Abstract

The importance of catalytic after-treatment for automotive emissions is not neglectable concerning current environmental protection discussions. A reasonable and time efficient catalyst model can help to reduce the necessity of time consuming experimental investigations on physical parameters for catalytic converter construction. It can further support the preparation of necessary experimental setups to analyze physical and chemical phenomena in catalysts. Physical parameter and/or chemical kinetic optimizers can be an advanced tool to support computational models in terms of adjustment to an experiment. In this work a physical parameter optimizer is developed and validated against a transient three-way catalyst experiment. The modeling results are compared to the measured data in terms of temperature and emission conversion behavior and show a good agreement.

Abstract

Highly fuel-efficient Diesel engines, combined with effective exhaust aftertreatment systems, enable an economic and low-emission operation of heavy-duty vehicles. The challenge of its development arises from the present engine complexity, which is expected to increase even more in the future. The approved method of test bench measurements is stretched to its limits, because of the high demand for large parameter variations. The introduction of a physics-based quasi-dimensional stochastic reactor model combined with tabulated chemistry enables the simulation-supported development of these Diesel engines. The stochastic reactor model mimics mixture and temperature inhomogeneities induced by turbulence, direct injection and heat transfer. Thus, it is possible to improve the prediction of NOx emissions compared to common mean-value models. To reduce the number of designs to be evaluated during the simulation-based multi-objective optimization, genetic algorithms are proven to be an effective tool. Based on an initial set of designs, the algorithm aims to evolve the designs to find the best parameters for the given constraints and targets. The extension by metamodels improves the prediction of the best possible Pareto-front, while the time of optimization is kept low. This work presents a methodology to couple the stochastic reactor model and the multi-objective genetic algorithm. First, the stochastic reactor model is calibrated for 10 medium and high load operating points at low engine speeds. Second, each operating point is optimized to find the lowest fuel consumption and specific NOx emissions. Further, it was ensured that the maximum peak cylinder pressure and turbine inlet temperature are not exceeded. This enables a safe operation of the engine and exhaust aftertreatment system under the optimized conditions. The results reveal two major outcomes. First, the selection of the optimization and space filler algorithm is crucial to find the best possible Pareto-front. In this work, the NSGA-II genetic algorithm coupled with metamodels in conjunction with the Latin Hypercube space filler algorithm has proven to be the best choice. Second, the EGR rate and compression ratio are found to be the most effective measures to reduce fuel consumption and NOx emissions for the selected
engine design.

Abstract

There has been several NOx mechanism developed in recent years often dedicated to particular condition of interest. Unlike hydrocarbons, there has not been enough studies for nitrogen related chemistry on experimental and numerical side. Still today we lack reliable rate parameter values, thermochemistry and experimental data base for validating the NOx model. Our main objective is to develop kinetic model which covers a wide range of experiments (laminar flame speed, oxidation in jet stirred reactor (JSR), oxidation in plug flow reactor (PFR), shock tube ignition delay time and speciation in burner stabilized flame) for different fuels. An updated kinetic mechanism of H2 and C1 system along with NOx prediction mechanism has been developed on the basis of an extensive literature review and including wide set of experimental data. The kinetic model has been developed on a hierarchical basis.

Abstract

The conversion effects of a three-way catalyst are simulated in previous works using single and multiple representative channel approaches with detailed surface kinetic models. In addition, this article introduces global gas phase chemistry to the model. This allows reflecting ongoing reactions due to incomplete combustion products in low temperature regime. The 1D single-channel model representative for the catalyst is used here. Next to the comparison of the catalyst outlet emissions with and without gas phase chemistry, the transient temperature increase is simulated in order to model the catalysts light off temperature. Additionally, the transient inlet emissions are enhanced to show the influence of water and hydrogen on the modeling results. The heat transfer is modeled by wall heat losses to provide proper heat dissipation out of the catalyst. The modeling results show a good agreement to the experimental data with low computational cost.

