摘要
为了实现齿轮箱健康状态早期识别以及定量化评估,提出了一种基于自组织特征映射(SOM)神经网络的齿轮箱健康评估方法。该方法首先对齿轮箱振动信号进行小波降噪处理,减小噪声对信号分析的干扰;然后提取可以描述信号特征的时域参数,归一化后输入到SOM神经网络,计算输入数据与正常数据最佳匹配单元之间的欧氏距离;构造一个用于定量化表征齿轮箱健康状况的无量纲参数——健康值CV,其取值范围为[0,1],CV值越接近1表明系统越健康,反之则表示系统处于故障状态。同时进行数据包络分析,实现状态识别,并基于采集的变速箱振动加速度数据进行验证。试验结果表明,该方法计算所得健康值CV可以较好反映变速箱健康状况,且可以准确识别变速箱状态,具有工程意义。
To realize the early recognition and quantitative evaluation of gearbox health status,a gearbox health assessment method based on self-organizing feature map(SOM)neural network was proposed.Wavelet de-noising was applied to the vibration signal of gearbox to reduce the interference of noise on signal analysis.Then,time-domain indexes which can describe signal features were extracted.After normalization,these features were input into the SOM neural network to calculate the Euclidean distance between the input data and the optimal matching unit of the normal data.A dimensionless parameter based on Euclidean distance was constructed to quantitatively characterize the health status of gearbox,which is named Confidence Value(CV).The CV is in the range of[0,1].The closer the CV is to 1,the healthier the system is;otherwise,it means that the system is in fault status.At the same time,data envelopment analysis(DEA)was carried out to realize the status recognition,and the vibration data of a gearbox was used to verify.The test results show that the CV calculated by the method can reflect the health status of the gearbox well.And it can accurately identify the status of the gearbox,which has certain engineering significance.
作者
邵旋
康兴无
曹向荣
王旭平
袁晓静
孙磊
SHAO Xuan;KANG Xingwu;CAO Xiangrong;WANG Xuping;YUAN Xiaojing;SUN Lei(Rocket Force University of Engineering,Xi’an 710025,China;Beijing Institution of Space Launch Technology,Beijing 100076,China;The No.96864 th Troop of PLA,Luoyang 471000,China)
出处
《兵器装备工程学报》
CSCD
北大核心
2021年第7期246-251,共6页
Journal of Ordnance Equipment Engineering
关键词
小波降噪
时域参数
归一化
SOM神经网络
健康值CV
数据包络分析
wavelet denoising
time domain parameters
normalization
SOM neural network
Confidence Value(CV)
data envelopment analysis(DEA)
