基于蜣螂优化算法和K近邻算法的轴承特征选择和故障诊断  

Bearing Feature Selection and Fault Diagnosis Based on Dung Beetle Optimization Algorithm and K-Nearest Neighbor Algorithm

作　　者：徐福斌 杨洪武 陆晔 张伟 朱静 邓艾东[4] Xu Fubin;Yang Hongwu;Lu Ye;Zhang Wei;Zhu Jing;Deng Aidong(National Energy Group Taizhou Power Generation Co.,Ltd.,Taizhou 225327,China;National Energy Xinzhu Interconnection Technology Co.,Ltd.,Beijing 102200,China;School of Vehicle and Transportation Engineering,Henan University of Science and Technology,Luoyang 471003 China;School of Energy and Environment,Southeast University,Nanjing 210096,China)

机构地区：[1]国家能源集团泰州发电有限公司,泰州225327 [2]国能信控互联技术有限公司,北京102200 [3]河南科技大学车辆与交通工程学院,洛阳471003 [4]东南大学能源与环境学院,南京210096

出　　处：《信息化研究》2024年第4期8-19,共12页INFORMATIZATION RESEARCH

摘　　要：针对风电机组传动系统滚动轴承在故障诊断中容易受到冗余特征的影响,从而导致故障诊断的准确率和效率不高的问题,本文提出了一种基于蜣螂优化(DBO)算法和K最近邻(KNN)算法的轴承特征选择和故障诊断方法。该方法首先通过时域、频域分析提取了与滚动轴承故障相关的20个特征数据,然后对特征进行包括归一化处理和特征集划分在内的特征处理;接着以DBO算法的适应度为评价参数,利用DBO算法的路径选择能力寻求最优特征子集;最后通过测试验证所选的特征子集对于KNN分类准确率的优化效果。实验结果表明,标准差(SD)和平均绝对差值(MAD)这两个特征参数作为KNN分类器的输入数据时可以达到95.10%的分类准确率。此方法在大幅度减少特征数量的同时,提高了轴承的故障诊断准确率。Aiming at the problem that rolling bearings of wind turbine drive train system are easily affected by redundant features in fault diagnosis,which leads to poor accuracy and efficiency of fault diagnosis.In this paper,a bearing feature selection and fault diagnosis method based on Dung Beetle Optimization(DBO)and K-Nearest Neighbors(KNN)is proposed.The method firstly extracts 20 feature data related to rolling bearing faults through time and frequency domain analysis,and then performs feature processing including normalization and feature set partitioning on the features;then seeks the optimal feature subset using the DBO algorithm with the adaptation degree of the DBO-KNN network as the target parameter;and finally verifies the optimization effect of the selected feature subset on the accuracy of the KNN classification through testing.The results of the The test results show that the two feature parameters,standard deviation and mean absolute difference,can achieve a classification accuracy of 75.10%when they are used as the input data of the KNN classifier;this method improves the fault diagnosis accuracy of the bearings while significantly reducing the number of features.

关 键 词：风电机组 传动系统滚动轴承 特征选择 故障诊断 蜣螂优化算法 K最近邻算法 

分 类 号：TH133.33[机械工程—机械制造及自动化]