Prediction of creep degradation in Fe-Cr-Ni single-crystal alloys for high-temperature applications: a molecular-dynamics and machine-learning approach

In this paper, we investigate the creep-deformation behavior of Fe-Cr-Ni single-crystal alloys, a crucial factor in the longevity and safety of materials in high-temperature applications. Using molecular-dynamics (MD) simulations, we generate the creep-strain data on the creep behavior of Fe-Cr-Ni single-crystal alloy. To predict creep curves under various temperatures and stress conditions, we employ random forest (RF) and convolutional neural network (CNN) models. These models are trained, tested, and validated on creep data at 300 K, 750 K, 950 K, and 1150 K, achieving deviations within 20% of simulation values. The RF model demonstrates strong predictive capabilities, with correlation coefficients of 0.96, 0.96, 0.94, and 0.98 at the respective temperatures. In contrast, the CNN model shows correlation coefficients of 0.92, 0.99, 0.99, and 0.99. The results of this investigation show that both models are capable of accurately predicting creep behavior. As compared to the CNN model, which performs better at higher temperatures and with larger datasets, the RF model works better at lower temperatures and with smaller datasets. These results enhance our understanding of creep properties and improve predictive modeling under varying conditions.