基于EMD与GA-PLS的特征选择算法及应用

针对振动信号非平稳性和特征优化选择的问题，提出一种基于EMD和GA-PLS的特征选择算法。在该算法中，首先，采用EMD方法将振动信号分解成多个固有模态函数(Intrinsic Mode Function, IMF)，对IMF分量建立自回归（AR）模型，以AR模型系数和残差作为初始特征向量，然后，遗传算法与偏最小二乘法相结合(GA-PLS)的算法对初始特征向量进行筛选得到新的特征向量，最后，以新的特征向量为输入，建立分类器，用来识别手动换向阀的工作状态和判断故障类型。实验结果表明，采用该特征选择算法能准确地选择出特征，并能应用于手动换向阀的故障诊断

Feature selection algorithm based on EMD and GA-PLS and its application

In order to solve problems of the nonstationarity of vibration signal and the optimization of feature selection,feature selection algorithm based on EMD and GA-PLS was proposed.In the algorithm,EMD method was used to decompose the vibration signals into a number of intrinsic mode function(IMF) components,and the auto-regressive(AR) model of each IMF component was established.The main auto-regressive parameters and the loss function were regarded as original feature vectors.Then,GA-PLS algorithm was used to selecte new feature vectors,which are highly correlated with fault information,from original feature vectors.With these new feature vectors used as inputs,classifiers were established for identifying the conditions and fault patterns of manually operated directional valves.Experimental results show that all the feature vectors are selected correctly,and the proposed algorithm can be well used in fault diagnosis.