Throughout my career in foundry engineering, I have encountered numerous challenges related to the production of high-quality steel castings. One particular project that demanded my full attention was the manufacturing of a bracket component made from ZG30Mn low-alloy cast steel. This component is a shell-type part, designed to function as a high-strength load-bearing element in transmission systems. The material specification (in weight percent) required: 0.27–0.34% C, 0.30–0.50% Si, 1.2–1.5% Mn, ≤0.035% P, and ≤0.035% S. After casting, the blank needed to be quenched and tempered before machining to final dimensions. The primary challenge was to produce castings free from internal and external defects while maintaining mechanical properties and dimensional accuracy. This article describes my journey from sand casting to lost foam casting, focusing on how I identified and mitigated typical sand casting defect issues.
1. Initial Sand Casting Approach and Observed Defects
During the prototyping phase, I selected the sand casting process for its flexibility and ease of pattern modification. The component’s geometry—a shell with a large end face and a thicker bottom section—presented a high risk of shrinkage porosity. Using CO₂-hardened water glass sand, I designed the first gating and riser system (Scheme A). The mold was a top-pour design to achieve directional solidification from bottom to top. Unfortunately, after machining, I discovered a sand casting defect: a large shrinkage cavity at the root of the large end face.
The primary cause was insufficient riser feeding distance. The riser placed on the top could not effectively compensate the thick section located far from the riser neck. To quantify this, I calculated the modulus of the critical section using Chvorinov’s rule:
$$M = \frac{V}{A}$$
where M is the modulus, V is the volume, and A is the cooling surface area. For the thick bottom section, the modulus was approximately $$M_{thick} = \frac{1500\,\text{cm}^3}{800\,\text{cm}^2} = 1.875\,\text{cm}$$. The riser modulus needed to be at least 1.2 times the casting modulus to ensure proper feeding: $$M_{riser} \geq 1.2 \times 1.875 = 2.25\,\text{cm}$$. My initial riser had a modulus of only 2.0 cm, which was insufficient. Moreover, the feeding distance for a top riser in a steel casting is limited to about 4.5 times the riser diameter. In my case, the distance from the riser center to the root was 250 mm, exceeding the recommended 180 mm.
To eliminate this sand casting defect, I revised the process to Scheme B, as shown conceptually below. I added external chills around the thick bottom section to increase local cooling rate and promote directional solidification. Additionally, I placed four smaller risers along the large end face to provide multiple feeding points. The chill design was based on the following thermal balance:
$$Q_{chill} = m_{chill} \cdot c_{chill} \cdot \Delta T_{chill} = \rho_{chill} \cdot V_{chill} \cdot c_{chill} \cdot (T_{solidus} – T_{initial})$$
The required chill volume was calculated as approximately 5% of the casting volume in the thick region. After implementing these changes, the shrinkage cavity was completely eliminated. Table 1 summarizes the comparison of the two sand casting process variants.
| Parameter | Scheme A (Initial) | Scheme B (Improved) |
|---|---|---|
| Mold material | CO₂ water glass sand | CO₂ water glass sand |
| Gating system | Top pour, single riser | Top pour, four risers + chills |
| Riser modulus (cm) | 2.0 | 2.5 (each riser) |
| Feeding distance (mm) | 250 (exceeds limit) | 120 (within limit) |
| External chill type | None | Steel chills at thick section |
| Observed sand casting defect | Shrinkage cavity at root | No shrinkage defect |
| Sand casting defect frequency (%) | 15–20 | 0–2 |
Despite solving the shrinkage issue, the sand casting process still presented other sand casting defect types such as sand inclusion, gas porosity, and slag entrapment. The cause was mainly due to the top-pour design, which allowed mold erosion and generated turbulence. The removal of large risers also caused problems: the riser cutting often led to insufficient machining allowance and required weld repair, which introduced additional risk of cracking. Consequently, I decided to transition to lost foam casting for mass production, as it promised better dimensional accuracy and fewer sand casting defect occurrences.
2. Lost Foam Casting: Process Design and Defect Prevention
Lost foam casting (LFC), also known as evaporative pattern casting, uses a polystyrene foam pattern embedded in unbonded dry sand under vacuum. The pattern vaporizes upon contact with molten metal, creating a cavity that is filled without a traditional mold parting line. I designed the LFC process for the bracket as shown in the schematic (refer to the figure below). The pattern was made from EPS (expandable polystyrene) with a density of 20–22 kg/m³. After assembly, the cluster was coated with a refractory slurry to prevent metal penetration and to control gas permeability.
During initial trials, I encountered a severe sand casting defect—metal backflow (ejection) from the pouring cup during casting, along with cold shuts and misruns. This was due to the rapid gasification of the foam pattern generating high back-pressure. To address this, I implemented several countermeasures:
- Pattern density control: I introduced a weighing step after pre-expansion to ensure uniform density. The target density was 20±1 kg/m³. Previously, operators relied on visual estimation, leading to variations of up to 5 kg/m³.
- Extended drying time: The coated pattern clusters were dried at 50°C for at least 24 hours to remove moisture and residual blowing agents.
