DNN Based Fault Tolerant Control of Nonlinear Structural Vibration with Actuator Faults

Abstract

Recently, structural control systems have been widely applied in civil engineering. However, there are still many important issues to be solved such as the design of controller with sensor and actuator faults. In this paper, an innovative fault tolerant controller using dynamic neural network (DNN) is developed for a nonlinear structure with actuator faults. Firstly, a DNN based identification method is proposed to estimate nonlinear structural responses with different actuator faults. Training algorithms of the neural network were derived utilizing the Lyapunov's first method. Secondly, based on the above DNN model, a fault tolerant controller is proposed to control the nonlinear structural vibration with actuator faults. Numerical simulation was conducted on a 3-story benchmark nonlinear structure to demonstrate the effectiveness of the proposed intelligent algorithms. Simulation results show that the proposed DNN based identification method was able to identify nonlinear structural responses with different actuator faults. Moreover, the fault tolerant controller was able to maintain stability and an acceptable degree of control performance not only when the structure was fault-free but also when there were actuator faults.