Evolutionary computing-based models for predicting seismic shear strength of RC columns

Abstract

A number of regression-based models have been proposed to quantify the seismic shear strength of reinforced concrete (RC) columns. However, most of these models suffer from a high degree of uncertainty as a result of the limited datasets used in the development and/or the classic approaches used to capture the nonlinear interrelationships between the shear strength and influencing factors. To address these issues, this study harnesses the power of multi-gene genetic programming (MGGP), guided by mechanics, to identify the primary influencing factors and subsequently develop efficient shear capacity predictive models for rectangular and circular RC columns. Published comprehensive datasets for the shear strength of cyclically-loaded RC columns were compiled and employed to develop the MGGP-based models. The efficiency of the developed models was assessed, and their performances were also compared with that of relevant existing predictive models. The results demonstrated the ability of the mechanics-guided MGGP approach to produce more accurate and conssistant predictive models, compared to those available in relevant design standards and literature, that can describe the complex shear behavior of RC columns under cyclic loading.