Are you looking for opportunities to see EXCELLERAT in action or meet our staff EXCELLERAT personally? Click through our event calendar to see at which trade fairs EXCELLERAT will be represented with a booth, which (research) conferences we participate in, and further events organised by and with EXCELLERAT.
From 08:00 until 17:00
At Germany - Sankt Augustin - Fraunhofer-Gesellschaft e.V. Institutszentrum Birlinghoven (IZB) Schloss Birlinghoven 1
Germany - Sankt Augustin - Fraunhofer-Gesellschaft e.V. Institutszentrum Birlinghoven (IZB) Schloss Birlinghoven 1
The workshop will be structured into two parts. The first part provides a general introduction to Machine Learning (ML) aiming at the analysis and post-processing of engineering data extracted from simulations. Supervised and unsupervised ML techniques are discussed with application in classification and clustering as well as model and predict the physical behavior of the underlying dynamical system. As engineering data is usually characterized by a very high formal dimension, the workshop will also address (non-)linear techniques of dimension reduction, such as manifold learning. These methods try to identify intrinsic structures in the data that can be described in a much lower dimension then the original data. In the context of CFD, ML has successfully been applied to support the simulation, e.g. by modeling turbulence, to cluster and monitor flow behavior or construct a cheap-to-evaluate surrogate, e.g. for a design optimization or uncertainty quantification.
The second part of the workshop aims to instruct, how to perform a modal decomposition of unsteady three-dimensional simulation data using dynamic mode decomposition (DMD) and proper orthogonal decomposition (POD). First, a short introduction into the theoretical background of the DMD and POD analysis is given and the parallel DMD/POD software provided by EXCELLERAT is introduced. In a hands-on workshop, the participants independently perform a DMD and POD analysis, which will be performed using a provided example data set, or, if available, on the participants own simulation data.