LQG CONTROL WITH RECONFIGURABLE STATE ESTIMATOR UNDER SENSOR AND ACTUATOR FAILURES

Occurrences of sensor / actuator failures can lead to significant degradation in the closed loop performance when conventional feedback controllers are used. In this work, we propose an active failure tolerant LQG (FTLQG) control scheme, which employs model based fault diagnosis for on-line reconfiguration of state estimator on diagnosis of failures. Generalized likelihood ratio (GLR) method proposed in the literature is extended for diagnosis of sensor failures. Recurrence relationships are derived for diagnosing sensor failures, which are amenable for on-line computations. The efficacy of the proposed FTLQG scheme is demonstrated using simulation and experimental studies on a laboratory scale heater-mixer setup.