摘要
针对风机齿轮箱采用振动信号进行故障诊断时故障准确率较低的问题,提出了基于北方苍鹰优化算法(northern goshawk optimization,NGO)的变分模态分解(variational mode decomposition,VMD)算法(NGO-VMD)和结合一维复合混沌映射(sine map and piece wise linear chaotic map,SPM)与正余弦优化算法(sine cosine algorithm,SCA)的改进NGO随机森林(random forest,RF)优化算法(SSNGO-RF)。首先,采用NGO对VMD的分解层数和惩罚因子进行寻优,避免VMD分解出现模态混叠及模态丢失问题;然后,对分解得到的若干本征模态分量(intrinsic mode function,IMF)逐层计算原始振动信号的相关系数,选取具有相关性的IMF层数进行重构并提取特征向量;最后,将特征向量输入由SPM和SCA改进的NGO优化算法,优化子树棵数和分类特征数的RF模型,进行故障分类。实验结果表明,该算法的故障诊断准确率可达0.9917,能够有效提高采用振动信号对齿轮箱进行故障诊断的准确率。
To solve the problem of fault diagnosis of low accuracy when using vibration signals for fault diagnosis of wind turbine gearbox,a variational model decomposition optimization model based on northern goshawk optimization(NGO-VMD)and one-dimensional composite sine map and piece wise linear chaotic map(SPM),sine cosine algorithm(SCA)and random forest(RF)optimization model based on improved NGO(SSNGO-RF)are proposed.Firstly,NGO is used to optimize the decomposition levels and penalty factors of VMD,in order to avoid modal aliasing and loss issues in VMD decomposition.Afterwards,the intrinsic mode functions(IMF)obtained from the decomposition are calculated layer by layer to obtain the correlation coefficients with the original vibration signal.The IMF layers with correlation are selected for reconstruction and feature vectors are extracted.Finally,the eigenvector is input into a random forest(RF)model that optimizes the number of subtrees and the number of classified features by means of improved SPM and SCA for NGO(SSNGO).The experimental results show that compared with traditional machine learning fault diagnosis methods,the proposed method has significantly improved classification accuracy,with a recognition accuracy of 0.9917.
作者
陈萱
杨永超
袁博洋
郭光华
钟建伟
CHEN Xuan;YANG Yongchao;YUAN Boyang;GUO Guanghua;ZHONG Jianwei(College of Intelligent Systems Science and Engineering,Hubei Minzu University,Enshi 445000,China;Hubei Energy Group New Energy Development Co.,Ltd.,Wuhan 430073,China)
出处
《湖北民族大学学报（自然科学版）》
CAS
2023年第4期520-529,共10页
Journal of Hubei Minzu University:Natural Science Edition
基金
湖北省自然科学基金项目(2022CFB264)。
关键词
风机齿轮箱
故障诊断
北方苍鹰优化
变分模态分解
随机森林
wind turbine gearboxes
fault diagnosis
northern goshawk optimization
variational mode decomposition
random forest
