Bearing residual life prediction method and device and medium

Abstract

The invention provides a bearing residual life prediction method and device and a medium. The method comprises the following steps: collecting a vibration signal during rotation of a bearing through a vibration sensor; obtaining a training set containing training samples and a test set containing test samples based on the collected vibration signals; extracting a training sample from the training set, inputting the training sample into an initialized hybrid model comprising a convolutional deep belief network (CDBN) model and a gated loop unit (GRU) model, and training the hybrid model; wherein the CDBN model comprises a multi-layer convolution restricted Boltzmann machine (CRBM), a training sample in the training set is firstly input into the CDBN model, and the output of the CDBN model is input into the GRU model; and inputting a test sample into the trained hybrid model, and outputting a prediction result of the residual life of the bearing. According to the CDBN-GRU-based bearing residual life prediction method and device and the medium provided by the invention, the self-learning capability is realized, and the prediction intelligence can be greatly improved.

Description

technical field [0001] The invention relates to the technical field of fault diagnosis, in particular to a bearing residual life prediction technology, and more particularly to a bearing residual life prediction method, device and medium based on a convolutional deep belief network-gated recurrent unit (CDBN-GRU) . Background technique [0002] Bearings are widely used in various fields of national economy and national defense. They are the basic parts of mechanical equipment and are easily damaged. They need regular inspection and replacement. Their reliability will directly affect the performance and life of the equipment. By monitoring the bearing state, the fault point and type can be accurately judged and the prediction of the remaining life of the bearing can be realized. Therefore, the research on the prediction of the remaining life of the bearing has great engineering value. [0003] At present, there are two main methods for predicting the remaining life of beari...

AI Technical Summary

Problems solved by technology

The time complexity and space complexity of this model method are not very ideal, and the prediction accuracy is not high; (2) The application of backpropagation neural network in the prediction of bearing remaining life

With the rapid development of modern industry, the data resources are constantly increasing, and the data resources of bearing faults are also becoming more and more abundant, which makes the traditional backpropagation neural network very prone to the problem of dimensionality disaster in the process of remaining life prediction, and the faults obtained on-site Data is often unlabeled

Therefore, it is difficult to establish an effective diagnostic model with traditional methods, and a diagnostic algorithm with self-learning ability is urgently needed for model optimization and research.

Images