Mechanics-Guided Genetic Programming Expression for Shear-Strength Prediction of Squat Reinforced Concrete Walls with Boundary Elements

Abstract

Squat reinforced concrete shear walls with boundary elements (SRCSW-BE) are used in special structures (e.g., nuclear facilities) to resist lateral seismic loads. However, several studies have demonstrated the inaccuracy of the relevant current shear strength prediction expressions (e.g., ASCE/SEI 43-05). Specifically, expressions originally developed based on empirical or experimentally calibrated analytical models (using different datasets) showed discrepancies when their predictions were compared with experimental results from other datasets. This situation is mainly attributed to the complex shear behavior and failure mechanisms of SRCSW-BE in addition to the wide ranges of their interdependent design characteristics. To address this issue, the current study utilizes genetic programming (GP), a form of artificial intelligence, to develop an elegant shear strength prediction expression using a dataset of 254 SRCSW-BE. Guided by mechanics, the key factors governing wall shear strength were first identified, and the GP-based expression was subsequently developed, trained, validated, and tested. The accuracy of the developed GP-based expression was assessed through different performance evaluation measures. The analyses showed that the developed expression can provide better predictions with significantly higher accuracy compared to other shear strength prediction expressions available in relevant design standards and literature. Further robustness assessment also demonstrated the conformity of the GP-based expression with known underlying behavior mechanics of SRCSW-BE, which, along with its elegant form, makes the developed expression adoption-ready by relevant design standards (e.g., ACI 318 and CSA A23.3). Overall, the current study is expected to demonstrate the ability of GP-based approaches in addressing other complex behaviors of structural components/systems and tackling relevant challenges pertaining to the latter’s behavior predictions.