Desenvolvimento de sistema de tempo real baseado em FPGAS para processamento de sinais de descargas parciais

Abstract

In this work it was developed an implementation in FPGA of the necessary routines to process signals from occurrence of partial discharges. The primary purpose of the system is the elimination of noise components to allow the possibility of interpretation operations on the sampled signals. In addition to the attempted use of parallelization and pipelining, a set of approaches was developed aiming at maximizing the sampling rates. An interesting innovation was obtained in the procedure for signal identification by using a classifier based on concepts of neural networks and distance metrics, and defining parameters for identifying the patterns of interest. The procedures made use of the Discrete Wavelet Transform (DWT). The proposed approaches maintained the effectiveness of other procedures available in the literature while allowing significant simplification of the necessary operations. Thus, it obtained considerable decrease in demanded resources for implementation. Moreover, the proposed approaches can be used in further operations to identify patterns associated with equipment failures, based on the partial discharge sampled signals.