Data-driven modeling of rotating detonation waves Article Swipe
YOU?
·
· 2021
· Open Access
·
· DOI: https://doi.org/10.1103/physrevfluids.6.050507
· OA: W3048716856
The direct monitoring of a rotating detonation engine (RDE) combustion\nchamber has enabled the observation of combustion front dynamics that are\ncomposed of a number of co- and/or counter-rotating coherent traveling shock\nwaves whose nonlinear mode-locking behavior exhibit bifurcations and\ninstabilities which are not well understood. Computational fluid dynamics\nsimulations are ubiquitous in characterizing the dynamics of RDE's reactive,\ncompressible flow. Such simulations are prohibitively expensive when\nconsidering multiple engine geometries, different operating conditions, and the\nlong-time dynamics of the mode-locking interactions. Reduced-order models\n(ROMs) provide a critically enabling simulation framework because they exploit\nlow-rank structure in the data to minimize computational cost and allow for\nrapid parameterized studies and long-time simulations. However, ROMs are\ninherently limited by translational invariances manifest by the combustion\nwaves present in RDEs. In this work, we leverage machine learning algorithms to\ndiscover moving coordinate frames into which the data is shifted, thus\novercoming limitations imposed by the underlying translational invariance of\nthe RDE and allowing for the application of traditional dimensionality\nreduction techniques. We explore a diverse suite of data-driven ROM strategies\nfor characterizing the complex shock wave dynamics and interactions in the RDE.\nSpecifically, we employ the dynamic mode decomposition and a deep Koopman\nembedding to give new modeling insights and understanding of combustion wave\ninteractions in RDEs.\n
