| 91 | 0 | 431 |
| 下载次数 | 被引频次 | 阅读次数 |
针对生产线故障诊断中高维长时序特征提取困难、故障样本稀缺及特征耦合性强导致的特征提取不充分问题，本文提出了一种融合卷积神经网络（convolutional neural networks,CNN）、双向长短期记忆网络（bidirectional long short-term memory networks,BiLSTM）和自注意力机制的故障诊断模型.首先，提取动态故障数据区间并构建标签化的初始数据，以增强数据表征能力，缓解故障样本不均衡问题.其次，采用CNN进行初始化局部特征提取，并通过BiLSTM进一步提取时序特征.在此基础上，引入注意力机制自适应优化故障特征学习权重，以增强特征表示.随后，优化Dropout抑制网络在训练过程中存在的过拟合问题，从而构建CNN-BiLSTM-Transformer故障诊断模型.最后，在公开数据集上进行实验，以验证模型性能，通过基线实验和系统消融实验，进一步验证了模型的优势和各模块的有效性.
Abstract：To address the challenges of high-dimensional long-sequence feature extraction,scarce fault samples, and strong feature coupling in production line fault diagnosis, which lead to insufficient feature representation, a fault diagnosis model is proposed by integrating convolutional neural network(convolutional neural networks, CNN), bidirectional long short-term memory(bidirectional long short-term memory networks, BiLSTM), and self-attention mechanism.First, dynamic fault data intervals are extracted to construct labeled initial datasets, enhancing data representation and mitigating sample imbalance. Second, CNN is employed for initial local feature extraction, followed by Bi LSTM for further temporal feature learning. Third, an attention mechanism is introduced to adaptively optimize feature weights, strengthening discriminative representation. Additionally, Dropout is optimized to suppress overfitting during training, forming the proposed CNN-BiLSTM-Transformer diagnostic model. Finally, experiments on public datasets validate the model's performance. Baseline comparisons and systematic ablation studies further demonstrate the superiority of the proposed approach and the effectiveness of each module.
[1]沈南燕，武星，李静，等.自动化生产线中关键设备的预维护策略研究[J].机械工程学报，2020,56(21):231-240.
[2]谭帅，王一帆，姜庆超，等.基于不同故障传播路径差异化的故障诊断方法[J].自动化学报，2025,51(1):161-173.
[3] QIAN C H, ZHU J J,SHEN Y H,et al. Deep transfer learning in mechanical intelligent fault diagnosis:application and challenge[J]. Neural Processing Letters,2022,54(3):1-23.
[4] WANG N,YANG F,ZHANG R D,et al. Intelligent fault diagnosis for chemical processes using deep learning multimodel fusion[J]. IEEE transactions on cybernetics,2020,52(7):7121-7135.
[5] HOANG D T,KANG H J. A survey on deep learning based bearing fault diagnosis[J]. Neurocomputing,2018,335:327-335.
[6] YUAN Y , WANG J. End-to-end deep learning for fault diagnosis of rotating machinery[J]. IEEE Transactions on Industrial Electronics,2020,67(5):4123-4131.
[7] LECUN Y,BOTTOU L,BENGIO Y,et al. Gradient-based learning applied to document recognition[J].Proceedings of the IEEE,2015,86(11):2278-2324.
[8] GOODFELLOW I,BENGIO Y,COURVILLE A. Deep learning[M]. Cambridge:MIT Press,2016.
[9] ZHANG C Y,BENGIO S, HARDT M,et al. Understanding deep learning(still)requires rethinking generalization[J]. Communications of the Acm,2021,64(3):107-115.
[10] KUNIHIKO F. Artificial vision by deep CNN neocognitron[J]. IEEE Transctionson Systems,Man,and Cybernetics:Systems,2021,51(1):76-90.
[11] CHU Y D,FEI J T,HOU S X,et al. Adaptive global sliding-mode control for dynamic systems using double hidden layer recurrent neural network structure[J]. Engineering Applications of Artificial Intelligence,2023,123(4):1-13.
