基于EMD分解和广义相位排列熵的相近金属材料分类

Classification of metal materials with similar properties based on EMD decomposition and generalized phase permutation entropy

为辨识性质相近且种类不同的金属材料,提出了一种基于EMD分解和广义相位排列熵的算法。该算法首先利用经验模态分解方法对原始信号从高频到低频进行分解,之后计算分解后信号的广义相位排列熵,根据信号不同分量的广义相位排列熵值,选取差异显著的信号分量的广义相位排列熵作为特征量,采用KSVM(KNN-SVM)算法,实现对金属材料的分类。通过实验采集性质相近的40个金属材料的超声回波信号,利用提出的算法提取回波信号特征,结合KSVM分类器进行分类。结果表明:金属材料IMF1分量的广义相位排列熵差异最显著,以此作为对40个金属材料超声回波信号的特征进行分类,分类效果稳定且准确率高于96.3%。

In order to identified the similar property but different kinds of metallic material,a classification algorithm based on EMD(empirical mode decomposition)and generalized phase permutation entropy is proposed.The empirical mode decomposition algorithm is used to decompose the original signals,and the IMFs(Intrinsic Mode Functions)of the signals from high frequency to low frequency are obtained.The entropy values of the IMFs are calculated by the generalized phase permutation entropy.The entropy values which show obvious differences components of the signals,are selected as the characteristic ones.The KSVM clustering algorithm is applied to achieve the classification of the metal materials.The ultrasound echoes signal of forty kinds with similar properties are collected by experiments.The echo signal features are extracted by EMD decomposition and generalized phase permutation entropy algorithm,and classified by KSVM classifier.The results indicate that the generalized phase permutation entropy of IMF1 components of metal materials is the most significant difference,which is used to classify the characteristics of forty ultrasonic echo signals of metal materials.The classification results are stable and the accuracy is higher than 96.3%.