摘要
针对发动机转子的多种故障模式,提出了基于排列组合熵的SVM故障诊断方法。利用转子实验台,模拟了转子正常、转子不平衡、转子不对中、动静碰磨和基座松动5种典型振动模式,并使用振动传感器采集多路振动数据。计算振动数据的排列组合熵并将其作为故障特征向量,对特征向量样本集进行多级SVM分类诊断,并运用小波包能量特征提取方法提取信号特征。实例计算与结果对比表明,本文方法的正确率要高于基于小波包能量提取特征的SVM分类诊断方法,在提取转子振动信号的特征向量及在小样本下的故障分类诊断等方面,具有可行性和有效性。
For a variety of rotor failure modes, the SVM fault diagnosis method based on permutation entro?py was proposed. Five kinds of typical faults: normal rotor vibration, rotor unbalance, rotor misalignment, rubbing and base loosening were simulated and vibration failure data was collected in rotor experiments. The permutation entropy of vibration fault signal was calculated as the fault feature, multi-class SVM was used to classify and diagnose the feature vector sample sets, signal feature was extracted by wavelet energy feature extraction. By calculating and comparing, the accuracy rate of the method in this paper is higher than the SVM method based on the wavelet energy feature extraction, and the fault diagnosis method was verified to solve the rotor vibration faults signal feature extraction and small sample cases validly.
出处
《燃气涡轮试验与研究》
北大核心
2013年第3期38-42,共5页
Gas Turbine Experiment and Research
关键词
发动机
转子振动
排列组合熵
特征提取
支持向量机
engine
rotor vibration
permutation entropy
feature extraction
support vector machine