Detection of Single and Dual Incipient Process Faults Using an Improved Artificial Neural Network

Changes in the physicochemical conditions of process unit, even under control, may lead to what are generically referred to as faults. The cognition of causes is very important, because the system can be diagnosed and fault tolerated. In this article, we discuss and propose an artificial neural network that can detect the incipient and gradual faults either individually or mutually. The main feature of the proposed network is including the fault patterns in the input space. The scheme is examined through a sample unit with five probable occurring faults. The simulation results indicate that the proposed algorithm can detect both single and two simultaneous faults properly.