Graph-based Model Reduction of Machine System Digital Twins
Chakraborti, Ananda S. (2024)
Chakraborti, Ananda S.
Tampere University
2024
Teknisten tieteiden tohtoriohjelma - Doctoral Programme in Engineering Sciences
Tekniikan ja luonnontieteiden tiedekunta - Faculty of Engineering and Natural Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Väitöspäivä
2024-05-03
Julkaisun pysyvä osoite on
https://urn.fi/URN:ISBN:978-952-03-3440-6
https://urn.fi/URN:ISBN:978-952-03-3440-6
Tiivistelmä
Digital Twin technology is the talking point of academia and industry. When defining a digital twin, new modeling paradigms and computational methods are needed. Developments in the Internet of Things, Artificial Intelligence and advanced simulation and modeling techniques have provided new strategies for building such complex digital twins. The digital twin is a virtual entity representation of the physical entity, such as a product or a process. This virtual entity is a collection of computationally complex knowledge models that embeds all the information of the physical world. This definition of the virtual entity representation makes it immensely complex with several dependencies across various domains of the real-world object. This virtual entity must meaningfully represent information and deductions from these models.
To that end, this research work presents a graph-based representation of the virtual entity. This graph-based representation provides a method to build a knowledge model that embeds the interactions between several parameters across different modeling domains. For that, both traditional and newer methods of graph-based modeling of multivariate systems are researched and usability of these methods are identified. Thereafter, a new method for digital twin conceptualization with graph-based method is proposed by combining conceptual modeling mechanism known as dimensional analysis conceptual modeling and heuristic methods such as greedy equivalence search. Hence, the virtual entity could be represented as a directed graph. However, such virtual entity graph becomes inherently complex with multiple parameters for a complex multidimensional physical system. This research contributes to the body of knowledge with a novel Graph-based Model Reduction method that simplifies the virtual entity graph by preserving the important parameters in it. The graph-based model reduction method uses spectral decomposition method to segment the knowledge graph into structurally similar chunks. Then, investigation is performed to identify the important nodes of the knowledge graph with node importance algorithms such as weighted PageRank and eigenvector centrality. To consolidate the ranking scores, algorithms from the domain of artificial intelligence such as Dempster-Shaffer theory is applied.
The Graph-based Model Reduction method is validated with two case studies of complex machine systems: (1) grinding wheel wear digital twin and (2) turbo compressor digital twin. In both these case studies, the graph-based modeling method combines information from the physics-based models such as finite element models and system level simulation models, with data-driven models to create a hybrid graphical representation. Then, the graph-based model reduction method is applied on the hybrid graphical representation. The method is benchmarked against other model reduction methods in literature and the results are analysed.
This thesis work provides a detailed analysis and results of the graph-based modeling and model reduction methods of digital twins of complex systems. It is argued that the important area of model reduction has been overlooked by the digital twin community. The thesis work shares the learnings from application of the graph-based modeling and model reduction methods in traditional machine systems. This work promotes the application and integration of graph based methods in digital twin frameworks and software solutions for building efficient digital twins that provide efficient digital services.
To that end, this research work presents a graph-based representation of the virtual entity. This graph-based representation provides a method to build a knowledge model that embeds the interactions between several parameters across different modeling domains. For that, both traditional and newer methods of graph-based modeling of multivariate systems are researched and usability of these methods are identified. Thereafter, a new method for digital twin conceptualization with graph-based method is proposed by combining conceptual modeling mechanism known as dimensional analysis conceptual modeling and heuristic methods such as greedy equivalence search. Hence, the virtual entity could be represented as a directed graph. However, such virtual entity graph becomes inherently complex with multiple parameters for a complex multidimensional physical system. This research contributes to the body of knowledge with a novel Graph-based Model Reduction method that simplifies the virtual entity graph by preserving the important parameters in it. The graph-based model reduction method uses spectral decomposition method to segment the knowledge graph into structurally similar chunks. Then, investigation is performed to identify the important nodes of the knowledge graph with node importance algorithms such as weighted PageRank and eigenvector centrality. To consolidate the ranking scores, algorithms from the domain of artificial intelligence such as Dempster-Shaffer theory is applied.
The Graph-based Model Reduction method is validated with two case studies of complex machine systems: (1) grinding wheel wear digital twin and (2) turbo compressor digital twin. In both these case studies, the graph-based modeling method combines information from the physics-based models such as finite element models and system level simulation models, with data-driven models to create a hybrid graphical representation. Then, the graph-based model reduction method is applied on the hybrid graphical representation. The method is benchmarked against other model reduction methods in literature and the results are analysed.
This thesis work provides a detailed analysis and results of the graph-based modeling and model reduction methods of digital twins of complex systems. It is argued that the important area of model reduction has been overlooked by the digital twin community. The thesis work shares the learnings from application of the graph-based modeling and model reduction methods in traditional machine systems. This work promotes the application and integration of graph based methods in digital twin frameworks and software solutions for building efficient digital twins that provide efficient digital services.
Kokoelmat
- Väitöskirjat [4943]