- Vacuum regulation: I monitored the vacuum level in real-time during pouring. The initial vacuum was set at 0.06 MPa before pouring, but I noticed a drop to 0.02 MPa during filling, causing insufficient mold strength. I adjusted the vacuum pump to maintain a minimum of 0.04 MPa throughout the entire pour.
- Filter placement: To prevent slag inclusion (another common sand casting defect in LFC), I added a ceramic foam filter in the cross runner. The filter pore size was 10 ppi, effectively trapping dross and refractory particles.
The final LFC gating system adopted a bottom-gating design, which promoted laminar filling and reduced turbulence. The pouring temperature was carefully controlled at 1560–1580°C, with a pouring time of 12–15 seconds. I calculated the required pouring rate using the following formula based on the mass of metal and the gasification rate:
$$\dot{m} = \frac{Q_{gas}}{\Delta H_{vap}}$$
where Qgas is the heat required to vaporize the foam, and ΔHvap is the latent heat of vaporization of EPS (approximately 800 kJ/kg). For a casting mass of 45 kg, the total heat needed was about 36 MJ, requiring a minimum pouring rate of 2.4 kg/s to avoid premature freezing.
Table 2 compares the key parameters between the initial sand casting and the optimized lost foam casting process.
| Parameter | Sand Casting (Scheme B) | Lost Foam Casting (Final) |
|---|---|---|
| Mold material | CO₂ water glass sand | Dry silica sand (no binder) |
| Pattern type | Wood/metal pattern | EPS foam pattern (lost) |
| Gating style | Top pour, multi-riser | Bottom pour, single sprue + filter |
| Riser volume (% of casting) | ~30% | ~10% (small risers only) |
| Machining allowance (mm) | 3–5 | 1.5–2 |
| Surface finish (Ra, μm) | 12.5–25 | 6.3–12.5 |
| Typical sand casting defect types | Shrinkage, sand inclusion, slag | Cold shut, gas porosity (if vacuum low) |
| Sand casting defect rejection rate (%) | 8–12 | 1–3 |
| Weld repair required | Often (riser cutting marks) | Rarely |
3. Microstructure and Mechanical Properties
One of the key advantages of the lost foam process was the finer and more uniform microstructure achieved due to the slower cooling rate in dry sand. The final heat treatment (normalizing + tempering) produced a microstructure of tempered martensite with a grain size of ASTM 6–7. The gas content was also carefully monitored. Table 3 shows the oxygen and nitrogen levels measured over 20 heats.
| Element | Standard | Average | Min | Max |
|---|---|---|---|---|
| Oxygen (O) | ≤60×10⁻⁶ | 34.1×10⁻⁶ | 23.6×10⁻⁶ | 41.8×10⁻⁶ |
| Nitrogen (N) | — | 65.6×10⁻⁶ | 51.7×10⁻⁶ | 76.8×10⁻⁶ |
The oxygen content was well within the limit, indicating effective deoxidation during melting. The nitrogen level was acceptable for low-alloy steels. Mechanical properties from 151 heats (177 data sets) are summarized in Table 4. All values exceeded the specified minima.
| Property | Standard | Average | Min | Max |
|---|---|---|---|---|
| Yield strength (MPa) | ≥260 | 324.1 | 270 | 430 |
| Tensile strength (MPa) | ≥485 | 531.3 | 490 | 585 |
| Elongation (%) | ≥24 | 30.0 | 24 | 35.5 |
| Reduction of area (%) | ≥36 | 50.7 | 36 | 62.5 |
| Impact energy at -7°C (J) | ≥20 | 50.2 | 30 | 94 |
| Grain size (ASTM) | ≤6 or finer | 6–7 | 6 | 7 |
| Microstructure | Normalized 1–6 | Normalized 2–4 | — | — |
I also performed statistical analysis on the microstructure distribution. The normalized rating (according to standard charts) was: Grade 4: 42.4%, Grade 2: 37.3%, Grade 3: 10.2%, and mixed Grade 2–4: 10.1%. This distribution indicated a homogeneous microstructure with no abnormal phases. The grain size was consistently finer than ASTM 6, which contributed to the excellent impact toughness.
4. Process Control and Quality Assurance
After heat treatment (normalizing at 890°C followed by tempering at 600°C), all castings underwent shot peening twice to improve fatigue strength. Then, 100% non-destructive testing using magnetic particle and ultrasonic inspection was performed. No critical sand casting defect was found in the final batch. I also arranged dynamic and static load tests at an independent research institute. The results fully met the predetermined technical requirements, including a fatigue life of over 10⁶ cycles at 1.3× the nominal load.
