Seismic Shear Design of RC Structural Walls

Abstract

ACI 318–19 Eq. 18.10.4.1, which is used to predict shear strength of reinforced concrete (RC) walls, has remained essentially the same since it was introduced into ACI 318–83. Although this equation accounts for the influence of concrete compressive strength, web horizontal bars, and aspect ratio, it does not include other important parameters such as axial load, vertical reinforcement, and cross-section shape. Existing equations adopted in codes and standards, as well as those reported in the literature, have been derived using limited or biased datasets and thus do not generally provide accurate predictions when evaluated against comprehensive databases that include walls with a wide range of characteristics. To address these issues, a framework based on the generic steps of Machine Learning (ML) is proposed and used to establish target model performances for different model complexities, which are expressed in terms of mean and coefficient of variance of the true-to-predicted ratios. Lastly, an approach that combines ML and statistics methods is used to derive a predictive model with a code-oriented format and interpretable parameters. A comprehensive database of 333 RC shear-controlled walls is a key part of the study.