Acceleration of BEM using CUDA/GPU programming with application to marine renewable energy extraction

see the original item page
in the repository's web site and access all digital files if the item*



Acceleration of BEM using CUDA/GPU programming with application to marine renewable energy extraction (EN)

Koutsogiannakis, Panagiotis (EN)

ntua (EL)
Belibassakis, Kostas (EN)
Politis, Gerasimos (EN)
Athanassoulis, Gerassimos (EN)

bachelorThesis

2019-03-22
2019-02-14
2019-03-22T10:07:21Z


In the present thesis a biomimetic flapping-foil device is studied for the exploitation of marine renewable energy resources. As a first approximation, the lifting body is submerged far from the free surface, neglecting the interaction with the additional boundary. The device is semi-activated, i.e. a pitching motion is enforced and a heaving motion is induced by an alternating lift, produced in the presence of the current. For this purpose a Boundary Element Method (BEM) in its 3-dimensional version is applied. A Morino-type Kutta condition is imposed on the trailing edge and, by linearizing the trailing vortex sheet dynamics, a simplified wake model is used. However, the motion of the body and the initial shape of the wake are not linearized. The present method, after enhancement and further verification, can be applied to the design and control of such biomimetic devices extracting energy from waves and tidal currents nearshore. The performance of various integration quadratures for the calculation of singular integrals emerging in BEM are also investigated. Simple Gauss-Lobatto and Newton-Cotes quadratures, of arbitrary order, are embedded in an adaptive routine, enabling the treatment of singularities associated with the single/double layer potential induction factors. Different partitioning schemes are examined and the Richardson extrapolation technique is used to accelerate the convergence of the recursive quadrature routine. The developed numerical integration method is able to evaluate efficiently integrals with multiple singularities on N-dimensional hypercubes. The in-house GPU-accelerated computational code, developed by E.S. Filippas, is reprogrammed, by using object-oriented programming, extending the method to solve the problem of the semi-activated system. The parallelization parameters that affect performance are determined and a mixed precision arithmetic scheme is used to optimize the performance of the algorithm. The concepts of polymorphism and inheritance, incorporated in object-oriented programming, encourage the extension of the solver to treat different problems with similar structure in an elegant manner. (EN)


Μέθοδος Συνοριακών Στοιχείων (EL)
Μοντελοποίηση ανωστικών ροών (EL)
Lifting flows modeling (EL)
Υδροδυναμική μοντελοποίηση (EL)
Προγραμματισμός Καρτών Γραφικών (EL)
Θαλάσσια ανανεώσιμη ενέργεια (EL)
GPGPU programming (EN)
Boundary element methods (EN)
Marine renewable energy (EN)
Hydrodynamic modeling (EN)

English

Εργαστήριο Ναυτικής και Θαλάσσιας Υδροδυναμικής (EL)
Σχολή Ναυπηγών Μηχανολόγων Μηχανικών (EL)

Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα
http://creativecommons.org/licenses/by-nc-nd/3.0/gr/




*Institutions are responsible for keeping their URLs functional (digital file, item page in repository site)