Shear capacity assessment of perforated steel plate shear wall based on the combination of verified finite element analysis, machine learning, and gene expression programming

Abstract

In this study, two formulations have been suggested for the calculation of the shear capacity of stiffened steel plate shear wall (SSPSW) containing two rectangular openings by integrating verified finite element results, machine learning (ML) models, and gene expression programming. In this regard, a comprehensive nonlinear finite element analysis was conducted, which included 200 records with various values. Considered variables are the thickness and aspect ratio of the steel infill plate, yield strength of the infill plate and boundary frame as well as the ratio of opening area to the total area of the infill plate. Three machine learning (ML) models were employed namely Stochastic Gradient Descent (SGD), Decision Tree (DT), and Random Forest (RF). These models were evaluated on the test data, resulting in \(\:{R}^{2}\)scores of 0.96, 0.90, and 0.95, respectively. Among these models, SGD demonstrated superior performance and was identified as the best model for this dataset. Based on the SGD model, an equation was derived to predict the shear capacity of the shear wall. Furthermore, using gene expression programming (GEP) model, an accurate formulation was proposed to calculate the shear capacity of SSPSW system, which led to \(R^{2}\) of 0.98 on the same test data used in the ML models. In addition, by employing the SHapley values technique, the contribution of each characteristic to the final prediction values was explained. This technique showed that the prediction values were significantly influenced by the feature (L/h), while the mechanical characteristics of steel plate and boundary frame had the least impact. Overall, the study underscored the efficacy of the SGD model in predicting the shear capacity of the studied shear walls and provided insights into the relative importance of different features in the prediction process.