Data Assimilation for Detonation Simulations

Detonation waves in propulsion systems exhibit complex multidimensional dynamics, including unstable cellular structures, triple-point collisions, and strong shock-reaction coupling. Predictive modeling remains challenging due to high computational costs of resolving these features in simulations and experimental limitations where diagnostics are typically restricted to line-of-sight optical methods (Schlieren, chemiluminescence) that indirectly sample the flow field.

We address this by integrating sparse experimental observations with numerical model prediction using a data assimilation (DA) framework. This framework assimilates flow images to generate a full-field state reconstruction, improving the accuracy of indirectly observed quantities of interest as well as flow features. In addition, our method constrains detonation dynamics, enabling the study of rare events.