Advanced flamelet tabulation strategies for Large Eddy Simulations of single- and multi-phase turbulent jet flames

The constantly increasing demands placed on modern combustion processes in terms of efficiency and pollutant reduction make it necessary to optimize existing combustion processes, develop new ones and use alternative fuels. Numerical simulations help to develop and understand these processes. One efficient combustion modeling methodology is the pre-tabulation of flame structures. For the simulation of non-premixed flames, the so-called Flamelet-Progress-Variable (FPV) approach has been established. This involves the flame structure, which is not resolved in 3D CFD, being modeled using one-dimensional diffusion flames with different strain rates. The progress variable is used as a reactive scalar to describe the reaction progress of these structures. This approach, in combination with the Large Eddy Simulation (LES) of the turbulent flow typically occurring in technical systems, provides very good results for many fields of application. However, the pre-assumption of flame structures implies modeling assumptions that have to be constantly verified for new applications. For example, it is known how differential diffusion along the flame structure can be modeled. However, the damping of differential diffusion due to turbulent structures must be directly specified when modeling the diffusion within the flame structures. Thus, flames which are strongly influenced by differential diffusion and flames where that influence is not relevant can be modeled well. Experimental data from a non-premixed oxy-fuel jet flame with hydrogen admixture have shown, however, that the effect of differential diffusion can vary locally and for individual species. The first part of the present dissertation examines how suitable modern modeling approaches are for representing this effect. Existing diffusion modeling approaches describing the flame structure with different levels of complexity are systematically compared. The complexity of the resulting flame structure makes it necessary to develop a suitable table parameterization, which is presented in this dissertation. The different approaches are compared in a prior analysis of the flame structure and in coupled LES with experimental Raman/Rayleigh data. In this process, the potential of the individual modeling approaches is elaborated. The identified limitations indicate the need for further research in this area. Another technical field in which the tabulation of flame structures has become established itself as a valid approach is high-pressure spray combustion, which is relevant for diesel engines. The challenge in this area is to map the multitude of processes occurring during the usually two-stage ignition process up to the formation of pollutants. The tabulation of igniting transient diffusion flames (Unsteady Flamelet Progress Variable Approach, UFPV) provides very good results in a large number of studies. The modeling approaches found in the literature employ a wide range of strain rates. However, the model quality is often very similar with respect to global combustion characteristics. The present dissertation contributes in this respect within the framework of a systematic discussion of the influence of different strain rates in spray simulation. The single-hole injector Spray A defined in the Engine Combustion Network is used as a reference case. This dissertation shows that the strain rate significantly delays the ignition upstream of the flame lift-off length. Moreover, local extinction during ignition is identified for this case. However, its probability is comparatively low. In addition to ignition, the formation of pollutants in these spray flames is a problem of high technical relevance. The challenge here is to formulate the progress variables for the strongly differing time scales of ignition and the slow development of pollutant species such as nitrogen oxides and soot precursors. The present dissertation complements the approaches known in the literature by adding a new progress variable definition. This definition consists of two progress variables, which describe individual processes of ignition and pollutant formation. Their normalized value is added to obtain the progress variable used for the parameterization of the look-up table. The approach is therefore referred to as the Unsteady Flamelet Composed Progress Variable (UFCPV) approach in this dissertation. This approach is verified in a one-dimensional test case with respect to the representation of ignition, combustion and pollutant formation under conditions relevant to the diesel engine. Comparison with the original definition shows the improvement of the approach. The validation of the approach in the 3D simulation is again performed using ECN Spray A. For validation purpose, 355-nm PLIF data and recently published 355-nm high-speed PLIF data are used. These allow a detailed comparison of the spatially and temporally resolved structure of formaldehyde and soot precursors. Here, the simulation shows very good agreement with the experimental data. The validated model is also used to analyze experimentally obtained statements concerning the separation of the above-mentioned species and its dynamic detachment behavior for the first time in a numerical study and to relate them to periodic fluctuations of the mixture fraction field. Its application to the ECN Spray D also serves to investigate the suitability of the model when the nozzle hole diameter is varied. Here, an underestimation of the ignition delay time is found. However, it can be shown from comparison with so far unpublished 355-nm high-speed PLIF data that the resulting flame structure is well reproduced by the experimentally determined morphology when this time offset is included. This suggests that although chemical reactions are initiated too early, the resulting flame structure is valid. The validated model allows the development of a comprehensive overview of the structure of the spray flame, including all relevant variables along the cause-effect chain from injection to gaseous pollutant formation. This is presented for the ECN Spray A and the ECN Spray D