Acceleration of high-fidelity reactive flow simulations based on computational singular perturbation with adaptive hash mapping

Simulating chemically reactive flows remains a major computational challenge due to the stiffness and high dimensionality of the governing equations. In this work, we integrate the computational singular perturbation (CSP) framework into the KAUST adaptive reactive flow solver (KARFS) to accelerate direct numerical simulations (DNS) involving detailed chemical kinetics. The method applies operator splitting at each spatial location, where CSP dynamically eliminates stiffness by solving a reduced, non-stiff thermo-chemical system. This enables the use of explicit time-stepping with significantly larger step sizes, while preserving solution accuracy. To avoid the overhead typically associated with Jacobian decomposition, we introduce an efficient online hash-based mapping that reuses eigensystem information across space and time. The proposed strategy is demonstrated in DNS of laminar and turbulent ammonia-air premixed flames, showing strong potential for reducing the computational cost of high-fidelity reactive flow simulations.