Study the Efficiency of the XGBoost Algorithm for Squat RC Wall Shear Strength Prediction and Parametric Analysis

Abstract

Squat-reinforced concrete (RC) shear walls with an aspect ratio of less than two are considered effective structural members, where shear is the dominant failure mechanism. Squat shear walls are widely used in nuclear power plants and building construction and feature optimal cost and outstanding performance, due to their lateral strength and high rigidity to resist lateral loads. However, since the accurate evaluation of the shear strength of squat shear walls must meet the design specifications, its calculation may be very complex, challenging, and inaccurate using experimental and theoretical equations due to many influential and overlapping design factors, so it takes more time and higher cost to determine it. This study uses machine learning (ML) methods to build a shear strength prediction efficient model for squat RC walls to address these issues. First, a huge dataset of 1424 RC squat wall test specimens gathered from the literature is utilized for developing an ML model, by employing XGBoost, to predict the shear strength. Results verified that the XGBoost model had the best accuracy and least error while assessing the squat walls' strength at shear. Moreover, an XGBoost optimum algorithm fared better than the empirical models based on mechanics, with a 99.2% accuracy. Finally, to prove that the model can identify the most important variables that significantly affect the shear strength, parameter and sensitivity analyses were performed and the results showed that the wall length is the factor that contributes most to the ultimate shear strength of the squat shear wall as a percentage (7.62%), followed by the yield strength. For the web as a ratio. (6.88%), concrete strength (6.75%), reinforcement ratio information (6.56%), and geometric properties (6.01%), while the axial load represents the smallest contribution, reaching (4.16%).

Keywords

• Machine Learning (ML)
• Squat RC walls
• Shear strength
• XGBoost Model
• Empirical Model