Machine learning-based estimation of energy dissipation capacity of RC shear walls

In current practice, the seismic design of reinforced concrete shear walls primarily relies on the wall shear strength as well as proper reinforcement detailing to ensure sufficient ductility. However, experimental results have shown that the seismic (input) demands are met by the energy dissipation capacity in the structural members; therefore, the influence of hysteretic behavior on seismic behavior is considerable. An energy-based approach that reflects the effect of repeated loads in seismic performance – which is typically neglected in the seismic codes – would serve as a supplemental index in the design process. With this motivation, a predictive model is proposed to estimate the energy dissipation capacity of reinforced concrete shear walls. A comprehensive database consisting of 312 shear walls tested under cyclic loading and a widely used and powerful machine learning method, namely Gaussian Process Regression (GPR), is used to investigate the effects of wall design parameters (e.g. wall geometry, reinforcement details) on energy dissipation capacity and to develop the predictive model as a function of such parameters. Eighteen design parameters are shown to influence the energy dissipation, the most important of which are identified by applying sequential backward elimination and feature selection methods. The ability of the proposed model to make robust and accurate predictions is validated based on unused data with a prediction accuracy (the ratio of predicted/actual values) of around 1.00 and a coefficient of determination (R²) of 0.93. The outcomes of this study are believed to contribute to the wall design process by (i) defining the most influential wall properties on the seismic energy dissipation capacity of shear walls and (ii) providing predictive models that can enable comparisons of different wall design configurations to achieve higher energy dissipation capacity.