Development of an efficient solver for detailed kinetics in reactive flows

Doctoral thesis, 2021

The use of chemical kinetic mechanisms in CAE tools for reactive flow simulations is of high importance for studying and predicting pollutant formation. However, usage of complex reaction schemes is accompanied by high computational cost in both 1D and 3-D CFD frameworks. The combustion research community has addressed such challenge via two main approaches: 1) tailor made mechanism reduction strategies; 2) pre-tabulation of the chemistry process and look-up during run-time. The present work covers both topics, although much of the methodology development and validation efforts focused on tabulation.

In the first phase of the PhD work, an isomer lumping strategy based on thermodynamic data was developed and applied to a detailed three component reaction mechanism for n-decane, alpha-methylnaphthalene and methyl decanoate comprising 807 species and 7807 reactions. A total of 74 isomer groups were identified within the oxidation of n-decane and methyl decanoate via the assessment of the Gibbs free energy of the isomers. The lumping procedure led to a mechanism of 463 species and 7600 reactions, which was compared against the detailed version over several reactor conditions and over a broad range of temperature, pressure and equivalence ratio. In all cases, excellent agreement between the predictions obtained using the lumped and the detailed mechanism has been observed with an overall absolute error below 12%.

In the second phase of the PhD work, a tabulated chemistry approach was developed, implemented and validated against an on-the-fly chemistry solver across different simulation frameworks. As a first attempt, a flamelet-based tabulation method for soot source terms was coupled to the stochastic reactor model and tested against a well stirred reactor-based approach under Diesel engine conditions. The main purpose was to assess and quantify benefits of tabulation within the 0-D SRM framework with respect to soot formation only. Subsequently, a latent enthalpy (h298) based approach was developed and implemented within the SRM model to predict both combustion and emission formation. This approach was widely validated against the detailed on-the-fly solver solutions under 0-D reactor conditions as well as Diesel engine conditions for a wide range of operating points. Good agreement was found between the two solvers and a remarkable speed-up was obtained in terms of computational costs of the simulation. As a last step, the same tabulated chemistry solver was coupled to a commercial CFD software via user defined functions and performances were assessed against the built-in on-the fly chemistry solver under Diesel engine sector simulations. The tabulated chemistry solver proved to be within an acceptable level of accuracy for engineering studies and showed a consistent speed-up in comparison to the online chemistry solver.

Across all the investigated frameworks, the developed tabulated chemistry solver was found to be a valid solution to speed-up simulation time without compromising accuracy of the solution for combustion and emissions predictions for engine applications. In fact, the much-reduced CPU times allowed the SRM to be included in broader engine development campaigns where multi-objective optimization methods where efficiently used to explore new engine designs.