In the realm of metal casting, the accuracy of solidification and cooling simulations for sand castings is paramount for predicting defects, optimizing riser design, and ensuring final product quality. The fidelity of these simulations hinges critically on the precision of the thermophysical parameters assigned to the molding sand. These parameters, primarily thermal conductivity and specific heat, govern the heat extraction rate from the casting into the mold. However, molding sand is a complex, porous, and heterogeneous material whose properties vary significantly with composition, binder type, compaction density, and temperature. Relying on generalized literature values often leads to substantial errors in simulation outcomes, thereby diminishing their practical utility for guiding production. This study, therefore, was undertaken to determine the temperature-dependent thermophysical properties of the specific molding sand system used in our foundry for producing large steel sand castings. Through a combination of in-situ temperature measurements and iterative numerical simulation, I have derived accurate thermal conductivity and specific heat curves, the application of which significantly enhances the predictive capability of our casting process models.
The foundational principle behind this investigation is the heat conduction equation, most commonly described by Fourier’s law. For one-dimensional heat flow, which is a reasonable approximation near a flat casting interface, the law is expressed as:
$$ q = -k(T) \frac{\partial T}{\partial x} $$
Here, \( q \) represents the heat flux (W/m²), \( k(T) \) is the temperature-dependent thermal conductivity (W/m·K), and \( \frac{\partial T}{\partial x} \) is the temperature gradient. The energy storage within the sand is governed by its specific heat capacity, \( c_p(T) \) (J/kg·K), and density, \( \rho \) (kg/m³). The transient heat conduction equation, which forms the core of any solidification simulation, is given by:
$$ \rho c_p(T) \frac{\partial T}{\partial t} = \nabla \cdot (k(T) \nabla T) $$
It is evident that both \( k(T) \) and \( c_p(T) \) are crucial inputs. For sand castings, the mold’s ability to absorb and dissipate heat directly controls the solidification rate, grain structure, and the development of thermal stresses.
Materials and Experimental Methodology
The molding sand system studied is a multi-layer system typical for large steel sand castings. It consists of a facing layer of chromite sand and a backing system of silica sand. This design leverages chromite sand’s higher chilling power and resistance to metal penetration for the critical casting surface. The composition and particle size distribution of the sands are detailed in Table 1.
| Sand Type | SiO₂ (wt.%) | Fe₂O₃ (wt.%) | Cr₂O₃ (wt.%) | Others (wt.%) | Particle Size Distribution (wt.%) |
|---|---|---|---|---|---|
| Chromite Sand | ≤ 1 | 20-25 | ≥ 46 | ≤ 20 | 40-70 mesh: ≥ 75; 30-70 mesh: ≥ 85; 30-100 mesh: ≥ 95 |
| Silica Sand | ≥ 85 | ≤ 0.2 | ≤ 2 | ≤ 10 | 40-70 mesh: ≥ 75; 30-70 mesh: ≥ 85; 30-100 mesh: ≥ 95 |
The binder system was a urethane resin cured with an isocyanate hardener. For the chromite layer, the resin addition was 1.0 wt.% of the sand, with the hardener at 25 wt.% of the resin. For the silica sand backing, the resin addition was 0.65 wt.% of the sand, with the hardener at 40 wt.% of the resin. After mixing, the sand was compacted to an average hardness of 80-90 MPa for the overall mold, with the chromite layer achieving 2.5-4.5 MPa and the silica layer 1.5-2.5 MPa. The average bulk density of the compacted mold was between 2.5 and 2.9 kg/dm³. These parameters are critical as they influence the effective thermal contact and the porosity of the mold, both affecting heat transfer in sand castings.
The test casting was a large, low-carbon steel (ACM1506) mining machinery component. To accurately measure the mold’s thermal response, a simplified but representative geometry was also used for detailed analysis: a steel slab measuring 6400 mm × 500 mm × 700 mm, with an elliptical riser. The pouring temperature was 1575°C. K-type thermocouples were embedded in the mold at predetermined distances from the casting-mold interface during molding. For the main test, thermocouples were placed at distances of 50 mm, 100 mm, 150 mm, and 200 mm vertically above a specific point (Point A) on the casting surface. For the simplified slab casting, thermocouples were placed at 20 mm (Point A1) and 50 mm (Point A2) from the interface. Temperature data was recorded at a frequency of 0.5 Hz from the start of pouring until the mold cooled significantly.
