Abstract

Within this thesis, a detailed multicomponent gasoline surrogate reaction scheme was developed and reduced to a four component scheme of skeletal size. The main target is to cover the most important features for typical spark ignited (SI) combustion - flame propagation, emission formation and the tendency to auto ignite and subsequently cause engine knock. To achieve this a variable mechanism concept was developed to include sub models for different fuels as needed. Using this approach a detailed mechanism describing the oxidation of n-heptane, iso-octane, toluene and ethanol was compiled and compared against various experiments published in literature. Furthermore, correlations were developed to suggest four component gasoline surrogates based on typical fuel data sheets. The correlation method is validated against measurements in Cooperative Fuel Research (CFR) engine from various groups and further compared against correlations between octane numbers (ON) and predicted 0D ignition delay times. These correlations are used to identify and discuss the impact of the uncertainty of two reactions on ignition delay time of a multicomponent fuel. To be able to reduce the detailed scheme in a time efficient way existing reduction concepts where improved and applied to different schemes and targets. Since various reduction techniques are available, an optimal sequence of those was worked out. Using this sequence of reduction steps two multicomponent schemes were compiled: one scheme for the prediction of laminar flame speeds and one for the prediction of major emissions and auto-ignition. To underline that the suggested reduction procedure is universal it was also applied to n-heptane as single fuel surrogate for diesel fuel and to a large two component fuel from another work group.

Abstract

The present work is concerned with the simulation of combustion, emission formation and fuel effects in Diesel engines. The simulation process is built around a zero-dimensional (0D) direct injection stochastic reactor model (DI-SRM), which is based on a probability density function (PDF) approach. An emphasis is put on the modelling of mixing time to improve the representation of turbulence-chemistry interactions in the 0D DI-SRM. The mixing time model describes the intensity of mixing in the gas-phase for scalars such as enthalpy and species mass fraction. On a crank angle basis, it governs the composition of the gas mixture that is described by PDF distributions for the scalars. The derivation of the mixing time is based on an extended heat release analysis that has been fully automated using a genetic algorithm. The predictive nature of simulations is achieved through the parametrisation of the mixing time model with known engine operating parameters such as speed, load and fuel injection strategy. It is shown that crank angle dependency of the mixing time improves the modelling of local inhomogeneity in the gas-phase for species mass fraction and temperature. In combination with an exact treatment of the non-linearity of reaction kinetics, it enables an accurate prediction of the rate of heat release, in-cylinder pressure and exhaust emissions, such as nitrogen oxides, unburned hydrocarbons and soot, from differently composed fuels. The method developed is particularly tailored for computationally efficient applications that focus on the details of reaction kinetics and the locality of combustion and emission formation in Diesel engines.

Abstract

Three different methods for Reynolds-averaged navier-Stokes computational fluid dynamics modeling of non-premixed ignition and combustion using tabulated chemsitry have been developed. All methods make use of flamelet libraries, where the flamelet auto-ignition process is parameterized using a progress variable. the progress variable parametrization of the autoignition chemsitry allows for using arbitrarily large chemical mechanisms, at constant computational costs and for modeling of turbulence-chemsitry interactions.

In the first method a coordinate transform from time and space to a space described by mixture-fraction and a progress variable is made. the method was shown to be capabale of predicting the response of injection pressure and nozzle diameter on lift-off length. It was shown that it was possible to apply the method for use in computational fluid dynamics simulations of compression-ignited engine combustion.

In the second method, the transient flamelet libraries were directly used in an interactive flamelet setting. It was investigated if it was possible to generate tables by computing homogeneous adiabatic constant-pressure reactors instead of igniting flamelets. It was found that omitting the effect of scalar dissipation rate during the tabulation process leads to an error in prediction of ignition delay.

In the third method a simplified conditional moment closure approach was developed. By using tabulated chemsitry, and by making the conditional moment closure for the progress variable only, it was possible to use the same computational grid as used by the flow solver for the spatial transport of the conditionally averaged scalars. This method was tested for a simple autoigniting spray configuration and it was found that it was able of capturing the response of the ignition dlay and lift-off length due to changed ambient oxygen level. Software technical improvements from the transient flamelet library based approaches were carried over the stationary flamelet library based on soot source term model, and further model updates yielded a model capable of predicting soot emissions for a light-duty diesel engine.

Abstract

The virtual development of future Spark Ignition (SI) engine combustion processes in three-dimensional Computational Fluid Dynamics (3D-CFD) demands for the integration of detailed chemsitry, enable - additionally to the 3D-CFD modeling of flow and mixture formation - the prediction of fuel-dependent SI engine combustion in all its complexity. the conflict of goal arising in coupling 3D-CFD calculations with detailed chemistry is to keep computational costs low while achieving accurate results.

