摘要
随着互联网技术的快速发展,在智能制造过程中会伴随着出现海量的工艺知识数据,为了提升对工艺数据的充分利用和掌握,提出一种知识推理框架下的改进自组织映射算法。在协同训练的思想下,对于知识库当中的工艺知识数据进行自组织映射网络下的筛选优胜,提高优胜单元的抗局部最优能力;利用改进自组织映射算法对特征优胜单元进行知识推理准则判断,在向量空间的映射下,通过双曲空间距离公式优选出置信度高的样本数据进行更新子代样本集;为了进一步提升特征信息的利用率,在知识推理框架下多次循环筛选提高工艺知识数据的有效预测。通过对铣削过程中真实数据进行建模仿真,验证了所提方法在面对多样本数据情况下的良好预测优化的性能。
In the rapidly developing world of Internet technologies,a vast amount of process knowledge data is gener-ated in the smart manufacturing process.To effectively utilize this data and master it,we propose an improved self-or-ganizing mapping algorithm within the framework of knowledge reasoning.The improved algorithm screens and optim-izes the process knowledge data in the knowledge base using collaborative training,enhancing the anti-localization abil-ity of the winning unit.The improved self-organizing mapping algorithm judges the knowledge reasoning criteria of the winning feature unit and selects sample data with high confidence through the mapping of vector space and the use of the hyperbolic space distance formula.Multiple cycles of screening are carried out to further improve the utilization of feature information and enhance the effective prediction of process knowledge data.Through the modeling and simula-tion of actual milling process data,the proposed method demonstrates its strong predictive and optimized performance when faced with multi-sample data.
作者
杨伟凯
王艳
YANG Weikai;WANG Yan(School of Internet of Things Engineering,Jiangnan University,Wuxi 214122,China)
出处
《智能系统学报》
CSCD
北大核心
2023年第5期926-935,共10页
CAAI Transactions on Intelligent Systems
基金
国家重点研发计划项目(2018YFB1701903)。
关键词
知识推理
预测
自组织映射
智能制造
图匹配
置信度
双曲空间
优胜单元
knowledge reasoning
predicting
self-organized mapping
smart manufacturing
map matching
confidence
hyperbolic space
winning unit©《智能系统学报》编辑部版权所有
