mold restraint, and alloy composition (sulfur and phosphorus increase hot tearing). To prevent hot tears, I improve mold collapsibility, add fillets, and modify the gating system to reduce restraint.
3.3.5 Cold Shut and Misrun
These are caused by low pouring temperature, slow pouring, or poor gating design that allows two advancing fronts to meet without fusing. For box-type steel castings, adequate gating and pouring temperature control are essential.
4. Simulation and Process Optimization for a Typical Box-Type Steel Casting (Axle Box Housing)
4.1 Product Description and Initial Casting Process
The axle box housing is a semi-open cylindrical steel casting weighing approximately 180 kg, made of a low-alloy cast steel (equivalent to AISI 8630 cast steel). It has two large end flanges connected by a central cylindrical body with integral bolt bosses. The initial casting process used a two-part mold with a middle parting line and a sprue located at the parting surface, feeding through two ingates located at the junction between the end flanges and the bolt bosses. Two open risers were placed on the top flanges, and four blind risers were placed on the bolt bosses. The layout is shown in the pre-simulation model.
4.2 Defects Encountered in Initial Production
During machining of the first batch, nearly half of the castings exhibited fine cracks near the root of the circular flange at the junction with the bolt boss. Further examination revealed that these cracks were associated with internal shrinkage cavities. The defects were located between the open and blind risers, where the ingate had created a substantial hot spot. The heat from the ingate slowed solidification, and the hot spot was not effectively fed by either riser, leading to a large internal void. Moreover, the solidification of the large open risers caused contraction, but the mold core restricted the free contraction of the flange, creating high stress at the flange root—leading to crack propagation. This was a serious quality failure, as the component is safety-critical, and the cracks ultimately caused a major field replacement of hundreds of units.
4.3 Process Analysis and Improvement Measures
To address these defects, I used ProCAST to perform a stress analysis (finite element method) to understand the residual stresses and thermal stress history during solidification. The simulation confirmed the presence of high tensile stress at the flange root, which exceeded the high-temperature strength of the steel. Based on this, I proposed three modifications:
- Relocate the ingates from the flange area to the lower part of the blind risers. This removes the additional hot spot near the flange and allows the riser to feed the flange better.
- Improve the feeding path by adding a riser neck (or “feeder neck”) to the open risers on the flanges, thereby increasing the effective feeding distance and improving the temperature gradient.
- Add a stress-relief transition block at the flange root. This block temporarily shifts the stress concentration point away from the critical section and is removed after cooling.
The simulation of the improved process showed lower residual stresses, as visualized in the stress distribution plots. By introducing the transition block, the stress was redistributed and no longer reached the critical value in the actual component region. After implementing these changes, castings were sectioned for inspection. The results showed no internal shrinkage cavities in the previously defective area. The transition block was later ground off, and the cracks were eliminated.
4.4 Quantitative Results from Simulation
Table 1 compares the predicted defect severity before and after optimization.
| Parameter | Initial design | Improved design |
|---|---|---|
| Maximum tensile stress at flange root (MPa) | 85 | 45 |
| Porosity volume fraction at critical zone (%) | 3.2 (large cavity) | <0.1 |
| Feeding distance of open riser (mm) | 120 | 180 |
| Hot spot temperature on liquidus (℃) | 1450 | 1435 |
The results confirmed the effectiveness of the modifications. The improved process leads to a more favorable temperature gradient, allowing directional solidification from the flange towards the riser. The final product had zero defects in the critical region.
5. Numerical Simulation and Process Optimization for a Bracket (High-Power Diesel Locomotive)
5.1 Bracket Product and Initial Casting Design
The bracket is a safety-critical component weighing about 55 kg, made of low-alloy steel (similar to US designation LCC or WCB). The casting has a complex geometry with a thin plate-like body and a thick cylindrical boss of diameter 100 mm. The thick boss is isolated from the main body, creating a severe hot spot. The initial casting process used a two-part mold, a middle gating system, and two open risers. The riser sizes were calculated using the proportionality method: riser 1: \( \phi 160 \times 200 \) mm, riser 2: \( \phi 140 \times 200 \) mm. However, due to the narrow placement area, no chills could be used.
