【作者】 黎凌峰；

【导师】 陈吉红；

【作者基本信息】 华中科技大学 ， 机械工程， 2021， 硕士

【摘要】 新世纪以来,新一代信息技术的飞速发展,为数控机床的智能化升级提供了重大机遇。数控机床智能化主要体现在数控系统上。从技术路径的角度研究智能化技术与数控系统的技术融合过程,有助于揭示数控系统智能化发展趋势和动向,识别融合发生的具体子领域和关键技术范畴,以对后续的技术发展趋势研判、技术选择等工作进行决策支持。智能化技术与数控系统技术融合路径识别,既需要专家知识的指导,又需要大量相关领域的理论研究、技术应用等成果作为数据支撑,因此对专家知识和数据的链接提出了需求。作为传统知识工程在人工智能和大数据时代下的延续,知识图谱具有联合多维度的领域数据和高质量的专家知识的能力。并且知识图谱还能通过自动化的知识获取,来应对传统制造业知识工程面临的知识更新困难问题。因此知识图谱能够为智能化技术与数控系统技术融合研究提供更丰富、更及时的数据基础。图神经网络则基于同时利用图的结构信息和节点信息进行计算推理的特点,为知识图谱的应用提供了功能强大的方法。智能化技术与数控系统的技术融合过程具有技术跨领域复杂性和发展多维度性,难以通过传统制造业知识工程结合现有技术融合分析方法进行识别。首先,针对专家知识和数据链接的需求,本文构造面向智能化技术与数控系统技术融合研究的知识图谱。此知识图谱以智能化技术和数控系统的论文和专利构造数据层;以凝结了专家知识的数控机床技术体系构造模式层。将数据层和模式层关联,以进行专家知识和数据链接。并以数据驱动的方式对技术体系进行补充和细化,丰富技术体系内涵。其次,针对智能化技术与数控系统技术融合中体现的技术跨领域复杂性,本文构造的知识图谱数据层,包含了体现跨领域交流特征的引用网络数据和具有更丰富、更广泛的技术内涵特征的文本数据。以此为数据基础,本文提出一种基于图神经网络的技术融合早期识别方法,该方法同时使用引用网络和文本数据,能够实现具有跨领域复杂特征的技术融合早期识别。最后,针对智能化技术与数控系统技术融合中体现的发展多维度性,结合知识图谱的“论文+专利+关键词”数据层与“数控机床技术体系”模式层,本文提出一种基于知识图谱表示学习的双层技术融合路径识别方法,识别出论文层和专利层的技术融合路径及其跨层连接,以揭示技术融合路径在两个视角下的关联与差异。并且结合数控机床技术体系进一步挖掘出双层技术融合路径中隐含的智能化应用发展趋势。本文将所提出的核心方法与现有文献中方法进行了对比,验证了本方法的优越性。研究结果发现,智能化技术与数控系统的技术融合趋势主要体现在两个方面:一方面是智能算法通过状态检测、误差补偿、运动控制等方面的应用提升加工质量;另一方面是大数据、物联网、云计算等技术在数据采集、传输、计算等多个方面的应用,以达到能量效率优化、远程监控等目标。本文研究结果可用于支撑专家进行技术发展趋势研判、技术选择决策等工作。更多还原

【Abstract】 In the new century,the rapid development of the new generation of information technology provides an important opportunity for the intelligent upgrade of NC machine tools.The intellectualization of NC machine tools is mainly embodied in NC system.The study of the technology convergence of intellectualized technology and NC system through technology trajectory helps reveal the trends of the intelligent development of NC system,identify the specific sub-fields and key technologies where the convergence occurs,and provides decision support for the research and judgment of technology development trend and technology selection in the subsequent stage.The technology convergence trajectory identification of intellectualized technology and NC system requires not only the guidance of expert knowledge but also the data support of a large number of theoretical research and technical applications in related fields.Therefore,the requirements for expert knowledge and data linking are put forward.As a continuation of traditional knowledge engineering in the era of artificial intelligence and big data,the knowledge graph can combine multi-dimensional data and high-quality expert knowledge.And,the knowledge graph can deal with the problem of knowledge updating that traditional knowledge engineering faces through automatic knowledge acquisition.So the knowledge graph can provide a richer and more timely data basis for the research on the technology convergence of intellectualized technology and NC system.Graph neural networks are powerful methods for the application of the knowledge graph,which makes use of the structure information and node information of the graph simultaneously to calculate and reason.The technology convergence development of intellectualized technology and NC system reflects the cross-domain complexity of technology and the multi-dimensionality of development,which is difficult to identify through traditional manufacturing knowledge engineering combined with existing technology convergence analysis methods.Firstly,in response to the demand for expert knowledge and data linking,the knowledge graph for the technology convergence of intellectualized technology and NC system is constructed.This knowledge graph takes the papers and patents of intellectualized technology and NC system as the data layer;takes the NC machine tools technology system containing expert knowledge as the schema layer.This thesis associates the data layer with the schema layer for expert knowledge and data linking.And this thesis supplements and refine the technology system in a data-driven way to enrich the connotation of the technical system.Secondly,in view of the cross-domain complexity of technology embodied in the technology convergence of intellectualized technology and NC system,the knowledge graph data layer constructed in this thesis contains the citation network that reflects the characteristics of cross-domain communication and the text data that contains richer and broader technical connotation features.Based on the data layer,this thesis proposes an early identification method of technology convergence based on graph neural networks,which simultaneously use the citation networks and text to realize the early identification of technology convergence with complex cross-domain features.Finally,in view of the multi-dimensionality of development embodied in the technology convergence of intellectualized technology and NC system,combining the "papers + patents+ keywords" data layer and "NC machine tool technology system" schema layer of the knowledge graph,this thesis proposes a two-layer technology convergence trajectory identification method based on the knowledge graph representation learning.The technology convergence trajectories of the papers layer and the patents layer,and their cross-layer connections are identified,to reveal the correlations and differences between the two perspectives of technology convergence trajectories.And this thesis combines the NC machine tools technology system to further extract the hidden information of the development trend of intelligent application in the two-layer technology convergence trajectory.By comparing with the existing methods,the advanced performance of the proposed methods is verified.The results show that the technology convergence of intellectualized technology and NC system is mainly reflected in two aspects.On the one hand,intelligent algorithms improve the processing quality through the application of state detection,error compensation,motion control,etc.On the other hand,big data,the Internet of Things,cloud computing and other technologies are used in the data collection,transmission,calculation and other aspects to achieve energy efficiency optimization,remote monitoring and so on.The research results of this thesis can be used to support experts and scholars in researching and judging technology development trends,making technology selection decisions and so on.更多还原