Machine learning optimization strategy for the inelastic buckling modelling of un-corroded and corroded reinforcing plain bars

Many concrete structures and infrastructure reinforced with plain steel reinforcement were designed in the 1960s and 1970s. Such structures are deteriorating due to corrosion, which has exposed reinforced concrete (RC) components to the spalling of the concrete cover and the subsequent inelastic buckling of steel longitudinal rebars. This paper presents an optimization strategy based on genetic algorithms to simulate the cyclic response of steel plain bars. A refined finite element model of the steel bar is adopted for the numerical analysis. Thus, the genetic algorithm optimizes the main parameters of the most adopted constitutive model for steel rebars. The optimization procedure compares the available experimental tests with the numerical results by reducing a pre-defined objective function. Regression analyses are then performed for each calibrated model parameter. Therefore, taking advantage of the comprehensive numerical procedure, a parametric analysis is conducted to include the effects of corrosion on the inelastic buckling of plain rebars. The parametric study aims to develop accurate and adequate constitutive models for steel reinforcing bars in robust seismic analyses for RC structures. A case study of a typical old RC column with longitudinal plain bars under cyclic loading is presented