Curriculum Vitæ

Pierre KERFRIDEN Doctor in Mechanics Professeur agrégé in Mechanics ATER at ENS de Cachan Born July 28, 1981 in La Teste, France Nationality: French

LMT-Cachan 61 avenue du Président Wilson 94235 CACHAN CEDEX Phone number: +33 (0) 1 47 40 22 26 Cell number: +33 (0) 6 64 76 09 48 E-mail: [email protected]

Research activities 2009

ATER ( TEMPORARY ASSISTANT PROFESSOR ) AT LMT C ACHAN (Laboratoire de mécanique et technologie), France (ENS de Cachan / CNRS UMR 8535 / UPMC / Pres Universud Paris), “Structures and Systems” team. • Research interest: domain decomposition methods, parallel computing, multiscale strategies, resolution of large nonlinear problems, application to the computation of the evolution of damage in structural components (composite laminates in particular) until final failure, coupling of buckling and damage. • Referees : Pr. O. Allix ([email protected]), Dr. ([email protected]).

2005-2008

Pierre Gosselet

P H . D. THESIS AT LMT-C ACHAN (ENS de Cachan / CNRS UMR 8535 / UPMC / Pres Universud Paris), “Structures and Systems” team. • Title: “A three-scale domain decomposition method for the analysis of the debounding in laminates (Stratégie de décomposition de domaine à trois échelles pour la simulation du délaminage dans les stratifiés)”. • Application field: computational mechanics. • Key words: homogenization, multiscale strategies, domain decomposition methods, parallel computing, nonlinear intensive computations, composite laminates, debounding, damage theory. • Supervisor : O. Allix / Advisor : P. Gosselet. • committee: J.-L. Chaboche (committee foreman), N. Moës (secretary), A. Corigliano (secretary), S. Maison-le-Poc, O. Allix, P. Gosselet. • Distinction: with high honors (thesis defended on November 24, 2008).

2005

MS C 2 RESEARCH PROJECT AT TU D ELFT (Netherlands), faculty of Mech., Maritime and Materials Eng., dpt. of Engineering Dynamics (5 months) • Title: “The FETI method applied to multibody dynamics”. • Application field: computational mechanics. • Key words: mechanics of rigid multibodies, nonlinear dynamics, domain decomposition methods. • Advisor: D. Rixen.

Teaching activities 2008-2009

ATER ( TEMPORARY ASSISTANT PROFESSOR ) AT THE M ECHANICAL E NGINEERING FACULTY OF THE ENS DE C ACHAN (Ecole Normale Suprieure). • Teaching hours: 192. • Teaching topics: – tutorials in numerical methods applied to mechanical systems (grade: MSc1 and Bachelor): numerical approximation of boundary value problems (finite difference, finite element), direct and iterative solvers, resolution of eigenvalues problems, finite element theory and technology. – tutorials in computational mechanics and sizing of structures (grade: MSc1): modelling of deformable structures, mechanical design using the finite element method. – tutorials in CAD (grade: MSc1).

2005-2008

M ONITEUR ( TEACHING des Arts et Métiers).

ASSISTANT ) AT

ENSAM PARIS (Ecole Nationale Supérieure

• Teaching hours per year: 64. • Teaching topics: tutorials in Mechanical design for engineers (grade: MSc1).

Work experience 2003

MS C 1

PROJECT AT

EADS CCR, Suresnes, France (4 months).

• Title: “Numerical simulation of the hydraulic ram generated by high velocity impacts on tanks”. • Application field: computational mechanics. • Key words: high velocitiy dynamics, impact problems, fluid-structure interaction, explicit time integration scheme, finite element simulations. • Advisor: S. Maison-le-Poëc.

Education 2004-2005

MS C 2 “A DVANCED T ECHNICS IN C OMPUTATIONAL S TRUCTURAL M ECHANICS ” ENS DE C ACHAN (ENS Cachan, Paris VI, ENSAM).

AT

• Main topics: CAD, finite element modelling and technology, parallel computing, multiscale and multiphysic problems, nonlinear mechanics (plasticity, damage, contact, large displacements), dynamics (shocks, vibrations and wave propagations), homogenization of heterogeneous materials, stochastic approaches. 2003-2004

P REPARATION OF Agrégation (competitive examination for the recruitement of high school teachers in mechanics). Acceptance in 2004.

2002-2003

MS C 1

IN

M ECHANICAL

ENGINEERING AT

ENS

DE

C ACHAN .

