群智制造环境下语义驱动的物联制造服务情境感知与协作优化技术研究

Faced with new trend of integration of heterogenous resource, real-time processing of big data and effective collaboration of service under the circumstance of Internet of Things (IoT) based manufacturing, this proposal presents some key techniques to advance the crowdsourced manufacturing, including large-scale knowledge graph modeling, edge computing based architecture for distributed processing, real-time acquisition and analytics of manufacturing context, and effective management of enterprise-scale service, thereby reaching a holistic approach of semantic driven based paradigm for crowdsourced manufacturing, which forms a close loop from semantic modeling, programming embedding, active acquisition to intelligent application. A prototype system and testbed will be created and deployed into real scenarios under a variety of contexts..This work establishes fundamental models and solutions that incorporate context information, edge computing and data stream programming for manufacturing under real time and optimization requirements. As a meaningful extension and new method for Internet manufacturing and network manufacturing research area, it provides flexible and effective methods for IoT based resource allocation and integration, which shall empower crowd intelligence to create a paradigm shift for manufacturing and pave the way for "Made in China 2025" action plan as well.

本课题针对群智制造环境下泛化物联、海量规模、群智协同所提出的资源整合、实时业务、服务协作等需求，引入群智计算、边缘计算、数据流处理等新计算范式，探索众智驱动的制造服务知识图谱构建技术、去中心化的群智制造系统架构、制造服务情境实时获取与分析，以及制造服务高效管理机制与协同方法，构成制造语义由静态建模、编程嵌入、动态获取与智能化应用闭环，建立一个面向物联制造服务协作优化理论体系及应用框架，并通过研制原型系统完成应用验证。本课题充分发挥集众智、汇众力效应，为物联制造环境下制造服务协作优化提供了新的思路和解决途径，对于盘活存量制造资源，优化制造资源配置，深化制造资源配套服务，具有重要意义，为“中国制造2025”计划向群智制造服务生态系统的深度协作模式转移提供理论框架和技术方法支撑。

本项目针对群智制造环境下泛化物联、海量规模、群智协同所提出的资源整合、实时业务、分布处理、服务协作等需求，开展了众智驱动的制造服务知识图谱构建技术、群智制造系统架构及物端~边缘~云端侧交互协作机制、制造服务情境获取与实时分析，以及制造服务高效管理机制与协同方法研究，并重点探索了制造信息结构化抽取方法、隐式属性映射与功能知识自动补全、轻量化制造语义实时分析、深度多任务学习制造任务语义分析、制造服务组合优化模型，以及制造设备隐式控制冲突自动检测等关键技术。.本项目共计发表学术论文14篇，其中在《IEEE Internet of Things Journal》、《Enterprise Information Systems》、《International Journal of Production Research》、《Journal of Manufacturing Systems》等国际学术期刊发表SCI论文11篇，国内一级期刊1篇，国际学术专著Book chapter 1篇，申请发明专利6项，软件著作权登记4项，培养研究生13人。.本项目的研究成果将有助于充分发挥集众智、汇众力效应，盘活存量制造资源，优化制造资源配置，深化制造资源配套服务，为“中国制造2025”计划向群智制造服务生态系统的深度协作模式转移提供理论框架和技术方法支持，促进中小制造企业服务化能力的转型和升级。进而，本项目在课题研究中积累的语义模型轻量化、语义模型预训练优化、可变精度浮点函数实时计算等成果，将有助于赋能基于国产自主芯片的制造设备端侧推理，促进复杂神经网络模型结构简化、制造语义深度理解，优化运算密集型语义实时处理效率。