An improved incipient fault detection method based on Kullback-Leibler divergence. (August 2018)

• Record Type: Journal Article
• Authors: Chen, Hongtian; Jiang, Bin; Lu, Ningyun
• Is Part Of: ISA transactions. Volume 79(2018)
• Page Start: 127
• Page End: 136
• Publication Date: 2018-08
• Subjects: Incipient fault detection -- Kullback-Leibler (KL) divergence -- Principal component analysis (PCA) -- Electrical drive systems; Engineering instruments -- Periodicals; Engineering instruments; Periodicals; Electronic journals; 629.805
• DOI: 10.1016/j.isatra.2018.05.007
• Languages: English
• ISSNs: 0019-0578

Abstract

Abstract: This paper presents an improved incipient fault detection method based on Kullback-Leibler (KL) divergence under multivariate statistical analysis frame. Different from the traditional multivariate fault detection methods, this methodology can detect slight anomalous behaviors by comparing the online probability density function (PDF) online with the reference PDF obtained from large scale off-line data set. In the principal and residual subspaces obtained via PCA, a symmetric evaluation function is defined for both single variate and multivariate cases. The uniform form of probability distribution and fault detection thresholds associated with all eigenvalues are given. In addition, the robust performance is analyzed with respect to a wide range of Signal to Noise Ratio (SNR). Case studies are conducted with three types of incipient faults on a numerical example; combining with two nonlinear projections, the proposed scheme is successfully used for incipient fault detection in non-Gaussian electrical drive system. The results can demonstrate the superiority of the proposed method than several other methods. Highlights: No assumptions on unchanged mean or variance of incipient faults are needed in the proposed method. It is sensitive to the incipient faults, and is robust to noises as well. Precise limiting distribution of the evaluation function is given and proved by theoretical derivation, which can help determine rejection region in principal subspace accurately. In order to achieving a remarkable computational cost of probability density estimation, the KL divergence between reference probability density function (PDF) and online PDF are simply to calculate mean and variance in principal and residual subspaces. The proposed method can be applied and extended to non-Gaussian electrical systems. And its effectiveness is tested on a real experimental setup.