Generalzed Interfaces and non intrusive Coupling: R&D software embedded into industrial software for computational analysis of non linear structures – ICARE

In order to meet the non-intrusive requirement, any modification of the generalist industrial software used for solving the simplified global model is prohibited. This results in banning any adjustment of the global mesh. The un-availability of results on the complementary part of the local model is a major technical issue. Furthermore, within a context of probabilistic analysis, the requirement results in considering determinist global models, computable by standard FE software, while handling uncertainties at local stage only.

An essential constitutive block of such methodology is a media that ensures efficient communications between global and local models (through interfaces). Any adjustment of the global simplified mesh is forbidden. In addition, the method should be able of considering models with different dimensions, such as a mixture of global models with structural elements (beams, shells) with local models made of solid elements.

Starting date: 01/01/2013
Kick-off meeting: 17-18/01/2013
Periodic meeting n°1: 25/09/2013

During the first semester, computations with preliminary implementations of the non-intrusive coupling methodology have been performed within the numerical framework selected for ICARE (Code_Aster). The feasibility of the implementation has been demonstrated, and the necessary adaptations of the method to the numerical framework have been identified. A set of use-cases has been collected and documented, representative of industrial configurations (simplified but still representative): crack propagation within homogeneous media and complex geometry (2D and 3D), notched plates with local non-linearity in stress concentration areas (2D and 3D), coupling between global 2D and local 3D models, management of uncertainties localized within an heterogeneous media.

The objective through ICARE is to develop methodologies and derive numerical implementations into the open FE environment Code_Aster, which meets industrial requirements in terms of maintenance, documentation and backward compatibility. Industrialization of methods is out of ICARE’s scope, but evaluation during the project on test cases derived from actual industrial experiences will arouse opportunities for transfer towards EADS and EDF operational environments; such transfer towards industry will demand additional practices within development programs that will be assumed by EADS and EDF afterwards, within a delay of one to some years after ICARE. During ICARE, elementary blocks constitutive of the non-intrusive methodology (field transfer, re-meshing), will be integrated into DISTENE’s commercial offer towards software editors, thus towards end-users of software for structural analysis end pre-, post-processing.

1. M. Chevreuil, A. Nouy, E. Safatly, “A multiscale method with patch for the solution of stochastic partial differential equations with localized uncertainties”, Computational methods in Applied Mechanical Engineering, 255 (2013) 255-274
2. G. Guguin, O. Allix, P. Gosselet, S. Guinard, “non-intrusive coupling between 3D and 2D laminated composite models based on finite element 3D recovery”, To be published in International Journal for Numerical Methods in Engineering, 2013
3. M Duval, “couplage non intrusif et propagation de fissures”, rapport de stage INSA, cotutelle UTO/EADS, 2013

The project aims at developing new computational methods for the analysis of large and complex structures.

The challenge is to investigate local areas, with potential severe influence on the overall strength, but embedded in large structures with geometrical complexity. Usually, when addressing global and local modeling, a set of models, discretizations, behavior laws and numerical tools with increasing complexity are used. Whereas standard software are used for global analysis (ABAQUS, NASTRAN…), with topological idealization (plate theory, geometrical simplification), behavior idealization (homogenization), statistical idealization (deterministic models), local analysis demand for specialized software (routines, dedicated software) when handling faithful models of the geometry and the physics (contact, heterogeneity, damage, uncertainty).

Though computational methods exist that can connect different models, there still remains heavy constraints that make it unaffordable for industry ; indeed, bridging models of different natures demands for significant efforts when designing these models : a geometry of the connection area should be prepared, its discretization also, then quantities of interest should be computed in this area, and finally a special post-processing of these quantities should be needed (when dealing with shell to solid couplings for instance).

Non-intrusive coupling is a scientific breakthrough that will enable a brand new way of connecting the models together, with characteristics as follows:

• Constraints on topology and meshes of global models will be minimized: transfer towards industry will be made easy, since the global model will not need any specific processing ;
• Quantities of data exchanged will be minimized (thus CPU times will be optimized) ;
• Specific developments of standard software will be minimized; global and local software remain independent, particularly as for meshes and solvers are concerned.

As a consequence, a major outcome of ICARE will be to accelerate the dissemination of R&D production towards industry: once made easy, coupling generic models with specialized ones will become compatible with an industrial framework.

With EADS and EDF carrying the most critical needs regarding simulation, the consortium groups three academic partners, specialists of computational mechanics (GeM, LMT, UTO) and one SME DISTENE, acknowledged by major finite element method software editors: DISTENE has for long being acting as a key player in integrating CAD, analysis and post-processing into one unified numerical suite.