A Reusable and Maintainable Digital Twin Framework for State Estimation Using Physics-Based Models of Mechanical Systems: Predictive Maintenance Case Studies

One key trend in smart manufacturing (SM) is the move toward state estimation (SE) and predictive technologies, which leverage advanced data analytics. To address this SM trend, the concept of the Digital Twin (DT) is rapidly emerging. A DT is a sophisticated software representation of a real-world counterpart within the cyber world that utilizes real-time data together with models to make predictions or estimations that can be used to positively impact the physical environment. Despite the potential for advancements using DTs, most examples from the literature lack demonstrations of how DTs can meet important requirements, such as maintainability (capacity of the DT to adapt to changes in the physical twin) and re-usability (capacity of the DT to be used in multiple systems). To bridge this gap, this paper presents a framework for building DTs for SE applications ensuring maintainability and re-usability by using physics-based models updated as data is collected from the physical twin. The proposed DT framework provides a generalized set of steps to convert physics-based models into recurrently updated physics-based models for a class of systems governed by similar physical mechanisms. To demonstrate the effectiveness of the proposed framework, two case studies were conducted in predictive maintenance (PdM) of rotating machinery. The case studies demonstrated the ability of the DT framework to provide high-confidence estimations for two different systems experiencing varying levels of imbalance or mechanical looseness. The results indicate that the proposed framework yields maintainable and reusable DTs for PdM of rotating machines. While the case study is limited to rotating machinery, they provide proof of concept of the proposed methodology. All the experimental data sets generated for this contribution are public and can be downloaded from GitHub .