A governing equation based fault detection and diagnosis algorithm and its application in a chemical plant

This paper describes a governing equation fault detection and diagnosis algorithm, which is based on Kramer's (1987) approach and extended it to deal with multiple faults. The algorithm is specifically designed to overcome disturbances and uncertainties in the process which might result in diagnostic instability. The approach is used to detect and/or diagnose the following faults: sensor faulty status of fails-high and fails-low, sticking valve, plant leaks, and other unanticipated or novel faults. The algorithm was installed on a distributed control system of a PET plant in Dec. 1995, and continues to be used thus demonstrating the success of the fault detection and diagnosis approach.