Data-Driven ANN-Based Predictive Modeling of Mechanical Properties of 5Cr-0.5Mo Steel: Impact of Composition and Service Temperature

Abstract

The mechanical properties of steel are intricately connected to their composition and service temperature. Predicting these properties across different work temperatures using traditional statistical methods, algorithms, and equations is highly challenging due to these complex interdependencies. To address this, we developed an artificial-neural-network (ANN) model to elucidate the relationships between composition, temperature, and mechanical properties of 5Cr-0.5Mo steels. Our model demonstrated high accuracy, with minimal percentage errors in predicting YS, UTS, and El (%)-3.5%, 0.97%, and 1.9%, respectively. The ANN predictions are realistic and closely match the experimental results. We propose an easy-to-use model's GUI to predict steel composition to achieve desired properties at any temperature. The ANN model's findings offer valuable insights for researchers and designers, aiding in developing steel components with optimized properties. This technique is expected to significantly enhance the planning of practical experiments and improve material performance overall.