Verification of vibration data in electromachine systems using neural-network software

Abstract

Verification of vibration data in electromachine systems using neural network software is described. Due to a number of factors, the data are often corrupted and therefore it has been usually verified by a visual inspection. The developed neural networks were trained to screen out glitched and clipped waveforms. Satisfactory results were obtained, indicating the feasibility of neural networks for vibrational diagnostics.<>