Identification of contributing variables using kernel-based discriminant modeling and reconstruction
Expert Systems with Applications: An International Journal
Dynamic independent component analysis approach for fault detection and diagnosis
Expert Systems with Applications: An International Journal
A novel fault diagnosis system using pattern classification on kernel FDA subspace
Expert Systems with Applications: An International Journal
Computers and Industrial Engineering
| Hi-index | 0.01 |
Data-driven fault diagnosis technique has exhibited its wide applications in industrial process monitoring. However, how to recognize fault pattern based on process dataset is still a difficult problem in data-driven fault diagnosis field. In this paper, a novel nonlinear fault recognition method is proposed based on statistics kernel principal component analysis similarity factor (SKPCASF), which combines nonlinear similarity factor and statistics pattern analysis. Principal component analysis similarity factor (PCASF) is firstly reviewed which measures the similar degree of two datasets by comparing their principal component subspaces. In order to deal with nonlinear characteristics of process dataset, kernel principal component analysis (KPCA) is applied to build a nonlinear similarity factor, referred to as KPCA similarity factor (KPCASF). Moreover, for well utilizing the statistical information of data variables, statistics pattern analysis is used to compute data statistics for substituting for the original measured variables in fault recognition. Based on the statistics, a new similarity factor method called as statistics KPCA similarity factor (SKPCASF) is lastly built for fault pattern recognition. Simulations on a simple nonlinear system and the benchmark Tennessee Eastman process show that the proposed SKPCASF method is more effective than PCASF and KPCASF in terms of fault pattern recognition performance.