The exceptional combination of high comprehensive mechanical properties, good wear resistance, excellent machinability, and relatively low manufacturing cost has solidified ductile iron’s position as an ideal material for critical components like railway wheels. The performance and reliability of these components are intrinsically linked to their as-cast microstructure, which is predominantly governed by the thermal history experienced during solidification and subsequent cooling. For heavy-section ductile iron castings with significant variations in wall thickness, such as full-size wheels, the differing cooling rates at various locations lead to microstructural and consequently, mechanical property gradients. This non-uniformity presents a significant challenge for manufacturing consistency. Therefore, understanding and predicting the microstructure distribution within a complex casting like a wheel is paramount for process optimization and quality assurance.
This study focuses on establishing a methodology for predicting the as-cast microstructure of full-size ductile iron wheels. The approach integrates controlled physical experiments with finite element numerical simulation. By investigating the fundamental relationship between cooling conditions and microstructure in simplified castings, and then applying this knowledge through simulation to a complex industrial component, we aim to provide a predictive tool that can reduce development time and cost for ductile iron castings.
1. Introduction: The Critical Link Between Cooling Rate and Microstructure in Ductile Iron Castings
The final microstructure of ductile iron castings is a direct consequence of its solidification and solid-state transformation kinetics. During the eutectic solidification, the cooling rate significantly influences graphite nodule count, size, and shape (nodularity). A higher cooling rate generally promotes a larger number of smaller, more spherical graphite nodules. Subsequently, during the eutectoid transformation, the cooling rate determines the ratio of ferrite to pearlite in the matrix. Slower cooling allows for greater carbon diffusion from the austenite to the existing graphite nodules, favoring ferrite formation. Faster cooling suppresses this diffusion, leading to a predominantly pearlitic matrix or even the formation of carbides.
For a monolithic casting like a wheel, regions near chills or thin sections (like spokes) cool rapidly, while thermal centers (like the hub or rim core) cool slowly. This inherent variation makes predicting the final microstructure distribution complex. Numerical simulation software like ProCAST can accurately calculate the temperature field throughout the casting process. However, translating this thermal data into a quantitative microstructural prediction requires robust empirical or semi-empirical models that link thermal parameters (e.g., local solidification time, cooling rate) to microstructural features.
Previous studies have established qualitative trends, but quantitative models applicable to specific alloy compositions and casting geometries are often lacking. This work aims to bridge that gap by:
- Experimentally determining the quantitative effect of cooling rate on microstructure using sand-cooled ductile iron (SCDI) samples of varying thickness.
- Developing mathematical prediction models for graphite characteristics and matrix phase fractions based on simulated thermal parameters.
- Applying these models to a simulated full-size wheel casting to predict its as-cast microstructure distribution.
2. Experimental Methodology and Numerical Simulation Setup
2.1 Material and Sample Design for Thermal Analysis
The base material used throughout this investigation is a ferritic-pearlitic ductile iron. Its chemical composition is detailed in Table 1.
| C | Si | Mn | P | S | Cu | Ni | Mg | Ceq |
|---|---|---|---|---|---|---|---|---|
| 3.68 | 2.16 | 0.13 | 0.05 | 0.02 | 0.61 | 0.63 | 0.05 | 4.40 |
To systematically study the effect of cooling rate, a series of Sand-Cooled Ductile Iron (SCDI) samples were designed. The key variable was the mold wall thickness, which directly governs the cooling intensity. Six different mold thicknesses were employed: 3 mm, 12.5 mm, 25 mm, 50 mm, 75 mm, and 100 mm. All other sample dimensions were kept constant and sufficiently large to ensure that heat transfer occurred predominantly through the thickness direction. High-strength furan resin sand was used for molding to withstand the expansion pressures during ductile iron solidification and minimize mold wall movement, thereby reducing the risk of shrinkage porosity. Side risers were used to ensure adequate feeding.
K-type thermocouples were embedded at the geometric center of each SCDI sample to record the temperature-time curves during casting. Due to furnace capacity limitations, the samples were produced in two separate melts, labeled SCDI-1 and SCDI-2.
2.2 Numerical Simulation of a Full-Size Wheel Casting
The target component is a full-size ductile iron wheel for urban rail transit, with a diameter of 840 mm, a height of 170 mm, and an approximate weight of 350 kg. Considering the heavy sections at the hub and rim connected by thinner spokes, a simultaneous solidification casting process was designed. The gating system is a closed shower type with eight ingates evenly distributed around the wheel spokes. Chills are placed around thermal centers (hub and rim), and small risers are positioned on the top surfaces of the hub and rim to compensate for shrinkage.
