An advanced machine learning framework for predicting and experimentally validating Nb and Ta effects on the mechanical behavior of high-entropy alloys with reduced experimental dependency

Abstract

Abstract This study employs machine learning (ML) models such as Decision Tree (DT), Extra Tree (ET), K-Nearest Neighbors (KNN), Support Vector Regressor (SVR) and Lasso to predict how Ta and Nb content effects the mechanical behavior of the CoCrFeNi HEAs at diverse compositions. ET and KNN models developed as the most effective forecasters, accomplishing high R 2 values of 0.985 and 0.984 respectively. The valuation of the mechanical behaviour with experimental data is very precise with the help of these models. The advanced methods like Hybrid Computational Variable Creation Method (HCVCM), Symbolic Variable Creation Method (SVCM), Data Structure Method (DSM) was further used which has enhance the prediction performance in terms of R 2 value upto 0.991 for ET and 0.99 for KNN model. A novel stress-strain behaviour was produced for a new composition with novel Ta and Nb content, where ET and KNN demonstrated robust R 2 values of 0.984 and 0.98, respectively. The strategy streamlines exploration of HEAs by cutting down extensive trial-and-error experiments, thereby conserving significant resources in terms of time, cost, and energy.