基于MC方法和BP网络的印刷质量控制图模式识别研究  

Pattern Recognition for Printing Quality Control Chart Based on MC Method and BP Neural Network

作　　者：初红艳[1] 李鹏[1] 蔡力钢[1] 李风光[1] 

机构地区：[1]北京工业大学北京市先进制造技术重点实验室,北京100124

出　　处：《北京工业大学学报》2011年第6期816-821,共6页Journal of Beijing University of Technology

基　　金：国家'十一五'科技支撑计划课题(2006BAF03B01);北京市教委科研计划项目(KM200910005006)

摘　　要：建立了印刷质量控制图模式的数学模型,基于Monte Carlo(MC)方法模拟印刷质量数据,生成仿真样本,并使用标准变换和线性编码方法进行预处理,既不失样本数据的内在特征,又大大降低了数据复杂度.通过实验确定了结构为24-18-16-4的4层反向传播(back propagation,BP)网络模型,并采用比例共轭梯度训练算法,提高了网络的稳定性和收敛速度.在对控制图模式识别时,采用不同训练样本容量的实验方案,模式识别正确率达95.87%.结果表明,该方法可以提高印刷企业的质量控制水平和自动化程度.A mathematical model for patterns of the printing quality control chart is established,and the data of printing quality is simulated based on Monte Carlo method,Then the complexity of the sample data is reduced by using the method of standard transformation and linear encoding.A 4-layer BP neural network model,as 24-18-16-4,is established through the experiments,and a scaled conjugated gradient training algorithm is adopted to enhance the stability and convergence of the network.The paper uses different capacity of training samples in pattern recognition for control chart,and the recognition accuracy achieves 95.87%.Results of experiments show that this method can improve the level of quality control and degree of automation for printing enterprise.

关 键 词：印刷质量控制图 模式识别 MONTE CARLO方法 反向传播神经网络 

分 类 号：TS805.3[轻工技术与工程]