Electromechanical equipment few-sample degradation trend prediction method of unsupervised meta-learning network

Abstract

The invention discloses an electromechanical equipment few-sample degradation trend prediction method of an unsupervised meta learning network, relates to the technical field of service performance evaluation and prediction of electromechanical equipment, and solves the technical problem that an existing meta learning method generally depends on label sample support and is difficult to be directly applied to historical data with scarce labels. According to the technical scheme, the method is characterized in that by aggregating the training process of each inner loop, cross-task outer loop optimization and training are carried out on model parameters obtained by training a support set of each training set through a support set of a test set, and finally an unsupervised meta-learning agent model is generated; and the classic deep circulation network is effectively reconstructed, the classic deep circulation network has remarkable generalization ability under excitation of few samples, connection is established between historical large sample data and insufficient prediction samples, and the problem of labeling of historical label-free data is effectively solved.

Description

technical field [0001] The present application relates to the technical field of service performance evaluation and prediction of electromechanical equipment, and in particular to a method for predicting the degradation trend of electromechanical equipment with few samples using an unsupervised meta-learning network. Background technique [0002] Electromechanical equipment widely exists in high-end intelligent manufacturing application scenarios such as aviation equipment, satellite manufacturing and application, rail transit equipment manufacturing, marine engineering equipment manufacturing, numerical control processing and manufacturing, and process industries. The accidental downtime caused by its failure often causes major economic losses and bad society. Effective degradation trend prediction and health assessment can avoid unknown risks and reduce economic and property losses, which has great scientific research value. [0003] In recent years, data-driven methods ha...

AI Technical Summary

Problems solved by technology

However, existing meta-learning methods generally rely on labeled samples, and it is difficult to directly apply them to historical data with scarce labels.

Images