Prediction Modeling of Low Alloy Steel Based on Chemical Composition and Heat Treatment Using Artificial Neural Network

Abstract

The utilization of machine learning methods in modern material science enables the design of more efficient and innovative materials. This research aims to develop a machine learning model using the Artificial Neural Network (ANN) algorithm to predict the mechanical properties of low alloy steel. The dataset used consists of 15 input variables and 2 output variables, namely Yield Strength (YS) and Tensile Strength (TS). In this study, three ANN architectures were designed and their performance was compared using evaluation metrics such as Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and R-squared. During the search for the best parameters for the ANN model, variations were made in the optimizer, learning rate, and batch size. The evaluation was conducted using cross-validation technique with k=10. The evaluation results indicate that the model with the best performance in predicting YS had MAE of 18.197, RMSE of 23.552, and R-squared of 0.969. For predicting TS, the model achieved MAE of 27, RMSE of 36.696, and R-squared of 0.907. The research results demonstrate that the ANN model can be used to predict the mechanical properties of low alloy steel based on alloy chemical composition and heat treatment temperature with reasonably high accuracy