Opendata, web and dolomites

Report

Teaser, summary, work performed and final results

Periodic Reporting for period 1 - ROMSOC (Reduced Order Modelling, Simulation and Optimization of Coupled systems)

Teaser

The development of high quality products and processes is essential for the future competitiveness of the European economy. In most key technology areas product development is increasingly based on simulation and optimization via mathematical models that allow to optimize...

Summary

The development of high quality products and processes is essential for the future competitiveness of the European economy. In most key technology areas product development is increasingly based on simulation and optimization via mathematical models that allow to optimize design and functionality using free design parameters. Best performance of modelling, simulation and optimization (MSO) techniques is obtained by using a model hierarchy ranging from very fine to very coarse models obtained by model order reduction (MOR) techniques and to adapt the model and the methods to the user-defined requirements in accuracy and computational speed.
Within the ROMSOC project a common framework is derived and the next generation of researchers is trained in this highly interdisciplinary field, in particular for SMEs, in areas such as optical and electronic systems, economic processes, and materials. A joint training programme has been developed which builds on the strengths of the academic and industrial partners and their strong history of academic/industrial cooperation. ROMSOC educates highly skilled researchers in mathematical MSO that will become facilitators in the transfer of innovative concepts to industry.

Work performed

The starting phase involved organizational tasks, the recruitment of the fellows, and the organization of network-wide training activities. The intensive training of the ESRs has been completed, and they had the opportunity to present their work in Workshops and at international conferences. The concept of creating, for a complete industrial product or process, a model hierarchy ranging from fine grain detailed models to very coarse and simplified models is a highly innovative concept. In ROMSOC it is combined with the emphasis of retaining the parameter dependence, the coupling of different physical phenomena, and to adapt to user defined accuracy and efficiency desires. This requires a close interaction of academia and industry and the training of qualified professionals that can operate on this interface. First steps in the research projects have been taken and first results including a selection of benchmark cases have been delivered.

ESR1: In a feasibility study possible algorithms for the European Extremely Large Telescope instrument MAORY in terms of computational performance were analysed. A parallel GPU implementation was achieved.
ESR2: A coupled model for porous-fluid interactions in no-flow conditions was developed. A compatibility study of different porous models with different materials is performed.
ESR3: Using the Optimal Transportation formulation of the far field/point source reflector problem, the Entropic regularization methods and the companion Sinkhorn algorithm have been adapted. The efficiency of the method could be enhanced.
ESR4: A network implementation has been achieved to deploy the deep image prior/ regularization by architecture approach to Magnetic Particle Imaging (MPI).
ESR5: Considering the objectives set out by ST Microelectronics, a benchmark case has been constructed for the coupling of multirate and MOR techniques for circuit simulation. A more general framework is under construction for a wider array of circuit components and physical aspects.
ESR6: A MOR approach has been developed for high dimensional convection-diffusion reaction PDEs arising in computational finance. For efficient parameter sampling, an adaptive greedy-POD approach has been introduced.
ESR7: Together with the industry partner DB the requirements for modelling crew and vehicle scheduling have been collected. An optimization model to cover all necessary application aspects has been compiled. Preliminary computational results were achieved for real-world planning data.
ESR8: A methodology for the real time solution of inverse heat transfer problems was identified and adapted to continuous casting molds. An ad hoc MOR technique for the mentioned methodology was developed.
ESR9: Numerical approaches for fluid-structure interaction (FSI) problems in blood pump systems have been studied as well as the membrane-based technology developed at CorWave Inc. A software in the FEM library LIFEV for 3D fluid-structure interaction simulations has been developed.
ESR10: The modelling of thermo-hydro-mechanical phenomena arising in blast furnaces is studied with particular focus on MOR and numerical simulation. New MOR techniques for coupled problems and the application to coupled thermo-mechanical problems in the blast furnace hearth process have been proposed.
ESR11: The analysis and numerics of Navier-Stokes equations and turbulence models to model high Reynolds numbers fluid flow has been studied. Fluid flow simulations with real-world data were conducted as a basis to run shape optimization subject to fluid flow.

Final results

Various progress beyond the state-of-the-art could be made in the specific subprojects, in particular
• parallel implementation of the proposed algorithm FEWHA on a GPU;
• application of the new deep learning algorithms for the analysis of MPI data;
• derivation of a model that covers compatibilities between drivers, locomotives and trains, particular Polish labour code provisions, and long-term maintenance requirements;
• application of the XFEM numerical approach to solve a 3D fluid-structure interaction problem in a real scenario;
• design of a model hierarchy for the numerical simulation of acoustic wire-screens;
• application of inverse problem techniques to the in-situ characterization of porous materials by means of free-field experimental measurements;
• application of Entropic regularization and the Sinkhorn algorithm for the solution of an optimal transport problem;
• new MOR methods for the solution of inverse problems arising in steel castings including non-regular data;
• new MOR methodologies for non-linear thermo-mechanical problems in 3D structures and real-time 3D simulation of a blast furnace hearth.

The following results are expected until the end of the project:
• a complete set of benchmark cases of model hierarchies that will be used for interdisciplinary research and future training programmes;
• new MSO and MOR methods for specific industrial applications;
• software packages based on the new methods;
• a-posteriori error estimates to be used in different industrial applications.

The ESRs benefit from an increased set of skills leading to improved employability and career prospects. The institutions within ROMSOC benefit from an enhanced cooperation and better transfer of knowledge between academia and industry. The academic partner institutions obtain new and improved training programmes. The beneficiaries of the ROMSOC project gained boosted R&D capacity. The ROMSOC project has increased international mobility of researchers in Europe.

Potential socio-economic impact could be made, addressing various societal challenges such as health and wellbeing, secure, clean and efficient energy, or smart, green and integrated transport, e.g:
• improved utilization of production factors and transportation capacity at DB;
• support of CorWave in the optimization of device design and performance by means of a software for 3D simulations;
• reduction in the need of in vivo trials in animals and in the risks in human clinical trials;
• increased ability to analyse the risk of financial instruments via efficient testing methods;
• improvement in the industrial design process of free-formed reflectors.

Website & more info

More info: http://www.romsoc.eu.