摘要
针对齿轮箱振动信号中的背景噪声过大影响故障特征质量,进而降低故障识别准确率的问题,提出了一种基于改进自适应噪声完备集成经验模态分解(ICEEMDAN)、改进多尺度加权排列熵(IMWPE)、利用线性判别分析(LDA)、蝴蝶优化算法(BOA)优化支持向量机(SVM)的齿轮箱故障诊断方法(ICEEMDAN-IMWPE-LDA-BOA-SVM)。首先,采用ICEEMDAN对齿轮箱振动信号进行了分解,生成了一系列从低频到高频分布的本征模态函数分量;接着,基于相关系数筛选出包含主要故障信息的本征模态函数分量,进行了信号重构,降低了信号的噪声;随后,提出了改进多尺度加权排列熵的非线性动力学指标,并利用其提取了重构信号的故障特征,以构建反映齿轮箱故障特性的故障特征;然后,利用线性判别分析(LDA)对原始故障特征进行了压缩,以构建低维的故障特征向量;最后,采用蝴蝶优化算法(BOA)对支持向量机(SVM)的惩罚系数和核函数参数进行了优化,以构建参数最优的故障分类器,对齿轮箱的故障进行了识别;基于齿轮箱复合故障数据集对ICEEMDAN-IMWPE-BOA-SVM方法进行了实验和对比分析。研究结果表明:该方法能够较为准确地识别齿轮箱的不同故障类型,准确率达到了99.33%,诊断时间只需5.31 s,在多个方面都优于其他对比方法,在齿轮箱的故障诊断中更具有应用潜力。
Aiming to address the issue of excessive background noise in gearbox vibration signals affecting the quality of fault features and thereby reducing the accuracy of fault identification,a gearbox fault diagnosis method(ICEEMDAN-IMWPE-LDA-BOA-SVM)based on improved adaptive noise complete ensemble empirical mode decomposition(ICEEMDAN),improved multi-scale weighted permutation entropy(IMWPE),linear discriminant analysis(LDA),butterfly optimization algorithm(BOA),and support vector machine(SVM)optimization was proposed.Firstly,ICEEMDAN was used to decompose the gearbox vibration signal and generate a series of intrinsic mode function components distributed from low frequency to high frequency.Next,based on the correlation coefficient,the intrinsic mode function components containing the main fault information were selected for signal reconstruction to reduce signal noise.Subsequently,a nonlinear dynamic index for improving multi-scale weighted permutation entropy was proposed,and it was used to extract fault features of the reconstructed signal to construct fault features that reflect the fault characteristics of the gearbox.Then,linear discriminant analysis(LDA)was used to compress the original fault features to construct a low dimensional fault feature vector.Finally,the butterfly optimization algorithm(BOA)was used to optimize the penalty coefficients and kernel function parameters of the support vector machine(SVM),in order to construct a fault classifier with the optimal parameters,achieving fault identification of the gearbox.Experimental and comparative studies were conducted on the ICEEMDAN-IMWPE-BOA-SVM method based on the gearbox composite fault dataset.The research results show that the method can accurately identify different fault types of gearboxes,with an accuracy rate of 99.33%and a diagnosis time of only 5.31 s.It is superior to other comparative methods in multiple aspects and has more potential for application in fault diagnosis of gearboxes.
作者
王洪
张锐丽
吴凯
WANG Hong;ZHANG Ruili;WU Kai(Intelligent Manufacturing College,Yibin Vocational and Technical College,Yibin 644003,China;School of Electromechanical Engineering,Chengdu University of Technology,Chengdu 610059,China)
出处
《机电工程》
CAS
北大核心
2023年第11期1709-1717,共9页
Journal of Mechanical & Electrical Engineering
基金
国家自然科学基金资助项目(51875479)。
关键词
故障特征提取
信号分解及信号重构
特征降维
改进自适应噪声完备集成经验模态分解
改进多尺度加权排列熵
线性判别分析
蝴蝶优化算法
支持向量机
fault feature extraction
signal decomposition and signal reconstruction
feature dimension reduction
improved complete ensemble empirical mode decomposition with adaptive noise(ICEEMDAN)
improved multi-scale weighted permutation entropy(IMWPE)
linear discriminant analysis(LDA)
butterfly optimization algorithm(BOA)
support vector machine(SVM)