The finite element simulation was performed using ProCAST software. The thermophysical properties of the ductile iron were calculated using JMatPro software. The mold material was defined as silica sand (SAND_Silica from the ProCAST database), and the chills were defined as gray iron. The following boundary conditions and parameters were set:
- Interfacial heat transfer coefficient (HTC): 700 W/(m²·°C) for casting/mold and mold/chill interfaces; 1000 W/(m²·°C) for casting/chill interface.
- Pouring temperature: 1400 °C.
- Pouring rate: 19.29 kg/s.
- Ambient cooling: Natural air convection with an HTC of 100 W/(m²·°C).
- Initial temperature for all materials: 20 °C.
3. Results and Discussion: From Experimental Data to Predictive Models
3.1 Influence of Cooling Intensity on As-Cast Microstructure
The recorded cooling curves for the SCDI samples confirmed the designed gradient in cooling intensity. As the mold thickness decreased, the local solidification time (LST) – the duration of the eutectic plateau – shortened significantly, and the actual eutectic temperature was depressed. For the 12.5 mm sample, the cooling was so rapid that a clear eutectic plateau was not detectable. The 3 mm sample was too thin for reliable thermocouple measurement but exhibited the fastest cooling.
Metallographic analysis of the SCDI samples revealed clear trends. The graphite morphology and matrix structure were quantitatively characterized, with the results summarized in Table 2.
| Mold Thickness (mm) | Nodularity (%) | Avg. Nodule Diameter (μm) | Nodule Count (nodules/mm²) | Ferrite Content (%) | Pearlite Content (%) |
|---|---|---|---|---|---|
| 3 | 93.75 | 12.29 | 549.51 | 11.61 | 88.39 |
| 12.5 | 91.42 | 17.85 | 332.14 | 18.23 | 81.77 |
| 25 | 89.15 | 22.34 | 185.62 | 37.94 | 62.06 |
| 50 | 86.02 | 26.78 | 124.33 | 39.08 | 60.92 |
| 75 | 83.55 | 28.91 | 89.45 | 35.12 | 64.88 |
| 100 | 80.71 | 30.67 | 74.77 | 33.85 | 66.15 |
The data shows that as mold thickness decreases (cooling intensity increases):
- Graphite Nodularity increases.
- Average Nodule Diameter decreases.
- Nodule Count increases substantially.
- Ferrite Content first remains relatively stable for thicker molds (>25 mm) and then drops sharply for the fastest-cooled samples. Conversely, Pearlite Content follows the opposite trend.
These observations are consistent with solidification and phase transformation theory. Higher undercooling during eutectic solidification increases nucleation rate, leading to more, smaller nodules. The shorter time for growth also aids in maintaining sphericity. The shift from a ferritic to a pearlitic matrix at very high cooling rates is due to the suppressed diffusion of carbon during the eutectoid transformation, favoring the cooperative growth of ferrite and cementite (pearlite) over the divorced eutectoid transformation that forms ferrite envelopes around graphite.
3.2 Development of Microstructure Prediction Models
To create a predictive framework, the experimental SCDI geometries were simulated in ProCAST to obtain precise thermal parameters. From the simulated cooling curves, two key parameters were extracted for each sample:
- Local Solidification Time (ts): Directly related to the solidification cooling rate.
- Cooling Rate (Rc): The average cooling rate between 800°C and 900°C, relevant for the eutectoid transformation.
The relationship between these simulated thermal parameters and the experimentally measured microstructural features was then analyzed. Using least-squares regression, empirical prediction models were developed. The goodness of fit was evaluated using the Adjusted R-squared (R²adj) statistic, defined as:
$$ R^2_{adj} = 1 – (1 – R^2)\frac{n-1}{n-k-1} $$
where \(R^2\) is the coefficient of determination, \(n\) is the number of observations, and \(k\) is the number of predictors.