Analysis of Experimental Temperature Data
The temperature-time histories recorded at various distances from the casting surface in sand castings provide the direct experimental basis for inverse calculation of thermal properties. The data is summarized conceptually in Table 2 and graphically in the subsequent analysis.
| Distance from Casting Surface (mm) | Initial Slow Heating Phase (min) | Rapid Heating Phase (min) | Approximate Maximum Temperature (°C) | Time to Reach Near-Maximum (min) |
|---|---|---|---|---|
| 50 | 0-90 | 90-200 | ~500 | ~200 |
| 100 | 0-400 | 400-600 | ~230 | ~600 |
| 150 | Continuous slow rise | Not distinct | ~120 | >750 |
| 200 | Continuous slow rise | Not distinct | ~80 | >750 |
The key observation is the profound influence of distance on the thermal transient. Points closer to the casting experience a sharper temperature rise after an initial latency period, reaching much higher peak temperatures. This indicates a significant thermal resistance within the mold material itself. The heat flux, \( q \), escaping the casting can be related to the temperature gradient measured between two points. For instance, between the 50 mm and 100 mm points, the average gradient over a time interval can be estimated. Using the fundamental heat equation, we can derive an effective thermal diffusivity, \( \alpha \), which combines conductivity, density, and specific heat:
$$ \alpha = \frac{k}{\rho c_p} $$
The slower temperature rise at greater distances is consistent with a relatively low thermal diffusivity for the molding sand compared to metals. This low diffusivity is why sand castings generally cool slower than equivalent permanent mold castings, influencing the microstructure and mechanical properties of the final component.
Inverse Determination of Thermophysical Parameters
To extract the specific functional forms of \( k(T) \) and \( c_p(T) \), an inverse parameter estimation technique was employed. The process involved using a commercial casting simulation software (Magma) to model the solidification of the test slab casting. The computational domain was discretized with a graded mesh, finer near the interface (5 mm cells around A1/A2) and coarser farther away (20 mm cells). The initial simulations used constant or generic temperature-dependent property sets from the software’s database.
The core of the inverse method is to iteratively adjust the sand’s property curves until the simulated temperature histories at points A1 and A2 match the experimentally measured curves. To ensure convergence and computational efficiency, a sequential approach was adopted. First, a reasonable constant value for specific heat was assumed (e.g., \( c_p = 1250 \, \text{J/(kg·K)} \)), and the thermal conductivity function \( k(T) \) was varied. The objective function, \( S \), to be minimized was the sum of squared errors between measured and simulated temperatures at all time steps for both points:
$$ S(k(T)) = \sum_{i=1}^{N} \left[ (T_{sim, A1}(t_i) – T_{exp, A1}(t_i))^2 + (T_{sim, A2}(t_i) – T_{exp, A2}(t_i))^2 \right] $$
Once a satisfactory \( k(T) \) curve was obtained, it was held constant, and the \( c_p(T) \) function was then iteratively adjusted to further minimize the error. This process was repeated until a self-consistent set of \( k(T) \) and \( c_p(T) \) was found that yielded an excellent match between simulation and experiment. The final derived properties are presented in Figure 1 (conceptualized in the formulas below) and tabulated in Table 3.
The thermal conductivity exhibited a distinct “V-shaped” trend with temperature. It decreased from room temperature up to a critical point around 400°C, after which it increased. This non-monotonic behavior can be attributed to several competing mechanisms. At lower temperatures, the dominant heat transfer mechanism is conduction through the solid sand grains and points of contact. As temperature rises, differential expansion may slightly reduce inter-grain contact pressure, potentially lowering effective conductivity. Furthermore, the pyrolysis of the organic resin binder in the temperature range of 200-400°C creates gases and voids, increasing thermal resistance. Above 400°C, radiative heat transfer across pores begins to become significant, and the conductive contribution from the gaseous phase inside pores increases with temperature, leading to a net rise in effective thermal conductivity. This behavior is crucial for accurate simulation of sand castings, especially during the later stages of cooling when the mold interior is at high temperature.
