目的 提升包装装备在高负荷连续化生产场景下电机系统多状态故障诊断的精度和效率,解决传统方法对复杂动态特征捕捉不足及模型轻量化不足等问题。方法 基于时频域多模态信号特征融合,构建选择性状态空间模型（Mamba）。首先,采用快速傅里叶变换提取电机振动信号的全局频域特征,通过连续小波变换生成时频域图像,捕捉局部动态特性;然后,设计基于结构化状态空间建模的特征融合框架,建立电机健康状态动态演化轨迹的微分方程模型;最后,构建轻量化分类器,实现多模态特征协同推理。结果 在CWRU和TBVD电机数据集上的实验结果表明,通过时频域特征融合,使得故障分类准确率达到99.88%,相较于单模态方法提升了7.2%;Mamba模型的参数量仅需2.7×106,比传统诊断模型减少了39.3%以上,推理速度提升了3.8倍。结论 提出的多状态空间建模方法有效实现了包装装备电机故障特征的动态表征与高效融合,在保持模型轻量化的同时,显著提升了诊断精度,为智能维护系统提供了可工程化部署的解决方案。
Abstract
The work aims to enhance the accuracy and efficiency of multi-state fault diagnosis for motor systems in high-load continuous packaging equipment, addressing the limitations of traditional methods in capturing complex dynamic features and achieving model lightweighting. A selective state space model (Mamba) was constructed based on time-frequency domain multimodal signal feature fusion. Fast Fourier Transform was firstly applied to extract global frequency-domain features from motor vibration signals, while time-frequency domain images were generated by continuous wavelet transform to capture local dynamic characteristics. A structured state space modeling framework was then designed to establish differential equation models for dynamic evolution trajectories of motor health states. Finally, a lightweight classifier was developed for multimodal feature collaborative inference. Experiments on CWRU and TBVD datasets demonstrated that time-frequency feature fusion achieved 99.88% fault classification accuracy, showing 7.2% improvement over single-modal methods. The Mamba model required only 2.7×106 parameters, representing over 39.3% reduction compared to traditional diagnosis models, with 3.8× faster inference speed. The proposed multi-state space modeling method effectively realizes dynamic characterization and efficient fusion of motor fault features in packaging equipment, significantly improving diagnosis accuracy while maintaining model lightweighting, providing an engineering-deployable solution for intelligent maintenance systems.
关键词
时频域特征融合 /
包装装备故障诊断 /
选择性状态空间模型 /
轻量化模型 /
智能化故障诊断
Key words
time-frequency feature fusion /
packaging equipment fault diagnosis /
selective state space model (Mamba) /
lightweight model /
intelligent fault diagnosis
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 张秋昕, 周进军, 陈锋, 等. 基于DS-PCA模型的包装设备滚动轴承故障诊断方法研究[J]. 包装工程, 2023, 44(9): 224-231.
ZHANG Q X, ZHOU J J, CHEN F, et al.Fault Diagnosis Method of Rolling Bearings of Packaging Equipment Based on DS-PCA Model[J]. Packaging Engineering, 2023, 44(9): 224-231.
[2] HUANG M, SHENG C X, RAO X.Conditional Variational Autoencoder and Generative Adversarial Network-based Approach for Long-tailed Fault Diagnosis for the Motor System[J]. Measurement, 2025, 242: 116116.
[3] WANG J J, ZHANG S H, SHEN H, et al.A Review of Multi-Source Fault Diagnosis and Fault-Tolerant Control Methods for Hybrid Electric Vehicles[J]. International Journal of Green Energy, 2025: 1-15.
[4] LEI Y G, YANG B, JIANG X W, et al.Applications of Machine Learning to Machine Fault Diagnosis: A Review and Roadmap[J]. Mechanical Systems and Signal Processing, 2020, 138: 106587.
