Digital twin computing for predictive maintenance of industrial systems

工业系统的预测性维护（PdM）是一种使用监控工具和预测分析来确定在役设备状况并估计何时应该执行维护的策略。它可以优化资源使用并减少计划外维护，因此业务可以实现更高的生产率、减少故障和降低维护成本。然而，PdM系统开发面临着处理来自不同来源的异构数据、将解决方案外推到具有多个组件的复杂系统以及缺乏系统设计方法的挑战。数字孪生是一种虚拟的表现形式，作为工业系统中物理对象或过程的实时数字对应物。数字双胞胎的诊断和预测能力使其成为预测性维护的完美伴侣。这种虚拟世界和物理世界的结合，允许对数据和监控系统进行分析，在问题发生之前阻止问题，防止停机，开发新的机会，甚至通过使用模拟来规划未来。近年来，数字孪生技术得到了广泛的研究，但一个通用的系统模型尚未完全建立并被广泛接受。此外，大多数数字双胞胎应用程序都侧重于单个双胞胎。整合多个数字孪生镜像复杂产业系统的研究尚未见文献报道。理解单个组件及其之间的相互作用对于复杂系统的预测性维护至关重要。该研究计划旨在推进工业系统预测性维护的数字孪生技术，覆盖资产或过程的整个生命周期，并形成互联产品和服务的基础。长期目标是通过新的数字孪生计算范式实现数字孪生生态系统的可扩展性。在短期内，我们将为工业预测性维护开发这样一个数字孪生生态系统。我们的研究将首先集中在基于工业物联网（IoT）平台的数字孪生架构，以及多源数据/信息融合。此外，通过多智能体建模和强化学习，将通过数字双胞胎计算方法推导出多个数字双胞胎之间相互作用的见解。这项研究的结果将使完全控制资源和了解每项资产及其组成部分的状况成为可能。因此，它将帮助加拿大工业，特别是中小企业部门，提高操作可用性，生产力和机器正常运行时间。此外，拟议的研究将为HQP提供一个独特的多学科培训机会。HQP将获得多个学科的深入知识和经验，并在不考虑学科界限的情况下解决现实生活中的挑战。

Predictive maintenance (PdM) of industrial systems is a strategy that uses monitoring tools and predictive analytics to determine the condition of in-service equipment and estimate when maintenance should be performed. It can optimize resource usage and reduce unscheduled maintenance, and thus the business can achieve increased productivity, reduced breakdowns, and lower maintenance costs. However, the PdM system development faces the challenges of dealing with heterogeneous data from different sources, extrapolating solutions to complex systems with multiple components, and lacking a systematic design approach. A digital twin is a virtual representation that serves as the real-time digital counterpart of a physical object or process with an industrial system. The diagnostic and prognostic capabilities of the digital twin make it a perfect pair with predictive maintenance. This pairing of the virtual and physical worlds allows analysis of data and monitoring systems to head off problems before they even occur, prevent downtime, develop new opportunities and even plan for the future by using simulations. Although digital twin has recently received much research attention, a general system model has not been fully established and widely accepted yet. Moreover, most digital twin applications focus on an individual twin. The research on integrating multiple digital twins to mirror a complex industry system has not been reported in the literature yet. Understanding individual components and the interactions between them is critical to the predictive maintenance of a complex system. This research program aims to advance the digital twin technology for the predictive maintenance of industrial systems, covering the entire life cycle of an asset or process and forming the foundation for connected products and services. The long-term objective is to achieve the scalability of a digital twin ecosystem with the novel digital twin computing paradigm. In the short term, we will develop such a digital twin ecosystem for industrial predictive maintenance. Our research will first focus on the digital twin architecture based on an industrial Internet of Things (IoT) platform, and multi-source data/information fusion. Moreover, the insights of the interactions among multiple digital twins will be derived with the digital twin computing methodology through multi-agent modelling and reinforcement learning.      The outcomes of this research will make the complete control of resources and awareness of the condition of each asset and its components possible. Thus, it will help the Canadian industry, especially the SME sector, improve operational availability, productivity and machine uptime. In addition, the proposed research will offer a unique multidisciplinary training opportunity for HQP. The HQP will gain in-depth knowledge and experience of more than one discipline and address real-life challenges without regard to disciplinary boundaries.