Dynamic adaptive chemistry via species time-scale and Jacobian-aided rate analysis

Abstract

A new on-the-fly mechanism reduction method is presented for dynamic adaptive chemistry (DAC), in which chemical species are categorized into active, directly coupled, and inconsequential species through time-scale and Jacobian-aided rate analysis (TSRA) with given instantaneous local composition. Chemistry integration is simplified by solving only parts of species with the inconsequential species being treated frozen. The method applies intuitive error control on species rates-of-change for the active species and requires no problem-dependent parameters such as starting species. In addition, it can guarantee element conservation and directly take transport flux into account for chemistry reduction. The new method has been validated in auto-ignition, unsteady perfectly stirred reactor (PSR) and 1-D flame propagation of methane/air mixture with the 53-species GRI-Mech 3.0 and n-heptane/air mixture with the 561-species LLNL mechanism. It shows that the proposed Jacobian analysis can successfully identify important low-concentration slowly-varying intermediate species as active species during the ignition process. Compared to directed relation graph (DRG) based DAC, TSRA yields more accurate predictions of ignition delay time and composition together with slight improvement in efficiency for methane/air auto-ignition. For the n-heptane/air mixture, TSRA accurately reproduces the challenging two-stage ignition and negative temperature coefficients over a wide range of pressures and initial temperatures. The importance of transport flux on dynamic reduction is investigated in unsteady PSR, showing that TSRA dynamically adjusts the local mechanism size with the relative significance of chemical reaction and it offers greater reduction in regions where transport processes are significant. The 1-D flame propagation validates the generality of TSRA method for inhomogeneous reacting flows. Results also show that TSRA has effective and more universal error control. For TSRA, a single threshold ε=1×10−4 applies to all the tests considered. In contrast, the appropriate reduction threshold for DRG is problem-dependent.