EEMD模糊熵和变量预测模型的转子故障诊断新方法  

Fault Diagnosis of Rotor System Based on EEMD-Fuzzy Entropy and VPMCD

作　　者：崔心瀚[1] 马立元[1] 魏忠林[1] 李世龙[1] 王天辉[1] 

机构地区：[1]军械工程学院四系,石家庄050003

出　　处：《内燃机工程》2016年第4期84-91,共8页Chinese Internal Combustion Engine Engineering

基　　金：军队通保科研基金项目(装通[2012]80号)

摘　　要：提出一种基于总体经验模态分解(ensemble empirical mode decomposition,EEMD)模糊熵和变量预测模型的转子故障诊断新方法,并将其应用于某型燃涡发动机转子的非平稳振动信号分析及故障诊断。将基于变量预测模型的模式识别方法引入转子故障模式识别中,利用其较强的非线性问题处理能力,通过变量内部特征值之间的内在关系建立预测模型,并以预测误差平方和最小作为故障模式判别依据。首先利用EEMD将转子振动信号分解成若干个模式分量;然后分别计算各个分量的指标能量,筛选出包含主要故障信息的分量并提取模糊熵组成特征向量;最后采用基于变量预测模型的模式识别方法进行故障识别和分类。对某型燃涡发动机转子正常、不平衡、不对中三种不同状态下的振动信号进行分析,结果表明所提方法能够有效识别转子工作状态。与神经网络、支持向量机算法的对比分析证明,所提方法能更准确、更高效地完成转子故障诊断。A rotor system fault diagnosis was proposed based on ensemble empirical mode decomposition （EEMD） with fuzzy entropy and variable predictive model based on class discriminate （VPMCD）. It was applied to an analysis of unsteady state vibration signals collected from a combustion turbine engine rotor monitoring. A pattern recognition method based on the variable prediction model was introduced for rotor fault pattern recognition. By using the method＇s strong ability of nonlinear problem, a prediction model was established through the inherent relationship between characteristic values of variables with the minimum sum of squares of prediction errors of failure modes as basis. Firstly, a rotor vibration signal was decomposed into a series of intrinsic mode functions （IMFs） by EEMD to calculate the weight index of energy. Then, fuzzy entropy was extracted from the IMFs which contain the maximum original information. Finally, VPMCD was used to recognize and classify rotor faults. The vibration signals of a certain type of combustion turbine engine rotor in the three different conditions of normal, imbalance and misalignment were analyzed. The results show that the proposed method can effectively identify rotor working condition, and accomplish the diagnosis of the faults of a rotor system more effectively and more efficiently compared with the artificial neural network and support vector machine algorithm.

关 键 词：内燃机 变量预测模型 总体经验模态分解 模糊熵 指标能量 故障诊断 

分 类 号：TK401[动力工程及工程热物理—动力机械及工程]