Comparison of envelope demodulation methods in the analysis of rolling bearing damage

Abstract

In rotating machinery, especially in rolling bearing diagnostics, early and effective fault diagnosis is essential to ensure the reliability of rolling bearings. The commonly used vibration analysis gathers large amounts of information about the dynamics and the general condition of the bearings by recording vibrations generated by a rolling bearing. Within the vibration analysis, the location of the damage can be determined by using envelope demodulation. This is done by identifying the damage frequency within a damaged bearing. Peak-, Root-Mean-Square (RMS)-, and Hilbert-envelopes are the conventional demodulation methods in this field, but each of these methods is different and thus more or less suitable for particular applications. This paper presents a comparative study of these three methods. The selected envelopes are compared and evaluated through experiments, the evaluation of measurement data and the establishment of quality characteristics regarding efficiency and quality. In the process, a decision tool is created to help in the selection of a suitable envelope demodulation technique. A formula for calculating an optimal window size of the Peak-envelope is established, as well as a recommendation for selecting a window size of the RMS-envelope.

A New Method for Bearing Fault Diagnosis across Machines Based on Envelope Spectrum and Conditional Metric Learning

“…In the contemporary landscape of fault diagnosis, an array of algorithms have been developed based on the principles of mechanical fault theory. These methods encompass diverse techniques, such as resonance demodulation [7], envelope demodulation [8,9], generalized demodulation [10], and order ratio analysis [11]. The recent surge in the field of bearing fault diagnosis can be attributed to the continuous advancements in deep-learning technologies.…”

A hybrid approach for gearbox fault diagnosis based on deep learning techniques

“…This is possible in recent years, due to the great technological advancement in terms of industrial hardware and software. In this context, vibration analysis is widely used to monitor rotating machines because, the interpreted results are relatively easy, reliable and allows further manipulation using computers compared to other offline approaches [3]. The main difficulty is that reliable indicators are difficult to obtain from a complex machine, since the signals collected may be vibrations from one or several defective machine components [33].…”

Cross-domain bearing fault diagnosis method based on SMOTENC and deep transfer learning under imbalanced data

“…Currently, most methods for bearing fault diagnosis are based on vibration signals. Vibration signals contain periodic fault impulse impacts, and based on this characteristic, traditional fault diagnosis methods such as resonance demodulation [7], envelope demodulation [8,9], generalized demodulation [10], and order ratio analysis [11] have been proposed and achieved satisfactory results. With the development of artificial intelligence, fault diagnosis methods based on deep learning have emerged, including convolutional neural networks [12], autoencoders [13], recurrent neural networks [14], generative adversarial networks [15], and graph neural networks [16].…”