摘要
将有裂纹鸡蛋和无裂纹鸡蛋进行敲击激励后,采用柔性压电薄膜传感器获取时域信号和频域特征:无损蛋的频域特征曲线有明显的主频率值,峰值突出;有裂纹蛋频域特征曲线上没有明显的主频率值,峰值多而紊乱.在归一化功率谱中,分别按间隔频率、依次取最高幅值和幅值高低提取的频率等方法来提取前10个或取前20个特征值,采用间隔频率提取归一化幅值,用神经网络来判别裂纹蛋与完好蛋效果较差;用幅值高低提取归一化幅值作为特征值来神经网络判别的效果较佳;采用功率幅值高低提取频率作为特征值来判别的效果最佳.采用20个特征值后的判别效果分别不如采用10个特征值的判别效果.遗传优化神经网络的测试集判断正确率高于标准BP网络.
The eggshell cracks were detected by dynamical frequency response of an egg excited with light mechanical impacts on different locations of the eggshell using flexible piezoelectric film sensors. As a result, the dominant resonance frequency can be observed and the frequency value is lower in the intact eggs. From the cracked eggs, some peak frequencies were found, the magnitudes of those frequencies were approximated, and the values are higher. In normalization average of frequency domain, 10 or 20 magnitudes of interval frequency, 10 magnitudes in turn and frequencies by magnitude in turn were extracted as eigenvalues. In order to investigate whether the dynamical frequency response was able to distinguish between intact egg and eggshell crack, BP neural network （BPNN） and the combination of genetic algorithm and BP neural network （GA-BPNN） were used. The results show that the dynamical frequency response could better distinguish between intact egg and cracked egg using BPNN and GABPNN by magnitudes and frequencies, and more distinguishing effect can be obtained by frequencies and magnitudes,and by using GA-BPNN.
出处
《江苏大学学报（自然科学版）》
EI
CAS
北大核心
2009年第5期454-458,共5页
Journal of Jiangsu University：Natural Science Edition
基金
国家自然科学基金资助项目(30570449)
国家高技术研究发展计划"863"项目(2006AA10Z212)
关键词
鸡蛋
裂纹
检测
神经网络
频率
egg
crack
detection
neural network
frequency
