CFD and Field Calculation
On behalf of our customers, we map physical relationships with highly discretized field computation and CFD methods. In pre-processing, we prepare and mesh geometries. We select suitable modeling and solution methods, perform parallelized calculations, and providevisualizations of the results that facilitate their understanding. If necessary, we automate these steps and perform model reductions.
Our range of services includes the simulation of single-phase and multiphase flow processes. Heat transfer within and between solids as well as between solids and fluids is another of our central fields of work, as is modeling of the local electric current density and associated Joule heat production in volumes and at contact points between solids. On request, we supplement the models created for our customers with chemical reaction models, for example on the basis of an Arrhenius approach.
Typical applications of CFD and field calculation are:
We use the software Ansa for the preparation of geometries used in the CFD analysis.. The meshing is done either in Ansa or in an environment matching the solver. The solvers we use are Star CCM+, various Ansys products, or OpenFOAM. Post-processing is performed in the solver environment or in Python.
If necessary, we integrate peripheral models into CFD simulations. For this purpose we either use FMUs (CoSim) or we couple our solvers with other simulation environments using TISC. The solvers we use provide their own material data libraries. Depending on the requirements, we can additionally integrate our software package TILMedia, which offers an enormous selection of media for mapping thermophysical properties.
For model reduction we use different machine learning methods. High-resolution CFD models are transformed into models with a significantly reduced state space, while retaining the information of interest to the user. In this process, it is also possible to reconstruct the corresponding states of the high-resolution model, so that a detailed description is possible even with a short computation time.
The reduced-order models can be used to accelerate CFD simulations, for example, by using a very good approximation of the result to initialize a calculation. Additionally, reduced-order models can be provided for system simulations, for example for use in Modelica or Simulink (0D/1D methods). Thus, parameter studies, controller design, or system optimizations can be performed very efficiently.
If necessary, we automate various processes, from linking and calculation to evaluation and documentation. This is mostly done in Python or Java using the APIs of the software products used, with documentation being done in PowerPoint. Automating this process allows, for example, optimizations with regard to geometry or other model properties to be carried out very efficiently.
