基于LVQ网络的航空发动机气路故障特征提取方法研究

针对航空发动机预测与健康管理系统对其状态判断和故障诊断的需求,结合LVQ网络具有处理分类问题时能够识别信息内含有的重要聚类特征信息的优点,提出了基于LVQ神经网络的航空发动机故障特征提取方法。分析研究了LVQ神经网络的结构和学习算法,以及某型航空发动机的测量参数、数据预处理和故障样本选取方法。并以其设计点为例进行了系统仿真。通过与BP网络的分类器对比试验,表明了该算法的可行性和有效性。

A Health Feature Extract Algorithm Based on LVQ Neural Network for Aero-Engine

According to the requirement of prognostics and health management in aero-engine,a novel approach of fault diagnosis named the health feature extract algorithm is presented,which is based on LVQ neural network.And the LVQ neural network has the important advantages that the classification can be identified when the health feature information is contained.The structure and learning algorithm of LVQ neural network are analyzed.The selection methods of measurement parameters,data preprocessing and fault sample in aero-engine are discussed.Finally,the example for engine is established in its designed point.Simulation results from the application to a turbofan model show that the classification has perfect performance than BP neural network.And the simulation results demonstrate the effectiveness of this method,particularly for turbofan engine health feature extracted.