Abstract
Representing the entire operational range of an ocean-going vessel with linear equations is often a formidable task. In this research study, a datadriven localized model is presented for ship performance prediction as a part of the digital twin development. For this purpose, different operational conditions of the vessel, i.e., data clusters, are identified using the Gaussian Mixture Models (GMM) coupled with the Expectation Maximization (EM) algorithm. Subsequently, Singular Value Decomposition (SVD) as a part of the Eigensystem Realization Algorithm (ERA) is applied to each cluster to establish the relationships between different operational and navigational variables and capture the system dynamics in localized operational conditions in each cluster.