Fault tolerant control with indentified models: an experimental study

Abstract

Experimental evaluation of the active fault tolerant control scheme developed with state space models (innovations form) identified purely from input-output data is carried out in the present work. A key problem in fault diagnosis with identified models is the isolation of abrupt changes in unmeasured disturbances. A novel approach was developed to model and isolate such changes (classified as unknown input fault) with identified models. Simulation results for a heavy oil fractionator process (Shell Control Workshop challenge problem) was demonstrated. The efficacy of the developed FDI method and the fault tolerant control strategy were demonstrated. The present work experimentally evaluates the developed scheme on a laboratory scale two tank temperature control system for a variety of faults.