A reconstruction strategy for fault diagnosis based on fault direction

Abstract

In the present work, a new iterative selection strategy and reconstruction modeling method is proposed for fault diagnosis. The proposed algorithm can extract those informative fault directions which are responsible for the concerned alarming monitoring statistic. First, the fault effects are decomposed in two different monitoring subspaces, principal subspace (PCS) and residual subspace (RS). Then, the significance of each fault direction in both PCS and RS is evaluated by the proposed iterative fault directions selection strategy, and those informative ones are chosen for reconstruction modeling. The proposed method provide an effective way of modeling the fault effects so as to correct the fault more efficiently. Its feasibility and performance are illustrated with numerical simulation and the Tennessee Eastman (TE) benchmark process.