基于能量熵和自适应神经模糊推理系统的齿轮故障诊断  

Gear fault diagnosis based on energy entropy and adaptive neural fuzzy inference system

作　　者：高淑婷 刘裕鹏 王满意 GAO Shuting;LIU Yupeng;WANG Manyi(Nanjing University of Science and Technology,Nanjing 210094,China;The 32381st Unit of the Chinese People’s Liberation Army,Beijing 100000,China)

机构地区：[1]南京理工大学机械工程学院,南京210094 [2]中国人民解放军第32381部队,北京100000

出　　处：《兵器装备工程学报》2024年第4期294-300,共7页Journal of Ordnance Equipment Engineering

基　　金：国家杰出青年科学基金项目(61701237)。

摘　　要：为了解决齿轮振动信号中出现噪声污染严重,故障特征信息提取困难的问题,提出了基于能量熵和自适应神经模糊推理系统(ANFIS)的齿轮故障诊断方法。将预处理后的振动信号作自适应噪声完备经验模式(CEEMDAN)分解,可获得不同尺度的本征模态函数(IMF);由于各IMF包含主要故障特征信息,且不同故障状态下各IMF分布有明显不同,因此通过计算能量熵来量化故障时频域特征,构造出表征模态分量信息的特征向量;以此输入ANFIS进行样本的学习和训练,在自适应调整网络参数和隶属函数后,获得最优ANFIS。实验结果表明:该方法诊断结果准确率近乎100%,可有效地识别故障类型。In order to solve the problem of severe noise pollution and difficulty in extracting fault feature information in gear vibration signals,a gear fault diagnosis method based on energy entropy and adaptive neural fuzzy inference system(ANFIS)is proposed.By decomposing the preprocessed vibration signal into adaptive noise complete empirical mode decomposition(CEEMDAN),eigenmode functions(IMFs)of different scales can be obtained;Due to the fact that each IMF contains main fault feature information and the distribution of each IMF is significantly different under different fault states,the time-frequency domain features of faults are quantified by calculating energy entropy to construct feature vectors that represent modal component information;Using this input ANFIS for sample learning and training,the optimal ANFIS is obtained by adaptively adjusting network parameters and membership functions.The experimental results show that the diagnostic accuracy of this method is almost 100%,and it can effectively identify fault types.

关 键 词：齿轮 能量熵 经验模态分解 自适应神经模糊推理系统 故障诊断 

分 类 号：TH132.41[机械工程—机械制造及自动化] TP311.1[自动化与计算机技术—计算机软件与理论]