This work presents an approach which constitutes a coupled solution for flame propagation, autoignition and emission formation modeling incorporating detailed chemsitry, while exhibiting low computational costs.

For modeling the regular flame propagation, a laminar flamelet approach, the G-equation is used. This approach describes the flame propagation based on the turbulent flame speed, which is determined by the turbulence and the fuel-specific laminar flame speed. The latter one is incorporated using an adequate fitting function.

Auto-ignition phenomena are addressed using an integrated flamelet approach, which bases on the tabulation of fuel-dependant reaction kinetics. By introducing a progress variable for the autoignition - the Ignition progress Variable (IPV) - detailed chemistry is integrated in 3D-CFD. the tabulation approach only demands for the soltuion of the IPV transport equation, thus keeping the computational demand low, while allowing the consideration of local effects on auotigntion chemsitry on cell level.

The modeling of emissions formation bases on an interactively coupled flamelet approach, the Transient interactive Flamelet model. By transforming the species balance equations into a one-dimensional form, the numerical effort incorporated with the solution of small chemical time-scales is separated from the 3D-CFD flow field solution. Thus, the emission formation is calculated under representative boundary conditions. the description of the soot formation bases on a detailed soot model, and the properties of the soot Particle Size Dsitribution Function are calculated using the method of moments.

The coupling between the G-equation, integrated flamelet, and interactive flamelet models is done based on the IPV. The functionality of the combined approach to model the variety of SI enigne combustion phenomena is prooved first in terms of fundamentals and standalone sub-model functionality studies. For standalone and model coupling functionality studies, a simplified test case is introduced, representing an adiabatic pressure vessel without moving meshes. the vessel is initialised homogeneously, allowing the selective investigation of different parameters on combutsion process and direct comparison with direct numerical solution of the detailed chemistry in 0D homogeneous reactor calculations. Following the basic functionality studies, the standalone and combined sub-model functionalities are investigated in adequate engine test cases.

Abstract

An N-Heptane mechanism and a Methane/Propane mechanism have been reduced by an Automatic Reduction Tool (ART) and simulated with two different solver combinations, which solve the set of ordinary differential equations governing the time evolution of the species simultaneously with solving algebraic equations for species that can be considered to be in quasi steady state. The most successful of the two solver combinations is an optimized combination of Newton solvers. The algebraic part of the solver is based on a Newton solver and is given a speed-up by using the fact that the sparseness pattern of the Jacobian is constant in time. This allows for automatically written source code and an optimization of the sparseness pattern in a preprocessing step. The optimization method is based on a simulated annealing procedure that minimizes the number of operations in the algebraic part of the solver. The speed-up of the Newton solver for the algebraic equations is one of the major developments presented in this thesis. The other one is the development of the ART and the reduction of the N-Heptane and the Methane/Propane mechanisms using the ART. A reduction down to 37 out of 110 species and 23 out of 118 species is achieved for the N-Heptane and Methane/Propane mechanism respectively, while the accuracy of the solution is maintained and the CPU time is significantly lower than that of the detailed mechanism. Less, but still greatly reduced mechanisms are generated for larger ranges of physical conditions. Also, the two solver combinations were implemented into a commercial Computational Fluid Dynamics (CFD) code. CFD simulations were then performed for a detailed and reduced mechanism. The implementation involving the optimized combination of Newton solvers resulted in a speed-up for the reduced mechanism compared to the detailed mechanism, while the accuracy of important species for the reduced mechanism was well within acceptable limits.

Abstract

The aim of the thesis work is the further development of practical engine simulation tools based on Stochastic Reactor Models, SRMs. Novel and efficient implementations were made of a variety of SRMs adapted to different engine types. The models in question are the HCCI-SRM, the TwoZone SI-SRM and the DI-SRM. The specific models developed were incorporated into two different interfaces: DARS-ESSA, which is a stand-alone tool, and DARS-ESM through which all the models can be operated in a simple and effective manner with use of several commercial 1-D engine simulation tools. The tools and couplings to commercial 1-D codes were successfully developed and employed to simulate such complex combustion processes as of HCCI engines with NVO combustion.

SRMs are able to model cyclic variations, but these may be overpredicted if discretization is too coarse. It was found that for studies of cyclic variations in HCCI engines, by using the HCCI-SRM, discretization needs to have a level of resolution of 500 particles and of 0.5 CAD time steps, to provide the correct range of the cyclic variations. To get correct predictions of average values, of for example the pressure, temperature and species mass fractions, as few as 10 cycles are usually required, even when employing as coarse discretization of 100 particles and time steps of 0.5 CAD.

