Data driven approach to monitoring and fault detection in process control plants

Detecting faults and monitoring is a very important activity in process control plants for increasing the efficiency of the plant. Data driven approach is particularly helpful for fault detection, if the underlying mathematical model of the process is very complex. It can be used to create automatic systems which accurately predict whether the operating condition of the plant is normal or faulty. This paper compares different supervised learning algorithms, in order to detect fault in process control plant. The algorithms are tested in Matlab environment. Finally, all the models give satisfactory accuracy while detecting two different types of faults as well as normal operating condition.