Controlling Mechanisms And Uncertainty Quantification In Low-Temperature Combustion Of Large Hydrocarbons

Air pollution is one of the major issues in China.Low-temperature combustion（LTC）is a promising solution to it through reducing the emissions of nitrogen oxides and particulate matter.Unlike conventional combustion modes,LTC is dominated by chemical kinetics.Consequently,understanding LTC chemistry,particularly its reaction mechanisms and the associated uncertainties,is essential for developing LTC technology.This forms the central theme of the present dissertation.This thesis first presents results on an investigation of the turnover states in the transition from low-temperature to high-temperature chemistry,i.e.,the Negative Temperature Coefficient（NTC）regime.A dimensionless numberαis derived from a detailed mechanism to quantify the competition between the low-temperature chain branching reactions and the high-temperatureβ-scission reactions.Both numerical and experimental results show that the NTC upper turnover states correspond to a constant value ofα.The reduced models based onαare consistent with simulations using the detailed mechanism and provide the basis for further analysis of the controlling mechanisms of the NTC regime.Subsequently,an efficient approach for computing the kinetic sensitivity of the ignition delay time is developed to facilitate the sensitivity analysis of the high-dimensional kinetic mechanism.The approach is substantially faster than the brute force approach by three to four orders of magnitude and is validated with the mechanisms of various large hydrocarbons.Furthermore,the approach works for the autoignition in both constant volume simulation and Homogeneous Charge Compression Ignition（HCCI）simulation.Finally,a subspace-based framework is proposed for propagating kinetic uncertainty in combustion simulations.Inspired by the reduced models derived earlier,a gradient-based subspace method is employed to identify the linear combinations of important kinetic parameters,which contributes most of the variations of the prediction.Then a response surface can be built within the low-dimensional subspace with much fewer samples.One to five-dimensional subspace is identified in the ignition delay time and laminar flame speed for various kinds of fuels.Furthermore,a one-dimensional subspace is identified in the map from kinetic parameters to the liftoff height of the Cabra H₂/N₂flame,and the subspace reveals that the flame stabilization is dominated by autoignition process.In addition,a heuristic approach is proposed for uncertainty propagation in computationally expensive combustion simulations.The approach estimates a low-dimensional subspace for the target expensive simulation by fitting a subspace to multiple relevant inexpensive zero-and one-dimensional simulations,and a shared subspace method for approximating multiple outputs is developed for the approach.