Optimal sensor selection for Actuator Fault Detection and Identification Using Dynamic PCA-ANN in the Tennessee Eastman Process

Data driven fault detection methods are highly impressed by data quality gathered in distributed sensor network. So by increasing statistical properties in central processing unit, false alarm and miss detection will be reduced significantly in data based FDI method. In this paper a novel approach has been proposed for actuator fault detection in which a dynamic BP-ANN selects optimal set of sensors, then sensor data has been refined by PCA to highlight the variation caused by actuator fault. Simulation result achieved on Tennessee Eastman Process data justified the proposed approach performance.Keywords— Fault Detection and Identification, Dynamic PCA, Tennessee Eastman Process

کلیدواژه ها:

Fault Detection and Identification ، Dynamic PCA ، Tennessee Eastman Process