Benchmark application of fuzzy observers to fault detection on a ship propulsion system

This article is a contribution to the benchmark problem in the framework of the COSY (Control of Complex Systems) project and deals with fault detection on a simulated ship propulsion system. Our approach is to make use of a fuzzy output observer for the generation of the fault-reflecting residual signals. For this purpose, we first have to design suitable fuzzy models of the system under consideration. During the model-building stage, a lot of design parameters have to be chosen, such as for example the linguistic variables of the fuzzification and the composition operators in fuzzy space. On the other hand, using a fuzzy relational structure does not require the complete knowledge of all parameter values appearing in the system. So, we can take advantage of this fact during this design stage.