Investigations to study the effects of turbulence and heat transfer in HCCI combustion were performed. In the case of high levels of turbulence and evenly distributed heat transfer, the in-cylinder conditions become homogeneous more quickly. The results indicate that in HCCI engines, inhomogeneties tend to promote earlier ignition and lower pressure rates as well as more stable operating conditions with lesser cyclic variations. Turbulence and the heat transfer distribution had little impact on the duration of combustion or on the amount of HC and NO at EVO.

The calculated concentrations of hydroxyl radicals and formaldehyde were compared with LIF-measurements made in an optically accessed iso-octane / n-heptane fuelled HCCI engine. The averaged and distributed concentrations of CH2O and OH could be predicted with quite high accuracy by the SRM. This clearly proves the validity of the stochastic reactor model. The formation of exothermic centers was modeled with the SRM to investigate their impact on HCCI combustion. By varying the exhaust valve temperature, and thus assigning more realistic wall temperatures, the formation of exothermic centers and the ignition timing was shifted in time. It was shown that promoting exothermic centers provide more inhomogeneous conditions before ignition, and lead to earlier ignition. This in turn leads to more homogeneous conditions after combustion, counteracting emissions of hydrocarbons and CO which are a problem in HCCI engines.

Studies involving the use of a novel approach with adaptive chemistry, POSM, were performed. Incorporated into the Two-Zone SI-SRM code, calculations showed almost no accuracy to be lost, while there was a decrease in calculation time by a factor of 3. For a further gain in calculation speed of a factor of 12, clear losses in accuracy were experienced, although the global conditions were well captured.

Simulations of diesel engine combustion, DICI, using the newly developed DI-SRM coupled with a 1-D full engine simulation tool were found to agree well with the results of experiments that were conducted. Parametric studies were performed to indicate the sensitivity of the modeling parameters. The DI-SRM behaved as predicted, and even with use of coarse discretization the results were comparable to those of the experiments.

Abstract

The aim of this work is to generate detailed and simplified kinetic models for the oxidation of the Primary Reference Fuels (PRF) n-heptane, iso-octane and their mixtures, with low numbers of species and reactions. These mechanisms are consistent in terms of the choice of kinetic parameters for the different reaction classes. The further aim is to validate the kinetic models for a wide range of different combustor operating conditions, such as engines, shock tube, jet-stirred reactors, flow reactors, laminar flames and burners, to cover the full range of temperatures, pressures and air-fuel equivalence ratios. In the validation procedure the focus is on both extensive tests for the low temperature ignition and a validation against flame experiments. This development is motivated by the fact that the broadly validated hydrocarbon fuel oxidation mechanisms that have been developed previously are large and thus consumed more CPU-time than is acceptable by commercially available CFD-tools, because of their complexity, which prohibits the direct incorporation in simulations of complex reactor models, e.g. with turbulent flow.

For the simplification and the reduction of mechanisms a stepwise efficient and simple lumping strategy for different reaction types has been developed and subsequently combined with a necessity analysis .The lumping of species with the same functional groups helps to reduce the different complex pathways into one lumped reaction pathway, which reduces the mechanism in terms of both numbers of species and reactions, causing only a negligible loss of information of the detailed mechanism. This is controlled by comparing predicted concentration profiles of lumped species with the added profiles of the isomeric species. Then further reduction of the mechanism is achieved by using necessity analysis (reaction flow combined with a sensitivity analyses) which eliminated less necessary species and their corresponding reactions.

The developed mechanisms are validated against ignition delay times measured in shock tube experiments from Fieweger et al., Ciezki et al. and Davidson et al. for temperatures from 600?1300 K, for fuel air equivalence ratios in the range of 0.5 to 3.0 and pressures from 3.2?40 bar. It is also validated against shock tube data from Horning et al. and Smith et al. and Vermeer et al. for high temperatures from 1250?1800 K, for fuel air equivalence ratios in the range of 0.5 to 2.0 and pressures from 1 4 bar. The detailed, lumped and skeleton mechanisms are further validated against the laminar flame speed experimental data of Davis and Law and for flame structure data of El Bakali et al.. Predictions for the low and medium temperature range are also tested against species and heat release profiles obtained in plug flow and jet stirred reactors at pressures between 3 and 12.5 bar against the experimental data of Callahan et al., Held et al., and Dagaut et al.

The mechanisms are further validated and applied in HCCI (Homogeneous Charge Compression Ignition) engine simulations using a zero dimensional ignition model. Mechanism show good agreement in cylinder pressure, temperature and also heat release rate with the experiments of Tsurushima et al.

