Probabilistic Design of Fault Tolerant Control via Parameterization

Abstract

This paper studies the design of fault tolerant control systems (FTCSs) by considering random faults and two categories of design objectives. The FTCSs are modeled in a stochastic framework, resulting from the random fault process. The design objectives include a stability requirement and a probabilistic performance index. Such an index is chosen as a reliability-related criterion, which can be evaluated from a numerical procedure only but lacks analytical expressions. A parameterization procedure together with a randomization-based optimization method is developed to find a statistically optimal controller that can stabilize the system and achieve the highest reliability.