基于LMD云模型与PSO-KELM的齿轮箱故障诊断  

Gearbox Fault Diagnosis Based on the LMD Cloud Model and PSO-KELM

作　　者：赵小惠[1] 谭琦 胡胜 杨文彬 郇凯旋 张智杰 Zhao Xiaohui;Tan Qi;Hu Sheng;Yang Wenbin;Huan Kaixuan;Zhang Zhijie(School of Mechanical and Electrical Engineering,Xi´an Polytechnic University,Xi´an 710048,China)

机构地区：[1]西安工程大学机电工程学院,陕西西安710048

出　　处：《机械传动》2023年第2期157-163,共7页Journal of Mechanical Transmission

基　　金：国家自然科学基金(72001166);陕西省科技计划项目(2022JQ-721);陕西省社会科学界联合会项目(20ZD195-59)。

摘　　要：由于齿轮箱故障振动信号具有非平稳性与不确定性的特点,导致齿轮箱故障诊断精度较低。针对该问题提出一种基于局部均值分解(LMD)云模型特征提取结合粒子群优化(PSO)核极限学习机(KELM)的齿轮箱故障诊断方法。首先,将故障振动信号用LMD分解得到若干PF分量,并利用相关系数原则筛选出相关性较高的PF分量;其次,在云模型中输入筛选后的PF分量,采用逆向云发生器对特征向量进行提取并输入到PSO-KELM中进行故障诊断;最后,利用QPZZ-Ⅱ实验台齿轮箱实测数据对该方法进行了性能分析。结果表明,该方法识别精度为97.65%,与多种方法进行对比,该方法具备最佳识别性能。The characteristics of non-smoothness and uncertainty of gearbox fault vibration signal lead to the low accuracy of gearbox fault diagnosis.To address this problem,a gearbox fault diagnosis method based on local mean decomposition(LMD)cloud model feature extraction combined with particle swarm optimization(PSO)kernel extreme learning machine(KELM)is proposed.Firstly,the fault vibration signal is decomposed by LMD to obtain several PF components,and the PF components with higher correlation are screened out using the correlation coefficient principle.Secondly,the screened PF components are input into the cloud model,and the feature vectors are extracted using the inverse cloud generator and input into PSO-KELM for fault diagnosis.Finally,the performance of the method is analyzed using the measured data of the QPZZ-Ⅱtest-bed gearbox.The results show that the recognition accuracy of the method is 97.65%,and compared with various methods this method has the best recognition performance.

关 键 词：齿轮箱 故障诊断 局部均值分解 云模型 粒子群优化核极限学习机 

分 类 号：TH132.41[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]