首页|AOA-CEEMDAN和融合特征在齿轮箱故障诊断中的应用

AOA-CEEMDAN和融合特征在齿轮箱故障诊断中的应用

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
  • 维普
自适应噪声完备集成经验模态分解(CEEMDAN)的参数由于是人为设置的,从而会导致其信号的分解不彻底.针对这一问题,提出了一种基于算术优化算法(AOA)优化CEEMDAN、融合特征和随机森林(RF)的齿轮箱故障诊断方法.首先,采用AOA算法对CEEMDAN方法的关键参数进行自适应选取,并采用优化后的CEEMDAN方法对齿轮箱振动信号进行了分解,生成若干个本征模态函数(IMF);随后,利用相关系数准则选择了前4 阶IMF分量作为故障敏感分量;接着,利用由注意熵和散度熵组成的融合特征提取方法挖掘了故障敏感分量的故障特征,得到了故障敏感特征样本;最后,将表征齿轮箱故障特性的故障特征输入至RF多故障分类器中,建立了故障分类模型,完成了齿轮箱的故障识别;利用QPZZ-Ⅱ型齿轮箱数据集进行了实验,并将其结果与采用其他方法所得结果进行了对比.研究结果表明:相较于原始CEEMDAN,优化后的CEEMDAN能够更加准确地分解非线性齿轮箱振动信号,故障识别准确率提高了4％;相较于单一的故障特征,融合特征能够更加准确地表征齿轮箱的故障状态,故障识别准确率分别提高了3.2％和8％.基于AOA-CEEMDAN和融合特征提取以及RF分类器的故障诊断方法为齿轮箱的故障特征提取和故障诊断提供一种可行的思路和方案.
AOA-CEEMDAN and fusion features and its application in gearbox fault diagnosis
Aiming at the defect of incomplete signal decomposition caused by artificial setting of parameters of complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN),a gearbox fault diagnosis method based on arithmetic optimization algorithm(AOA)optimized CEEMDAN,fusion features and random forest(RF)was proposed.Firstly,AOA algorithm was used to adaptively select the key parameters of CEEMDAN method,and the optimized CEEMDAN method was used to decompose the gearbox vibration signal to generate several intrinsic mode functions(IMF).Then,the first 4 IMF components were selected as fault sensitive components by correlation coefficient criterion.Next,a fusion feature extraction method composed of attention entropy and diversity entropy was used to mine the fault features of fault sensitive components and obtain fault sensitive feature samples.Finally,the fault features representing the gearbox fault characteristics was input into RF multi-fault classifier to establish a fault classification model and complete the fault identification of the gearbox.Experiments were conducted using the QPZZ-Ⅱ gearbox dataset,and comparisons were made with other methods.The research results show that the optimized CEEMDAN method can more accurately decompose nonlinear gearbox vibration signals compared to the original CEEMDAN method,and the fault identification accuracy is improved by 4％.Comparing with a single fault feature,fused features can more accurately represent the fault status of the gearbox,and the fault recognition accuracy is respectively improved by 3.2％and8％.The proposed method provides a feasible idea and scheme for fault feature extraction and fault diagnosis of gearbox.

gearboxintrinsic mode functions(IMF)arithmetic optimization algorithm(AOA)complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)random forest(RF)

马卫东、刘子全、姚楠、朱雪琼

展开 >

西北师范大学 物理与电子工程学院,甘肃 兰州 730070

兰州石化职业技术大学 电子电气工程学院,甘肃 兰州 730087

国网江苏省电力有限公司电力科学研究院,江苏 南京 210000

齿轮箱 本征模态函数 算术优化算法 自适应噪声完备集成经验模态分解 随机森林

中央高校基本科研业务费专项

2022TS027

2024

10.3969/j.issn.1001-4551.2024.05.007

机电工程
浙江大学　浙江省机电集团有限公司

机电工程

CSTPCD北大核心
影响因子：0.785
ISSN：1001-4551
年,卷(期)：2024.41(5)
  • 23