MpCCI - self-learning and robust coupling methods

Application background

The modelling of manufacturing processes is a current research topic in the field of coupled simulations. Examples are:

  1. Optimization of the use of coolants in machining processes: In a joint project with the Fraunhofer Institutes IWU and ITWM, the simulation tools MESHFREE and MSC.Marc) are used to investigate the influence of a coolant jet on a single chip in the running milling process.
  2. Virtual drop tests: In the CEC FORTISSIMO and ITEA VMAP projects, filled containers (e.g. chemical drums) are subjected to virtual drop tests. The code combination Abaqus and OpenFOAM is used here.
  3. Joining and adhesive modelling in the automotive industry: In a project financed by a German automotive company, an FSI model for the bonding of car body components is created (Abaqus combined with OpenFOAM).
  4. Cooling processes in continuous casting: In a project with RIST (Research Institute of Industrial Science and Technology, Korea), the cooling process was modelled in a novel continuous casting process.

The current MpCCI version supports the implementation of such applications. In this project, however, the required computing times are to be reduced.

Methodical problems

The stabilization and acceleration of the convergence of the applications described above is a current research topic. Current solutions are based on different variants of the quasi-Newton method [1]. The solution of the coupled system is regarded as a fixed point; a quasi-newton method is used to determine the fixed point. Since commercial structure and fluid solvers are only available as "black box solvers", the Jacobi matrix cannot be calculated exactly, but only approximated using several iterations.

A minimization problem is solved to approximate the Jacobi matrix. As described in [7] and [8], there are several strategies to solve this minimization problem. The determination of the parameter options (reuse of information, matrix conditioning, choice of relaxation factor, choice of time step size, etc.) depends on certain parameters of the respective coupling approach or the individual models.

For such a solution to be efficient, it must be automated as much as possible. Current methods of machine learning are thus investigated and implemented - always with the aim of keeping the user effort for model adaptation as low as possible.

Project goals

The goal of this project is the realization of application-specific MpCCI extensions for the modeling of manufacturing processes. These solutions will be realized as separate add-on modules and made available to interested customers.

Literature

  1. Quasi-Newton Methods for Unstable Partitioned Fluid-Structure Interactions - Marcel Koch, Masterarbeit an der Universität Bonn November 2016 (Supervisors Jochen Garcke, Daniel Peterseim)
  2. Analysis and optimization of flow around flexible wings and blades using the standard co-simulation interface MpCCI – Nadja Wirth, et.al. in Recent Progress in Flow Control for Practical Flows : Results of the STADYWICO and IMESCON Projects, Springer International Publishing, 2017
  3. Numerische und experimentelle Untersuchungen zur Vorhersage der integrierten Wärmebehandlung mittels Spraykühlung – Dissertation Zhuo Yu at the Leibnitz University Hannover 2014
  4. Full Coupled Numerical Simulations of the Continuous Casting Process with Electromagnetic Braking for Slabs and Thin Slabs – Martin Barna in Proceedings of Conference: 4th International Conference on Modelling and Simulation of Metallurgical Processes in Steelmaking, 2011 Düsseldorf
  5. Development and validation of a CAE chain for unidirectional fibre reinforced composite components - Kärger, Oeckerath, et.al. in Composites Structures S.350-358, 2015
  6. A new Approach for a Thermo-Mechanical Coupled Simulation of the hot Stamping Process – Koutaiba Kassem et.al. in Proceedings of MpCCI User Forum 2009
  7. Robust Quasi-Newton Methods for Partitioned Fluid-Structure Simulations – Klaudius Scheufele, Master’s Thesis, Uni Stuttgart, 2014
  8. Effiziente Algorithmen zur partitionierten Lösung starkgekoppelter Probleme der Fluid-Struktur-Wechselwirkung – Thomas Georg Gallinger, Dissertation, TU München, 2010

Project duration

December 2018 – November 2021