Finally detailed and lumped mechanisms for the oxidation of PRF fuel blends n-heptane/iso-octane and the alternative fuel blend n-heptane/toluene are presented and validated under conditions relevant for HCCI engines. For mixtures of n-heptane/toluene the mechanism was validated against shock tube experimental data from Burcat et al. and for mixtures of iso-octane/n-heptane the mechanism has been validated against the shock tube experimental data of Fieweger et al. for different octane numbers. In addition, the mechanism has been tested and applied using a zero dimensional engine model for HCCI. Results were compared against the experiments from Kalghatgi et al. and Christtensen et al. under a range of different initial operating conditions and varying fuel blends.

Abstract

Two different mathematical methods of implementing a detailed kinetic soot model have been employed in this work. The theoretical description of the soot modeling employed in this work is divided into three parts. Initially a detailed kinetic soot model is described from a chemistry and physics perspective. The soot model is then elaborated into mathematical form, starting with the formulation of the method of moments. Later on a thorough description of the sectional mathematical formulation of the detailed kinetic soot model is given. The sectional method is a new addition to the toolbox developed and used by the kinetic workgroup at Lund University.

The sectional method and the moment method are compared, with the focus of doing a theoretical validation of the sectional method. Where discrepancies between the models exist, due to choices of approximations and discretizations, these are investigated and explained. The validation is carried out in the framework of a 0-dimensional code usually used for describing the process of ignition in perfectly stirred combustion reactors. A 0-dimensional reactor tool is also used, in which precalculated or premeasured chemistry profiles are read in. Based on the read-in profiles soot formation is calculated. Features as well as limitations of the sectional method are investigated.

The sectional method is also validated using experimental data. A laminar premixed flame is modeled and the calculated profiles of the soot particle size distribution function are compared to experimentally measured distributions. Comparisons are made for different flames with different temperatures. An investigation on how the sectional method performs with another well known soot model is also performed.

The moment method and the sectional method are both applied in different turbulent non-premixed combustion cases. In most of the work the soot models are used within the general turbulent combustion modeling approach called the flamelet model. Turbulent non-premixed combustion is also modeled using a stochastic reactor model.

Turbulent diffusion flames are simulated using the flamelet model. The flamelet model describes the turbulent flame as an ensemble of 1-dimensional laminar diffusion flames, so-called flamelets. The interaction between flowfield and chemical reactions is described, while decoupling the actual calculations of chemistry and flowfield. Both the moment method and the sectional method have been applied within the flamelet model to study turbulent freely propagating diffusion flames. By applying the sectional method in a turbulent flame, spatially detailed information on the evolution of the soot particle size distribution was obtained. This is novel and has taken the study of soot formation and specifically the study of the evolution of the soot particle size distribution into a new area. The moment method in combination with the flamelet model was also used to investigate diesel engine combustion.

A stochastic reactor tool is used to model carbon black (i.e. soot) formation in a carbon black reactor. This tool is a 0-dimensional model, assuming spatial homogeneity can be replaced by statistical homogeneity.

Abstract

The modeling work developed in this thesis can be divided in two main areas of investigations: autoignition related to spark ignition engine and combustion and emissions formation in relation to diesel engines.

A first version of a detailed kinetics engine simulation program, extended to handle full cycle calculations, was employed in order to demonstrate the strong effect that nitric oxide from the residual gas has on the autoignition onset. It was found that a concentration of about 500 ppm in the intake gas determines a maximal promotion of autoignition.

Increased simulation capabilities were achieved by integrating the in house detailed kinetic code into a commercial 1-D engine program. By this it was possible on one hand to have a simulation tool able to handle detailed kinetics and thus have control over combustion and emissions, and on the other hand to monitor global engine operation conditions.

Further accuracy on autoignition was achieved by employing a stochastic reactor model (SRM) for spark ignition engine calculations. Based on the probability density function (PDF) this approach was able to model phenomena with strong influence on engine knock as: turbulent mixing and inhomogeneities, phenomena usually neglected by regular existing engine programs. While keeping computational time low and still using detailed chemistry, good correlative results were obtained with the stochastic approach.

For the second part of the work, the stochastic reactor model was applied for diesel engine combustion and emissions investigations. The purpose of this implementation was primarily the calculation of NOx emissions and soot within diesel engines. Soot calculations were based on the method of moments. Comparisons of soot production with measured data from a carbon black reactor indicate good agreement. The diesel SRM model was applied on a FIAT car engine and on a low speed marine engine. The model is capable to correlate the ignition timing and to indicate the trends in NOx and soot generation.

