Robot fault diagnosis based on wavelet packet decomposition and Hidden Markov Model

Abstract

Fault diagnosis has great significance in industrial robots. The Selective Compliance Assembly Robot Arm (SCARA) is a widely used robot in the industry. In this paper, SCARA robot is taken as an example to do fault diagnosis. The electromechanical actuator model of SCARA was built to simulate typical faults, laying the foundation for the diagnose work. Then based on Wavelet Packet Decomposition and Hidden Markov Model (HMM), a new fault diagnosis method is proposed. A maximum likelihood estimator is derived to evaluate the fault. Finally, experiment is done to verify the accuracy of the fault diagnosis method.