Rolling bearing fault diagnosis method based on two-way memory cycle neural network

Abstract

The invention discloses a rolling bearing fault diagnosis method based on a two-way memory cycle neural network. An existing rolling bearing fault diagnosis method does not consider a single logical structure characteristic of data after characteristic extraction and a fault type can not be integrally determined from the data when fault data is processed. Aiming at the above defects, the method of the invention comprises the following steps of firstly, acquiring a program data sample, carrying out standardized preprocessing on vibration acceleration data, making the collected data accord with standard normal distribution, and then using a time-frequency domain characteristic extraction algorithm to obtain 512 time-frequency domain characteristic vectors; then, constructing an improved two-way memory type cycle neural network fault diagnosis model, using an idea of a simple design, and then using sample data to train a neural network weight parameter, after iteration training, generating a model which can map a relationship between bearing data and a fault type, wherein the designed memory-type cycle neural network includes a forgetting gate, an input gate and a cellular state; and finally, using the model to carry out fault analysis so as to achieve accurate diagnosis of a rolling bearing fault.

Description

technical field [0001] The invention relates to a rolling bearing fault diagnosis method, in particular to a rolling bearing fault diagnosis method based on a two-way memory cycle neural network. Background technique [0002] With the rapid development of modern industry and science and technology, production equipment is gradually developing towards high speed, automation and intelligence. However, problems such as aging of equipment components and changes in the application environment are often unavoidable. Due to the aging of components, mechanical equipment Various failures occur during operation, causing huge economic losses. Therefore, it is of great significance to collect massive data for equipment status monitoring, and to study and use innovative theories and methods to efficiently mine information from equipment big data for equipment status fault diagnosis. . [0003] Rolling bearings are the core components of rotating machinery. The fault detection of rolling...

AI Technical Summary

Problems solved by technology

[0004] The design of the traditional fault diagnosis model does not take into account the unsteady characteristics of the bearing during data collection. It is necessary to design a bearing fault diagnosis method for this feature to improve the accuracy of rolling bearing fault diagnosis. At the same time, due to structural limitations, the length of single processing data is limited in traditional methods (one sample data needs to be divided into multiple batches for processing ), cannot fully combine the data as a whole to judge the fault type

Images