Subspace-based Fault Detection and Isolation Methods - Application to Vibration Monitoring
Abstract
We address the problem of detecting and isolating faults modeled as changes in the eigenstructure of a linear dynamical system. The purpose of the paper is to describe and analyze new fault detection and isolation algorithms, based on recent stochastic subspace-based identification methods and the statistical local approach to the design of detection algorithms. The application to vibration monitoring of mechanical structures and rotating machines is discussed. A conceptual comparison is made with another detection algorithm based on the instrumental variables identification method, and previously proposed by two of the authors.