Reduced model reanalysis for path-dependent multifield problems

In many applications in Computer Aided Engineering (CAE), like parametric studies, structural optimization or virtual material design, a large number of almost similar models have to be simulated. Although individual simulation results may differ only marginally, the same effort is invested for every new simulation with no account for experience and knowledge from previous simulations. Therefore, we have combined data-based Model Order Reduction (MOR) and computational Reanalysis to a Reduced Model Reanalysis (RMR) approach to accelerate simulations by systematic reuse of simulation data. While MOR allows to reduce model fidelity in space and time without significantly deteriorating accuracy, reanalysis uses results from previous computations as a predictor. The expected overarching scientific innovation is a better establishing of the yet scarcely pursued idea to reuse simulation data to accelerate computations.

The aim of the currently ongoing project is to advance and further develop the Reduced Model Reanalysis (RMR) method developed in the first project phase to transient, path-dependent multi-field problems. Standard and isogeometric finite elements are used for discretization in space. Feasible model reduction methods, surrogate models and regression algorithms have to be identified for the involved history variables and physical fields. Beyond the required algorithmic and methodical developments, one central research question is, how the quality of path-dependent quantities predicted by the surrogate model can be guaranteed and verified. Identification of indicators for feasible time step sizes that are computable from the current solution are problem dependent and have to be developed on a case by case basis, which requires corresponding domain knowledge. It will therefore be investigated whether and how meta-data from previous simulation runs can be used to serve this purpose.