Health condition diagnoses of power plants turbines aided by neural networks and vibration tools

Abstract

In Brazil, a way of changing the aggressive exploitation of hydraulic potential resources, in order to produce electrical energy, was found by the new application of hydropower plants. A crucial characteristic is that a unique operator is responsible for several hydropower sites far away from each other. Faced with this geography problem, an intranet architecture has been developed and from this skilful application it is possible to use some intranet channels for transmission of a special data from a new technique of signal processing. Basically, this technique is a type of spectrum which uses fixed frequency bands and vibration severity levels. The special spectrum's data is issued to a neural network system which detects the fault in its early stages and a quickly and reliably automatically a diagnosis is obtained. The intranet system uses this diagnosis to transmit the real health condition of the machine in real-time, optimising both management maintenance and production.