摘要
蚕茧无损检测中的核心问题是蚕茧干壳量的测定。利用虚拟仪器技术和神经网络集成技术研究了一种无损检测蚕茧干壳量的方法,并实现了数据采集和信号处理等功能。系统首先提取并选择蚕茧振动信号中与蚕蛹质量相关的特征值,再将选择的特征值训练BP神经网络和RBF神经网络,用训练得到的这两种类型网络作为神经网络集成的输入,以蚕蛹质量作为神经网络集成的输出。检测试验的结果表明该方法有效可行,最高检测准确率达到85.6%。
In this paper, a non-destructive testing method for dried shell weight of cocoon is successfully built by means of technology of virtual instrument and neural network ensemble. It realized data collection, signal procession and so on, which was based on the system of software and hardware platform. Firstly, the characteristic parameters of cocoon vibration signal were extracted and selected, which were related to cocoon weights. Then, use the characteristic parameters to train BP model and RBF model, respectively. The input of neural network ensemble is two kinds of models which inputs include six simplex neural networks. The output of neural network ensemble represents cocoon weight. The test results show that the method is effective and feasible.
出处
《蚕业科学》
CAS
CSCD
北大核心
2008年第4期781-785,共5页
ACTA SERICOLOGICA SINICA
基金
国家“十五”重点科技攻关计划项目(编号2001BA502B0-1-01-03-02)
湖北省科技攻关计划项目(编号2001AA20-8B02)
湖北省自然科学基金项目(编号2006ABA050)
关键词
蚕茧
干壳量
无损检测
虚拟仪器
神经网络集成
数据采集
信号处理
Cocoon
Dried shell weight
Non-destructive test
LabVlEW
Neural Network Ensemble
Data collection
Signal procession