Abstract: To improve the utilization efficiency of manufacturing resources,it is required to connect and dynamically match manufacturing services of discrete distributed manufacturing resources rapidly and effectively under intelligent manufacturing environment.The feature description and vector extraction of manufacturing resources and user demands were carried out through word vector modeling on the cloud platform,and the word vectors of manufacturing resources and user demands were mapped to the public space with vector matching basis by using Joint Embedded Convolutional Neural Network (JE-CNN).The objective function was constructed based on the matching distance of two groups of word vectors,and the objective function was optimized by Adaptive time estimation (Adam) algorithm,and then the matching degree was judged according to the binary classification Area Under Curve (AUC) model,so as to realize the high quality and high efficiency matching of manufacturing resources.The feasibility of the proposed method was verified by an example.

Key words: manufacturing resource, cloud platform, term vectors, joint embedded convolutional neural network, adaptive time estimation algorithm, area under curve model

摘要： 为提高制造资源利用效率,智能制造环境下要求将离散分布的制造资源快速有效地连接并动态匹配制造服务。在云平台上通过词向量建模对制造资源和用户需求进行特征描述以及向量提取,并利用联合嵌入卷积神经网络(JE-CNN)将制造资源和用户需求词向量映射到具备向量匹配基础的公共空间。以两组词向量匹配距离构建目标函数,采用自适应时刻估计法(Adam)优化该目标函数,再根据二分类(AUC)模型判断匹配度是否满足要求,从而实现制造资源的高质量、高效率匹配。最后,通过实例验证了该方法的可行性。

关键词: 制造资源, 云平台, 词向量, 联合嵌入卷积神经网络, 自适应时刻估计算法, 二分类模型