On the efficient implementation of numerical solvers for the simulation and control of shallow flows on graphical processing units
- Autores: Asier Heradio Lacasta Soto
- Directores de la Tesis: Pilar García Navarro (dir. tes.), Pilar Brufau García (dir. tes.)
- Lectura: En la Universidad de Zaragoza ( España ) en 2017
- Idioma: español
- Tribunal Calificador de la Tesis: Francisco Reinaldo García Martínez (presid.), Javier Murillo Castarlenas (secret.), Ignacio Villanueva Lacabrera (voc.)
- Materias:
- Texto completo no disponible (Saber más ...)
- Resumen
This work is focused on the design, development and verification of a GPU based computational technique for the simulation and control of the two dimensional shallow water equations. These equations are widely used for modeling free-surface phenomena and they usually require an important computational effort to be solved in practical cases. Traditionally, this effort has been reduced by means of large computational facilities with many computational processing units. The Graphical Processing Units (GPUs) are computing devices which contain a large number of smaller computing elements that, adequately programmed, can perform the same computations as the common CPU units in less time. Thanks to this advance, the simulation tools become attractive also for inverse design purposes. The Shallow Water Equations (SWE) have been traditionally used for forecasting and predictive purposes, but the reduction in computational effort can be used for changing the perspective of their application. In this work, an adjoint based mathematical model is developed and implemented on GPUs. Its application in the one dimensional and two dimensional mathematical model has demonstrated that it can be a useful tool for recovering missing information as well as for control purposes.
