基于长短期记忆网络时序数据趋势预测及应用

doi:10.13675/j.cnki.tjjs.200394

推进技术 ›› 2021, Vol. 42 ›› Issue (3): 675-682.DOI: 10.13675/j.cnki.tjjs.200394

基于长短期记忆网络时序数据趋势预测及应用

杨柯，范世东

武汉理工大学能源与动力工程学院，湖北武汉 430063

出版日期:2021-03-15 发布日期:2021-08-15
作者简介:杨　柯，博士生，研究领域为运维与保障。E-mail：ake1231@163.com
基金资助:
国家自然科学基金（51679178）。

Long Short-Term Memory Network Based Method and Its Application in Time-Series Data Trend Prediction

School of Energy and Power Engineering，Wuhan University of Technology，Wuhan 430063，China

Online:2021-03-15 Published:2021-08-15

摘要/Abstract

摘要： 为了研究状态监测大数据对设备运行状态的估计和预测，提出了一种人工经验与主成分分析相结合的长短期记忆网络方法（AEPCA-LSTM），利用运行过程中的监测时序数据对设备运行趋势进行预测。通过基于人工经验的主要成分分析方法（AEPCA）从状态监测系统中提取与目标变量最相关的状态变量作为输入；利用长短期记忆网络（LSTM）对目标变量趋势变化进行预测，并考虑运行过程中新数据样本的持续产生，对模型进行定期更新，以提高模型的动态适应性。将所提出的方法应用于船舶副机系统的涡轮增压器转速预测中，结果表明，该方法相对于传统的PCA-LSTM和LSTM具有更小的预测平均误差0.18037，展现了其在时序数据趋势预测的优势。

关键词: 主成分分析;长短期记忆网络;人工经验;时序数据;趋势预测

Abstract: In order to study the estimation and prediction of equipment operation state by condition monitoring big data， a novel method of Long-short Memory Network integrating Principal Component Analysis is proposed based on Artificial Experience （AEPCA-LSTM）， which uses the monitoring time series data during operation to predict the equipment health trends. Firstly， the Principal Component Analysis method based on Artificial Experience （AEPCA） is used to extract the state parameters most relevant to target variable from the state monitoring system as input. Secondly， the Long short-term Memory Network （LSTM） is used to predict the trend changes of the target variable considering the continuous generation of new data samples during operation， the model is regularly updated to improve the dynamic adaptability of the model. Finally， the proposed method is applied to the turbocharger speed prediction of marine auxiliary engine system. The results show that the method has a lower prediction loss of 0.18037 compared with PCA-LSTM and LSTM， which indicates its advantages in the prediction of trend in time series data.

Key words: Principal component analysis;Long short-term memory network;Artificial experience;Time series data;Trend prediction

杨柯，范世东. 基于长短期记忆网络时序数据趋势预测及应用[J]. 推进技术, 2021, 42(3): 675-682.

YANG Ke, FAN Shi-dong. Long Short-Term Memory Network Based Method and Its Application in Time-Series Data Trend Prediction[J]. Journal of Propulsion Technology, 2021, 42(3): 675-682.

参考文献

[1] Pecht M. Prognostics and Health Management of Electronics［M］. Hoboken： John Wiley & Sons， 2008.
[2] 胡昌华，施权，司小胜，等. 数据驱动的寿命预测和健康管理技术研究进展［J］. 信息与控制， 2017， 46（1）： 72-82.
[3] Yan H， Wan J， Zhang C， et al. Industrial Big Data Analytics for Prediction of Remaining Useful Life Based on Deep Learning［J］. IEEE Access， 2018， 6： 17190-17197.
[4] 雷亚国，贾峰，周昕，等. 基于深度学习理论的机械装备大数据健康监测方法［J］. 机械工程学报， 2015， 51（21）： 49-56.
[5] Deutsch J， He D. Using Deep Learning Based Approaches for Bearing Remaining Useful Life Prediction［C］. Denver： Annual Conference of the Prognostics and Health Management Society， 2016.
[6] Deutsch J， He D. Using Deep Learning-Based Approach to Predict Remaining Useful Life of Rotating Components［J］. IEEE Transactions on Systems Man & Cybernetics Systems， 2017， 48（1）： 11-20.
[7] Babu G， Zhao P， Li X. Database Systems for Advanced Applications［M］. Berlin： Springer， 2016.
[8] Li X， Ding Q， Sun J. Remaining Useful Life Estimation in Prognostics Using Deep Convolution Neural Networks［J］. Reliability Engineering and System Safety， 2018， 172： 1-11.
[9] Chow T， Fang Y. A Recurrent Neural-Network-Based Real-Time Learning Control Strategy Applying to Nonlinear Systems with Unknown Dynamics［J］. IEEE Transactions on Industrial Electronics， 1998， 45（1）： 151-161.
[10] Heimes F. Recurrent Neural Networks for Remaining Useful Life Estimation［C］. Denver： International Conference on Prognostics and Health Management， 2008.
[11] Malhi A， Yan R， Gao R. Prognosis of Defect Propagation Based on Recurrent Neural Networks［J］. IEEE Transactions on Instrumentation & Measurement， 2011， 60（3）： 703-711.
[12] Wu Y， Yuan M， Dong S， et al. Remaining Useful Life Estimation of Engineered Systems Using Vanilla LSTM Neural Networks［J］. Neurocomputing， 2017， 275： 167-179.
[13] Miao H， Li B， Sun C， et al. Joint Learning of Degradation Assessment and RUL Prediction for Aeroengines via Dual-Task Deep LSTM Networks［J］. IEEE Transactions on Industrial Informatics， 2019， 15（9）： 5023-5032.
[14] Elsheikh A， Yacout S， Ouali M S. Bidirectional Handshaking LSTM for Remaining Useful Life Prediction［J］. Neurocomputing， 2019， 323： 148-156.
[15] Zhou F， Park J H， Wen C， et al. Average Accumulative Based Time Variant Model for Early Diagnosis and Prognosis of Slowly Varying Faults［J］. Sensors， 2018， 18（6）.
[16] Wang Y， Li H. Complex Chemical Process Operation Evaluations Using a Novel Analytic Hierarchy Process Model Integrating Deep Residual Network with Principal Component Analysis［J］. Chemometrics and Intelligent Laboratory Systems， 2019， 191： 118-128.
[17] A?t-Sahalia Y， Xiu D. Principal Component Analysis of High-Frequency Data［J］. Journal of the American Statistical Association， 2019， 114（5）： 287-303.
[18] Wang Y， Yang K， Li H. Industrial Time-Series Modeling via Adapted Receptive Field Temporal Convolution Networks Integrating Regularly Updated Multi-Region Operations based on PCA［J］. Chemical Engineering Science， 2020， 228（11）.
[19] Hochreiter S， Schmidhuber J. Long Short-Term Memory ［J］. Neural Computation， 1997， 9（8）： 1735-1780.

基于长短期记忆网络时序数据趋势预测及应用

Long Short-Term Memory Network Based Method and Its Application in Time-Series Data Trend Prediction

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 0

编辑推荐

Metrics

本文评价