Research on sensor fault diagnosis technology of dynamic positioning vessel based on filter and Support Vector Machine

Precise and reliable position and heading information are critical to the dynamic positioning vessel. The position and heading information relies on the system's measuring sensor. In order to solve the sensor fault problem, in this paper, the states estimation of the system is proposed by using the Extended Kalman Filter (EKF). The faults are detected by the residual error generated by the predicted value output by EKF and the actual measured value output by sensors, and then the faults are classified by the method combining Support Vector Machine (SVM) and binary tree. Additionally, the parameters of SVM are optimized by Particle Swarm Optimization (PSO) to make the best multi-classification. In this paper, the simulation of the method is carried out by using the gyro-compass in the dynamic positioning vessel. The experimental results show that the method can effectively be used in the fault diagnosis of the sensor.