基于贝叶斯优化改进胶囊网络的轴承故障诊断  

Bearings Fault Diagnosis based on Bayesian OptimizationImproved Capsule Network

作　　者：杜思宇 李丽君[1] 董增寿[1] DU Si-yu;LI Li-jun;DONG Zeng-shou(School of Electronic Information Engineering,Taiyuan University of Science and Technology,Taiyuan 030024,China)

机构地区：[1]太原科技大学电子信息工程学院,太原030024

出　　处：《太原科技大学学报》2025年第1期7-13,共7页Journal of Taiyuan University of Science and Technology

基　　金：国家自然科学基金青年科学基金(61703297);山西省回国留学人员科研资助项目(2021-134,2021-135);山西省基础研究计划(202303021211205)。

摘　　要：由于传统卷积神经网络存在特征提取不完整的问题,提出一种基于贝叶斯优化改进胶囊网络(BO-CapsNet)的轴承故障诊断策略。该模型首先在胶囊层前面加入一个大卷积核,拓宽了胶囊网络的感受野;其次使用贝叶斯优化算法优化胶囊网络的权值。文中将采集的轴承振动信号作为该模型的输入,通过对胶囊网络的不断迭代训练,实现了该模型在准确率方面的提升。选取凯斯西储大学的轴承数据对模型以及算法进行评估,并将该方法与宽核卷积神经网络、胶囊网络以及Adam优化算法进行比较。通过实验验证了该方法可以更加全面提取特征信息并且提高准确率,在轴承故障诊断中具有良好的泛化能力。Due to the problems of incomplete feature extraction in traditional Convolutional Neural Network,a new fault diagnosis strategy of rolling bearings based on Bayesian Optimization and Improved Capsule Network(BO-CapsNet)is proposed.Firstly,a large Convolution kernel is added in front of the CapsNet to widen the receptive field.Secondly,the weight of the CapsNet is optimized by Bayesian Optimization algorithm.In this paper,the collected bearing signals are taken as the input of the model,and the accuracy of the model is improved through the continuous iterative training of the CapsNet.In this paper,bearing data from Case Western Reserve University is selected to evaluate the model and algorithm,the diagnostic results of the proposed method are compared with the those of Wide-kernel Convolution Neural Network,Capsule Network and Adam optimization algorithms.The experimental results show that this method can extract the characteristic information more comprehensively and improve the accuracy of the model,which has a good generalization ability in bearing fault diagnosis.

关 键 词：故障诊断 胶囊网络 贝叶斯优化 特征提取 

分 类 号：TH133.33[机械工程—机械制造及自动化]