摘要
在生产香烟包装纸的过程中,需要对包装图案质量进行在线检测。本文针对香烟包装图案的特点,提出了一种基于BP网络的在线检测和诊断包装纸质量的方法。首先,采集典型的包装图案样本,建立足够数量的样本特征集。然后精心选择一个BP网络结构,利用样本集对该网络反复进行训练、测试和优化,最终得到一个合适实用的BP诊断网络。实验表明,基于该BP网络的诊断系统能够满足质量检测的要求,可用于包装纸质量的在线诊断。
Cigarette packing paper quality need to be detected online in the course of production. In view of the features of cigarette packing design, this paper introduces a kind of on-line detection and diagnosis method of cigarette packing paper quality based on BP network. First, we must gather some typical samples and establish enough sample feature sets. Then we design a BP network. The network is trained, tested and optimized repeatedly by using the sample feature sets until it is appropriate and available. The experiment has showed that the system based on BP network can satisfy the request of quality detection, and may be used in on-line detection and diagnosis of cigarette Dacking paper quality.
出处
《微计算机信息》
北大核心
2006年第10X期98-101,共4页
Control & Automation
基金
校引进人才基金(zk043093)
(国家自然科学基金)No.60574083
关键词
神经网络
BP算法
包装质量
特征提取
在线检测与诊断
Neural network,Backpropagation algorithm (BP),Packing paper quality,Feature extraction, On-line detection and diagnosis