• Citer

Optimization and uncertainty quantification of kinetic mechanisms for renewable fuels and sustainable combustion technologies

par Bertolino, Andrea
Président du jury Coussement, Axel
Promoteur Parente, Alessandro ;Frassoldati, Alessio
Publication Non publié, 2022-02-22

Thèse de doctorat

• Accès en ligne
• Détails
• Contenu
• Statistiques

Résumé :

In the present thesis, a data-driven approach for the optimization of detailed kinetic mechanisms is proposed. This methodology is based on heuristic optimization algorithms. The curve matching (CM) index is proposed as alternative error function to classical norms. In CM calculation the similarities between model’ responses and experimental data is measured quantitatively and qualitatively, considering also the first derivatives and shapes of corresponding splines. A novel protocol for the optimization of PLOG reactions is established. The interdependencies between Arrhenius expressions at different pressures were accounted for by handling three random variables for each PLOG, regardless of the number of discrete pressures specified in the mechanism. The Cumulative Sensitivity Function (CSF) and a Cumulative Impact Function (CIF) were introduced to make reaction selection automatic, fast, and efficient. The development above-mentioned methodology represents the underlying functioning of the OptiSMOKE++, a new C++ toolbox for the optimization of detailed kinetic mechanisms. OptiSMOKE++ is an flexible interface for the communications between other open source softwares like OpenSMOKE++, DAKOTA, Curve Matching and SciExpeM. This framework enables the simultaneous use of experimental targets from different facilities, i.e. Batch Reactors, Plug Flow Reactors (PFR), Perfectly Stirred Reactor (PSR), Shock Tubes (ST), Rapid Compression Machines (RCM) and 1D flames. Using this methodology, an optimized mechanism for ammonia combustion is obtained over a wide range of operating conditions. The approach involved all 101 kinetic parameters simultaneously. The role of diluents like H2O and CO2 in operating conditions relevant to applications for MILD combustion of hydrogen and syngas was discussed and analysed by means of a virtual species analysis (VSA), Global Sensitivity Analysis (GSA), and Optimization. The VSA suggests that measurements of ignition delay time (IDT) and species concentrations in perfectly stirred reactors (PSR) in diluted conditions are ideal candidates for the estimation of collision efficiencies as their physics is significantly third body driven. Eventually, the errors introduced by the replacement of the TROE formulation for fall-off reactions with PLOG were quantified for hydrogen combustion, and a method to extract information from data about third body efficiency of strong colliders in PLOG formulation was proposed in case high-level ab-initio calculations are not available. Finally, an experimental campaign was performed to investigate options for optimal operating conditions for the Université Libre de Bruxelles (ULB) flameless furnace fired with ammonia/hydrogen blends. In particular, the campaign aimed at identifying trade-off between NOx emissions and ammonia slip. In Reynolds-averaged Navier-Stokes simulations, substantial differences between mechanisms predictions and experimental data were observed in terms of NOx emissions. A combination of Uncertainty Quantification (UQ) and Chemical Reactor Network (CRN) was adopted to propagate the uncertainty of NOx kinetics to the CFD simulations