A machine-learning-based alloy design platform that enables both forward and inverse predictions for thermo-mechanically controlled processed (TMCP) steel alloys

Abstract

Predicting mechanical properties such as yield strength (YS) and ultimate tensile strength (UTS) is an intricate undertaking in practice, notwithstanding a plethora of well-established theoretical and empirical models. A data-driven approach should be a fundamental exercise when making YS/UTS predictions. For this study, we collected 16 descriptors (attributes) that implicate the compositional and processing information and the corresponding YS/UTS values for 5473 thermo-mechanically controlled processed (TMCP) steel alloys. We set up an integrated machine-learning (ML) platform consisting of 16 ML algorithms to predict the YS/UTS based on the descriptors. The integrated ML platform involved regularization-based linear regression algorithms, ensemble ML algorithms, and some non-linear ML algorithms. Despite the dirty nature of most real-world industry data, we obtained acceptable holdout dataset test results such as R2 > 0.6 and MSE < 0.01 for seven non-linear ML algorithms. The seven fully trained non-linear ML models were used for the ensuing ‘inverse design (prediction)’ based on an elitist-reinforced, non-dominated sorting genetic algorithm (NSGA-II). The NSGA-II enabled us to predict solutions that exhibit desirable YS/UTS values for each ML algorithm. In addition, the NSGA-II-driven solutions in the 16-dimensional input feature space were visualized using holographic research strategy (HRS) in order to systematically compare and analyze the inverse-predicted solutions for each ML algorithm.

A Generative Approach to Materials Discovery, Design, and Optimization

“…They observed that the nonlinear algorithms outperformed the linear algorithms, with many of the algorithms producing acceptable results. 131 …”

“…This presents the perfect opportunity to use machine-learning techniques to speed up the structural materials discovery process. , Researchers have had some success in structural material property prediction using conventional machine-learning algorithms. For example, Lee et al were able to predict both the ultimate tensile strength and yield strength of over 5000 thermomechanically controlled processed steel alloys using 16 chemical descriptors. They employed 16 machine-learning algorithms including linear regression, RFs, SVMs, and gradient-boosted trees.…”

“…They employed 16 machine-learning algorithms including linear regression, RFs, SVMs, and gradient-boosted trees. They observed that the nonlinear algorithms outperformed the linear algorithms, with many of the algorithms producing acceptable results …”

“…They observed that the nonlinear algorithms outperformed the linear algorithms, with many of the algorithms producing acceptable results. 131 As opposed to organic drug-like materials and even energy materials, structural materials have not been explored much using generative models with the exception of high-entropy alloys (HEAs). HEAs are engineered materials with five or more principal component elements, each with an atomic percent of 5−35 at %.…”

XGBoost automates the characterisation of reversibly actuating planar-flow-casted NiTi shape memory alloy foil

“…The system was tested using a blind validation procedure. The blind validation was highly successful showcasing the high potential and effectiveness of the new methodology in the eld of SMA foil-based applications [34][35][36][37][38][39][40].…”