Adjoint-based sensitivity analysis
Adjoint-based sensitivity analysis is an efficient approach that helps the understanding of models with several thousand parameters. It is applied to investigate complex processes such as pollutant- or ion-formation in flames. We use this approach for all parameters in combustion models, i.e., kinetics, thermodynamics, and transport data. In collaboration with the group for Software and Tools for Computational Engineering, the analysis is implemented in the open-source project FlameMaster using the dco++ library that enables algorithmic differentiation. Models for the nitrogen oxide and soot formation, as well as ion formation, are analyzed.
Example: Optimization potentials of model parameters of two different kinetic mechanisms for laminar burning velocities of ammonia/air flames with unburned conditions of 1 bar, 298 K, and a fuel-air equivalence ratio of 1.1 [1]
[1] R. Langer, J. Lotz, L. Cai, F. vom Lehn, K. Leppkes, U. Naumann, H. Pitsch, Adjoint sensitivity analysis of kinetic, thermochemical, and transport data of nitrogen and ammonia chemistry, Proceedings of The Combustion Institute (2021).