Accéder directement au contenu Accéder directement à la navigation
Communication dans un congrès

Regularized Artificial Neural Networks for Predicting the Strain of Traction-aged Polymer Systems Part I

Abstract : The Liquid Resin Infusion (LRI) is a process that has the greatest development and cost reduction potential for the manufacture of large complex parts which made of composite materials. The viscosity/temperature pair is the essential criterion for the smooth running of the infusion in order to obtain composite parts of quality. However, humidity is a threatening factor for composite materials. Therefore, aging factors and a predictive model of durability were investigated on a new polymer B and second time on A-150, A-185 polymer systems already certified for use in the aircraft and aerospace industry. Tensile tests were carried out at temperatures T = −40 • C, 25 • C, 70 • C. In this paper, an initial small experimental dataset of 33 samples is used to analyze the strain of polymers systems as a function of aging time, temperature, Young modulus and the breaking stress. In the view of the very small dataset, the strain of polymers systems is predicted by training Levenberg-Marquardt (LM), Bayesian regularization (BR), and Broyden-Fletcher-Goldfarb-Shanno (BFGS) algorithm with a regularized cost function algorithms. This paper is considered a first part of the problem. In this part, we give the setting of the experimental problem and we approach the theoretical part concerning Bayesian regularization and the BFGS algorithm. The article of part II will present the numerical results as well as its analysis.
Liste complète des métadonnées
Contributeur : helene canot Connectez-vous pour contacter le contributeur
Soumis le : lundi 25 avril 2022 - 11:45:26
Dernière modification le : vendredi 29 avril 2022 - 03:25:49


Fichiers produits par l'(les) auteur(s)


  • HAL Id : hal-03650820, version 1



Helene Canot, Philippe Durand, Emmanuel Frenod, Bouchera Hassoune-Rhabbour, Valerie Nassiet. Regularized Artificial Neural Networks for Predicting the Strain of Traction-aged Polymer Systems Part I. International Conference of Applied Ingineering Mathematics, Jul 2022, Londres, France. ⟨hal-03650820⟩



Consultations de la notice


Téléchargements de fichiers