The Noise Effects on Signal Processors Used for Fault Detection Purpose

Abstract

Signal Processing plays a crucial role in addressing failures in electrical machines. Experimental data are never perfect due to the intrusion of undesirable fluctuations unrelated to the investigated phenomenon, so-called noise. Noise has disturbing effects on the measurement data, and in the same way, could diminish or mask the fault patterns in feature extraction using different signal processors. In this paper, fault detection in a custom made 100 kVA synchronous generator under an inter-turn short circuit fault is studied by using measurements of the air gap magnetic field. Signal processing tools like a Fast Fourier Transform (FFT), Short Time Fourier Transform (STFT), Discrete Wavelet Transform (DWT), Continuous Wavelet Transform (CWT), and Time Series Data Mining (TSDM) are used to diagnose the faults with a central focus on noise impacts on processed data. Moreover, some useful methods are presented for hardware noise rejection.