In modern manufacturing, particularly within the foundry industry, the quest for high-quality, dimensionally accurate cast components is perpetual. Among the various sand casting processes, resin sand casting stands out for its ability to produce complex molds with excellent dimensional stability and surface finish. As a major supplier of engine castings, my company has extensively adopted the cold-box resin sand casting process to meet stringent quality and productivity demands. This process utilizes a gas-cured resin binder system, offering rapid cycle times and high core strength. However, despite its advantages, this method is frequently plagued by a persistent and costly surface defect known as burning-on, or sand adhesion. This defect manifests as a layer of sand grains tenaciously bonded to the casting surface, compromising its integrity and aesthetics.
The impact of burning-on defects in resin sand casting extends beyond mere appearance. It adversely affects the dimensional precision of critical features, leading to increased labor and cost for post-casting cleaning, fettling, and rework. In severe cases, inadequate cleaning can result in non-compliance with cleanliness specifications, while aggressive grinding to remove the adhered sand may induce micro-cracks, ultimately leading to warranty claims and financial losses. Therefore, developing a systematic understanding and a robust control strategy for this defect is paramount for enhancing the efficiency and reliability of the resin sand casting process.
Mechanism of Burning-On Defects in Resin Sand Casting
Burning-on is a complex phenomenon arising from high-temperature interactions at the metal-mold interface. It is broadly classified into two primary types: mechanical penetration and chemical bonding.
Mechanical Penetration (Mechanical Burning-On): This occurs when molten metal, under the influence of metallostatic pressure and its inherent fluidity, infiltrates the intergranular pores of the sand mold or core. Upon solidification, the metal mechanically interlocks with the sand grains, forming a strongly bonded composite layer. The propensity for mechanical burning-on increases with higher mold permeability, larger sand grain size, elevated pouring temperature (which reduces metal viscosity), and insufficient mold surface density or coating integrity.
Chemical Bonding (Chemical Burning-On): This is a more intricate process predominant in resin sand casting systems, especially with silica sand. At high temperatures, components of the molten metal (particularly iron oxides, FeO) can react with the silica (SiO₂) in the sand and the decomposed products of the organic resin binder. These reactions can form low-melting-point silicate phases, such as fayalite (2FeO·SiO₂), which have a glassy nature. These liquid silicates wet and cement the sand grains together and to the casting surface, creating a hard, dense, and often glossy layer that is extremely difficult to remove. The chemical reactions are exacerbated by higher pouring temperatures and the presence of oxides.
In our specific case, involving the production of N-series engine blocks, chronic burning-on was observed in challenging areas like water jacket passages, tappet bores, and sharp “bull-horn” features. Metallographic examination of sectioned castings confirmed the defect to be primarily of the chemical bonding type, characterized by a fused silicate layer between the cast iron and the sand.
Systematic Problem Analysis and Factor Selection for DOE
Addressing defects in resin sand casting often involves troubleshooting individual process parameters. However, burning-on is a multivariate problem where factors interact in non-linear ways. Isolated adjustments frequently yield suboptimal or inconsistent results. A holistic, data-driven approach is required. The Design of Experiments (DOE) methodology provides a powerful framework for this purpose, allowing for the efficient investigation of multiple factors and their interactions with a minimal number of trials.
Prior to designing the experiment, a thorough screening of the production system was conducted to isolate noise variables. The base sand (washed silica sand) was confirmed to have consistent properties: SiO₂ content ~91%, angularity factor of 1.3, pH 5.6, three-screen concentration 93.1%, loss on ignition (LOI) ≤0.5%, and sintering point of 1460 °C. Operator skill was stable, and equipment was well-maintained. Environmental factors were also accounted for. This screening allowed the team to focus on the most likely critical process variables influencing the chemical burning-on mechanism in our resin sand casting operation.
Three key control factors were identified for investigation:
- Pouring Temperature (Tp): Directly influences metal fluidity and the intensity of the thermal/chemical attack on the mold surface. A higher temperature increases both the potential for metal penetration and the kinetics of silicate-forming reactions.
- Coating Thickness (Ct): The refractory coating acts as a primary barrier between the molten metal and the resin-bonded sand. Its thickness is critical in preventing both metal infiltration and substrate chemical reactions.
- Resin Addition Level (Ra): Expressed as a weight percentage of the sand. This controls the strength and the gas evolution characteristics of the core. Higher resin content increases the amount of carbonaceous material available for decomposition, potentially affecting the local atmosphere and the sand’s resistance to metal penetration.