Abstract

A chemical species lumping approach for reduction of large hydrocarbons and oxygenated fuels is presented. The methodology is based on an a priori analysis of the Gibbs free energy of the isomer species which is then used as main criteria for the evaluation of lumped group. Isomers with similar Gibbs free energy are lumped assuming they present equal concentrations when applied to standard reactor conditions. Unlike several lumping approaches found in literature, no calculation results from the primary mechanism have been employed prior to the application of our chemical lumping strategy. An 807 species and 7807 individual reactions detailed mechanism comprising n-decane, α-methylnaphthalene and methyl- decanoate has been used. The thermodynamic data have been analysed and 74 isomer groups have been identified within the oxidation of n-decane and methyl-decanoate. The mechanism reduction has led to a mechanism size of 463 species and 7600 reactions. Thereafter the lumped mechanism has been checked under several reactor conditions and over a broad range of temperature, pressure and equivalence ratio in order to quantify the accuracy of the proposed approach. In all cases, very good agreement between the predictions obtained using the lumped and the detailed mechanism has been observed with an overall absolute error below 12%. Effects of the lumping procedure on sensitivities and on isomer concentrations were taken into account to further demonstrate the validity of the proposed approach.

Abstract

The major innovation in this work can be described in two parts: a multiphase reactor model and the reduction of a detailed biomass scheme to skeletal size. The result is a CPU efficient tool, suitable for analysis and design of biomass processing research and applications. Model prediction is compared against experiments for wood pyrolysis in a flow reactor. The total CPU time for a calculation was reduced to less than 5 minutes using a 60 species reaction scheme.

Abstract

High-fidelity soot models can reveal detailed soot information in flame simulations, but the complexity of the models renders them computationally impractical for three dimensional engine simulations with detailed chemistry. In the present study, a detailed soot model, where the soot particle size distribution function is treated using a sectional method was implemented into CONVERGE CFD solver. The use of detailed soot models requires higher order PAH chemistry in the chemical reaction mechanism, thereby increasing the size of the mechanism. Thus, there is a need for acceleration of chemistry calculations. Three different acceleration strategies have been applied in the solver to speed up the calculations. The first strategy is to use adaptive zoning, in which combustion cells with similar properties are grouped into a zone. The second strategy is to construct the Jacobian matrix based on analytically computed derivatives. The third strategy is to solve the linear system of chemical reactions using the GMRES based iterative solver with a suitable pre-conditioner. The proposed acceleration strategies were tested in a 0-D homogenous system and a 3-D diesel engine sector. It was observed that the computational time of the simulations can be reduced by a factor of more than 6 while very good agreement in important soot parameters was retained.

Abstract

Reliable quantitative information on the thermodynamic properties of organic compounds especially for radical and chain-hydrocarbons is quite limited. Experimentally, it is difficult to obtain the equilibrium data directly. In the situation where experimental data are not available, most often these data are estimated or predicted. In view of this, many different theoretical techniques, have been proposed, that can compute and evaluate enthalpies, entropies, and other properties.

However, due to the different techniques available, the values computed may vary for the difference sources. New developed computational methods for properties predictions have been developed and new values of thermodynamic data are available [1].

In this study, the influence of thermodynamic data on propane and propene chemistry for shock tube experiments [2,3] has been investigated. The species sensitive for enthalpy change over ignition delay time for propene were detected with the help of a sensitivity analysis. The thermodynamic data of the C3 sensitive species from our database were replaced using the values from Goos et al. [1] database. The differences established through the comparison of these quantities shows a positive influence on the prediction of the propene and propane ignition delay times. All calculations were performed with the latest version of the DARS program [4] and the current version of the base chemistry from BTU Cottbus [5], which is based on the work of Hoyermann et. al 2004 [6] was used.

