Solving partial differential equations for aerodynamic problems is a costly- and computationally-intensive challenge that is gradually pushing modern HPC systems to their limits when employing classic algorithms.
As part of the DLR project Machine Learning and Quantum Computing - Digitalization of Aircraft Development 2.0 led by the Institute for Aerodynamics and Flow Technology, the scientists are now investigating whether and how these highly precise aerodynamic simulation solutions can be carried out with the help of methods from the field of machine learning.
In addition to the algorithmic challenges, it is also the aim of the project group to go one step further and to work out how these artificial neural networks, which have been specifically adapted for aerodynamic applications, could be implemented on quantum computers as soon as they become available.
The use of perfect or error corrected quantum computers, which are not yet available today, promises a new level of computing power in this context. With the help of hybrid quantum-classical algorithms, the project team wants to find out how the computations can be evaluated orders of magnitude faster on quantum computers in order to be able to use them efficiently for engineering problems in the future. To do this, they are developing customized quantum algorithms.
At the Institute for Aerodynamics and Flow Technology, the focus of work is on the methods of machine learning for solving partial differential equations from the field of flow physics.