Skip to Main content Skip to Navigation

Contribution au développement de méthodes numériques destinées à résoudre des problèmes couplés raides rencontrés en mécanique des matériaux

Abstract : This work presents the development of the Proper Generalized Decomposition (PGD) method for solving stiff reaction-diffusion equations in the framework of mechanics of materials. These equations are particularly encountered in the oxidation of polymers and it is therefore necessary to develop a tool to simulate this phenomenon for example for the ageing of organic matrix composites in aircraft application. The PGD method has been chosen in this work since it allows a large time saving compared to the finite element method. However this family of equations has never been dealt with this method. The PGD method consists in approximating a solution of a Partial Differential Equation with a separated representation. The solution is sought under a space-time separated representation for a 1D transient equation.In this work, a numerical tool has been developed allowing a flexibility to test different algorithms. The 1D Fickian diffusion is first evaluated and two numerical schemes, Euler and Runge-Kutta adaptive methods, are discussed for the determination of the time modes. The Runge-Kutta method allows a large time saving. The implementation of the numerical tool for reaction-diffusion equations requires the use of specific algorithms dedicated to nonlinearity, couplingand stiffness. For this reason, different algorithms have been implemented and discussed. For nonlinear systems, the use of the Newton-Raphson algorithm at the level of the iterations to compute the new mode allows time saving by decreasing the number of modes required for a given precision. Concerning the couplings, two strategies have been evaluated. The strong coupling leads to the same conclusions as the nonlinear case. The linear stiff systems are then studied by considering a dedicated method, the Rosenbrock method, for the determination of the time modes. This algorithm allows time saving compared to the Runge-Kutta method. The solution of a realistic nonlinear stiff reaction-diffusionsystem used for the prediction of the oxidation of a composite obtained from the literature has been tested by using the various implemented algorithms. However, the nonlinearities and the stiffness of the system generate differential equations with variable coefficients for which the Rosenbrock method is limited. It will be necessary to test or develop other algorithms to overcome this barrier.
Document type :
Complete list of metadata
Contributor : ABES STAR :  Contact
Submitted on : Monday, September 26, 2016 - 2:20:08 PM
Last modification on : Wednesday, November 3, 2021 - 5:56:50 AM
Long-term archiving on: : Tuesday, December 27, 2016 - 1:04:29 PM


Version validated by the jury (STAR)


  • HAL Id : tel-01371753, version 1



Andrea Ramazzotti. Contribution au développement de méthodes numériques destinées à résoudre des problèmes couplés raides rencontrés en mécanique des matériaux. Autre. ISAE-ENSMA Ecole Nationale Supérieure de Mécanique et d'Aérotechique - Poitiers, 2016. Français. ⟨NNT : 2016ESMA0007⟩. ⟨tel-01371753⟩



Record views


Files downloads