UCLA

A large number of models to predict shear strength of structural walls have been proposed in the literature to replace models adopted in codes and standards. Evaluation of the predictive performance of new models relative to existing models is often difficult because the models were developed using different databases and a model may have substantially different performance (high, versus low variance) when evaluated against a different database. In addition, more complex models are expected to have less variance than relatively simple models and target performance metrics for models of different complexity do not exist. To address these issues, a study was conducted applying statistical and machine learning approaches to establish target model performance for different model complexities. The methodology is demonstrated by addressing the problem of assessing wall shear strength using a comprehensive database of 333 walls reported to have failed in shear.

Wall shear strength equations reported in the literature and used in building codes are assessed using a comprehensive database of reinforced concrete wall tests reported to have failed in shear. Based on this assessment, it is concluded that mean values varied significantly, and coefficient of variations were relatively large (> 0.30) and exceeded the target error for a code-oriented equation defined in the companion paper. Therefore, a methodology employing statistical and machine learning methods was used to develop a new equation with form similar to that currently used in ACI 318-19. The proposed equation is applicable to walls with rectangular, barbell, and flanged cross-sections and includes additional parameters not considered in ACI 318-19, such as axial stress and quantity of boundary longitudinal reinforcement. Parameter limits, e.g., on wall shear and axial stresses, and an assessment of the relative contributions to shear strength also are addressed. Finally, a reliability analysis is performed to study the relationship between probability of failure and strength reduction factor applicable to the proposed equation.