The specific heat capacity showed a continuously increasing trend with temperature, which is typical for most solid materials. The rate of increase was more pronounced below 600°C and became gentler above this temperature. This rise can be modeled empirically. A piecewise linear or polynomial function can be fitted to the data. For instance, in the lower temperature regime (20-600°C), the relationship can be approximated by a steeper linear function, while above 600°C, a shallower slope applies. The increase in \( c_p \) means the mold can store more energy per degree of temperature rise at higher temperatures, which moderates the cooling rate of the casting.
| Temperature Range (°C) | Thermal Conductivity, \( k \) (W/m·K) – Trend | Specific Heat, \( c_p \) (J/kg·K) – Trend | Proposed Empirical Correlation |
|---|---|---|---|
| 20 – 400 | Decreasing | Rapidly Increasing | \( k(T) \approx k_0 – a(T-20) \), \( c_p(T) \approx c_{p0} + b(T-20) \) |
| 400 – 600 | Increasing | Rapidly Increasing | \( k(T) \approx k_{400} + c(T-400) \), \( c_p(T) \approx c_{p400} + d(T-400) \) |
| 600 – 1000 | Increasing | Slowly Increasing | \( k(T) \approx k_{600} + e(T-600) \), \( c_p(T) \approx c_{p600} + f(T-600) \) |
Note: \( k_0, a, k_{400}, c, k_{600}, e, c_{p0}, b, c_{p400}, d, c_{p600}, f \) are positive constants determined from the inverse analysis. The exact values are proprietary to the foundry process but the trends are universally significant for sand castings simulation.
The final, optimized property curves can be represented by the following piecewise formulations for use in simulation codes:
For Thermal Conductivity (W/m·K):
$$ k(T) = \begin{cases}
k_{20} – \alpha (T – 293) & \text{for } 293 \leq T < 673 \, \text{K} (20 \leq T < 400^\circ\text{C}) \\
k_{min} + \beta (T – 673) & \text{for } T \geq 673 \, \text{K} (T \geq 400^\circ\text{C})
\end{cases} $$
where \( k_{20} \) is the conductivity at 20°C, \( \alpha \) and \( \beta \) are positive coefficients, and \( k_{min} \) is the minimum conductivity near 400°C.
For Specific Heat (J/kg·K):
$$ c_p(T) = \begin{cases}
c_{p,20} + \gamma (T – 293) & \text{for } 293 \leq T < 873 \, \text{K} (20 \leq T < 600^\circ\text{C}) \\
c_{p,600} + \delta (T – 873) & \text{for } T \geq 873 \, \text{K} (T \geq 600^\circ\text{C})
\end{cases} $$
where \( c_{p,20} \) is the specific heat at 20°C, \( \gamma \) and \( \delta \) are positive coefficients with \( \gamma > \delta \), and \( c_{p,600} \) is the value at 600°C.
Validation and Impact on Simulation Accuracy
The ultimate test of the derived parameters is their performance in predicting the thermal history of sand castings. Two simulation cases were compared for the slab casting: CASE1 used the newly derived \( k(T) \) curve but a constant \( c_p = 1250 \, \text{J/(kg·K)} \), while CASE2 used both the derived \( k(T) \) and \( c_p(T) \) curves. The results were stark. CASE2 produced temperature profiles at points A1 and A2 that were in remarkable agreement with the measured data, both in terms of the timing of temperature rises and the absolute peak temperatures. CASE1, on the other hand, showed significant deviations, particularly in the later stages of cooling where the specific heat’s temperature dependence becomes critical. This validation confirms that both parameters must be accurately characterized as functions of temperature to achieve reliable simulations.
The implications for the production of sand castings are substantial. With accurate thermophysical data, simulation software can now more reliably predict:
- Solidification Time and Risering: Accurate cooling curves allow for precise calculation of solidification time, enabling optimal riser size and placement to eliminate shrinkage porosity in critical sections of sand castings.
