Industrial Process Fault Diagnosis Algorithm Research Based On Pca

Multivariate statistical analysis has been considered as an important method of industrial process fault diagnosis. Principal component Analysis (PCA) is one of the most widely application methods in the field of fault diagnosis. However, there is obvious deficiency in the traditional PCA method. Such as the principal components (PCs) approximately equal eigenvalues after dimensionless standardization, so it is more difficulty to select PCs effectively. The paper based on the theory of traditional PCA, and on the base of kinds of improved PCA, research on the PCA in detail. The paper gives following results:(1) Introduce the traditional PCA and the improved PCA based on the traditional PCA and wavelet analysis and RPCA in short.(2) In search of one new method to reduce the rate of false alarm. The paper suggested the self-adaption RPCA method based on self-adaption PCA and RPCA. And demonstrate by the TEP. But the result is not so satisfactory. Even worse than the existing methods such as self-adaption PCA.(3)The self-adaption RPCA is intensive studied, an new algorithm is proposed,that is the self-adaption RPCA based on wavelet, the single self-adaption RPCA method is not so satisfactory, the reason is that when relative transformation the noise is increased meantime. The wavelet analysis theory is used to remove noise, and combine the self-adaption RPCA to demonstrate by TEP. The percent of fault alarm is dramatic decline. Thereby the availability and feasibility of this new algorithm is verified by simulation.