Applications of fault-tolerant supervisory control for discrete event systems

Failure-recovery supervisory control for discrete event systems (DES) is concerned with the recovery from faults that impair the desired system behaviors. Hereby, it is required to detect the occurrence of faults using fault diagnosis and then adapt the system operation such that the system can continue its operation with a potentially degraded performance. In addition, it is generally desired to resume the original system operation after a faulty component is repaired. As the first contribution of this thesis, a new fault diagnosis method is implemented and evaluated. Different from existing approaches, this method allows detecting the repeated occurrence of faults and is particularly useful when considering systems where faults can repeatedly occur after repair. As the second contribution of the thesis, a new method for the fault-recovery supervisory control is developed. Similar to existing approaches, this method assumes that the system follows its nominal behavior as long as the system is non-faulty. If a fault occurs, the system should at least obey a degraded specification until the desired behavior under fault is achieved in a bounded number of steps. As an extension, the proposed method also allows returning to the nominal behavior after system repair. In addition, our approach is based on the idea of modular supervisory control and hence scalable to large-scale systems. To the best of our knowledge, there is no other modular approach for the fault-recovery supervisory control. The applicability of the developed method is demonstrated by a medium-size laboratory model of a manufacturing system.