基于神经网络集成的蚕茧干壳量无损检测方法研究  

Research on the Non-destructive Testing Method for Dried Shell Weight of Cocoon Based on Neural Network Ensemble

作　　者：胡兴明[1] 吴恢[1] 叶楚华[1] 邓文[1] 杨光友[2] 周国柱[2] 

机构地区：[1]湖北省农业科学院经济作物研究所,武汉430070 [2]湖北工业大学机械工程学院,武汉430068

出　　处：《蚕业科学》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)

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.

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 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置] TP183[自动化与计算机技术—控制科学与工程]