An Improved Sensor Fault Diagnosis Scheme Based on TA-LSSVM and ECOC-SVM

Monitoring the operational state of sensors promptly and the accurate diagnosis of faults are essential. This paper proposes an improved fault diagnosis scheme for sensors, which includes both fault detection and fault identification. Firstly, trend analysis combined with least squares support vector machine (TA-LSSVM) method is proposed and implemented to detect faults. Secondly, an improved error correcting output coding-support vector machine (ECOC-SVM) based fault identification method is proposed to distinguish different sensor failure modes. To demonstrate the effectiveness of the proposed scheme, experiments are conducted with an MTi-series sensor, and some comparisons are made with other fault identification methods. The experimental results demonstrate that the proposed fault diagnosis scheme offers an essential improvement with detection real-time property and better identification accuracy.