The objective levels for these factors in the resin sand casting study were set based on production experience and process boundaries, as detailed in Table 1.
| Factor | Symbol | Low Level (-1) | High Level (+1) | Center Point (0) |
|---|---|---|---|---|
| Pouring Temperature | Tp | 1420 °C | 1460 °C | 1440 °C |
| Coating Thickness | Ct | 0.4 mm | 0.6 mm | 0.5 mm |
| Resin Addition | Ra | 1.1 % | 1.3 % | 1.2 % |
Design of Experiments (DOE) Implementation and Procedure
A full factorial design with three factors at two levels would require 2³ = 8 runs. To assess curvature and improve the estimate of pure error, one center point replicate was added. Furthermore, to enhance the statistical power and reliability of the conclusions in this resin sand casting study, the experiment was conducted with two replicates (blocks) of the full factorial design, bringing the total number of experimental runs to 17 (8×2 + 1). A statistical power analysis confirmed this design had an 86.9% probability of detecting significant effects, which is considered adequate for industrial experimentation. The response variable was quantitatively defined as the Percentage of Burning-On Area (PBA) on the critical surfaces of the cast engine block. The orthogonal array for the designed experiment is presented in Table 2, along with the observed results.
| Run Order | Block | Tp (°C) | Ct (mm) | Ra (%) | PBA (%) | Primary Defect Location |
|---|---|---|---|---|---|---|
| 1 | 1 | 1420 | 0.4 | 1.1 | 12 | Water Jacket, Bull-horn |
| 2 | 1 | 1460 | 0.4 | 1.1 | 15 | Water Jacket, Tappet, Bull-horn |
| 3 | 1 | 1420 | 0.6 | 1.1 | 9 | Water Jacket |
| 4 | 1 | 1460 | 0.6 | 1.1 | 13 | Water Jacket, Tappet, Bull-horn |
| 5 | 1 | 1420 | 0.4 | 1.3 | 13 | Water Jacket, Bull-horn |
| 6 | 1 | 1460 | 0.4 | 1.3 | 14 | Water Jacket, Bull-horn |
| 7 | 1 | 1420 | 0.6 | 1.3 | 8 | Water Jacket Passage |
| 8 | 1 | 1460 | 0.6 | 1.3 | 12 | Water Jacket, Bull-horn |
| 9 | 2 | 1420 | 0.4 | 1.1 | 11 | Water Jacket, Bull-horn |
| 10 | 2 | 1460 | 0.4 | 1.1 | 16 | Water Jacket, Tappet, Bull-horn |
| 11 | 2 | 1420 | 0.6 | 1.1 | 10 | Water Jacket |
| 12 | 2 | 1460 | 0.6 | 1.1 | 14 | Water Jacket, Bull-horn |
| 13 | 2 | 1420 | 0.4 | 1.3 | 13 | Water Jacket, Tappet, Bull-horn |
| 14 | 2 | 1460 | 0.4 | 1.3 | 13 | Water Jacket, Tappet, Bull-horn |
| 15 | 2 | 1420 | 0.6 | 1.3 | 9 | Water Jacket |
| 16 | 2 | 1460 | 0.6 | 1.3 | 14 | Water Jacket, Tappet, Bull-horn |
| 17 | – | 1440 | 0.5 | 1.2 | 11 | Water Jacket, Bull-horn |
Statistical Analysis and Model Development
The data from the resin sand casting experiment was analyzed using Analysis of Variance (ANOVA). The initial model included the three main effects (Tp, Ct, Ra), their two-way interactions (Tp*Ct, Tp*Ra, Ct*Ra), and the block effect. A stepwise model reduction procedure was employed, removing non-significant terms (p-value > 0.05) to arrive at a more parsimonious and predictive model. The final ANOVA confirmed that the main effects of Pouring Temperature (Tp) and Coating Thickness (Ct), as well as their interaction (Tp*Ct), were statistically significant. The Resin Addition (Ra) and its interactions were found to be non-significant within the studied range for this particular burning-on response in resin sand casting.
The final fitted regression model, derived from the standardized factor levels, is expressed as:
$$ P_{BA} = \beta_0 + \beta_1(T_p) + \beta_2(C_t) + \beta_{12}(T_p \cdot C_t) $$
Where:
$$ P_{BA} $$ is the predicted Percentage of Burning-On Area.
$$ \beta_0 $$ is the overall mean intercept.
$$ \beta_1, \beta_2, \beta_{12} $$ are the coefficients for the main effects and interaction.
$$ T_p $$ and $$ C_t $$ are the coded variables for Pouring Temperature and Coating Thickness, respectively.
