Pareto-Optimal Observers for Ship Propulsion Systems by Evolutionary Computation

Abstract

Fault detection for a marine-vehicle propulsion system is considered. In particular, the system of an engine and a propeller is supervised. A Pareto-optimal state observer allows for generating residual signals, which can be utilised for fault detection. This observer is designed by means of evolutionary algorithms. The applied algorithm solves in the sense of Pareto the task of multiobjective optimisation of a vector quality index, which describes the influence of faults, noises and modelling uncertainty on the considered system.