• Main topics: mechanics of solids and fluids, vibrations, numerical methods, material science, CAD, advanced structural mechanics (nonlinear behavior of structures, heterogeneous materials), technology of mechanical systems.

2001-2002

BACHELOR

IN

M ECHANICAL

ENGINEERING AT

ENS

DE

C ACHAN (with honors)

• Main topics: see MSc1 (first year) in Mechanical engineering. 1999-2001

Preparation in mathematics, physics and technology for entrance to French Grandes Ecoles at Lycée J.-B. Say in Paris, ENTRANCE TO ENS DE C ACHAN (Best school for teachers and researchers training in France) in 2001.

1999

S CIENTIFIC "B ACCALAURAT " (french equivalent to the high-school-leaving certificate) at Lycée J.-B. Say (with honors).

Skills • French: native language Languages • English: fluent (toeic: 890) Programming

C/C++, MPI, Python, Java, PHP, MySQL, HTML, Maple, Matlab, Scilab • CAD: Catia, SolidWorks, Abaqus/Simulia (Dassault Systemes)

Softwares

• Finite Element: Cast3M (CEA), Samcef (Samtech), Radioss (Altair Engineering) • Office: Microsoft Office, OpenOffice.Org, Latex

Operating syst.

Linux, Mac OS, Windows

Other interests Music

Guitar and Scottish bagpipe playing, active member of the celtic music association "Bagad Keriz".

Sports

Basket-ball playing, sailing

Detailed research activities

Research interest and experience Damage prediction in composites using multiscale methods The aim of my Ph.D. thesis is to develop efficient numerical tools to solve debonding problems in composite laminates. The targeted problems are, typically, laminates structural components such as joints in aircrafts. The chosen scale for the description of the material is the scale of the elementary ply (“meso” description of the laminates) as, in the past decades, it has been shown to be the coarser scale that allows to perform predictive simulations. We use the cohesive interface model from [1]. It is a simplified version of the model from [4], which describes all the potential micro and macro defects in the laminates at the meso-scale. Such a desciption leads to very large numerical problems, whose resolution requires intensive computation tools. In order to overcome this difficulty, we consider the multiscale method presented in [5] as a starting point for our numerical developpments. It includes an homogeneization procedure, which results in the construction of a so-called “macroscopic” problem. In the case of debonding analysis, we choose the ply’s thickness as a length scale for the homogenization of the laminates (i.e : at most one homogenized cell within the ply’s thickness). Although this choice is reasonnable in the low-gradient zones, this homogenization procedure is not relevant in the crack’s front region (because of the localized very high gradients). We thus improve the initial multiscale strategy to perform a valid scales separation in this particular zone [2]. More recently, our research focuses on the control of the implicit “time” discretization scheme used to solve the quasi-static debonding problem. In fact, the solution that is reached using the classical incremental procedure strongly depends on the chosen time increment, especially when sharp global limit-points are involved in the analysis (due to the integration of the softening local behavior in the case of debonding). In addition to the use of a local arc-length algorithm [7], we develop a strategy to adapt automatically the time increment to the “level of nonlinearity” of the problem to solve at each time step. Our last research topic is the damage / buckling coupling in a multiscale framework, which is the subject of a MSc2 research project that I am currently supervising.

Domain decomposition methods and parallel computing for intensive nonlinear computations My first contact with domain decomposition methods dates back to my MSc2 research project. The purpose of this work is to use a Schur-based domain decomposition method [6] in order to solve rigid multibody dynamics problem. An implicit time integration scheme is used for this analysis and, at each time step, the solution is obtained by using a Newton-Raphson resolution scheme (the non-linearities are, in this framework, caused by the large displacements of the rigid bodies). The prediction steps of this Newton scheme are then solved using the chosen Schur-based domain decomposition method. During my Ph. D. thesis, this knowledge in intensive computation is exploited to solve large debounding problems described at a fine scale. The nonlinearities are, in this case, due to the choice of a cohesive behavior to describe the damage evolution in the structure. The challenge is here to use a unique iterative algorithm to solve both the nonlinear problem and the condensed interface problem coming from the domain decomposition. The framework chosen is the LaTIn method [3, 5], which allows to prescribe nonlinear constitutive laws as constraints between sub-domains (contact, cohesive law,...). An important part of this work consists in stabilizing and enhancing the chosen solver, whose efficiency, in its standart version, strongly depends on the damage state of the structure. As explained in previous section, this domain decomposition method includes the resolution of an homogenized problem. In our cases, this coarse grid problem becomes extremely large when dealing with industrialist’s structures (the number of homogeneized cells depends on the number of plies and of the inplane dimensions of the laminates structure), which prohibits its direct resolution. The BDD [6] method is then employed to solve in parallel the linear coarse grid problem. This Schur-based domain decomposition strategy is based on the use of a projected Krylov algorithm (conjugate gradient, GMres). Consequently, a third length’s scale is naturally introduced in our enhanced multiscale method (coarse grid problem associated to the projector of the Krylov algorithm). A multiple right hand side strategy has recently been developped to increase the efficiency of this solver: the meaningfull macroscopic information captured during previous resolutions is reused for the current inversion (a full resolution

