AMG-2-INDUSTRY · Novel algebraic multigrid solver for real-life industrial applications
6РП — Действия „Мария Кюри“
- Период
- 2006-10-01 → 2007-12-31
- Финансиране от ЕС
- 151 521 €
- Участници
- 1
- Схема
- IIF
Линиите свързват координатора с партньорите.
Накратко на български
Разработват се нови алгоритми за по-бързо решаване на сложни математически уравнения, използвани например при проектирането на протези. Това помага за по-ефективното симулиране на физически процеси в индустрията, когато стандартните методи са твърде бавни или неефективни.
Кратко обяснение, генерирано от езиков модел по текста на CORDIS. Оригиналът е по-долу.
Цел на проекта
An important constraint on our ability to numerically simulate important physical processes is our ability to solve the linear and non-linear systems of equations that result from discretization of continuous mathematical models (partial differential equations). A useful computational model of a physical process is one that is accurate (includes features and scales that are important to the scientific goals) and that can be solved efficiently (acceptable computational solution time on available resources). Multiscale methods, such as multigrid, provide optimal order solution techniques for a wide range of these discrete models; however, efficient modelling of many important physical phenomena remains intractable. We propose to continue the recent development of adaptive algebraic multigrid methods and, thus, develop efficient solvers for industrial applications, such as seismic wave migration, electromagnetics, and prosthesis design. Recent advances in multigrid methodology, implemented in the adaptive algebraic multigrid algorithm, reduce the reliance of multigrid on limiting assumptions about the performance of its component parts. Complementarity between the chosen relaxation algorithm and the coarse-grid correction stage is, instead, achieved through a careful construction of the grid-transfer operators that explicitly depends on the discrete model and performance of relaxation. While this, in principle, allows efficient multigrid solution of a much larger class of problems, other challenges arise in the implementation of these ideas, particularly when the mathematical model includes complex-valued coefficients, a large near null space, or a system of PDEs. Such models, however, are ubiquitous in modern industrial applications, and so we seek to extend the adaptive algebraic multigrid algorithm to overcome such challenges.
Оригинален текст от CORDIS (на английски).
Участници
- TECHNISCHE UNIVERSITEIT DELFT · DELFTКоординаторНидерландия
Връзки
Данни: CORDIS, © Европейски съюз