Multi-parameter identification of earthquake simulation shaking table based on BP neural network

Abstract

Since the model parameters of the shaking table exist in a non-linear form, this leads to distortion of the reproduced waveforms and can even lead to bias in the ground vibration test results. Therefore, the selection of the controller is particularly critical. Multi-variable (MVC) controllers are often used in shaking table control, to improve the control effect of MVC controllers. In this paper, a multi-parametric (BP-MVC) controller based on BP neural network is proposed. The BP neural network is applied to the multi-parameter (MVC) controller to identify the shaking table model, adjust the parameters in real-time, accelerate the convergence speed, and reduce the system error. The simulation results show that the correlation coefficient (CC) of the BP-MVC controller is greater than 0.985, and the root-mean-square error (RMSE) and mean absolute error (MAE) are less than 0.04 and 0.25, respectively, in a nonlinear, time-varying hydraulic system. This suggests that the BP-MVC controller has a better control performance and parameter adaptivity, which can provide a reference for the subsequent ground vibration tests.