Robust design optimization with respect to model and data uncertainties in computational mechanics

Evangéline Capiez-Lernout (1) , Christian Soize (1)

1 LaM - Laboratoire de Mécanique

In this paper, a probabilistic approach is proposed to solve the robust design optimization problem of complex dynamical systems not only with respect to system parameters uncertainties but also to model uncertainties. The possible designs are represented by a numerical finite element model whose parameters belong to an admissible set of design parameters. A recent nonparametric probabilistic model of uncertainties is used for taking into account model uncertainties and data uncertainties. The robust design optimization problem is formulated as a multi-objective optimization problem which consist to minimize a cost function including a target with respect to an admissible set of design parameters. The theory is presented followed by a numerical application.

Mots clés: robust optimization; structural dynanamics; vibration; uncertainty quantification; nonparametric probabilistic model; modeleing errors; model uncertainties; random matrix; random uncertainties

Domaines: Vibrations [physics.class-ph]; Vibrations [physics.class-ph]; Optimisation et contrôle [math.OC]; Probabilités [math.PR]

Evangéline Capiez-Lernout, Christian Soize. Robust design optimization with respect to model and data uncertainties in computational mechanics. 5th Computational Stochastic Mechanics Conference (CSM), 2006, Jun 2006, Rhodes, Greece. pp.Pages : 139-146. ⟨hal-00689696⟩