Bearing fault edge diagnosis method and system based on wavelet improved MobileNet network

Abstract

The invention discloses a bearing fault edge diagnosis method and system based on a wavelet improved MobileNet network. The method comprises the following steps of publishing collected bearing vibration data, inputting the collected historical vibration data into the improved MobileNetV3-Small network for training, and acquiring a diagnosis model, and inputting vibration data acquired in real time into the diagnosis model for fault diagnosis. According to the method, a large convolution model trained in a Tensorflow environment is simplified by using a wavelet-based improved MoblieNetV3-small network, then the simplified deep learning model is put into a Raspberry Pi which is relatively poor in computing power and is cheaper than a computer by dozens of times for real-time diagnosis, and compared with the prior art, the method has the advantages that on the premise of ensuring the diagnosis accuracy, the size and the calculation amount of the model are greatly reduced, the operation speed of the model is greatly improved, the requirement of real-time response is met, and the operation cost is effectively controlled.

Description

technical field [0001] The invention relates to a method and system for diagnosing the edge of a bearing fault based on a wavelet-improved MobileNet network, and belongs to the field of fault diagnosis of mechanical equipment. Background technique [0002] In the field of fault diagnosis of mechanical equipment, convolutional neural network (Cvolutional neural network, CNN) has been widely used in equipment fault diagnosis and condition monitoring, and the accuracy of diagnosis is getting higher and higher. It has also made great achievements in the field of bearing faults. A qualitative leap has been made. However, in order to overly pursue the accuracy of diagnosis, the depth of the convolutional neural network is getting deeper and deeper, and its model is becoming more and more complex. For example, the number of layers of the residual network (ResNet) has reached 152 layers, and the number of parameters is hundreds of millions. If it is calculated as a unit, more than ...

AI Technical Summary

Problems solved by technology

Such a large-scale diagnostic model is generally deployed on a CPU / GPU with high performance and sufficient computing power to run, but it also takes a certain amount of time

Due to the limitations of hardware resources and computing power in mobile devices and embedded systems, it is difficult to directly deploy complex CNN models

In addition, the current low-latency and fast-response requirements for real-time evaluation of equipment status make it difficult for large-scale model diagnosis and training time to meet real-time requirements; therefore, how to reduce model volume and improve model running speed while ensuring model accuracy , and being able to deploy on embedded platforms has become an inevitable requirement for the application and promotion of neural networks in engineering

Images