An improved algorithm for the design of testable subsystems

In complex industrial plants, there are usually many sensors and the modeling of plants leads to lots of mathematical relations. Before using classical tools for fault detection, the first problem to solve is: what sensors and mathematical relations should be selected for the design of a detection test such as a state observer or a parity space based detection algorithms. This paper presents a general method for finding all the possible testable subsystems i.e. sets of relations, than can lead to detection tests, taking into account actuator and sensor locations. This method, which is based on a structural analysis, provides the constraints that have to be used for the design of each detection test and manages situations where constraints contain non deductible variables. Thanks to these results, it becomes possible to select the most interesting testable subsystems regarding detectability and diagnosability criteria.