A device lifetime prediction method based on convolutional Bayesian recurrent neural networks

By combining convolutional Bayesian recurrent neural networks, the problem of insufficient uncertainty management in equipment remaining life prediction is solved, resulting in more accurate and reliable prediction results and enhancing the scientific decision-making value of equipment maintenance.

CN116992750BActive Publication Date: 2026-06-30NORTHWESTERN POLYTECHNICAL UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
NORTHWESTERN POLYTECHNICAL UNIV
Filing Date
2023-03-30
Publication Date
2026-06-30

AI Technical Summary

Technical Problem

Existing methods for predicting the remaining useful life of equipment are inadequate in handling uncertainty, making it difficult to effectively manage multiple sources of uncertainty and individual differences, resulting in insufficient reliability and guidance value of the prediction results.

Method used

A method based on convolutional Bayesian recurrent neural network (CB-LSTM) is adopted, which combines convolutional neural network (CNN) and Bayesian recurrent neural network (B-LSTM). By performing deep feature extraction and degradation trend fitting on sensor data, uncertainty is quantified and a confidence interval for predicting the remaining life of the equipment is provided.

Benefits of technology

It improves the accuracy and reliability of equipment remaining life prediction, effectively quantifies uncertainty, and provides more reliable prediction results and maintenance guidance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116992750B_ABST
    Figure CN116992750B_ABST
Patent Text Reader

Abstract

This invention discloses a device lifetime prediction method based on a convolutional Bayesian recurrent neural network, relating to the field of remaining lifetime prediction. The method mainly consists of three parts: data preprocessing, building and training a CB-LSTM network, and model evaluation. The data preprocessing stage includes normalizing the collected sensor data, determining the true Remaining Life (RUL) labels, and performing sliding window processing on the data. The network model building and training stage mainly involves inputting the preprocessed data into the CB-LSTM network for training, selecting the best-performing model, and determining the hyperparameters. The model evaluation stage uses the obtained best model to perform RUL prediction and recognize uncertainty on a test machine. This method uses CNN to extract deep feature information within the window, which helps generate features with more semantic information to be fed into the CB-LSTM network, making the input data more abstract and robust. The effectiveness of the model is evaluated through RMSE and Score, and the cognitive uncertainty and random uncertainty caused by data noise in the model are quantified, providing a more effective solution for RUL prediction and uncertainty quantification.
Need to check novelty before this filing date? Find Prior Art