Offers

Fraunhofer Institute for Algorithms and Scientific Computing SCAI

Services

Our research and development primarily serves the goal of improving and opening new fields of application for simulation software.

We offer

  • Analysis of the robustness of crash simulation results
  • Methods and tools for the analysis and evaluation of simulation results
  • Many years of numerical, algorithmic and informatics experience for increasing the computational efficiency of simulation software

Crash Simulation

Crash simulation belongs to the most demanding disciplines in numerical simulation of industrial applications.

Bifurcations and resulting instable behaviour of simulation models are commonplace. Our DIFF-CRASH software tool allows for scatter and stability analysis of the simulation models. In particular, finding and backtracking instabilities is facilitated.

In addition, SCAI offers an analysis of whole process chains, taking production processes (forming, casting) and the resulting material history into account. Sensitivity and robustness analysis, and the solution of robust multi-objective optimization problems for whole process chains are enabled - even for highly resolved simulation grids.

Fast Solver Technology

For full-spectrum industrial simulations, the solving of extremely large systems of equations plays a central role. For these, the solution time determines whether a method is practicable. Researching and developing numerical algorithms which solve such systems of equations at an optimal speed (scalable solvers) has been one of our central themes for many years. We have extensive experience with modern hierarchical algorithms (multigrid and general multilevel algorithms), and in particular with their deployment in industrial simulation software. The highly efficient solver modules of our programme package SAMG have already been integrated into many commercial software products. Their areas of application include fluid dynamics, oil reservoir and ground water simulation, casting and molding, process and device simulation in semiconductor physics, structural mechanics and circuit simulation.

Data Mining for Simulations

Algorithms for data mining are used routinely in commercial applications. For engineering applications, we are currently developing modules which process simulation results to obtain meta-data which are suitable for data mining algorithms.

Concrete topics are:

  • Sensitivity analysis for simulation results (DIFF-CRASH) including topologically incompatible models
  • The influence of certain variants on CAE parameters and
  • Employing existing models to obtain constructions for new designs or questions

Compression Techniques for Simulations

The simulation of dynamical processes (e.g. crash simulation, weather forecast or casting) easily generates files with several gigabytes of data. Standard compression techniques reduce their size only insignificantly. The techniques used to compress graphic and video files may also be used for the results of simulations. Results indicate that a compression factor of 5 is obtainable without any degradation of the results. We have developed the toolbox FEMZIP, which supports compression of simulation results based both on structured and unstructured grids and in dependence of the accuracy required.