Abstract

The simulation of physical phenomena occurring in chemical reactors requires the description of the kinetics involved in the underlying combustion process. Kinetic models are developed for this purpose. A software package, BlueBellMouse, has been developed to facilitate a deeper automatization and ease of simulation, reduction and optimization of kinetic models. A kinetic model is made of a list of chemical reactions and their reaction rate parameters. The range of application of these models is limited by the set of validation cases and physical parameters that have been taken into account during the compilation. Often, their use under different conditions demands a re-optimization. A systematic optimization technique has been developed in the past that consists in adjusting the reaction rate parameters in a mathematically rigorous way using a set of experimental data as constraints. This approach, used so far for the compilation of “general purpose” detailed kinetic models, applies as well to the development of chemical mechanisms for specific tasks, like for example engine simulations. Homogeneous Charge Compression Ignition (HCCI) engine has found in recent years the interest of the scientific community and automotive industry for their ability to provide high thermal efficiencies and low NOx and particulate emissions. The next step in HCCI engine research is to transfer the accumulated knowledge to industrial applications. Some shortcomings are still to be solved to make these engines suitable for commercialization like, for example, the difficulty in control. Various techniques have been investigated in order to overcome those problems; from variation of the fuel composition to exhaust gas recirculation. The need to change fuel characteristics requires the availability of continuously updated kinetic models optimized under engine conditions. In an operative environment, the calculation speed becomes an essential feature. A key factor for the computational time is the dimension of the kinetic model. One way to achieve reasonable dimensions is to strongly reduce the detailed mechanism. Through the optimization techniques, it is feasible to over reduce the original mechanism and re-optimize the coefficients a posteriori to regain accuracy in the model predictions. Using BlueBellMouse a natural gas fuel reference model as developed by Warnatz et al. has been optimized for a set of HCCI engine experimental cases. The model has then been reduced eliminating redundant species and the corresponding reactions until just the most essential components were left and the model predictions showed high discrepancy with respect to the experimental data. The so obtained skeleton mechanism has then been re-optimized to regain the required accuracy. In addition the optimization of a gasoline fuel reference mechanism containing mixture of n-heptane and iso-octane, toward a set of shock tube experimental cases.

Abstract

Calculations on a Homogeneous Charge Compression Ignition (HCCI) engine have been performed. Zero-dimensional models were used. The simplest model compressed the gas to auto-ignition, using temperature and pressure at a certain crank angle position obtained from engine experiments.

It was found that calculations with good agreement could be accomplished, if using correct temperature, pressure and air/fuel mixture composition. However, the calculations proved to be extremely sensitive to even small variations in temperature. Further, natural gas engine calculations showed a high sensitivity to the contents of higher hydrocarbons such as ethane, propane and butanes. The validity of the kinetic mechanism was also a crucial factor. Due to the assumption of total homogeneity in the combustion chamber, a too rapid heat release was predicted.

Two interfaces were developed, coupling the chemical kinetics code to existing engine simulation tools. These combined kinetics calculations and engine simulations proved to be an efficient tool for HCCI-engine analyses.

Methods for mechanism reductions were developed, and implemented in the kinetic code. This was a stepwise procedure where the first part was to apply the quasi steady-state assumption (QSSA) on HCCI-calculations, where a measure of the species life-time was used to determine which species should be considered as steady-state species. The method showed a good agreement compared to the original mechanism, even for a relatively large degree of reduction.

Sensitivity analysis and reaction flow analysis was combined in a semi-automatic method to generate skeletal mechanisms. The skeletal mechanisms give a good agreement, but only for a limited degree of reduction. A fully automatic method for reduction over a selected range of physical parameters was developed. It combines the two methods by applying QSSA on an automatically generated skeletal mechanism. The suggested method showed good agreement and an excellent potential for future tailor made reaction mechanisms, using a detailed reaction mechanism as a basis.

A reaction mechanism for formaldehyde, methane and methanol was developed. The aim for this work was to produce a C1 mechanism of general characteristics, covering formaldehyde, methane, and possible methanol, giving correct species profiles for intermediate products. The mechanism was capable of accurately predicting ignition delays for formaldehyde and methane over a wide range, gave decent methanol auto-ignition prediction, and could further accurately predict the species profiles for formaldehyde but was not capable of calculating flame speeds for methane.

A semi-detailed reaction mechanism for Primary Reference Fuels, mixtures of iso-octane and n-heptane, was developed. The predictions of ignition delay times showed a good agreement to experiments. The mechanism proved to be numerically efficient compared to mechanisms of equivalent accuracy.