Machine Learning Prediction of Stacking Fault Energy in Steel Alloys Based on Chemical Composition

Volume 17, Issue 1 (March 2026), pp. 2–10

Download PDF

@article{jocse-17-1-1,
  author={Ikponmwosa J. Iyinbor and Ken-ichi Nomura and Paulo S. Branicio},
  title={Machine Learning Prediction of Stacking Fault Energy in Steel Alloys Based on Chemical Composition},
  journal={The Journal of Computational Science Education},
  year=2026,
  month=mar,
  volume=17,
  issue=1,
  pages={2--10},
  doi={https://doi.org/10.22369/issn.2153-4136/17/1/1}
}

Stacking fault energy (SFE) is a critical parameter in the design of steels with desirable mechanical properties such as strength, ductility, and strain-hardening rate. SFE influences secondary deformation mechanisms like Transformation Induced Plasticity (TRIP) and Twinning Induced Plasticity (TWIP). This work involves creating a machine learning model to classify steel alloys into low, medium, or high SFE categories, aiding in the prediction of secondary deformation behaviors. Data from literature containing experimental and theoretical SFE values for various steel alloy compositions were compiled and preprocessed, resulting in a dataset of 374 observations. Using this dataset, several machine learning models, including Feedforward Neural Network (FFNN), K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Random Forest (RF), Gradient Boosting Regressor (GBR), and CatBoost Regressor (CAT), and Adaptive Boost Regressor (ADB) were trained and evaluated for SFE prediction accuracy. Two models, SVM and RF, emerged as the top-performing models. To enhance accuracy and reduce misclassification, threshold probabilities were applied, allowing fuzzy classification when model uncertainty was high. Validation against literature data showed strong agreement between predictions and reported SFE values. This study provides valuable insights into predicting SFE and guiding the development of austenitic steel alloys with tailored properties.