Received: 21 June 2018 Accepted: 26 August 2018 DOI: 10.1002/pamm.201800332 On efficient computation of 3-d simulation within TPM 2 -Framework Florian Bartel 1, * , Tim Ricken 2 , Jörg Schröder 3 , and Joachim Bluhm 3 1 Chair of Structural Analysis and Dynamics, TU Dortmund University, August-Schmidt-Str. 6, 44227 Dortmund 2 Institute of Mechanics, Structural Analysis, and Dynamics, University of Stuttgart, Pfaffenwaldring 27, 70569 Stuttgart 3 Institute of Mechanics, University of Duisburg-Essen, Universitätsstr. 15, 45141 Essen With this contribution we would like to communicate the state of the art of TPM 2 application to realistic engineering problems. First of all, a conceptional overview of TPM 2 is shortly given, secondly we illustrate the benefit of Computer Tomography (CT) technology to capture geometry and create finite element meshes. Further, the application of the domain decomposition (DD) method for parallel execution will be shown on an example of a fluid saturated porous unit cube and finally we give advise for additional acceleration of computational runtime via model order reduction (MOR) for the TPM 2 -Framework. c  2018 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 1 Briefly on TPM 2 -Method The TPM 2 -Method includes the Theory of Porous Media (TPM), cf. DE BOER & EHLERS [3], into the two-scale homogenization scheme of the FE 2 -Method, see SCHRÖDER [5]. Within this framework fluid saturated porous domains can be analyzed to determine large poro-elastic defor- mation, hydrostatic pressure, stress states and fluid flow quantities and directions. For this purpose, the material response is received from a homogenized discrete component distribution of a representative volume element (RVE) on a microstructural level. This Method leads to a class of two-scale, non-linear, coupled and time dependent problems. A further description and a 2-d example can be found in BARTEL ET AL. [4]. 2 Capture realistic FE-Geometric-Model with CT-Scanning technology The challenge in numerical simulation is to approximate the real physical behavior as close as possible. On the one hand the suitable theory and numerical treatment are essential, on the other hand the geometry approximation via the finite element discretization are also of great importance. One way to achieve this requirement is for instance taking a CT-Scan of the structure of interest. As a feasibility study, which is pictured in Fig. 1, we captured a Ultra High Performance Concrete (UHPC) specimen and successfully transferred the image data, after an idealization process, to a finite element model. Both the macro as well as the micro scale can be imaged for the purpose of engineering problems. It can be seen that a suitable finite element mesh for a realistic problem can easily require up to 10 5 elements. In order to solve these kind of large scale problems the mechanical theory has to be embedded in a high performance environment meaning usage of Fortran or C++ with Intel compilers and optimization levels, up to date processors, derivation of analytical tangents, parallelization strategies or model order reductions, just to mention few possibilities of computational runtime acceleration, which we are dealing with for the TPM 2 calculations. Fig. 1: a) UHPC specimen (@WdB, Prof. Orlowsky, TU Dortmund), b) CT-Scanning device (@WPT, Prof. Walther, TU Dortmund), c) .Raw image data captured with CT-Scanning device, d) Idealized geometry modeled with Mimics, e) Finite element mesh created with Mimics, f) Von Mises stress (compression test) computed with ANSYS. * Corresponding author: e-mail florian.bartel@tu-dortmund.de, phone +49 231 755 4682, fax +49 231 755 2532 PAMM · Proc. Appl. Math. Mech. 2018;18:e201800332. www.gamm-proceedings.com c  2018 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 1 of 2 https://doi.org/10.1002/pamm.201800332