Data-Driven-Based Mechanical Analysis and Design Processusing Supervised Machine Learning: Steel Cantilever Studycase

Abstract

This study deals with the integration of machine learning (ML) techniques into structural engineering to enhance prediction accuracy of structural analysis and optimize design processes. It sets out to explore supervised data-driven based decision making of structural analysis and design process. The research adopts classification-based and regression-based machine learning techniques to predict key parameters of a cantilever beam using input data. Specifically, the Decision Tree Classifier is used to predict the I-section design (W-shape) based on features such as modulus of elasticity, beam length, and applied load. Regression models are then employed to estimate vertical displacement, using the same input features in addition to W-shape. A case study was conducted to validated the approach through three steps implementation. First, a dataset of 2,601 entries was generated using ETABS, where structural parameters were computed based on varying input values. Then, training the models discussed previously. Finally, Models were evaluated using Mean Squared Error (MSE) and R-squared (R²) metrics. This research contributes to the advancement of structural analysis methodologies in civil engineering. Notably, machine learning models exhibit promising performance in accurately predicting displacement and I-section design, thereby offering insights into the behavior of cantilever beams under different conditions.