Unsteady Flamelet Progress Variable Modeling in a Discontinuous Galerkin Framework: Implementation, Optimization and Demonstration

We present the implementation of the Unsteady Flamelet Progress Variable (UFPV) model in the JENRE® Multiphysics Framework, a high-order discontinuous Galerkin (DG) solver for compressible reacting flows on fully unstructured meshes. The focus is on making detailed- chemistry flamelet manifolds practical for large-scale CPU and GPU simulations of high-speed propulsion configurations. To address the cost of manifold queries, we develop two GPU-aware strategies: (i) analytic bounding of non-uniform tabulated axes, which replaces global bisection searches with short, bounded refinements using exponential or quadratic inverse mappings; and (ii) a nearest-point indexing strategy that collapses multi-dimensional interpolation to a small number of table accesses per query, while retaining minimal stencils where finite-difference Jacobians are required. For a homogeneous reactor using Burke’s hydrogen mechanism, the optimized UFPV formulation removes stiffness from the time integration and achieves speedups of about two on CPUs and three on GPUs compared to finite-rate chemistry (FRC). Canonical CFD tests, including a one-dimensional periodic flame and a two-dimensional jet-in-supersonic- crossflow, show that UFPV reproduces the large-scale flame structure, heat release, and mixing behavior obtained with FRC, while introducing only modest local differences in radical species. These results demonstrate that UFPV can be an accurate, efficient, and scalable chemistry model for high-order DG simulations of high-speed reacting flows.