Evaluation Method of Fault Diagnosability Based on Recursive Plot and Tensor Decomposition

Diagnosability reflects the difficulty of diagnosing system faults, so diagnosability evaluation of the system is a prerequisite for fault diagnosis. Taking the signals of test points into consideration, a fault diagnosability evaluation method based on a recurrence plot and tensor decomposition is proposed. The phase space reconstruction technique is used to graphically represent the signals of measured points in different states to form a recurrence plot. Features are extracted through analyzing the recurrence plot, which are then regarded as the original signal features. Through tensor decomposition, the similarity between signal features in different states is calculated as a basic measure of fault diagnosis difficulty. Simulation experiments show that, compared with the previous D-matrix diagnosability evaluation method that only considers the system model, the newly proposed method has effectiveness and objectivity. © 2023 China Ordnance Society. All rights reserved.