基于改进卷积神经网络的旋转机械故障诊断研究  

Study on Fault Diagnosis of Rotating Machinery Based on Improved Convolutional Neural Network

作　　者：李梓煊 刘宁 张晓蕊 罗跃纲 LI Zixuan;LIU Ning;ZHANG Xiaorui;LUO Yuegang(Dalian Minzu University,College of Mechanical and Electrical Engineering,Dalian Liaoning 116000,China;Dalian Minzu University,College of Computer Science and Engineering,Dalian Liaoning 116000,China)

机构地区：[1]大连民族大学机电工程学院,辽宁大连116600 [2]大连民族大学计算机科学与工程学院,辽宁大连116600

出　　处：《辽宁科技学院学报》2025年第1期24-29,共6页Journal of Liaoning Institute of Science and Technology

摘　　要：随着工业化水平的日渐提升,大型旋转机械在生产中的使用愈加广泛。转子系统的设计在不断朝着大型、高速化的趋势发展,引发转子系统发生故障的可能性增加。为较好地提取强噪声覆盖下旋转机械振动信号的故障信息,提高故障诊断识别与分类精度,本研究提出了一种基于自适应白噪声平均经验模态分解(Complete Ensemble Empirical Mode Decomposition with Adaptive Noise,CEEMDAN)的故障诊断方法,该方法通过与样本熵(Sample Entropy,SE)的结合,有效利用了其在衡量数据集纯度方面的优势。样本熵能够快速评估数据集的信息量,从而为故障诊断提供了一种高效的数据预处理手段。进一步,本研究将经过预处理的特征数据集输入到一个经过优化的卷积神经网络(Convolutional Neural Network,CNN)模型中。实验结果表明,优化后的故障诊断模型有较高的故障识别能力。With the increasing level of industrialization,large rotary machinery is used more and more widely in production.The design of rotor system is developing towards large scale and high speed,which increases the possibility of rotor system failure.In order to better extract the fault information of rotating machinery vibration signal covered by strong noise and to improve the accu⁃racy of fault diagnosis,recognition and classification,this paper proposes a fault diagnosis method based on complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN).The method takes advantage of its strength in measuring the pu⁃rity of a dataset by combining it with sample entropy.In particular,sample entropy can quickly evaluate the information content of data set,thus providing an efficient data preprocessing method for fault diagnosis.Further,the preprocessed feature data set is in⁃put into an optimized convolutional neural network(CNN)model.The experimental results show that the optimized fault diagnosis model has high fault recognition ability.

关 键 词：旋转机械 特征提取 卷积神经网络 故障诊断 

分 类 号：TH17[机械工程—机械制造及自动化] TP277[自动化与计算机技术—检测技术与自动化装置]