COMPARATIVE ASSESSMENT OF METHODS FOR PREDICTING THE TECHNICAL CONDITION OF ELECTRIC DRIVES OF HAZARDOUS PRODUCTION FACILITIES

Abstract

The paper deals with the design of efficient and reliable systems for operational diagnostics and forecasting of the technical condition of adjustable megawatt electric drives. The statistics of failure of the most critical components of AC electrical machines – stator windings, bearings and ACS are presented. The methodology and architecture of artificial neural networks have been developed for obtaining predictive models of high-voltage synchronous machines. Examples of neuro-fuzzy prediction of the technical state and resource of stator windings and analysis of the spectral composition of the supply voltage by the series method are given. Tests of selected networks, the Box–Jenkins fuzzy model, models of the method for analyzing the dynamics of spectral components with predicting the values of current and stator temperatures are obtained. The comparative results of the analysis of the expected states of electric machines of high power, based on the consideration of various operational factors of the operation of electric drives, made it possible to develop recommendations for the use of new predictive methods.