Spectral analysis of stationary and transient signals for the monitoring of electrical machines

Abstract

This video tutorial was prepared for the 13th Edition of the IEEE International Symposion on Diagnostics for Electrica Machines (SDEMPED 2021). The use of electric machines is not only widespread in the industry, they are also used in railway traction, electric vehicles, wind power generation, to name a few sectors. Adequate condition monitoring of these machines helps to detect failures at an early stage. This prevents these incidents from evolving into catastrophic failures, which saves on maintenance and operating costs, protects companies' investments, and therefore helps their competitiveness and quality of service. Monitoring methods can be based on spectral analysis of variables such as electric current, vibrations, flux, speed or sound. Obtaining quality signatures related to the machine condition must be based on a correct spectral analysis, taking into account concepts such as spectral resolution or spectral leakage. Moreover, in many of these applications, the machine's operation is not stationary, which requires the use of advanced time-frequency analysis tools. This tutorial aims to provide a useful practice guide for a correct spectral analysis of stationary and transient signals – this will permit a correct fault detection and severity quantification, which are essential for the diagnosis of electrical machines – and to present to the audience those areas of research where the presenters believe there are still open problems to be solved. The analysis of non-stationary signals will also be covered, providing an overview of the most advanced techniques, such as those based on time-frequency atoms correlation. There are a wide range of available techniques but the signal analysis results are conditioned by their time and frequency resolution capabilities. This tutorial overviews the basic concepts of spectral analysis of both stationary and transient signals and how these affect the spectra quality, which in the end also affects the motor diagnosis. The tutorial also covers advanced topics such as speed estimation, tools for transient analysis, and an introduction to time-frequency atoms.There is also a part imminently practical that illustrates real cases of stationary and transient induction motor fault detection using the techniques and concepts described during the tutorial.