Fault Diagnosis in Continuous Dynamic Processes using Speech Recognition Methods

Faults or special events which occur occasionally in continuous processes give rise to dynamic patterns in a large number of process variables. Patterns arising from the same fault can exhibit different time durations, magnitudes and directions, yet a robust Fault Diagnosis method must be able to correctly classifying them. This paper presents an off-line Fault Diagnosis method based on Pattern Recognition Principles, applied to multivariate dynamic data. The method consist of a filtering-scaling step where the magnitude dependent information is removed, and a similarity assessment step via Dynamic Time Warping, a flexible pattern matching method used in the area of Speech Recognition. Case studies from the Tennessee-Eastman plant are used test the proposed method and the advantages, limitations and extensions of the approach are discussed.