Fault-tolerant control and intelligent instrumentation

Fault-tolerance is a highly desirable property for any control system. Fault-tolerant controllers typically rely on some sort of fault detection algorithm. In this paper it is argued that intelligent instrumentation represents an economically attractive alternative to high level FDI methods. Besides the economic aspect, SEVA devices have the further advantage of providing a measure of the level of reliability of an instrument and not just a healthy/faulty flag like most FDI methods. Then, the fault accommodation algorithm should not just switch between sensors or actuators when a fault develops; instead, it should minimize the effect of measurement errors and actuator inaccuracies/limitations for all levels of instrumentation performance. This, coupled with the fact the SEVA information is standardized, allows to define a new, general and mathematically rigorous formulation for the design of fault-accommodating controller algorithms. This framework has been applied in Oxford to problems concerning the determination of plant availability and the simultaneous optimization of reference tracking and fault-tolerance.