Single-Sensor Incipient Fault Detection

This study presents a new detection method based on wavelet transform and kernel principal component analysis (KPCA) for single-sensor incipient faults. Wavelet transform is first utilized to extract the featured information of process noise while the single-sensor fault is analyzed by KPCA subsequently. The fault data from flow meter and pressure sensor on a process device as well as other disturbing factors which may lead to inaccuracy have also been analyzed in this study. The experimental results demonstrate the accuracy and effectiveness of the proposed method to cope with single-sensor incipient faults.