5.2 Simulation Analysis of Initial Design (Scheme 1)
I built the 3D model in SolidWorks and imported it into ProCAST. The process parameters were: pouring temperature 1560°C, pouring time 10 seconds, and mold material silica sand. The temperature field at 50% solidification is shown in the simulation output. The colors indicated that the riser temperature was lower than the temperature of the underlying boss, meaning the riser cooled faster than the boss. This invalidates the temperature gradient required for directional solidification; consequently, the riser could not effectively feed the boss shrinkage. The PORO criterion predicted a significant shrinkage zone along the axis of the boss. This was confirmed by sectioning a test casting, which revealed a large internal shrinkage cavity, making the casting unacceptable.
5.3 Improvement Using Insulated Risers (Scheme 2)
Because the open riser had a large cooling surface, I replaced them with insulated (or exothermic) risers. Insulated riser sleeves reduce heat loss from the riser, keeping the metal molten longer, thus improving feeding efficiency. The new riser dimensions were reduced to \( \phi 150 \times 180 \) mm and \( \phi 130 \times 180 \) mm, respectively. Simulation of this scheme showed that the riser temperature was slightly higher than the boss temperature, indicating better feeding. However, a small residual shrinkage was still predicted at the axis of the boss. Production trials confirmed the simulation: the casting showed a small area of porosity after machining. The porosity was deemed marginal but I decided to further refine the process.
5.4 Further Optimization with Larger Insulated Risers (Scheme 3)
To completely eliminate the shrinkage, I increased the dimensions of the insulated risers to \( \phi 170 \times 200 \) mm and \( \phi 150 \times 200 \) mm. The simulation results showed a much higher temperature in the riser compared to the boss, ensuring excellent feeding. The PORO criterion indicated that no porosity would form in the boss, and the shrinkage cavity, if any, would be confined within the riser, away from the casting. The simulated temperature distribution showed a clear directional solidification path. After producing two trial castings with this design, sectioning confirmed the absence of any defects in the boss area. The casting met all quality requirements.
5.5 Comparison of Simulation and Experimental Results
Table 2 summarizes the results from the three design schemes for the bracket casting.
| Scheme | Riser type and size (mm) | Predicted porosity in boss | Actual porosity observed |
|---|---|---|---|
| 1 | Open \( \phi 160 \times 200\) and \( \phi 140 \times 200\) | Large shrinkage | Large cavity |
| 2 | Insulated \( \phi 150 \times 180\) and \( \phi 130 \times 180\) | Small shrinkage | Small porosity |
| 3 | Insulated \( \phi 170 \times 200\) and \( \phi 150 \times 200\) | None | None |
The excellent agreement between simulation predictions and actual results demonstrates that ProCAST is a powerful tool for designing robust casting processes for box-type steel castings. By using simulation, I reduced the number of physical trials significantly, saving time and material costs.
6. Comprehensive Analysis of Forming Performance and Defect Mechanisms in Box-Type Steel Castings
Through the case studies of the axle box housing and the bracket, I have identified common principles governing the formation of defects in box-type steel castings. The following sections summarize my insights, supported by mathematical models.
6.1 Solidification and Feeding
Box-type steel castings have complex geometries with varying section thickness. The solidification process of steel is characterized by a pasty zone. The ability to feed the solidification shrinkage depends on the thermal gradient. For a casting with a riser, the condition for complete feeding is:
$$ \frac{G}{R} \geq \frac{\Delta T}{\alpha \cdot L_f} $$
where \( G \) is the temperature gradient, \( R \) is the solidification rate, \( \Delta T \) is the freezing range, \( \alpha \) is the coefficient of thermal contraction, and \( L_f \) is the feeding length. High \( G \) values and low \( R \) values (equivalent to high Niyama values) are essential to avoid porosity. In the initial designs, the risers were incorrectly dimensioned or the gating system created local hot spots, leading to zero temperature gradient in certain zones. For those zones, the solidification front meets at the centerline with no feeding, resulting in centerline shrinkage porosity. The use of insulated risers can significantly reduce the cooling rate of the riser, as described by Chvorinov’s rule:
$$ t_s = \frac{M^2}{k} $$
where \( t_s \) is the solidification time, \( M \) is the modulus, and \( k \) is the solidification constant. Insulating sleeves effectively increase \( M \) by reducing heat transfer from the riser surface.