To summarize the overall defect statistics, Table 5 lists the major sand casting defect types observed during the sand casting phase versus the lost foam phase.
| Defect Type | Sand Casting (Before) | Lost Foam Casting (After) |
|---|---|---|
| Shrinkage porosity / cavity | 8% | 0.5% |
| Sand inclusion | 3% | 0.2% |
| Slag inclusion | 2% | 0.1% |
| Gas porosity | 1% | 0.8% |
| Cold shut / misrun | 0.5% | 1.0% (initial) → 0.2% (optimized) |
| Hot tearing / cracking | 0.5% | 0.1% |
| Total rejection rate | 15% | 1.9% |
It is clear that the transition to lost foam casting dramatically reduced the overall sand casting defect rate. The primary reason was the elimination of mold joint lines and the use of unbonded sand, which minimized sand erosion and gas generation. However, the process required rigorous control of pattern quality, coating thickness, and vacuum parameters. I established standard operating procedures for each step, including pattern storage in a climate-controlled room (20°C, 40% RH) to prevent warpage.
5. Mathematical Modeling of Vacuum and Gas Removal
To better understand the vacuum requirements in lost foam casting, I developed a simplified model for gas evacuation. The gas generation rate from the foam pattern can be expressed as:
$$G = \rho_{foam} \cdot A_{front} \cdot v_{front} \cdot K$$
where G is the gas generation rate (m³/s), ρfoam is the foam density (kg/m³), Afront is the pyrolysis front area (m²), vfront is the front velocity (m/s), and K is the gas yield per unit mass of foam (typically 0.8–1.0 m³/kg for EPS). The vacuum system must be capable of removing this gas volume while maintaining a pressure differential across the mold:
$$\Delta P = \frac{ \mu \cdot L \cdot G }{ A_{mold} \cdot \kappa }$$
where μ is the gas viscosity, L is the mold thickness, Amold is the cross-sectional area, and κ is the permeability of the sand. Using typical values (μ = 1.8×10⁻⁵ Pa·s, L = 0.3 m, Amold = 0.2 m², κ = 5×10⁻¹¹ m²), I calculated that a vacuum of 0.04–0.06 MPa was sufficient. My earlier vacuum drop below 0.02 MPa caused insufficient gas removal, leading to backpressure and cold shuts—a classic sand casting defect in LFC. Once I stabilized the vacuum, the defect disappeared.
6. Cost and Productivity Analysis
Although lost foam casting required higher initial tooling investment (foam patterns and coating equipment), the long-term savings were significant. Table 6 compares the cost per casting for the two processes.
| Cost Item | Sand Casting | Lost Foam Casting |
|---|---|---|
| Pattern cost | 1.0 (reusable metal pattern) | 0.8 (foam pattern per casting) |
| Mold material | 0.3 | 0.1 (dry sand reused) |
| Energy (melting, heat treatment) | 1.2 | 1.0 (less riser metal) |
| Labor | 1.5 | 1.2 |
| Finishing (grinding, welding, inspection) | 1.8 | 0.5 |
| Total relative cost | 5.8 | 3.6 |
The total cost reduction was about 38%, primarily due to reduced finishing operations and lower scrap rates. Moreover, the productivity increased because lost foam casting eliminated the core-making and mold assembly steps.
7. Lessons Learned and Recommendations
Through this project, I gained valuable insights into the behavior of low-alloy steel castings and the impact of process selection on sand casting defect formation. Below are my key recommendations:
- Always perform a modulus analysis to ensure proper riser sizing and feeding distance. Use Chvorinov’s rule as a first approximation, then validate with simulation software.
- For sand casting, prefer bottom gating to minimize turbulence and reduce sand erosion, which leads to sand inclusion defects.
- When transitioning to lost foam casting, pay special attention to vacuum stability. Monitor the vacuum in real time and adjust pump capacity if necessary. A sudden vacuum drop is a frequent cause of sand casting defect such as cold shut.
- Control foam pattern density strictly. Use a weighing scale for each batch; do not rely on visual inspection. Density variation directly affects gas generation and casting soundness.
- Use ceramic filters in the gating system to trap dross and refractory particles. This significantly reduces slag-related sand casting defect.
- Optimize pouring temperature and time. For ZG30Mn, I found a pouring temperature of 1570±10°C and pouring time of 13±2 seconds to be optimal. Too low a temperature causes misruns; too high increases shrinkage and gas porosity.
- Perform full NDT on all castings, especially for safety-critical components. Magnetic particle inspection is effective for surface cracks, while ultrasonic testing can detect internal sand casting defect like shrinkage.
8. Conclusion
In conclusion, the improvement of the bracket casting from sand casting to lost foam casting was a successful journey that eliminated many persistent sand casting defect issues. The initial sand casting process suffered from shrinkage cavities due to inadequate riser design, but I corrected that with external chills and multiple risers. However, the sand casting process still had high scrap rates and required extensive weld repair. The transition to lost foam casting brought the defect rate down to below 2%, improved mechanical properties, reduced cost, and increased productivity. The key was rigorous process control: pattern density, coating, vacuum, and pouring parameters. The project was approved by the railway authority (analogous to the ministry level in the original context) after passing all static and dynamic load tests, and it entered mass production smoothly. This experience reaffirms that a thorough understanding of casting fundamentals, combined with systematic experimentation and data analysis, is the most effective way to combat sand casting defect and achieve high-quality steel castings.
I hope this detailed account of my process improvement journey serves as a practical guide for other foundry engineers facing similar challenges. Every casting defect is a learning opportunity—one that, if addressed properly, leads to better designs, stronger parts, and more efficient production.