Doctorat en génie
2014 - Present
Universidad Nacional de Rosario, Rosario. CIFASIS-CONICET
Groupe: Simulation et contrôle de systèmes dynamiques
Thèse: Simulation efficace de systèmes discontinus rigides.
[
pdf en espagnol
]
Abstract:
In this Thesis, new numerical integration methods for ordinary differential equations are proposed. These methods combine the ideas of classic discrete time and state quantization-based methods.
As a first result, we propose an extension to linearly implicit quantized state system algorithms that avoids the appearence of spurious oscillations in the integration of certain stiff systems. Under the detection of oscillations, this algorithm performs a classic method step on a pair of states avoiding the problem. It is shown that the use of this algorithm is particularly useful for the simulation of switched circuits in presence of parasitic components.
The second contribution of the Thesis is a mixed algorithm that splits a model so that a subsystem is integrated using quantized state systems methods and the remaining subsystem is integrated with a classic algorithm. This algorithm then exploits the advantages of both approaches, improving the overall simulation performance in several applications, particularly in multidomain problems.
Theoretical convergence and stability results for the mixed numerical integration scheme are also obtained. In addition, the efficiency of the proposed methods is verified by comparisons of performance with respect to classical methods and methods based on quantification.
Ingénierie électronique
2007 - 2013
Universidad Nacional de Rosario, Rosario.