Mechanical bearing fault diagnosis technology based on wide residual network learning model

Abstract

The present invention discloses a mechanical bearing fault diagnosis technology based on a wide residual network learning model. The technology comprises the following steps: Step one, collecting vibration data; Step two, using a convolutional neural network to train and test sample data, and for data of each layer in the convolutional neural network structure, using the data visualization technology to complete the preliminary application scenario establishment of the combination of the vibration data and the convolutional neural network model; and Step three, introducing the wide residual network model, and achieving filtering and compromise of the high-frequency noise signal and the low-frequency characteristic signal of the vibration data by widening the convolution kernel. According to the technology disclosed by the present invention, the wide residual network is used, the core is to solve the side effects (degradation problem) caused by increasing the depth, and in the case of maintaining the accuracy of vibration data fault diagnosis, filtering and compromise of the high-frequency noise signal and the low-frequency characteristic signal of the vibration data can be achievedin the deep learning process.

Description

technical field [0001] The invention belongs to the technical field of mechanical vibration and relates to a fault diagnosis technology, in particular to a mechanical bearing fault diagnosis technology based on a wide residual network learning model. Background technique [0002] Swivel bearings are key components of automobile transmission systems, and their operating status directly affects the comfort and safety of the vehicle. Quantitative data analysis of rotating bearing vibration is a common method for bearing fault identification. Bearing vibration data is a typical time-series data. The traditional analysis method is to use various time-frequency feature engineering to manually select data features, which greatly increases the cost of data preprocessing. [0003] Most of the existing fault diagnosis of mechanical bearings is based on the "rule" method, which requires preprocessing of static data, and the processing process is complicated. Especially when processing...

AI Technical Summary

Problems solved by technology

By using a wide residual network, the core of this technology is to solve the side effects (degeneration problem) caused by increasing the depth, and at the same time, it can achieve a high accuracy of vibration data in the process of deep learning while keeping the accuracy of vibration data fault diagnosis basically unchanged. Filtering trade-off between high-frequency noise signal and low-frequency characteristic signal

Images