基于递归定量分析的液压泵故障识别方法

Fault Recognition Method Based on Recurrence Quantitation Analysis for Hydraulic Pump

提出了一种局部投影消噪和递归定量分析相结合的轴向柱塞泵故障识别方法。以轴向柱塞泵故障振动信号为研究对象,首先用局部投影消噪方法对振动信号进行消噪;其次对消噪后的振动信号绘制递归图,进而通过递归定量分析对递归图所反映出的系统动力学信息进行特征提取,选择确定率(DET)和递归熵(ENTR) 2个特征构成特征向量,构成故障特征样本;然后通过核模糊C均值聚类(KFCMC)方法对训练样本进行聚类,进而依据最小欧氏距离准则对测试样本进行故障识别;最后,将递归定量分析方法和相空间复杂网络定量特征方法进行对比。结果表明,基于递归定量分析的轴向柱塞泵故障识别方法具有更高的故障确诊率。

A fault recognition method for axial piston pump with local projective noise reduction method and recurrence quantitation analysis integrated is presented.Fault vibration signals of axial piston pump are taken as the analysis objects.Firstly,the vibration signal is denoised by local projective noise reduction method.Secondly,the recurrence plots of vibration signals after noise cancellation are drawn.The system dynamic behavior reflected in the recurrence plots is extracted by recurrence quantification analysis.The determinism(DET)and the recurrence entropy(ENTR)are selected to form the eigenvectors.The fault feature samples are constructed.Then,the fault training samples are clustered by the kernel fuzzy C-means clustering(KFCMC)method.The minimum Euclidean distance principle is used to identify the fault test samples.Finally,through comparing the recurrence quantitation analysis method with the quantitative feature method of phase space complex network,we find that the recurrence quantitation analysis-based fault recognition method of hydraulic pump has a higher fault diagnosis correct rate.