[12] BHARATHEEDASAN K,MAITY T,KUMARASWAMIDHAS L,et al. Enhanced fault diagnosis and remaining useful life prediction of rolling bearings using a hybrid multilayer perceptron and LSTM network model[J]. Alexandria Engineering Journal,2025,115:355-369.
[13] LI W X,SHANG Z W,ZHANG J,et al. A novel unsupervised anomaly detection method for rotating machinery based on memory augmented temporal convolutional autoencoder[J]. IEEE Access,2021,9(1):123456-123465.
[14] CAO L Y,LI H D,LIU X R , et al. Semantic segmentation of plant leaves based on generative adversarial network and attention mechanism[J]. IEEE Access,2022,10:76310-76317.
[15] CHEN Y Q,YIN X Y,ZHANG R D,et al. Reinforced convolutional neural network fault diagnosis of industrial production systems[J]. Chemical Engineering Science,2024,299:120466.
[16] LU Z Q,LINAG L Y,ZHU J,et al. Rotating machinery fault diagnosis under multiple working conditions via a time-series transformer enhanced by convolutional neural network[J]. IEEE Transactions on Instrumentation and Measurement,2023,72:1-11.
[17]陆继翔，张琪培，杨志宏，等.基于CNN-LSTM混合神经网络模型的短期负荷预测方法[J].电力系统自动化，2019,43(8):131-137.
[18]杨志淳，闵怀东，杨帆，等.基于卷积双向长短期记忆网络的微网继电保护故障诊断技术[J].太阳能学报，2025,46(1):420-428.
[19]杜先君，巩彬，余萍，等.基于CBAM-CNN的模拟电路故障诊断[J].控制与决策，2022,37(10):2609-2618.
[20]徐新，蒋波涛，曹雯．蝗虫优化神经网络在变压器故障诊断中的应用[J].电网与清洁能源，2021,37(5):17-23.
[21] XU X,JIANG B T,CAO W. Application of locust optimized neural network in transformer fault diagnosis[J]. Power Grid and Clean Energy,2021,37(5):17-23.
[22]孙巍伟，毛亦鹏，郑家春，等.基于改进Transformer-BiLSTM的人体活动识别模型[J].电子测量技术，2024,47(17):54-61.
[23]董方旭，咸日常，咸日明，等.基于细菌觅食算法优化的电力变压器故障诊断技术[J].电测与仪表，2018,55(19):34-40.
[24] SUBAKAN C,RAVANELLI M,CORNELL S,et al. Attention is all you need in speech separation[C].2021 IEEE International Conference on Acoustics,Speech and Signal Processing,2021,21-25.
[25]孟祥福，石皓源.基于Transformer模型的时序数据预测方法综述[J].计算机科学与探索，2025,19(1):45-64.
[26]张彼德，邱杰，娄广鑫，等.基于CNN和Transformer的轻量化电能质量扰动识别模型[J].电力工程技术，2025,44(1):69-78.
[27] ZHANG X , CHEN X , HE Z. An ensemble deep random forest for fault diagnosis of industrial systems[J]. IEEE Transactions on Industrial Informatics,2020,16(1):299-310.
[28] ZHAO R,YAN R,CHEN Z,et al. Deep learning and its applications to machine health monitoring[J].Mechanical Systems and Signal Processing,2019,115:213-237.
[29] JANSSENS O,SLAVKOVIKJ V,VERVISCH B , et al. Convolutional neural network basedfault detection for rotating machinery[J]. Journal of Sound and Vibration,2016,377:331-345.
基本信息:
中图分类号:TP277;TP183
引用信息:
[1]幸利莎,吴一帆,刘元源,等.基于注意力机制和卷积神经网络的生产线故障诊断[J].汕头大学学报(自然科学版),2025,40(04):38-50.
基金信息:
2024年国家级大学生创新创业项目（202410560029）