摘要
针对液压泵振动信号具有非线性、非平稳性,以及信噪比低等特点,提出了基于完备总体经验模态分解和模糊熵结合的液压泵性能退化特征提取方法。首先,使用完备总体经验模态分解方法对液压泵振动信号进行分解,得到若干个固有模态函数分量。其次,求取各个分量与原始信号的相关性,选取相关性较高的前几个分量作为有效分量并求其模糊熵,实现液压泵的退化特征提取,形成特征向量。最后,以液压泵不同退化状态下的实测数据为例,使用基于变量预测模型的模式识别方法对提取的特征向量进行验证。实验结果表明,该液压泵退化特征提取方法具有较高的精度,使退化状态识别的准确率提高到了100%。
Aiming at the fact that the vibration signals of hydraulic pumps have the characteristics of nonlinear,nonstationary and low signal to noise ratio, a feature extraction method based on complete ensemble empirical mode decomposition and fuzzy entropy integrated is proposed. Firstly,the vibration signals of the hydraulic pump are decomposed into several intrinsic mode functions by means of complete ensemble empirical mode decomposition. Secondly,the correlation coefficients between each intrinsic mode function and the original signal are calculated,and the components with higher correlation coefficients are selected to obtain their fuzzy entropies,then the degradation characteristics are obtained.Finally,taking the measured data in different degradation states of hydraulic pump as an example,the variable predictive model based class discriminate method is used to verify the effectiveness of the proposed feature extracted method. The experimental results show that the method has higher precision in extracting the hydraulic pump degradation characteristics,and the accuracy rate of degenerate state identification is increased to 100%.
作者
姜万录
孔德田
李振宝
佟祥伟
岳文德
JIANG Wan-lu;KONG De-tian;LI Zhen-bao;TONG Xiang-wei;YUE Wen-de(Hebei Province Key Laboratory of Heavy Machinery Fluid Power Transmission and Control,Yanshan University,Qinhuangdao,Hebei 066004,China;Heavy-duty Intelligent Manufacturing Equipment Innovation Center of Hebei Province,Yanshan University,Qinhuangdao,Hebei 066004,China)
出处
《计量学报》
CSCD
北大核心
2020年第2期202-209,共8页
Acta Metrologica Sinica
基金
国家自然科学基金(51875498,51475405)
河北省自然科学基金重点项目(E2018203339).
关键词
计量学
液压泵
状态识别
完备总体经验模态分解
模糊熵
退化特征提取
变量预测模型
metrology
hydraulic pump
state identification
complementary ensemble empirical mode decomposition
fuzzy entropy
degradation feature extraction
variable predictive model based class discriminate
