Method for predicting residual life of rolling bearings based on sparse coding

Abstract

The invention provides a method for predicting the residual life of complex rotating mechanical rolling bearings based on sparse coding. The method includes the steps: firstly, acquiring mechanical equipment monitoring vibration signals; secondly, performing time domain feature extraction, and building an over-complete dictionary serving as a base vector with time domain features as original signals according to a signal sparsity theory to guarantee validity of prediction information; thirdly, processing training data and testing data by regularization; fourthly, finding an optimal sparse weight; fifthly, performing failure trend prediction to omit training time of a traditional method, greatly reduce time burden and improve prediction accuracy and robustness of a system due to less parameter tuning; sixthly, performing residual life prediction to ensure residual life prediction accuracy of the bearings by high-precision prediction values.

Description

technical field [0001] The invention belongs to the field of equipment system failure prediction and health management, and in particular relates to a method for predicting the remaining life of a rolling bearing based on sparse coding. Background technique [0002] With the development of science and technology and the progress of society, various types of rotating machinery equipment have been widely used in engineering. Rolling bearings are key components in rotating machinery equipment. Once a failure occurs, it may cause property losses, and even cause casualties in serious cases. The traditional supervised rolling bearing prediction method not only has low prediction accuracy but also requires a large amount of historical information, which is not suitable for cross-working condition prediction, and often cannot make reliable preventive maintenance decisions. Therefore, it is necessary to develop a remaining life prediction method and system centered on unsupervised l...

AI Technical Summary

Problems solved by technology

The traditional fault prediction technology often can only make effective prediction evaluation based on a large amount of historical information. When the historical information is insufficient, the prediction model will lose its original accuracy; this kind of prediction model also requires a large number of parameter tuning. When the prediction After the object is changed, the prediction effect will become worse or even unable to make a maintenance strategy, resulting in economic losses

Images