Predicting drag on rough surfaces by transfer learning of empirical correlations

Recent developments in neural networks have shown the potential of estimating drag on irregular rough surfaces. Nevertheless, the difficulty of obtaining a large high-fidelity dataset to train neural networks is deterring their use in practical applications. In this study, we propose a transfer learning framework to model the drag on irregular rough surfaces even with a limited amount of direct numerical simulations. We show that transfer learning of empirical correlations, reported in the literature, can significantly improve the performance of neural networks for drag prediction. This is because empirical correlations include ‘approximate knowledge’ of the drag dependency in high-fidelity physics. The ‘approximate knowledge’ allows neural networks to learn the surface statistics known to affect drag more efficiently. The developed framework can be applied to applications where acquiring a large dataset is difficult but empirical correlations have been reported.

Meshing Strategi untuk Memprediksi Hambatan Total pada Kapal Planing Hull

A high-speed vessel has a range of Froude Number (Fr) > 1. A drag prediction method based on Fr > 1 has high complexity because it is influenced by trim and heave motions. Hence, a specific treatment is necessary to obtain accurate results. This study is using mesh density and mesh shapes to predict the total drag of a planing hull ship. The Computational Fluid Dynamic (CFD) results show good performance in predicting the drag, trim, and heave. Mesh density of 2300K shows the most stabilized result. The trimmed mesh type is more efficient to obtain accurate results because it has a smaller mesh size. The polyhedral mesh type is as good as trimmed mesh but is not as efficient as trimmed mesh and it has largely a time-consuming time.

Development of a Real-Time Simulation Model for an ASD Tug

A hydrodynamic digital twin of vessel can be used to replicate the behaviour and response of the vessel in a virtual environment. In this paper, a real-time simulation model (RTSM) for an azimuth stern-drive (ASD) tug has been developed for simulating the hydrodynamic performance of the vessel under a range of environmental conditions. Based on the framework of a 4-DoF MMG manoeuvring model, the RTSM comprises manoeuvring, propulsion and environmental loads which are parameterised using numerical results from a combination of computational fluid dynamics (CFD) modelling work, including virtual planar motion mechanism (vPMM), seakeeping analysis, wind drag prediction and propulsion modelling. The RTSM is used to demonstrate the manoeuvrability of the vessel in calm water and under external loads from waves, winds and currents.

Wake-model effects on induced drag prediction of staggered boxwings

For staggered boxwings the predictions of induced drag that rely on common potential-flow methods can be of limited accuracy. For example, linear, freestream-fixed wake models cannot resolve effects related to wake deflection and roll-up, which can have significant affects on the induced drag projection of these systems. The present work investigates the principle impact of wake modelling on the accuracy of induced drag prediction of boxwings with stagger. The study compares induced drag predictions of a higher-order potential-flow method that uses fixed and relaxed-wake models, and of an Euler-flow method. Positive-staggered systems at positive angles of attack are found to be particularly prone to higher-order wake effects due to vertical contraction of wakes trajectories, which results in smaller effective height-to-span ratios than compared with negative stagger and thus closer interactions between trailing wakes and lifting surfaces. Therefore, when trying to predict induced drag of positive staggered boxwings, only a potential-flow method with a fully relaxed-wake model will provide the high-degree of accuracy that rivals that of an Euler method while being computationally significantly more efficient. Keywords: wake-model; boxwing; induced drag; potential-flow theory

Wake-model effects on induced drag prediction of staggered boxwings

For staggered boxwings the predictions of induced drag that rely on common potential-flow methods can be of limited accuracy. For example, linear, freestream-fixed wake models cannot resolve effects related to wake deflection and roll-up, which can have significant affects on the induced drag projection of these systems. The present work investigates the principle impact of wake modelling on the accuracy of induced drag prediction of boxwings with stagger. The study compares induced drag predictions of a higher-order potential-flow method that uses fixed and relaxed-wake models, and of an Euler-flow method. Positive-staggered systems at positive angles of attack are found to be particularly prone to higher-order wake effects due to vertical contraction of wakes trajectories, which results in smaller effective height-to-span ratios than compared with negative stagger and thus closer interactions between trailing wakes and lifting surfaces. Therefore, when trying to predict induced drag of positive staggered boxwings, only a potential-flow method with a fully relaxed-wake model will provide the high-degree of accuracy that rivals that of an Euler method while being computationally significantly more efficient. Keywords: wake-model; boxwing; induced drag; potential-flow theory