Space rolling bearing residual life prediction method based on VETMRRN

Abstract

The invention discloses a space rolling bearing residual life prediction method based on VETMRRN, and the method comprises the following steps: S1, extracting time domain and frequency domain features from the original vibration acceleration data of a space rolling bearing, carrying out the shape value feature fusion, and taking the time domain and frequency domain features as the performance degradation features of the space rolling bearing; s2, inputting the performance degradation characteristics of the space rolling bearing into the VETMRRN to train hyper-parameters and network parameters of the VETMRRN; s3, utilizing VETMRRN to predict the performance degradation characteristic trend of the space rolling bearing in multiple steps; and S4, establishing a Weibull distribution reliability model, and predicting the precision failure threshold time point and the residual life of the space rolling bearing. According to the space rolling bearing residual life prediction method based on the VETMRRN, the VETMRRN is constructed and has good nonlinear approximation capability, generalization performance and calculation efficiency, so that the space rolling bearing residual life prediction method based on the VETMRRN has high prediction precision, good generalization performance and high calculation efficiency.

Description

technical field [0001] The invention relates to a method for predicting the remaining life of a space rolling bearing based on VETMRRN. Background technique [0002] Space rolling bearings are key components of space motion mechanisms, and their life and reliability largely affect whether the mechanical components of space vehicles such as space stations and satellites can operate normally, achieve predetermined functions, and achieve expected service life. Therefore, predicting the remaining life of space rolling bearings is of great significance to adapt to the high reliability and long life of space vehicles, and to avoid or reduce aircraft failures. However, because it is difficult to obtain bearing operating status information in the space on-orbit environment, and there is a big difference between the space environment and the conventional ground environment, the aerospace industry departments and related research institutions at home and abroad usually use the ground ...

AI Technical Summary

Problems solved by technology

It is difficult for IEGO to deal with the performance degradation data of space rolling bearings in the dynamic graph mode of a long time span, in other words, it cannot reliably predict the performance degradation trend and remaining life of a long time span (that is, more time steps)

[0004] Classic time recurrent neural networks (Sequence Recurrent Neural Networks, SRNNs) such as Long Short-Term Memory (LSTM) have the advantages of long-term memory, and are relatively better choices in solving remaining life prediction, but SRNNs There is a defect of long-term dependence in the prediction of time series, so the generalization ability is poor; at the same time, SRNNs increase the time complexity by traversing the entire training data set to achieve supervised learning

Images