PRINCIPAL COMPONENT ANALYSIS BASED APPROACH FOR FAULT DIAGNOSIS IN PNEUMATIC VALVE USING DAMADICS BENCHMARK SIMULATOR

This paper presents Artificial Neural Network (ANN) based classifier approach for fault diagnosis of pneumatic valve used in process industry. The proposed approach uses back propagation algorithm (BPN) to detect and diagnose the faults in pneumatic valve under normal and faulty operating conditions. Artificial Neural Network is trained using BPN algorithm to capture the relationship between the fault symptom and fault type. The required data for the development of ANN model were collected from the Development and Application of Methods for Actuator Diagnosis in Industrial Control Systems (DAMADICS) pneumatic valve simulator under normal and abnormal operating conditions. The performance of the proposed ANN model is improved by proposing suitable dimensionality reduction technique like Principal Component Analysis (PCA). The performance of the ANN model with PCA is compared with the performance of the ANN model without PCA. ANN model with PCA results in reduced mean square error during training and testing