Indexing metadata

Today, one of the main directions of industrial development is the digitalization of production processes. In order to achieve high production rates, the reliability of production equipment is necessary; more and more advanced means of its self-diagnosis are being developed. Thus, self-diagnosis, combined with a high level of automated analytics, makes it possible to predict a malfunction with a high degree of probability, warn about the timing of its occurrence and methods of preventive elimination. This article discusses existing methods of vibration diagnostics, including those that appeared during the fourth industrial revolution, namely in the conditions of widespread and high-quality application of machine learning systems, neural networks and artificial intelligence. Methods for collecting primary information about vibration and methods for analyzing data using the above algorithms are described. The results of experimental applications of various analytical mechanisms developed to determine the type of defects in parts rotating under mechanical load are considered, and the advantages and disadvantages of each method are listed. The purpose of the review is to determine the existing methods of vibration diagnostics, determine their properties and compare them. As a result of the analysis, it was found that the most developing direction in the field of vibration signal research is a combination of wavelet transformation and neural network learning.