Equipment for Predictive Maintenance in Hydrogenerators

Abstract

This paper presents an equipment for predictive maintenance in large hydrogenerators. This equipment uses techniques of digital signal processing of the information contained in the electrical variables involved in the operation of the generator. Basically, the current and voltage signals of the generator are monitored and applied the techniques of electric signature analysis. The central idea is to unite the techniques of current signature analysis (CSA), voltage signature analysis (VSA) and Enhanced Park's Vector Approach (EPVA), to separate the spectra of signals and detect frequencies related to electrical and mechanical defects of generator-turbine set. This is possible because the generator is basically a device handling magnetic fields, so it's believable to infer that any operating conditions of all, somehow, influences the behavior of the magnetic field, reflecting noticeably in variations in signs of tensions and currents provided by its. The problem is to detect these variations, because some of them are under existing noise signs, and relate them to defects which they represent. This paper presents a real implementation in a hydrogenerator at Itapebi Power Plant, Brazil.