Prof. Dr. Nicolas Gauger, Chair for Scientific Computing (SciComp), TU Kaiserslautern

SciComp Seminar Series

Please contact Prof. Gauger, if you want to register for an online talk in our SciComp Seminar Series or just to register for the seminar.

A list of the already scheduled talks can be found –> here:

Alexander Schlicher, RPTU Kaiserslautern-Landau

Title: Automated Reconstruction of Meshed Topology Optimization Results using Parametric Modeling

Abstract:

Motivated by the need for integration of optimization results into Computer Aided Design work-
flows, in this talk the automated reconstruction of Topology Optimization results by means of
a Parametric Modeling software is investigated.
In order to extract information about the geometry that is to be reconstructed, an input mesh
representing the optimization result is segmented to identify non-organic parts and a mesh
skeleton is calculated. A novel surface and curve skeletonization approach with the advantage
of noise independence and linear time complexity in the input mesh facet count is presented.
The gathered mesh information is then used to approximate different Topology Optimization
results by feature executions in a Parametric Modeling software, imitating the way a human
would reconstruct the given geometry.
The described algorithms are then applied to a number of input geometries and the results are
discussed. Finally, future research potential is outlined.

How to join online

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

Dr. Max Sagebaum, Chair for Scientific Computing (SciComp), University of Kaiserslautern-Landau (RPTU)

Title: New features in CoDiPack – Linear system handling and basic algorithm support

Abstract:

Linear system handling: CoDiPack now provides a helper for handling linear system solvers on the tape. The user has to implement a few functions for CoDiPack and then the linear system solve is automatically added to the tape. We will present the implementation details and how the helper can be used in software.

Algorithm support: Algorithmic Differentiation can be used to implement design optimization with the adjoint calculus. One algorithm the sensitivity computation is the reverse accumulation algorithm. A new module in CoDiPack provides implementations of the reverse accumulation algorithm and many others. A user has to implement an application interface that is used by the algorithms. We will present the implementation details, the interfaces, and the current set of algorithms and their properties.

How to join online

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

Mario Alfredo Carrillo Arevalo, RPTU Kaiserslautern-Landau

Title: Investigation of machine learning methods to monitor 3D stem cell culture in a bioreactor

Abstract:

The development and discovery of new drugs require a large number of tests to fulfill strict quality protocols. These tests are mainly performed on animals which involve an ethical debate, are expensive, and are time-consuming. The most relevant alternative to replace, reduce and refine this practice, is the use of cell-based model systems. Especially the cultivation of human induced pluripotent stem cells (hiPSC) in a 3D environment. This process is carried out in bioreactors with controlled conditions. However, the cultivation process can be prone to failures or unexpected behaviors. Therefore, it is necessary to establish a real-time monitoring mechanism that allows taking early decisions and enables large-scale production in the long term.

This talk presents an interdisciplinary work that investigates machine learning methods and other computational science tools to be applied to monitor the state of these processes. It also illustrates the results of the applied algorithms as well as their behavior when run on an embedded platform with limited resources.

How to join online

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

Dr. Andreas Schütte, Institute of Aerodynamics and Flow Technology, German Aerospace Center (DLR), Braunschweig

Title: HPC applications at DLR for the design, analysis and performance evaluation of highly agile aircraft configurations

Abstract:

The content of the lecture focuses on the use of HPC within the DLR project Diabolo on the development of methods and processes for the design and performance assessment of highly agile aircraft configurations. In Diabolo technology developments through a large spectrum of disciplines for aircraft design are considered. The main focus in Diabolo is the design of a virtual demonstrator to capture technological gaps and to demonstrate the design and assessment capabilities of the methods and processes developed at DLR.
The technology spectrum ranges from aerodynamics, structure and aeroelasticity, control system design, the layout of a control concept and flight mechanics analysis, technologies and the design of a propulsion concept and the determination of radar, infrared and acoustic signature properties.
The tasks of the Institute of Aerodynamics and Flow Technology in Braunschweig and Göttingen in particular are the development of the overall design environment as well as the validation of the DLR RANS method TAU for vortex-dominated flows in subsonic and supersonic speeds, experimental and numerical aerodynamic studies for the design and understanding of the fluid physics and aerodynamic performance of highly agile aircraft configurations, the numerical aerodynamic optimization, the design and evaluation of highly integrated engine intakes and thrusters, the investigation of dynamic control elements as well as further development of high-precision measurement techniques.

How to join online

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

Thomas Kühn, RPTU Kaiserslautern-Landau

Title: An introduction to the Recursive Projection Method for stabilization of fixed-point iterations and its application in adjoint sensitivity analysis

Abstract:

This seminar aims to give an overview of the recursive projection method, which may be used to accelerate and stabilize fixed-point iterations. The recursive projection method is motivated by examples of problems that may be solved using fixed-point iterations. Then the RPM is explained, which splits the fixed-point iteration into an unstable and stable part performed on two separate subspaces. Additionally, we will take a look at an approach to estimate projectors for both subspaces and how to increase or decrease the size of the basis during fix point iteration according to the number of problematic eigenvalues. Finally, an application of RPM to fix point iterations that arise from adjugate methods in sensitivity analysis is presented.

How to join online

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