Using the actual factor units from the resin sand casting process, the empirical model can be written as:
$$ P_{BA}(\%) = 80.8 – 0.0437 \cdot T_p(^{\circ}C) – 371 \cdot C_t(mm) + 0.250 \cdot T_p(^{\circ}C) \cdot C_t(mm) $$
The analysis of the main effects plot clearly shows the direction of influence: PBA increases with higher Pouring Temperature and decreases with greater Coating Thickness. The significant interaction between Tp and Ct indicates that the effect of coating thickness depends on the pouring temperature, and vice-versa. For instance, the benefit of increasing coating thickness is more pronounced at higher pouring temperatures. The statistical metrics for the final model (R-squared, Adjusted R-squared, and Predicted R-squared) were in close agreement, indicating a good model fit with reliable predictive capability for the resin sand casting process.
Process Optimization and Validation
With the validated model, the next step was to determine the optimal process window to minimize the burning-on defect in resin sand casting. The optimization goal was set to minimize PBA. A practical upper specification limit of 11% was defined, beyond which cleaning costs and risks become unacceptable. Using the response optimizer and contour plots generated from the model, the ideal combination of factors was identified.
The optimization analysis conclusively pointed to the following settings:
- Pouring Temperature (Tp): 1420 °C (the lower bound of the study).
- Coating Thickness (Ct): 0.6 mm (the upper bound of the study).
- Resin Addition (Ra): Set to the center point of 1.2%, as it was not a significant factor for defect minimization, allowing for a balance between core strength and cost.
The model predicted an optimal PBA with a 95% confidence interval between 7.8% and 10.0%, and a prediction interval between 6.4% and 11.4%. This represents a substantial improvement over the higher values observed in many of the initial trials (e.g., 14-16%).
To confirm the optimization results in a real production setting for resin sand casting, a verification run was conducted. Over three separate production batches, a total of 120 N-series engine blocks were cast using the recommended parameters: Tp = 1420 °C, Ct = 0.6 mm, Ra = 1.2%. The results, summarized in Table 3, confirmed the model’s effectiveness.
| Batch | Pouring Temp. (°C) | Coating Thickness (mm) | Resin Addition (%) | Observed PBA (%) |
|---|---|---|---|---|
| 1 | 1420 | 0.6 | 1.2 | 7 |
| 2 | 1420 | 0.6 | 1.2 | 10 |
| 3 | 1420 | 0.6 | 1.2 | 9 |
| Average | 1420 | 0.6 | 1.2 | 8.67 |
The average PBA of 8.67% fell well within the predicted confidence interval and was significantly below the 11% target. Visual and tactile inspection showed a dramatic reduction in the severity and frequency of burning-on in the water jacket passages, tappet bores, and bull-horn areas. The defect, when present, was less tenacious and easier to clean. This successful validation confirmed that the DOE-derived model provided a reliable and effective framework for solving the chronic burning-on problem in this specific resin sand casting application.
Conclusions and Future Perspectives
This study successfully demonstrated the application of a structured Design of Experiments (DOE) methodology to diagnose and mitigate a persistent burning-on defect in a cold-box resin sand casting production process. The systematic approach moved beyond trial-and-error and enabled a scientific understanding of the factor effects and interactions.
The key findings are:
- For the studied resin sand casting system and defect type (chemical burning-on), Pouring Temperature and Coating Thickness were identified as statistically significant factors, with a significant interaction between them. The resin addition level, within the tested range, did not show a significant direct effect on the burning-on area percentage.
- The derived empirical model, $$ P_{BA}(\%) = 80.8 – 0.0437 \cdot T_p – 371 \cdot C_t + 0.250 \cdot T_p \cdot C_t $$, accurately described the relationship between these key process inputs and the defect response.
- The optimal process parameters for minimizing burning-on were determined to be: a lower pouring temperature (1420 °C) combined with a maximized refractory coating thickness (0.6 mm), while maintaining a standard resin addition level of 1.2%.
- Production validation trials confirmed the optimization, reducing the average burning-on area percentage to 8.67%, which represented a major improvement in casting surface quality and projected downstream cleaning efficiency.
This research underscores the critical importance of a integrated process view in resin sand casting. While factors like sand quality and resin type form the foundation, the dynamic conditions at the metal-mold interface—controlled by pouring temperature and coating integrity—are often the decisive elements in defect formation.
The scope of this study was intentionally focused on three primary variables. Future work in resin sand casting optimization could expand this model by incorporating other potentially influential factors such as sand compaction density (hardness), coating refractory composition, pour time, and specific part geometry characteristics. More advanced DOE techniques, like Response Surface Methodology (RSM), could then be employed to map the process landscape with greater resolution, potentially identifying non-linear optimal points. Furthermore, linking these process parameters to quantitative measures of cleaning effort (e.g., time, tool wear) or directly to the mechanical properties of the interfacial layer could provide even more comprehensive cost-benefit insights for the resin sand casting industry.