of the “macroscopic” problem is required at each iteration of the fine scale LaTIn solver). The timevariations of the “macroscopic” fields being very slow, this problem reduction strategy drastically reduces the computation time in quasi-static analysis. Efficient computing tools have been used in order to ensure the numerical scalability of this threescale method (robust algebraic solvers and linear algebra functions, LAM/MPI librairies for the data exchanges between processors). Thanks to these efforts, we are now able to assess "meso" debonding problems in industrialists’s structural components in reasonnable computing time and memory [2].

Ph. D. details Title

A three-scale domain decomposition strategy dedicated to the prediction of the debounding in composite laminates.

Keywords

homogenization, multiscale strategies, domain decomposition methods, parallel computing, nonlinear intensive computations, composite laminates, debounding, damage theory.

Abstract

Composite laminates are widely used by the industrialists, as they are light and enable to optimize the design of structures for given applications, which results in a global drop of energetic costs. Nevertheless, performing predictive simulations of the global behavior and the degradation processes of the laminates requires a very fine description of the material. In particular the cohesive zone models have demonstrated their ability to predict the debonding of laminates. In order to perform the simulation of the debonding in industrialist’s structures described by such models, a three-scale domain decomposition method is used to solve, at each time step of the time integration scheme, the nonlinear global equilibrium problem. The introduction of three levels of resolution is the key point of the strategy, as it enables to deal specifically with the informations at the fine scale, the scale of the ply and the scale of the structure through the iterative process. This method is then adapted to handle efficiently the sharp nonlinearities which are caused by the softening behavior of the material, and which are confined in very small zones and may result in global instabilities.

Selected publications 2009

P. Gosselet, P. Kerfriden, O. Allix A three-scale strategy for the analysis of delamination of stratified composites , keynote lecture 10th US National Congress of Computational Mechanics (USNCCM X) July, 2009, Columbus, Ohio, USA

2009

O. Allix, P. Kerfriden, P. Gosselet A dedicated multiscale approach for the simulation of delamination propagation 10th International Conference on Computational Plasticity (Complas X) Septembre, 2009, Barcelona, Spain

2009

P. Kerfriden, O. Allix, P. Gosselet A three-scale domain decomposition method for the 3D analysis of debonding in laminates Computational Mechanics, Vol. 33, Num. 3, Pages 343-362

2008

P. Kerfriden, O. Allix, P. Gosselet Multiscale analysis of delamination in composites laminate 6th International Conference on Computation of Shell and Spatial Structures (IASSIACM 2008) May, 2008, Ithaca, USA

2007

O. Allix, P. Kerfriden, P. Gosselet Recent developments of a multiscale strategy for crack propagation and delamination 9th International Conference on Computational Plasticity (Complas IX) Septembre, 2007, Barcelona, Spain

References [1] O. Allix and P. Ladevèze. Interlaminar interface modelling for the prediction of delamination. Computers and structures, 22:235–242, 1992. [2] P. Kerfriden, O. Allix, and P. Gosselet. A three-scale domain decomposition method for the 3d analysis of debonding in laminates. Computational mechanics, DOI : 10.1007/s00466-009-03783, 2009. [3] P. Ladevèze. Sur une famille d’algorithmes en mécanique des structures. l’académie des Sciences, 300(2):41–44, 1985.

Compte rendu de

[4] P. Ladevèze and G. Lubineau. An enhanced mesomodel for laminates based on micromechanics. Composites Science and Technology, 62(4):533–541, 2002. [5] P. Ladevèze and A. Nouy. On a multiscale computational strategy with time and space homogenization for structural mechanics. Computer Methods in Applied Mechanics and Engineering, 192:3061– 3087, 2003. [6] J. Mandel. Balancing domain decomposition. Communications in Numerical Methods in Engineering, 9(233-241), 1993. [7] J.C.J. Schellenkens and R. De Borst. On the numerical integration of interface elements. International Journal for Numerical Methods in Engineering, 36(1):43–66, 1993.