Wavelet - Based principal component analysis for process monitoring with experimental application

PCA methods for Fault Detection (FD) use data collected from a steady-state process to monitor T² statistics with a fixed threshold. The fixed threshold method causes false alarms in the transient state of the system. To overcome the false alarms arising from the transient state the combination of the fixed and adaptive threshold (T_comb) based PCA method is used. But the measurements noise results in very high T_comb. This causes to produce the missing fault signal components. The problem can be solved by filtering the noisy measurement signals. The wavelet transform has been widely used in signal de-noising, due to its extraordinary time-frequency representation capability. In this paper, a new PCA method based on wavelet is proposed to overcome false alarms which occur in the transient states according to changing process conditions and the missing data problem. The proposed method is implemented and validated experimentally on an electromechanical system. Experimental results illustrate the much better fault detection performance of the proposed method in comparison with classical PCA monitoring and process controlling charts.