Fault diagnosis strategy for startup process based on standard operating procedures

Startup is one kind of common transitions in chemical process industry. Startup is normally executed by plant operators who follow predefined standard operating procedures so that the plant settles down to a different steady state. Therefore operator errors in addition to process faults occur frequently during startup. The existing diagnosis methods did not consider operating errors and did not make full use of the knowledge of SOP. This paper presents a fault diagnosis strategy for startup process based on SOP. The approach proposes behavior observer combined with dynamic PCA to estimate process faults and operator errors at the same time. This method has been implemented on UniSim and found successfully detect faults during the startup of debutanizer.