Recursive Correlative Statistical Analysis Method with Sliding Windows for Incipient Fault Detection

Abstract

This paper proposes a new combination of correlative statistical analysis and the sliding window technique to detect incipient faults. Compared with the existing monitoring methods based on principal component analysis and transformed component analysis, the combination fully uses the information from the process and quality variables. The sliding window, however, inevitably increases the computation burden due to the repeated window calculations. Therefore, a recursive algorithm is proposed in this paper, which has been shown to have less calculation complexity. Furthermore, a randomized algorithm is proposed to determine the width of sliding window. A numerical example and the thermal power plant process are presented to show the effectiveness and advantages of the proposed method.