摘要
精准预测印制电路板样板物料投入将减少超投浪费和补投成本,为此提出结合自组织映射(Self-organizing maps,SOM)-反向传播(Back propagation network,BPN)网络的预测机制。基于SOM对样本进行聚类分组;采用特征选择机制优选各分组样板报废率关键影响属性;对各分组构建基于BPN的报废率预测模型;将其转换为预测投入生产面板数,并开展模型训练与性能评估。与多种模型进行对比分析,结果表明该模型在降低均方误差、绝对平均误差、平均绝对百分比误差、车间余数入库率和补投率等方面具有较明显优势。
More accurate prediction of PCB template material input reduces over-investment waste and supplemental feeding cost.This paper proposes a prediction mechanism based on self-organizing maps(SOM)and back propagation network(SOM-BPN).The samples were clustered and grouped based on the SOM,and then the feature selection mechanism was used to optimize the key impact attributes of the scrap rate of each sample category;the BPN prediction model of scrap rate based on BPN was constructed for each group;it was converted to the number of predicted production panels,and model training and performance evaluation were carried out.Compared with various models,and the results indicate that our model has obvious advantages in reducing the mean square error,the absolute average error,the average absolute percentage error,the workshop inventory storage rate and the supplementary investment rate.
作者
郑彬彬
吕盛坪
李灯辉
冼荣亨
Zheng Binbin;Lu Shengping;Li Denghui;Xian Rongheng(Key Laboratory of Key Technology on Agricultural Machine and Equipment,Ministry of Education,South China Agricultural University,uangzhou 510642,Guangdong,China)
出处
《计算机应用与软件》
北大核心
2020年第8期57-63,共7页
Computer Applications and Software
基金
国家自然科学基金青年科学基金项目(51605169)。
关键词
印制电路板
投料预测
自组织映射
反向传播网络
PCB
Material feeding prediction
Self-organizing maps
Back propagation network
