FAULT DETECTION AND DIAGNOSIS BY SUPPORT VECTOR MACHINES: APPLICATION TO VINYL-CHLORIDE-MONOMER PROCESS

Monitoring process status and identifying process operational faults are essential for improving the process safety in petrochemical plants that interactions between various process streams and units are associated. This paper presents a deployment of a support vector machine technique for detecting and identifying operational fault cases with a case study of a vinyl chloride monomer plant. An integrated simulation environment between MATLAB and UniSim Design dynamic simulator is utilized for evaluating the performance of the proposed fault detection and identification framework. Under the real-time software-in-the-loop simulation, the confusion matrix results and receiver operating characteristics supported that the proposed framework provides high accuracy of fault classification.