3.2.1 Graphite Morphology Prediction Model
Graphite characteristics showed a strong correlation with the local solidification time \(t_s\). The derived models are:
- Nodularity (\(Nod\)): $$ Nod(t_s) = 81.38 + 14.84 \times e^{(-t_s / 95.63)} $$ (Adjusted R² ≈ 0.991)
- Average Nodule Diameter (\(D_{avg}\)): $$ D_{avg}(t_s) = 30.29 – 18.20 \times e^{((15.27 – t_s) / 143.99)} $$ (Adjusted R² ≈ 0.998)
- Nodule Count (\(N_c\)): $$ N_c(t_s) = 78.50 + 628.12 \times e^{(-t_s / 60.40)} $$ (Adjusted R² ≈ 0.998)
These equations indicate that nodularity and nodule count decrease exponentially with increasing local solidification time, while the average diameter increases, asymptotically approaching a maximum value.
3.2.2 Matrix Microstructure Prediction Model
The ferrite content (\(F_{c}\)) was correlated with the cooling rate \(R_c\) during the eutectoid transformation range. The model, which includes a data point for near-equilibrium cooling, is:
$$ F_c(R_c) = 12.66 + 86.44 \times e^{(-R_c / 0.095)} $$
(Adjusted R² = 0.926). The pearlite content (\(P_c\)) is simply derived as:
$$ P_c(R_c) = 100\% – F_c(R_c) $$
This model captures the trend where ferrite content drops sharply as cooling rate increases beyond a certain threshold, leading to a predominantly pearlitic matrix.
3.3 Microstructure Prediction for a Full-Size Ductile Iron Wheel
With the prediction models established, the next step was to apply them to the simulated full-size wheel. The thermal simulation provided the distribution of local solidification time \(t_s\) and cooling rate \(R_c\) throughout the wheel casting. The results along the central cross-section are particularly illustrative.
The thermal analysis revealed that locations near chills (rim and hub surfaces, spokes) had short local solidification times (75-150 s), while thermal centers (hub core, rim core) exhibited much longer times. In the subsequent cooling stage, the chills’ effect diminished, and heat extraction proceeded radially outward from the geometric center, leading to a gradual increase in cooling rate from the center towards the outer surfaces.
Applying the prediction models to these thermal field results yields the anticipated as-cast microstructure distribution for the ductile iron wheel casting:
- Graphite Morphology: The nodularity across the wheel section is predicted to be high, ranging between 80% and 90%. The average nodule diameter varies from 20 to 30 μm, with the largest nodules found in thermal centers. The nodule count inversely correlates with diameter, ranging from 75 to 300 nodules/mm². Overall, the designed simultaneous solidification process promotes a relatively uniform distribution of graphite nodules.
- Matrix Microstructure: The ferrite content is predicted to vary from approximately 13% to 41%, with pearlite making up the balance (59% to 87%). A clear radial gradient is predicted: the ferrite content decreases (and pearlite increases) from the wheel’s interior towards the outer surface of the rim, which experiences the fastest cooling during the eutectoid transformation. The core regions, cooling more slowly, retain a higher ferrite fraction.
4. Conclusions and Implications for Foundry Practice
This study successfully demonstrates an integrated methodology for predicting the as-cast microstructure in complex, heavy-section ductile iron castings. The key findings are:
- Cooling Rate Mastery: The design of sand molds with varying wall thickness is an effective method for generating a wide range of controlled cooling conditions, effectively simulating the diverse thermal histories found within a single large ductile iron casting like a wheel.
- Quantitative Relationships: Clear, quantifiable relationships exist between simulated thermal parameters (local solidification time, cooling rate) and final microstructural features (graphite nodularity, size, count, and matrix phase fractions) for a given ductile iron composition.
- Predictive Model Utility: Empirical prediction models derived from simplified experiments can be effectively coupled with finite element simulation results to forecast the microstructure distribution in industrial-scale ductile iron castings. This provides a powerful tool for virtual prototyping and process optimization.
- Wheel Casting Prediction: For the specific full-size wheel casting analyzed with a simultaneous solidification design, the models predict a uniformly good graphite morphology throughout. The matrix microstructure, however, exhibits a predictable gradient, transitioning from a more ferritic matrix in the slowly cooled interior to a more pearlitic matrix near the rapidly cooled outer surfaces, particularly at the rim.
The practical implication of this work is significant. Foundries producing heavy-section ductile iron castings can adopt this approach to anticipate and control microstructure-related properties. By adjusting the casting process design—such as the placement and size of chills, risers, and the gating system—based on simulation-guided predictions, manufacturers can achieve more consistent and reliable performance in critical components, reducing scrap rates and development cycles. Future work will focus on validating these predictions with physical wheel castings and extending the models to include mechanical property forecasts based on the predicted microstructure.