Fault Detection and Isolation in Dynamic Systems Using Principal Component Analysis - Application to an Heating System Benchmark

Abstract

Recently, fault detection and process monitoring using principal component analysis (PCA) were studied intensively and largely applied to industrial processes. Principal component analysis may reduce the dimensionality of plant models significantly by exposing linear dependencies among the variables. By using the equivalence between PCA and parity space (Gertler and McAvoy, 1997), all the powerful concepts of analytical redundancy may be transferred to PCA. Hence, using partial PCA (PPCA) (Gertler et al., 1999), a fault isolation scheme is formed according to a properly designed incidence matrix. This paper proposes a decomposition of a global system in different subsystems by means of PCA framework, in respect to the IFATIS european project (EU-IST-2001-32122) hierarchy. The presented FDI approach is illustrated with an heating system benchmark and future directions on Fault Detection and Isolation/Fault Tolerant Control are discussed.