- Thermal Stresses and Distortion: The temperature gradient history within both the casting and the mold can be calculated more faithfully, leading to better predictions of residual stress and potential distortion, which is vital for large, complex sand castings.
- Microstructure Prediction: Accurate cooling rates directly feed into models for predicting grain size, phase distribution, and mechanical properties, linking process parameters directly to product performance in sand castings.
To further generalize the findings, a sensitivity analysis can be performed. The impact of varying key sand parameters like compaction density (which affects effective conductivity) and binder content on the derived \( k(T) \) and \( c_p(T) \) can be explored. This leads to a more comprehensive understanding of how process variations in preparing molds for sand castings affect heat transfer. A conceptual framework for this is presented in Table 4, which can guide future studies.
| Process Parameter | Effect on Bulk Density (\( \rho \)) | Expected Impact on \( k(T) \) | Expected Impact on \( c_p(T) \) (per unit mass) | Overall Impact on Thermal Diffusivity (\( \alpha \)) |
|---|---|---|---|---|
| Increased Compaction | Increases | Increases (better grain contact) | Minimal change | Increases |
| Higher Binder Content | Slight decrease (more organics) | Decreases at mid-temps (pyrolysis), complex effect | Increases (organics have different \( c_p \)) | Generally decreases |
| Use of Different Base Sand (e.g., Zircon) | Depends on sand | Can be significantly higher or lower | Depends on mineral | Major change |
Conclusions and Future Perspectives
This integrated experimental and numerical study has successfully determined the temperature-dependent thermophysical properties—thermal conductivity and specific heat—of a production-grade molding sand system used for steel sand castings. The key findings are:
- The distance from the casting-mold interface profoundly affects the thermal transient within the mold, with regions within 100 mm experiencing rapid and high-temperature rises, underscoring the mold’s role as the primary thermal resistance in sand castings.
- The thermal conductivity exhibits a non-monotonic “V-shaped” dependence on temperature, decreasing up to approximately 400°C due to factors like binder pyrolysis and then increasing thereafter likely due to enhanced radiation and gas-phase conduction within the pores.
- The specific heat capacity increases monotonically with temperature, with a more rapid increase below 600°C and a gentler slope above this temperature.
- The inverse parameter estimation method, using measured in-mold temperatures and iterative simulation, is a powerful and practical tool for characterizing mold materials specific to a foundry’s practice.
- Implementing these accurate property curves into casting simulation software dramatically improves the predictive accuracy of thermal histories, solidification patterns, and related defect forecasts for sand castings, providing a robust tool for process optimization.
For future work, the methodology should be extended to other molding material systems, such as greensand, chemically bonded sands with different resins, and molds with insulating or exothermic sleeves. Furthermore, developing a standardized database of thermophysical properties for various sand-binder combinations would be immensely valuable for the wider casting industry. The continuous improvement of simulation accuracy is fundamental to advancing the quality, efficiency, and reliability of sand castings across all industrial sectors. The relationship between mold properties and casting quality can be encapsulated in a holistic performance index, \( \Pi \), for a sand casting process:
$$ \Pi = \int_{0}^{t_{solid}} \frac{k_{eff}(T_{interface}(t)) \cdot A}{\rho_{sand} c_{p,sand}(T_{avg}(t)) \cdot V_{casting}} \, dt $$
where \( k_{eff} \) is the effective interface conductivity, \( A \) is the interfacial area, \( \rho_{sand} \) and \( c_{p,sand} \) are mold properties, \( V_{casting} \) is the casting volume, and \( t_{solid} \) is the total solidification time. Accurate knowledge of \( k(T) \) and \( c_p(T) \) allows for the precise evaluation of such indices, enabling the scientific design of molding materials for specific sand castings applications.
In summary, the mastery of molding sand thermophysics is not merely an academic exercise but a critical engineering endeavor that bridges the gap between virtual simulation and physical reality in the production of high-integrity sand castings. The data and methodology presented here provide a template for foundries to calibrate their own simulation environments, leading to reduced prototyping costs, shorter development times, and enhanced product quality for a vast array of sand castings components.