WP 9. Application cases towards the exascale

The numerical methods and codes developed in previous WPs will be tested and benchmarked for accuracy and scalability on a selected number of applications representing practical problems currently being addressed by the partners in different FP7 projects where exaflop computing will constitute a significant breakthrough in the size of the problem that can be effectively addressed.

We remark that this set of test cases may be changed in consultation with the EU’s officer of this project. It represents, however, a cross-section of multiphysics challenging problems that demand exascale computing.

T9.1 Analysis of fluid –structure interaction problems

We will test the explicit and implicit solvers developed in NUMEXAS in a number of fluid –soilstructure interaction (FSSI) problems related to the destructive effect of water hazards on constructions. Examples of this class of problems are common in tsunami flows and floods acting on buildings and infrastructure. Simulations of this type for real problems involve the coupling of billions of finite elements and discrete particles in order to model the complex interactions that occur in the so called particulate flow problem and its effect on adjacent constructions. The tests will be defined in coordination with the work in the FP7 project: “New Computational Methods for Predicting the security of constructions to Water Hazards accounting for fluid-soil-structure interactions” (SAFECON, ERC-2010-AdG) developed at CIMNE in the period 2011-2016.

Task leader: CIMNE. Partners involved: All (CIMNE, CESCA, LUH-IKM, LUH-HLRN, NTUA, QUANTECH)

T9.2 Computational design of advanced structural materials

In this task the numerical methods and optimization procedures developed in previous tasks will be applied to the optimal design of the material meso/micro structure arrangement and topology in engineering materials.

Work will be carried out in conjunction with the research in the on-going FP7 projects “Computational design of advanced structural materials (COMP-DES-MAT: Advanced tools for computational design of engineering materials. ERC-2012-AdG)” developed at CIMNE in the period 2013-2018, and “Optimum design under uncertainties of CNT composites” (MASTER: Mastering the computational challenges in numerical modeling and optimum design of CNT reinforced composites. ERC-2011-AdG) developed at NTUA in the period 2011-2015.

Task leader: NTUA. Partners involved: All (CIMNE, CESCA, LUH-IKM, LUH-HLRN, NTUA, QUANTECH)

T9.3 Unified numerical approximation of physical phenomena in fusion reactors

The design of breeding blankets (BBs) in a fusion reactor is one of the most challenging technological problems for the development of fusion energy. The BB consists of a set of microchannels filled with a liquid metal (Pb-Li) that will act as a shield for the electrons, and will extract the energy out of the reactor, as well as Tritium, needed to maintain the plasma. Unfortunately, experimental set-ups will not be at our disposal during the next decade, and numerical tools that provide realistic simulations are necessary to aid in their design process.

This task focuses on the numerical simulation of liquid metals in BBs that surround the plasma; this phenomenon is governed by the thermally coupled incompressible viscoresisitive MagnetoHydroDynamic (MHD) equations, both in the full and inductionless formulation. The numerical solution in a moderate time of the discretized MHD equations arising in realistic production simulations requires (reasonable) weakly scalable balancing domain decomposition solvers to be mapped on clusters/massively parallel computers with up to several dozen (if not hundreds) of kilo computational cores.

Task leader: CIMNE. Partners involved: All (CIMNE, CESCA, LUH-IKM, LUH-HLRN, NTUA, QUANTECH)

T9.4 Optimization of industrial codes for the manufacturing industry

In this task we will implement and validate the explicit structural analysis technology and the optimization methods developed in NUMEXAS in the commercial code STAMPACK of QUANTECH for analysis and stochastic optimization of large scale sheet metal stamping operations. These problems require thousands of three dimensional (3D) analyses involving each many hundreds of thousands of finite elements of both, shell and solid type, as well as complex frictional contact interactions and non-linear material behaviour. The need of exascale computing for solving practical industrial problems of this type is evident.

A second objective of these tasks is to implement and validate the implicit structural analysis technology developed in NUMEXAS in the commercial codes VULCAN and Click2Cast of QUANTECH for analysis and stochastic optimization of large scale casting problems. These type of problems also fit well in the exascale type of problems addressed in the project as they require thousands of fluid flow analyses involving each one many hundreds of thousand finite elements for simulating the mould filling process, as well as a similar number of 3D thermal-mechanical analyses for simulating the solidification and cooling phases. For the work related to the fluid flow analysis of the mould filling problem, we will make use of the computational technology developed in the project: “Real Time Computational Mechanics Techniques for Multi-Fluid Problems“(REALTIME, ERC-2009-AdG) developed at CIMNE in the period 2010-2014.

Task leader: QUANTECH. Partners involved: All (CIMNE, CESCA, LUH-IKM, LUH-HLRN, NTUA, QUANTECH)

D9.1 Report of applications to fluid-structure interactions problems

Lead beneficiary: CIMNE

D9.2 Report of applications to advance materials design

Lead beneficiary: NTUA

D9.3 Report of applications to physical phenomena in fusion reactors

Lead beneficiary: CIMNE

D9.4 Report of applications to industrial codes for the manufacturing industry

Lead beneficiary: QUANTECH

Friday, April 28, 2017
  [ login ]

The research leading to these results has received funding from the European Community's Seventh Framework Programme under grant agreement n° 611636