OPTIMAL CONTROL UNDER DISTRIBUTED SENSOR FAILURE DETECTORS.

This paper presents an extension to a new technique for sensor failure detection in optimally controlled processes, proposed by Saif and Villaseca (1986). Their detection scheme uses a bank of observers to estimate the states of the system and a quadratic function of states, namely the expression for the optimal cost at any time instant, to detect sensor failures. A simpler but more restrictive alternative to this scheme is presented along with the effect of small and known plant parameter variations on the detection technique. Using a sensitivity analysis, it is shown that it is possible to detect sensor failures in spite of small plant parameter variations without redesigning the observers and the overall control system