Adaptation of Fuzzy Diagnostic Models In Real Time

Abstract

Currently, fuzzy diagnostic models are widely used in the problems of analysis and forecasting, classification, and management in various areas of industry and technology. A neuro-fuzzy network that consists of neurons-fuzzy granules performing data aggregation functions is proposed as a fuzzy diagnostic model. This network is an adaptive learning system in real time when new fuzzy data arrives. Fuzzy input data generated by inaccurate measurements and qualitative perceptions are set in a parametric form. Architecture and granular adaptive learning of a neuro-fuzzy network is considered. It is emphasized that adaptive learning is carried out by adjusting the fuzzy granules membership functions, as well as changing the knowledge base by including and excluding fuzzy rules. An example of industrial equipment diagnostics, as well as the research results are presented. The proposed model makes it possible to evaluate the performance of monitored devices during monitoring, as well as to reduce the time of monitoring systems for significant and non-significant changes in the data flow that characterize the initial stages of fault manifestation.