Prediction of Mechanical Properties of Cr-Mn-N Austenitic Stainless Steel Using Machine Learning Approach

In view of designing new alloys as well as optimizing the mechanical properties of annealed Cr-Mn-N alloys, a machine learning model has been developed using large industrial data. In this work, continuous multi-output regression models were developed to predict mechanical properties such as yield strength, tensile strength and elongation of Cr-Mn-N austenitic stainless steel based on the chemical composition, thickness and grain size. In the present study, the performance of several model such as KNN, ETR, XGboost, RF etc. has been compared. The ETR model outperformed other models and thereby it has been selected for further analysis. The relative importance of different parameters affecting the mechanical properties of the steels have been investigated. While the yield strength and ultimate tensile strength could be predicted very well but the percentage elongation showed slight deviation from actual data. The deviation has been explained in the light of metallurgical fundamentals. Further, the developed model was validated using separate data points outside the training data. The optimized model will be useful in designing and optimizing the composition of Cr-Mn-N austenitic stainless steels.