Bearing life prediction method based on hidden Markov model and transfer learning

Abstract

The invention discloses a bearing life prediction method based on a hidden Markov model and transfer learning. The method comprises the steps of (1) collecting a full-life original signal of a rollingbearing; extracting a feature set containing time domain, time frequency domain and trigonometric function features; (2) inputting the feature set into a hidden Markov model to predict a hidden state, and obtaining a fault occurrence moment; (3) forming a training set by feature sets from all source domains and part of target domains, inputting the training set into the constructed multi-layer perceptron model, obtaining domain invariant features and optimal model parameters through optimization target training, and substituting the optimal model parameters into the perceptron model to obtaina neural network life prediction model; and (4) inputting the remaining target domain feature set into a neural network life prediction model, and predicting the remaining life of the bearing according to the output value. The hidden Markov model is used for automatically detecting the fault occurrence moment, and then the transfer learning based on the multilayer perceptron is used for solving the distribution difference of a source domain and a target domain caused by different working conditions.

Description

technical field [0001] The invention relates to the technical field of operating state and life prediction of rolling bearings, in particular to a bearing life prediction method based on a hidden Markov model and transfer learning. Background technique [0002] Rolling bearings are the most widely used in industrial sectors such as aerospace, electric power, petrochemical, metallurgy and machinery, and are also one of the most easily damaged parts. The working state of rotating machinery is closely related to rolling bearings. According to statistics, about 30% of mechanical failures in rotating machinery using rolling bearings are related to bearing damage. Bearing life prediction is of great significance to ensure normal production process and avoid economic loss. [0003] Life prediction methods are generally divided into 4 categories: physical model-based methods, statistical model-based methods, artificial intelligence-based methods, and hybrid methods. Among them, th...

AI Technical Summary

Problems solved by technology

[0004]However, the current life prediction methods usually have two defects: 1) The failure time (FOT) is ignored or determined by experience

However, due to different working conditions, the distribution difference exists in the source domain and the target domain, resulting in a decline in the model's generalization ability

Images