Industrial & Engineering Chemistry Research, Vol.51, No.34, 11207-11217, 2012
A Multiple-Time-Region (MTR)-Based Fault Subspace Decomposition and Reconstruction Modeling Strategy for Online Fault Diagnosis
Time-varying fault characteristics have not yet been addressed by conventional fault-reconstruction-based modeling methods, which could affect fault diagnosis performance. In the present work, the multiple-time-region (MTR) nature, that is, the multiplicity of fault characteristics along with the process evolution, is proposed and efficiently analyzed for fault diagnosis. First, an automatic time-region-division algorithm is developed that can partition the whole fault process into different local regions according to the changes in fault characteristics. Different local fault characteristics are thus analyzed by building different representative fault feature models in multiple time regions. Following the changing relationships between the fault and normal operation statuses, different fault reconstruction actions are finally taken in different time regions. By a proper time-region division, the proposed method can better model the time-varying fault behaviors and capture the different fault-to-normal reconstruction relationships for fault diagnosis. The feasibility and performance of the proposed method are illustrated with the Tennessee Eastman process, revealing enhanced fault understanding and improved fault diagnosis performance.