Novel ARMF Integrated with Improved LSDA and Its Application in Fault Diagnosis

The industrial process is developing to be intelligent and complex, thus the process data presents high dimension, nonlinearity and highly coupled features. Facing these features, this paper proposes a signal pattern-matching fault diagnosis method based on the adaptive rank-order morphological filter (ARMF) integrated with improved locality sensitive discriminant analysis (LSDA) named ILSDA-ARMF. This proposed methodology first fully extracts the variable features related to the fault using the improved LSDA; then the data after dimensionality reduction (DR) is used for signal pattern matching by using ARMF to achieve fault classification. The main advantage of the improved LSDA is that the Mahalanobis distance considers the correlation between samples and their nearest neighbor points. Meanwhile, the Tennessee Eastman (TE) chemical process is experimented with to verify the performance of the proposed ILSDA-ARMF. The simulation results show that the method proposed in this paper achieves more satisfactory results compared with other related methods.