梯度修正Logistic回归模型的滚动轴承故障诊断  

Fault Diagnosis of Rolling Bearing Based on Gradient Adaptive Corrected Logistic Regression Model

作　　者：王钰烁 熊彬羽 印明昂[2] WANG Yushuo;XIONG Binyu;YIN Ming′ang(CRRC Changchun Railway Vehicles Co.,Ltd.Changchun,130062,China;School of Mechanical Engineering&Automation,Northeastern University Shenyang,110819,China)

机构地区：[1]中车长春轨道客车股份有限公司,长春130062 [2]东北大学机械工程与自动化学院,沈阳110819

出　　处：《振动．测试与诊断》2024年第6期1106-1111,1245,共7页Journal of Vibration,Measurement & Diagnosis

基　　金：中央高校基本科研业务费专项资金资助项目(N2103015)。

摘　　要：为准确解析高维信号数据,减少特征选择产生的信息损失,正确评估滚动轴承性能退化状态,提出一种基于梯度自适应修正(gradient adaptive corrected,简称GAC)Logistic回归模型的滚动轴承状态诊断方法,找出了GAC算法的收敛上界。首先,提取轴承检测信号的若干典型特征参量作为模型变量;其次,对信号特征数据进行降噪与归一化处理;最后,在保持数据维度的前提下建立GAC-Logistic回归模型,并对滚动轴承的性能状态进行评估。数据验证结果表明,该方法可提高模型的构建效率和对轴承状态的判定准确率,具有良好的鲁棒性,可有效降低信号随机波动对轴承状态评估产生的影响,GAC-Logistic回归模型的平均拟合正确率和平均验证正确率分别达到了99.08%和98.17%。In order to accurately analyze the high-dimensional signal data,reduce the information loss caused by feature selection,and correctly evaluate the performance degradation state of rolling bearing,a fault diagnosis method based on gradient adaptive corrected(GAC)Logistic regression model is proposed,and the upper bound of convergence of the GAC algorithm is found.Firstly,typical characteristic parameters of bearing detection signal are extracted as model variables;Secondly,denoise and normalize the signal characteristic data;Finally,the GAC-Logistic regression model is established on the premise of maintaining the data dimension,and the performance state of rolling bearing is evaluated.The experimental results show the proposed method can improve the efficiency of model construction and the accuracy of bearing state discrimination,and has good robustness,which can effectively reduce the influence of random signal fluctuation on bearing state evaluation.The average fitting accuracy and average verification accuracy of GAC-Logistic regression model reaches 99.08%and 98.17%,respectively.

关 键 词：LOGISTIC回归 自适应梯度 故障诊断 滚动轴承 

分 类 号：TH165.3[机械工程—机械制造及自动化] TH17