Process modelling and fault diagnosis using fuzzy neural networks

Process modelling and fault diagnosis using fuzzy neural networks are studied in this paper. In fuzzy network based modelling, the process operation is partitioned into several fuzzy operating regions. Within each region, a local model of simple form, for example a low-order linear model, is used to model the process. The global model output is obtained through the centre of gravity defuzzification which is essentially the interpolation of local model outputs. A fuzzy neural network can be used to implement such a fuzzy model. Process knowledge is used to initially divide the process operation into several fuzzy operating regions and to set up the initial fuzzification layer weights. Process input output data are used to train the network. Through training, membership functions of fuzzy operating regions are refined and local models are learnt. This technique has been successfully applied to the modelling of pH dynamics in a continuous stirred tank reactor (CSTR). The fuzzy network for fault diagnosis is obtained by adding a fuzzification layer to a conventional feed forward network. The fuzzification layer converts the increments in on-line measurements and controller outputs into three fuzzy sets: ''increase'', ''steady'' and ''decrease''. The following layers then classify the symptoms, represented by these fuzzy sets, into various fault categories. The fuzzification layer can compress training data, and thereby ease training effort. Robustness of the diagnosis system is enhanced by adopting a fuzzy approach in representing abnormalities in the process. The technique has been successfully applied to a CSTR system.