Kratkoročno predviđanje vibracionog ponašanja Fransis turbine nakon višedecenijske eksploatacije

Abstract

Contemporary approaches in forecasting and preventing accidents, reducing the number and duration of downtimes, detecting and monitoring the failures in the real decades-long operating conditions of units in hydropower plants, are based on vibrodiagnostics. The multidisciplinary character of such approaches is reflected in the identification of certain vibrations in the hydro-aggregates, the application of various vibrodiagnostic methods, the quality assessment of measured data, as well as analysis and prediction based on artificial intelligence. The paper presents one of the possible approaches to solving the complex problem of predicting the behavior of a hydraulic turbine that has been in operation for more than half a century. An appropriate algorithm using artificial neural networks has been developed for short - term prediction of absolute and relative vibrations. The obtained results are presented depending on various operating conditions.