| 380 | 0 | 15 |
| 下载次数 | 被引频次 | 阅读次数 |
涡轮机作为能源、航空、船舶等领域的核心设备，其性能退化和寿命问题可能导致生产效率下降和安全隐患。因此，准确预测涡轮机的健康状态(State of Health, SOH)和剩余使用寿命(Remaining Useful Life, RUL)对于实现预测性维护至关重要。针对关键机械设备——涡轮机的故障预测问题，提出了一种基于改进双向长短期记忆网络(Bidirectional Long Short-Term Memory, BiLSTM)的预测方法，提出的混合模型结合了卷积神经网络(Convolutional Neural Network, CNN)的局部特征提取、注意力机制(Attention Mechanism, AM)的权重分配和BiLSTM的双向时间序列处理能力，旨在提高故障预测的准确性和效率。通过对100台涡轮机的运行数据进行分析，实验结果表明，改进的BiLSTM模型在预测精度上优于其他主流模型，如CNN-BiLSTM、CNN-LSTM、循环神经网络(Recurrent Neural Network, RNN),具有更低的平均绝对误差(Mean Absolute Error, MAE),保持了较高的效率和准确度。
Abstract：Turbines, as core equipment in the fields of energy, aviation, and shipping, may experience performance degradation and lifespan issues, which can lead to reduced production efficiency and safety hazards. Therefore, accurately predicting the State of Health(SOH) and Remaining Useful Life(RUL) of turbines is crucial for achieving predictive maintenance. To address the fault prediction problem of critical mechanical equipment—turbines, a prediction method based on an improved Bidirectional Long Short-Term Memory(BiLSTM) network is proposed. The proposed hybrid model combines the local feature extraction capability of Convolutional Neural Network(CNN), the weight allocation mechanism of Attention Mechanism(AM), and the bidirectional time-series processing ability of BiLSTM to enhance the accuracy and efficiency of fault prediction. By analyzing the operational data of 100 turbines, the experimental results demonstrate that the improved BiLSTM model outperforms other mainstream models such as CNN-BiLSTM, CNN-LSTM, and Recurrent Neural Network(RNN) in terms of prediction accuracy, exhibiting lower Mean Absolute Error(MAE) while maintaining high efficiency and accuracy.
[1] 蔺瑞管，王华伟，车畅畅，等.基于 LSTM 分类器的航空发动机预测性维护模型[J].系统工程与电子技术，2022,44(3):1052-1059.
[2] 房崇鑫，盛震宇，夏明，等.基于CNN-BiLSTM混合神经网络的雷达信号调制方式识别[J].无线电工程，2024,54(6):1440-1445.
[3] 邵霞，段臣续，罗文宇，等.基于融合注意力Bi-LSTM的V2X通信阻塞预测方法[J].无线电工程，2024,54(5):1277-1285.
[4] VAN HOUDT G,MOSQUERA C,NáPOLES G.A Review on the Long Short-term Memory Model[J].Artificial Intelligence Review,2020,53(8):5929-5955.
[5] WANG W L,MAO W J,TONG X L,et al.A Novel Recursive Model Based on a Convolutional Long Short-term Memory Neural Network for Air Pollution Prediction[J].Remote Sensing,2021,13(7):1284.
[6] 黄茂庭，徐金明.使用 CNN (卷积神经网络)-LSTM (长短期记忆) 联合神经网络预测盾构隧道施工引起的地面沉降[J].城市轨道交通研究，2024,27(6):166-171.
[7] 衣思彤，刘雅浓，马耀浥，等.基于贝叶斯优化-卷积神经网络-双向长短期记忆神经网络的锂电池健康状态评估[J].电气技术，2024,25(5):1-10.
[8] 崔明勇，董文韬，卢志刚.基于密度聚类模态分解的卷积神经网络和长短期记忆网络短期风电功率预测[J].现代电力，2024,41(4):631-641.
[9] WANG H,ZHANG J Z,YANG J C.Time Series Forecasting of Pedestrian-level Urban Air Temperature by LSTM:Guidance for Practitioners[J].Urban Climate,2024,56(10):102063.
[10] CHEN J,HU W,ZHANG Y,et al.An Attention-based Teacher-Student Model for Multivariate Short-term Landslide Displacement Prediction Incorporating Weather Forecast Data[J].Journal of Mountain Science,2024,21(8):2739-2753.
[11] JIN R B,CHEN Z H,WU K Y,et al.Bi-LSTM-based Two-stream Network for Machine Remaining Useful Life Prediction[J].IEEE Transactions on Instrumentation and Measurement,2022,71:1-10.
[12] LI X,ZHOU S J,WANG F W,et al.An Improved Sparrow Search Algorithm and CNN-BiLSTM Neural Network for Predicting Sea Level Height[J].Scientific Reports,2024,14(1):45-60.
[13] ZHANG W,JIN S H,BIAN G,et al.A Method for Sound Speed Profile Prediction Based on CNN-BiLSTM-attention Network[J].Journal of Marine Science and Engineering,2024,12(3):414.
[14] SONG J W,PARK Y I,HONG J J,et al.Attention-based Bidirectional LSTM-CNN Model for Remaining Useful Life Estimation[C]//2021 IEEE International Symposium on Circuits and Systems (ISCAS).Daegu:IEEE,2021:1-5.
[15] SUN S,SUN J Z,WANG Z L,et al.Prediction of Battery SOH by CNN-BiLSTM Network Fused with Attention Mechanism[J].Energies,2022,15(12):4428.
[16] ZRAIBI B,OKAR C,CHAOUI H,et al.Remaining Useful Life Assessment for Lithium-Ion Batteries Using CNN-LSTM-DNN Hybrid Method[J].IEEE Transactions on Vehicular Technology,2021,70(5):4252-4261.
[17] HUANG W Q,CHEN Y ,CHEN Y Q,et al.Life Prediction Method of Rolling Bearing Based on CNN-LSTM-AM[J].Journal of Vibroengineering,2024,26(5):1027-1039.
[18] TIAN Y K,WEN J,YANG Y R,et al.State-of-Health Prediction of Lithium-Ion Batteries Based on CNN-BiLSTM-AM[J].Batteries,2022,8(10):155.
[19] WU Z Y,YU S Y,ZHU X N,et al.A Weighted Deep Domain Adaptation Method for Industrial Fault Prognostics According to Prior Distribution of Complex Working Conditions[J].IEEE Access,2019,7:139802-139814.
[20] SHERSTINSKY A.Fundamentals of Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM) Network[J].Physica D:Nonlinear Phenomena,2020,404:12306.
基本信息:
DOI：
中图分类号:TP183;U463
引用信息:
[1]袁灏诚,吴钦木,李佳恒.基于改进BiLSTM算法的车辆涡轮机寿命预测[J].无线电工程,2025,55(05):913-919.
基金信息:
国家自然科学基金(51867006,52267003)~~