Verification framework for continuous-time black-box models based on Gaussian process regression applied to co-simulation of marine propulsion system

Abstract

In this study, a black-box time-continuous simulation model verification framework was developed based on Gaussian process regression (GPR), which is a method of adaptive input space searching regardless of the first-principles equations behind a subject simulation model. Further, pre-developed simulation model verification methodologies and their applicability to a time-continuous black-box simulation model were investigated. An aggregated GPR model was utilised for the adaptive input space search algorithm, serving as a key sub-process of the simulation model verification framework. Finally, a case study was performed by employing the proposed framework to verify a hybrid marine propulsion system consisting of heterogeneously generated time-continuous simulation models.