基于CEEMDAN与信息熵的液压泵故障特征提取方法研究

由于液压泵故障振动信号微弱和不平稳的特性,造成特征向量提取和故障诊断困难。针对这些问题,提出一种CEEMDAN与信息熵结合的特征提取方法。将传感器测得的液压泵的故障振动信号进行CEEMDAN分解得到多个固有模态函数(IMF),并计算其信息熵,然后筛选出信息熵最小的3个IMF分量重构信号,计算重构信号的多域熵作为特征向量来训练决策树模型。液压泵故障诊断实验结果证明了该方法的有效性和优越性。

Research on Feature Extraction Method of Hydraulic Pump Based on CEEMDAN and Information Entropy

The characteristics that hydraulic pump fault vibration signal is weak and unstable, brings difficulty to vector extraction and fault diagnosis. To solve these problems, a method was put forward for vector extraction based on the combination of CEEMDAN and information entropy. The fault vibration signal of hydraulic pump was decomposed by CEEMDAN to obtain the intrinsic mode func-tions ( IMF) , the information entropy of them was calculated and 3 smallest were selected to rebuild new signal. The multi domain en-tropy of the new signal was calculated, which worked as vectors to train the decision tree. The results of hydraulic pump fault diagnosis experiment demonstrate the effectiveness and superiority of the method.