Evaluating the role of agentic digital twin in service lifecycle management : a comparison with real-time simulation-based systems
Larik, Sheeraz Hussain (2026)
Diplomityö
2026
School of Energy Systems, Konetekniikka
Kaikki oikeudet pidätetään.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe20260704109697
https://urn.fi/URN:NBN:fi-fe20260704109697
Tiivistelmä
As industrial assets become more complex, maintenance systems must move beyond human-supervised operation. This thesis compares two digital twin paradigms in Service Lifecycle Management (SLM) namely Agentic Digital Twins (ADTs) and Real-Time Simulation-Based Systems (RTSBS).
Drawing on eight peer-reviewed industrial case studies from the manufacturing, aerospace, energy, and smart factory sectors, the study developed and applied an original five-dimensional comparative framework evaluating ADTs and RTSBS across autonomy, adaptability, learning capability, decision-making logic, and human involvement. No primary experimental data were collected. The analysis is qualitative and comparative.
Three key trade-offs emerged: ADTs enabled faster, event-driven decisions, while RTSBS incurred higher latency due to simulations and human involvement. The RTSBS gave more deterministic and traceable decision support, while the ADTs were more flexible. But in all eight cases, full maintenance autonomy was lacking, and human authorizations for physical interventions were still required.
Two main contributions of the thesis are: the five-dimensional SLM comparison framework and a decision-allocation framework that maps task criticality to data certainty to help organizations decide when decision authority can be delegated to autonomous systems or by human operators. The results indicate that the most viable future path for next generation SLM is the hybrid digital twin approach with RTSBS and ADTs, that is, the use of both the interpretability and physics-based rigor of RTSBS and the adaptability and responsiveness of ADTs.
Drawing on eight peer-reviewed industrial case studies from the manufacturing, aerospace, energy, and smart factory sectors, the study developed and applied an original five-dimensional comparative framework evaluating ADTs and RTSBS across autonomy, adaptability, learning capability, decision-making logic, and human involvement. No primary experimental data were collected. The analysis is qualitative and comparative.
Three key trade-offs emerged: ADTs enabled faster, event-driven decisions, while RTSBS incurred higher latency due to simulations and human involvement. The RTSBS gave more deterministic and traceable decision support, while the ADTs were more flexible. But in all eight cases, full maintenance autonomy was lacking, and human authorizations for physical interventions were still required.
Two main contributions of the thesis are: the five-dimensional SLM comparison framework and a decision-allocation framework that maps task criticality to data certainty to help organizations decide when decision authority can be delegated to autonomous systems or by human operators. The results indicate that the most viable future path for next generation SLM is the hybrid digital twin approach with RTSBS and ADTs, that is, the use of both the interpretability and physics-based rigor of RTSBS and the adaptability and responsiveness of ADTs.