采用模糊RBF神经网络的直升机旋翼不平衡故障诊断  

Unbalance Fault Diagnosis of Helicopter Rotor Using Fuzzy RBF Neural Network

作　　者：廖文峰[1] 高亚东[1] 

机构地区：[1]南京航空航天大学直升机旋翼动力学国家级重点实验室,南京210016

出　　处：《南京航空航天大学学报》2015年第2期285-289,共5页Journal of Nanjing University of Aeronautics & Astronautics

基　　金：直升机旋翼动力学重点实验室基金(9140C4004010805)资助项目

摘　　要：提出一种利用模糊径向基函数(Radial basis function,RBF)神经网络进行直升机旋翼不平衡故障诊断的方法,建立了用于直升机旋翼不平衡故障识别的模糊诊断模型。基于直升机旋翼不平衡故障模拟实验,对采集于旋翼配重不平衡、桨距不平衡、后缘调整不平衡和正常状态下的试验台体振动信号进行功率谱分析,并采用主分量分析(Principal component analysis,PCA)的方法进行故障特征提取。采用模糊RBF神经网络诊断模型对旋翼不平衡故障进行了故障分类识别,同时分析了不同主分量累计贡献率和模糊子空间对故障分类精度的影响,并与RBF神经网络的诊断模型、支持向量机(Support vector machine,SVM)诊断模型进行了故障识别效果对比。结果表明,模糊聚类RBF神经网络的诊断方法对旋翼不平衡故障具有更好的识别能力。A method is presented for unbalance fault diagnosis of helicopter rotor by using fuzzy radial basis function （RBF） neural network. A diagnosis model based on fuzzy RBF neural network is estab- lished and a model test is conducted on the rotor. Three types of rotor system faults are considered： im- balanced mass, misadjusted pitch-control rod and misadjusted trim tab. Power spectrum is applied to the data processing, and the imbalance fault feature is extracted by using principal component analysis （PCA）. The fuzzy RBF neural network diagnosis model is employed to identify the rotor unbalance faults and compared with the diagnosis model based on RBF neural network and support vector machine （SVM）. The results show that the fuzzy RBF neural network diagnosis model is better than RBF-based model and SVM model in diagnosing the unbalance faults of helicopter rotor.

关 键 词：直升机旋翼 故障诊断 模糊RBF神经网络 累计贡献率 

分 类 号：V211.52[航空宇航科学与技术—航空宇航推进理论与工程]