Machine Learning-Assisted Thermo-Mechanical Stress Analysis of Steel IPE Beams Under Asymmetric Thermal Conditions

Abstract

Proper prediction of thermally induced stresses in the structural members is still a major task in structural engineering, particularly in the complex real-life scenarios. The research paper introduces a machine-learning-based framework revealing the stress analysis of IPE steel beams based on data. The preprocessing of the Materials and their Mechanical Properties dataset was performed by the inspection of data, the treatment of missing values, the encoding of labels, Minmax normalization and class balancing with SMOTE. A Decision Tree (DT) model was created to predict stress behavior and measured it on the basis of R², MSE, RMSE and MAE. The advanced performance of the proposed DT model was 99.4, MSE of 0.0001, RMSE of 0.0114 and a MAE of 0.0073 which means that the model has a strong predictive ability and low deviation to the actual values of the stress. It was found that DT was more effective than Linear Regression and Neural Network models in predicting the two variables in general. The strength and stability of the proposed method was further proven by regression, performance, and residual analysis, which prove the current method to be effective in estimating thermo-mechanical stress in steel structural elements.