Hybrid modeling of marine turbocharger engines with minimal data requirement: A versatile physics-data fusion approach toward digital twin applications

Abstract

To successfully implement a digital twin for marine turbocharger engines, simulation models must achieve real-time execution while maintaining sufficient predictive accuracy. Although conventional engine simulation tools, such as GT-POWER, provide high-fidelity results, their computational complexity prevents practical real-time applications. This study proposes a hybrid modeling approach that significantly reduces computational time without compromising accuracy, making it suitable for real-time digital twin integration. To successfully implement a digital twin for marine turbocharger engines, simulation models must achieve real-time execution while maintaining sufficient predictive accuracy. Although conventional engine simulation tools, such as GT-POWER, provide high-fidelity results, their computational complexity prevents practical real-time applications. This study proposes a hybrid modeling approach that significantly reduces computational time without compromising accuracy, making it suitable for real-time digital twin integration. Model validation is performed using operational data from two distinct marine engines, each with different turbocharger characteristics and cylinder configurations. The results demonstrate that, upon reaching steady-state conditions, simulation outputs differ by less than 5% from actual measurements across diverse operational scenarios. Furthermore, the model exhibits exceptional computational performance, requiring less than one second to simulate 160 s of engine operation. A key advantage of this research, distinguishing it from previous works, is its minimal data requirement. To apply the developed model to a different engine type, only turbocharger map data and limited factory test data are needed, enabling efficient adaptation across various engine platforms. Consequently, this approach demonstrates considerable versatility and scalability for widespread digital twin deployment. Future research will extend the validation under a wider range of operating conditions and further refine the model to serve as a foundational component within comprehensive digital twin systems.