基于匹配追踪的特征频率提取算法及其应用

Feature frequency extraction algorithm based on MP and its application

针对大型转子试验台轴心轨迹提纯的问题,提出了采用匹配追踪(Matching Pursuit,MP)算法提取特征频率合成轴心轨迹的新方法。算法中信号的稀疏分解采用MP算法实现,字典集由正弦波模型构建,通过对正弦波的伸缩、平移构造过完备字典库,该算法可实现对单个或多个特征频率的提取;将此方法应用于大型转子系统轴心轨迹的提纯上,并与主成分分析法、谐波小波算法和小波包算法的处理效果进行了比较。结果表明,采用匹配追踪算法提纯的轴心轨迹更为清晰。

Aiming at the problem of purifying axis orbit of a large rotor system test rig,a new method for extracting feature frequency and synthesizing axis orbit was proposed based on the matching pursuit(MP)algorithm.Sparse decomposition of signals in the method was realized with MP algorithm.The dictionary set was constructed using the sine wave model,and an over-complete dictionary library was constructed by stretching and translating sinusoidal waves.Extracting single feature frequency or multiple ones was realized with this method.This method was applied to purify axis orbit of a large rotor system,and the results were compared with those using PCA algorithm,the harmonic wavelet one and the wavelet packet one,respectively.Results showed that axis orbits purified with MP algorithm are much clearer.