Scientific Computing Seminar

Date and Place: Thursdays and hybrid (live in 32-349/online via Zoom). For detailed dates see below!

Content

In the Scientific Computing Seminar we host talks of guests and members of the SciComp team as well as students of mathematics, computer science and engineering. Everybody interested in the topics is welcome.

List of Talks

Event Information:

  • Thu
    17
    Nov
    2022

    SC Seminar: Guillermo Suarez

    11:45Hybrid (Room 32-349 and via Zoom)

    Guillermo Suarez, Chair for Scientific Computing (SciComp), TU Kaiserslautern

    Title: Non-Linear Surrogate Model Design for Aerodynamic Dataset Generation

    Abstract:

    One of the key components for designing an Air Vehicle is an Aerodynamic Dataset – a model representing the aircraft’s aerodynamic behavior throughout the entire flight envelope. One of the major benefits of Artificial Intelligence (AI) is the possibility to automate most parts of a design process in a highly reduced timeframe. It is therefore believed that one or several AI techniques show potential for automating parts of the current dataset generation process.

    In this talk we will present the latest work regarding the design and development of a non-linear surrogate model to adequately predict static stability and control parameters for the Unmanned Combat Air Vehicle DLR-F17 and the DLR-F19 aircraft. Within this scope, two different topologically optimized machine learning based surrogate modelling techniques are investigated in order to achieve the best generalization properties.

    The first architecture being investigated is Artificial Neural Networks (ANN), who have gained great popularity due to their ability to efficiently process large amounts of data. The second architecture to be investigated relies in Ensemble Learning (EL) methods. For the current work three different ensemble learning models have been assessed: Random Forests, Adaptive Boosting and Extreme Gradient Boosting.

    Additional lines of research also investigate machine learning algorithms with multiple outputs in order to explore whether a multi-output learning framework brings benefits in the form of increased predictive performance compared to a framework based on training multiple disjoint models. Furthermore, because of their outstanding performance in feature extraction, 1-dimensional Convolutional Neural Networks (CNN) will be also considered as a surrogate model.

    How to join online

    You can join online via Zoom, using the following link:
    https://uni-kl-de.zoom.us/j/63123116305?pwd=Yko3WU9ZblpGR3lGUkVTV1kzMCtUUT09