基于MUSIC功率谱和CPNN的鸡蛋散黄无损检测方法  

A nondestructive detection method of scattered eggs based on MUSIC power spectrum and CPNN

作　　者：丁天华[1] 卢伟[1] 张超[1] 杜健健[1] 丁为民[1] 王玲[1] 

机构地区：[1]南京农业大学工学院/江苏省现代设施农业技术与装备工程实验室,江苏南京210031

出　　处：《南京农业大学学报》2015年第6期1009-1015,共7页Journal of Nanjing Agricultural University

基　　金：国家自然科学基金青年基金项目(61401215);江苏省自然科学基金项目(BK20130696);中央高校基本科研业务经费专项资金项目(KYZ201427)

摘　　要：[目的]针对当前国内外对散黄鸡蛋较难无损检测的问题,对散黄鸡蛋的振动特性和无损检测方法进行了研究。[方法]通过对鸡蛋进行生理学解剖,据此进行ANSYS有限元分析;构建基于磁致伸缩振子扫频式振动的鸡蛋散黄检测系统,对采集的鸡蛋振动音频信号进行MUSIC(multiple signal classification)功率谱分析,再利用主成分分析法提取特征向量中的有用信息并构建基于对向传播神经网络(CPNN)的鸡蛋散黄检测模型。[结果]鸡蛋的ANSYS固液耦合有限元分析证明新鲜蛋与散黄蛋振动特性差异明显,为基于振动信息的鸡蛋散黄检测提供理论依据;基于磁致伸缩振子扫频式振动的鸡蛋散黄检测系统能有效增强鸡蛋振动信息,基于对向传播神经网络的鸡蛋散黄检测模型对300枚鸡蛋进行检测(训练集200枚,测试集100枚),结果新鲜蛋与散黄蛋的识别率分别达到98%和96%。[结论]新鲜蛋与散黄蛋振动存在差异,采用基于扫频振动式的MUSIC功率谱分析和对向传播神经网络的鸡蛋散黄检测是可行的。[Objectives]This article is directed at the problem that the scattered eggs are difficult to be detected at home and abroad,and the vitration characteristics and the nondestructive testing methods of the scattered yellow eggs were studied. [Methods]This article plans to analyse the structural differences of fresh eggs and scattered eggs by physiological anatomy,and based on this finite element analysis was carried out by ANSYS. The sweep frequency vibration detection system of scattered eggs based on the magnetostrictive vibrator was constructed,the MUSIC（multiple signal classification）power spectrum was used to analyse vibration audio signal of eggs,and the useful information in the feature vector was extracted by principal components analysis,and the scattered eggs detection model was built based on counterpropagation neural network（CPNN）. [Results]The solid-liquid coupling system of egg by ANSYS proves that the differences of the vibration of fresh egg and scattered egg are large,which provide a theoretical basis for the detection of scattered eggs by the vibration information. The sweep frequency vibration detection system of scattered eggs based on the magnetostrictive vibrator enhances the vibration information of eggs effectively. The scattered eggs detection model based on counterpropagation neural network detects 300 eggs（the training set was 200,and the test set was 100）,the discriminant rate of scattered eggs and fresh eggs reaches to 98% and 96% respectively in test set. [Conclusions]The vibration of fresh egg and scattered egg is different,using the MUSIC power spectrum and CPNN based on sweep frequency vibration for detecting scattered eggs shows its feasibility.

关 键 词：散黄蛋 无损检测 磁致伸缩 MUSIC功率谱 主成分分析 对向传播神经网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] O241.82[自动化与计算机技术—计算机科学与技术]