Gregor Döhner, M.Sc.

Ongoing Research:

The development of increasingly efficient and environmentally friendly combustion systems can lead to highly undesirable combustion instabilities. To predict the occurrence of such instabilities, a flame model is usually required. So far, neural networks (NNs) are the only system that can adequately represent the non-linear flame dynamics over a wide frequency range. 
However, large NNs require the calibration of thousands of parameters, making any attempt to physically interpret the resulting model impossible. An alternative modeling approach is to describe the flame by a physical equivalent system. Previous work (see publications below) has shown that a simple mass-spring damper with nonlinear damping is able to represent the most important flame dynamics quantitatively in the linear and qualitatively in the nonlinear domain for laminar and turbulent flames. The resulting system of ODEs is compact and can be physically interpreted.
Current research focuses on the further development of the ODE system, relying on so-called universal differential equations and regression techniques such as PySINDy.