Dynamic functional — Link neural networks genetically evolved applied to fault diagnosis

The paper addresses the development of neural observer schemes for process fault diagnosis. The design is based on a generalised functional-link neural network with internal dynamics. An evolutionary search of genetic type and multi-objective optimisation in the Pareto-sense is used to determine the optimal architecture of the dynamic network. Symptoms characterising the current state of the process are obtained based on prediction errors. The latter are further evaluated by a static artificial network. Experimental results regarding the detection and isolation of artificial sensor faults in an evaporation station from a sugar factory illustrate the approach.