Forecasting hysteresis behaviours of magnetorheological elastomer base isolator utilizing a hybrid model based on support vector regression and improved particle swarm optimization

Abstract

Due to its inherent hysteretic characteristics, the main challenge for the application of a magnetorheological elastomer- (MRE) based isolator is the exploitation of the accurate model, which could fully describe its unique behaviour. This paper proposes a nonparametric model for a MRE-based isolator based on support vector regression (SVR). The trained identification model is to forecast the shear force of the MRE-based isolator online; thus, the dynamic response from the MRE-based isolator can be well captured. In order to improve the forecast capacity of the model, a type of improved particle swarm optimization (IPSO) is employed to optimize the parameters in SVR. Eventually, the trained model is applied to the MRE-based isolator modelling with testing data. The results indicate that the proposed hybrid model has a better generalization capacity and better recognition accuracy than other conventional models, and it is an effective and suitable approach for forecasting the behaviours of a MRE-based isolator.