Sensor fault diagnosis based on least squares support vector machine online prediction

In order to solve the challenging problem of diagnosis for sensor bias and drift faults, a method of sensor fault diagnosis based on the least squares support vector machine (LS-SVM) online prediction is proposed. In the paper, the real-time outputs of the sensor are made full use to establish LS-SVM prediction model. Through the residual which is obtained by comparing the outputs of LS-SVM prediction model and the actual output of the sensor, the real-time detection of the sensor faults can be achieved. Based on the residual sequence, the on-line identification of sensor bias fault and drift fault can be achieved as well. A model of sensor faults is established by the toolbox of matlab simulink in this paper, the simulation results show that the approach proposed can not only improve the accuracy and time efficiency of fault diagnosis, but also identify the type, size and the time of sensor faults occurred accurately.