Abstract
The increasing digitization of the maritime industry has enabled the collection of large volumes of operational and performance data from vessels, offering new opportunities for data-driven modeling, monitoring, and optimization. Among these advancements, the development of Digital Twins (DTs) has emerged as a critical tool for enhancing energy efficiency, condition monitoring, and predictive maintenance in complex marine systems. However, capturing the highly nonlinear behavior of vessels under diverse operating conditions remains a significant challenge. To address this, a novel data-driven framework is proposed, integrating Multiple Model Adaptive Estimation (MMAE), clustering, and subspace system identification techniques to accurately estimate vessel dynamic states. The methodology consists of clustering the operational data with Gaussian Mixture Models (GMM), developing localized linear state-space models via the Eigensystem Realization Algorithm with Observer/Kalman Filter Identification (ERA-OKID) algorithm, and adaptively fusing model outputs using the MMAE algorithm. Validation using real-world vessel datasets demonstrates the framework's ability to adapt to system changes, estimate Fuel Consumption (FC) accurately, and maintain robustness across varying conditions. Modal analysis further enhances interpretability by linking internal dynamics to physical outputs. The proposed framework provides a scalable, interpretable, and robust solution for real-time vessel monitoring, performance prediction, and operational optimization in shipping.