A New Model for Remaining Useful Life Prediction Based on NICE and TCN-BiLSTM under Missing Data

The Remaining Useful Life (RUL) prediction of engineering equipment is bound to face the situation of missing data. The existing methods of RUL prediction for such cases mainly take “data generation—RUL prediction” as the basic idea but are often limited to the generation of one-dimensional test dat...

Full description

Saved in:
Bibliographic Details
Published inMachines (Basel) Vol. 10; no. 11; p. 974
Main Authors Zheng, Jianfei, Zhang, Bowei, Ma, Jing, Zhang, Qingchao, Yang, Lihao
Format Journal Article
LanguageEnglish
Published Basel MDPI AG 01.11.2022
Subjects
bidirectional long short-term networks
deep generative networks
Deep learning
Degradation
Fault diagnosis
Life prediction
Missing data
Multivariate analysis
multivariate degraded data
Neural networks
nonlinear independent component estimation
Prediction models
RUL
temporal convolutional networks
Time series
Useful life
Variables
Online AccessGet full text

Cover

More Information
Summary:The Remaining Useful Life (RUL) prediction of engineering equipment is bound to face the situation of missing data. The existing methods of RUL prediction for such cases mainly take “data generation—RUL prediction” as the basic idea but are often limited to the generation of one-dimensional test data, resulting in the extraction of the prediction network. Therefore, this paper proposes a multivariate degradation device based on Nonlinear Independent Components Estimation (NICE) and the Temporal Convolutional Network–Bidirectional Long Short-term Memory (TCN-BiLSTM) network for the RUL prediction requirements in the case of missing data. First, based on the NICE network, realistic data are generated through reversible sampling; then, the filling of multivariate missing data is completed. Next, the filled multivariate degradation data are processed to generate multivariate degradation data and predicted labels for constructing the training set and test set. Based on this, a residual life prediction model integrating TCN and the BiLSTM network is proposed. To evaluate the proposed method, this paper takes an example of the RUL prediction of aeroengines to perform multivariate degradation data-filling and prediction tasks. The results demonstrate the superiority and potential application value of the method.
ISSN:2075-1702
2075-1702
DOI:10.3390/machines10110974