Data-Driven Approaches for Digital Twins of Manufacturing Processes

这项合作研究旨在推进制造流程的数字孪生，以优化工厂分布，增强制造，并实现更好的决策。数字孪生是对物体或环境的数字化再创造。数据收集、分析、模拟和其他数据驱动的功能可以帮助优化其物理对应物。计算研究可以通过机器学习和现代软件工程技术进行，其中数据被用来指导决策和控制。该项目旨在全面了解数字孪生范式中的工程任务，以及如何使用数据分析，机器学习和深度学习来处理这些任务。这些研究包括设计和部署i）基于机器学习的流程、设备和环境舒适度控制系统; ii）能源和设备使用的数据驱动模拟; iii）从资产和环境的实时图像中可视化几何重建和环境变量映射，如温度和湿度; iv）数据集成，以管理和交换数字孪生软件组件之间的数据; v）使用历史数据进行预测性维护，以计算资产的健康状况和剩余使用寿命; vi）边缘雾方法，其中计算资源和存储以联合方式从本地处理器卸载到网络中的服务器，以改善响应时间。在这个合作研究项目中，加拿大团队将联合收割机与波尔图大学的葡萄牙团队结合其专业知识，以推进数字孪生领域的知识和技术。这项研究将通过提供一种渐进和结构化的方法来实现数字孪生自动化，从而使智能工厂和先进制造等多元化的加拿大行业在自动化过程中受益。拟议的研究还将提供巨大的HQP培训机会，并为加拿大工业提供制造过程工程数字孪生专家，这一领域在未来五到十年内需求量很大。

This collaborative research aims to advance the digital twins of manufacturing processes to optimize plant distribution, enhance manufacturing, and enable better decision-making. A digital twin is a digital re-creation of an object or environment. Data collection, analytics, simulation, and other data-driven capabilities can help optimize its physical counterpart. Computational studies can be performed with machine learning and modern software engineering techniques, where data are leveraged to guide decision-making and control. This project aims to comprehensively understand the engineering tasks within the digital twin paradigm and how to approach such tasks using data analytics, machine learning, and deep learning. Such studies include designing and deploying i) machine-learning-based control systems for processes, equipment, and environmental comfort; ii) data-driven simulation for energy and equipment usage; iii) visualization of geometric reconstruction and mapping of environment variables, such as temperature and humidity, from real-time images of assets and the environment; iv) data integration to manage and exchange data between software components of digital twins; v) predictive maintenance using historical data to calculate assets' health and remaining useful life; vi) edge-fog approaches, where computing resources and storage are offloaded from local processors to the servers in the network in a federated manner, to improve response time. In this collaborative research project, the Canadian team will combine its expertise with the Portuguese team at the University of Porto to advance knowledge and technology in the digital twin domain. This research will benefit a diverse Canadian industry, such as smart factories and advanced manufacturing, in their automation processes by providing a progressive and structured approach to realizing digital twin automation. The proposed research will also stage a tremendous HQP training opportunity and equip Canadian industries with experts in engineering digital twins for manufacturing processes, a sector believed to be in high demand in the next five to ten years.