An error-controlled adaptive chemistry method for reacting flow simulations

Abstract

Many technologically important processes in the chemical and mechanical industries involve coupled interactions of heat and mass transfer with chemical reactions - e.g. commercial burners, gas turbines, internal combustion engines, etc. However, detailed computational studies of such processes remain difficult at best, particularly due to the large reaction mechanisms that describe the chemical kinetics over the relevant range of reaction conditions. As a result, reduced models that contain fewer reactions and/or species while still capturing the "important" kinetics are often used in place of the full comprehensive reaction model in modeling complex reacting flows. "Adaptive Chemistry" - a method that uses several smaller locally-accurate reduced reaction models rather than a single "catch-all" model - has been shown in the combustion literature to be a viable option for improving computational efficiency in such studies. However, several outstanding challenges have prevented the adoption of this method in mainstream studies, most notably the difficulty of determining the accuracy of a solution obtained using Adaptive Chemistry. The focus of this research was to develop methods to enable efficient and accurate implementation of Adaptive Chemistry for reacting flow simulations.

(cont.) A method was developed for determining how much error may be tolerated in each reduced model in order to achieve a desired accuracy in Adaptive Chemistry solutions at steady-state. A novel model reduction method was also developed to obtain automatically reduced models (based on reaction elimination) that are guaranteed to satisfy the imposed error tolerances at all conditions in a user-specified range. In order to enable point-validated reduced models to be used accurately over ranges, an iterative method was developed for identifying ranges of reaction conditions over which such reduced models are guaranteed to remain valid. An Adaptive Chemistry method that demonstrates the application of these methods is presented. Efficient implementations of construction, storage and retrieval of reduced models that are appropriate for the reaction conditions encountered during Adaptive Chemistry simulations are presented, including an algorithm that adapts the library of reduced models to the solution trajectory "on the fly".

(cont.) The error-controlled Adaptive Chemistry method developed here is the first method that enables rigorous control of the model reduction error in steady-state Adaptive Chemistry solutions, as demonstrated in 1-D and 2-D premixed and partially premixed flame simulations. Results of a collaborative effort to facilitate engine research by developing the necessary cyberinfrastructure to provide remote access to the model reduction tools developed here are also discussed. Finally, methods are described for extending the error control criteria developed and demonstrated for reduced reaction models to reduced-species models and suggestions are made for future research in Adaptive Chemistry.