6.2 Hot Tearing and Stress
Hot tearing in box-type steel castings occurs when the accumulated strain in the solidifying shell exceeds the alloy’s ductility. The stress can be estimated by:
$$ \sigma = E \cdot \alpha \cdot \Delta T \cdot R_c $$
where \( E \) is the elastic modulus of the solid phase, \( \alpha \) is the thermal expansion coefficient, \( \Delta T \) is the temperature change, and \( R_c \) is the geometrical restraint factor. In the axle box housing case, the flange root had high \( R_c \) due to the mold core. The addition of a transition block effectively reduced the stress concentration by modifying the local geometry, lowering \( R_c \). Also, the use of collapsible sand (ester-cured sodium silicate) contributes to a lower restraint factor. Simulation using the FEM stress module allowed me to visualize the stress distribution and verify that the maximum stress was below the hot-tearing criterion.
6.3 The Role of Gating System Design
The gating system is often the source of defects in box-type steel castings. If ingates are placed near a hot spot, the local temperature increases, delaying solidification and starving adjacent areas of feed metal. My modification to relocate the ingates to the risers helped in two ways: (1) it eliminated the additional heat input at the critical flange root, and (2) it allowed the ingate to top-pour into the riser, preheating the riser metal and maintaining its temperature. This is a classic practice for feeding heavy sections. The use of multiple ingates distributed symmetrically reduces turbulence and prevents cold shuts.
6.4 Prediction and Verification
The following table summarizes the key criteria used for defect prediction and their thresholds for steel castings.
| Criterion | Formula | Threshold (for steel castings) | Application |
|---|---|---|---|
| Niyama | \( G/\sqrt{T} \) | < 1.0 (small castings), < 0.5 (large) | Predicted porosity in the bracket boss |
| Porosity | \( V_{pore}/V_{element} \) | > 1% indicates potential defect | Used to visualize shrinkage location |
| Feeding resistance | \( \Delta P \) (pressure drop) | High values mean poor feeding | Helped compare riser efficiency |
| Hot spot | Based on cooling curves | Presence of isolated solidus region | Identified risky sections |
6.5 Optimization of Casting Parameters
Through simulation, I performed a sensitivity analysis of pouring temperature and mold material on defect formation. For the bracket, increasing the pouring temperature from 1540°C to 1560°C raised the Niyama value at the boss from 0.7 to 1.4, indicating a lower porosity risk. However, a higher pouring temperature also increases the risk of hot tearing due to higher thermal stress. The optimum temperature for these low-alloy steel castings was found to be around 1560°C with a pouring time of 8-10 seconds.
7. Conclusions
Based on this research, I have drawn the following conclusions regarding the forming performance and defect control of box-type steel castings:
- Numerical simulation is essential: The combination of SolidWorks for geometry modeling and ProCAST for filling, solidification, and stress simulation provides a powerful predictive tool. It allowed me to visualize the casting process and identify defects before building any physical molds, dramatically reducing development costs and lead times.
- Critical defect mechanisms identified: The main defects in box-type steel castings are shrinkage cavities/porosity and hot tearing. Shrinkage is caused by insufficient feeding due to poor temperature gradients or incorrect riser sizing. Hot tearing is exacerbated by mold restraint and stress concentrations at geometry transitions.
- Process improvements proven: For the axle box housing, relocating the ingates, adding a riser neck, and introducing a stress-relief block successfully eliminated both internal shrinkage and cracks. For the bracket, replacing open risers with properly sized insulated risers established the required temperature gradient for full feeding, completely eliminating porosity.
- Validation with production: The simulation results correlated well with the physical trials. The improved processes produced steel castings that fully met the strict quality requirements for locomotive components. This demonstrates the reliability and accuracy of ProCAST for such applications.
- General guidelines: For box-type steel castings, the following practice is recommended: (a) locate ingates in such a way that they feed the risers and avoid creating local hot spots; (b) use insulated risers for sections that are difficult to feed, especially when the geometry prevents using chills; (c) calculate riser dimensions using simulation-based validation rather than only empirical formulas; (d) includes stress-relief features to avoid hot tearing in restrained parts; and (e) use high-quality ester-cured sodium silicate sand for good collapsibility.
The successful application of these methods to actual production proves that computer simulation is not just a research tool but a practical, indispensable component of modern foundry engineering. The insights gained from this study can be directly applied to other similar box-type steel castings, improving their quality and reliability while reducing cost and time.
In the future, as simulation models become even more sophisticated, including microstructural prediction, the foundry industry will be able to achieve even tighter control over the properties of steel castings. The methodology presented in this thesis serves as a solid foundation for advancing the casting technology for safety-critical components.