航空发动机转静径向碰摩位置智能识别技术研究

Intelligent recognition for radial rubbing location of an aero-engine rotor-stator

获取航空发动机转静碰摩位置对于诊断发动机碰摩故障和改进设计具有重要意义,基于航空发动机转子实验器的机匣振动加速度信号,研究基于拉普拉斯特征映射(Laplacian Eigenmaps,LE)结合球结构支持向量机的径向碰摩位置智能识别方法。采用拉普拉斯特征映射算法提取碰摩样本的特征信息,用网格搜索法优化拉普拉斯特征映射算法的相关参数;将特征样本输入到球结构支持向量机分类器,识别不同位置的碰摩样本;利用实测的碰摩数据对该方法进行验证,并与主成分分析法(PCA)所得特征样本分类结果进行比较,结果表明,该方法具有实用性和有效性。

： In order to diagnose rubbing faults and improve design, it is very important to acquire the radial rubbing location of an aero-engine rotor-stator. Based on the casing vibration acceleration signals of an aero-engine rotor experimental rig, a method for the radial rubbing location identification using Laplacian eigenmaps （LE） and sphere support vector machine was investigated here. Firstly, Laplacian eigenmaps were used to extract the rubbing samples＇ features, their parameters were optimized with grid search method. Then, the characteristics of the samples were input to a sphere support vector machine to identify different locations of rubbing samples. Besides, with the actual rubbing data, the method was verified and compared with the corresponding results using the principal component analysis （PCA）. The results showed the practicability and effectiveness of the method.