[5] WENG C Y, LU B C, CHEN L M, et al.A Novel Progressive Domain Separation Network with Multi-metric Ensemble Quantification for Open Set Fault Diagnosis of Motor Bearings[J]. Advanced Engineering Informatics, 2025, 64: 103060.
[6] GLOWACZ A, SULOWICZ M, ZIELONKA J, et al.Acoustic Fault Diagnosis of Three-Phase Induction Motors Using Smartphone and Deep Learning[J]. Expert Systems with Applications, 2025, 262: 125633.
[7] YOU K S, WANG P Z, GU Y K.Toward Efficient and Interpretative Rolling Bearing Fault Diagnosis via Quadratic Neural Network with Bi-LSTM[J]. IEEE Internet of Things Journal, 2024, 11(13): 23002-23019.
[8] 方群玲, 马智宇, 张锐, 等. MDCNet轴承智能故障诊断方法研究[J]. 包装工程, 2023, 44(9): 218-223.
FANG Q L, MA Z Y, ZHANG R, et al.Intelligent Fault Diagnosis Method of Bearing Based on MDCNet[J]. Packaging Engineering, 2023, 44(9): 218-223.
[9] YOU K S, WANG P Z, HUANG P, et al.A Sound-Vibration Physical-Information Fusion Constraint-Guided Deep Learning Method for Rolling Bearing Fault Diagnosis[J]. Reliability Engineering & System Safety, 2025, 253: 110556.
[10] 李准峰, 范磊, 赵旭东, 等. 切丝机轴承时频特性诊断方法研究[J]. 包装工程, 2022, 43(15): 322-329.
LI Z F, FAN L, ZHAO X D, et al.Diagnosis Method for Time-Frequency Characteristic of Bearing of Tobacco Cutter[J]. Packaging Engineering, 2022, 43(15): 322-329.
[11] YOU K S, QIU G Q, GU Y K.An Efficient Lightweight Neural Network using BiLSTM-SCN-CBAM with PCA-ICEEMDAN for Diagnosing Rolling Bearing Faults[J]. Measurement Science and Technology, 2023, 34(9): 094001.
[12] WANG P, SONG Y F, WANG X D, et al.MD-BiMamba: An Aero-Engine Inter-shaft Bearing Fault Diagnosis Method Based on Mamba with Modal Decomposition and Bidirectional Features Fusion Strategy[J]. Measurement, 2025, 242: 115870.
[13] CHEN Z Y, HUANG H Z, DENG Z W, et al.Shrinkage Mamba Relation Network with Out-of-Distribution Data Augmentation for Rotating Machinery Fault Detection and Localization under Zero-Faulty Data[J]. Mechanical Systems and Signal Processing, 2025, 224: 112145.
[14] TRUONG G B, TRAN T T, THAN N L, et al.SC-MambaFew: Few-Dhot Learning Based on Mamba and Selective Spatial-Channel Attention for Bearing Fault Diagnosis[J]. Computers and Electrical Engineering, 2025, 123: 110004.
[15] ZHANG Q, XU C, LI J, et al.LLM-TSFD: An Industrial Time Series Human-in-the-Loop Fault Diagnosis Method Based on a Large Language Model[J]. Expert Systems with Applications, 2025, 264: 125861.
[16] Case Western Reserve University. Bearing Data Center[EB/OL].[2022-10-11]. https://engineering.case.edu/bearingdatacenter/.
[17] YOU K S, QIU G Q, GU Y K.Rolling Bearing Fault Diagnosis using Hybrid Neural Network with Principal Component Analysis[J]. Sensors, 2022, 22(22): 8906.
[18] KUMAR D, MEHRAN S, SHAIKH M Z, et al.Triaxial Bearing Vibration Dataset of Induction Motor under Varying Load Conditions[J]. Data in Brief, 2022, 42: 108315.
基金
国家级技能大师工作室项目（人社部函[2019]197号）
PDF(3007 KB)
渝公网安备 50010702501716号 渝ICP备15012534号-2