Abstract

The influence of thermodynamic data on prediction of auto ignition chemistry for hydrocarbons such as propane and propene has been investigated. Thermodynamic data with high sensitivity on simulation results have been evaluated thoroughly.
It was found that for example a small change in the enthalpy of formation of propene has high impact on calculated ignition delay times for different propene and propane / oxygen mixtures.
On the other hand, thermochemical data of allyl has a lower sensitivity on the prediction of ignition delay times of propene and propane than thermochemical data of propene, but its positive influence on the prediction of this technical relevant property is still important. Recently, a value of 168.6 kJ mol-1 ± 1.8 kJ mol−1, obtained with the HEAT (High-accuracy Extrapolated Ab initio Thermochemistry) protocol, was recommended for allyl (C3H5) for the enthalpy of formation at 298 K [1].
The improvement of the prediction of a detailed reaction mechanism through using accurate thermochemical data, such as the Active Thermochemical Tables Results [2], the IUPAC critical evaluation of thermochemical properties of some free radicals [3] and own quantum chemical calculations with the benchmarked G3B3 composite method is shown.

Abstract

Polycyclic Aromatic Hydrocarbons (PAHs) are a group of chemicals that occur naturally in coal, crude oil and gasoline. The nearly endless growth process of PAHs, where the six carbon atoms form an aromatic ring, necessitates a detailed understanding of chemical reaction pathways responsible for their formation and sequential growth, followed by the nucleation or inception of small soot particles, soot growth by coagulation and mass addition from gas phase species etc. The transition process from planar PAHs to spherical nascent soot particles is accompanied with stocking of PAHs on top of each other. Because the growth of PAHs involves an unlimited number of different species and reactions, a detailed description of the entire process is yet not possible. But to still make a proposition about the amount of PAHs formed and their size distribution, the statistical method of Linear Lumping [1] that describes the chemical source terms of PAHs was proposed by Frenklach. Therefore the growth of PAHs can be approximately described as a rapid polymerisation [2] process. In this study, it is assumed that the rates of polymerisation of PAHs and that of particle formation from both PAHs and condensation of PAHs at the soot surface are faster than the rates of convection and diffusion. As a result, the Steady State Assumption can be introduced for every single PAH that leads to a systematical reduction and consequently a clear decrease of numerical effort. In general, this assumption is valid, if there are chemical reactions which consume a chemical specie faster than it is formed by other reactions. In multidimensional steady systems it means that the convective and diffusive terms in the conservation equations are essential smaller than the chemical terms.
According to the steady state condition, we investigate in this work the Particle Size Distribution Functions (PSDF) of the PAHs and incipient soot within a range of temperatures and compositions. Instead of endless conservation equations for every single PAH a limited number of the moments of the size distribution functions of the PAHs can be solved. In the HACA mechanism described by Frenklach [3] the accumulation of acetylenes and the aromatic ring closure is described in one reaction only, whereas in this study we divide it into two reactions in a modified HACARC mechanism [4]. This division is introduced, because the connection between the attached acetylenes and the soot surface element can be broken at high flame temperature before a new ring closure occurs. Furthermore in contrast to Mauss et al. [4], in this work, the ring closure is assumed to be irreversible.
In general, the well-known bimodality of the incipient soot PSDF can be described by using a complex log function [5]. In this work an algebraic description of the target particle diameters region based on one Difference Equation System is established and then solved with the help of Z-transform. The primary results have been evaluated against published experimental study from Wang et al. [5] and the satisfactory agreement in target range has been achieved.

Abstract

Soot formation in a turbulent jet diffusion flame is modeled using an unsteady flamelet approach. In the present work, we study the effects of the choice of the dependence of scalar dissipation rate on mixture fraction and agglomeration processes on the predicted soot volume fraction. it is found that good predictions of soot volume fraction can be obtained without considering preferential diffusion effects.

Abstract

A stochastic model based on a probability density function (PDF) was developed in a correlative tool for the investigation of different conditions that determine knock in spark ignition (SI) engine. By using a stochastic ensemble to represent the PDF of the scalar variables associated with the burned and the unburned gases it is possible to investigate effects as non-uniformities and turbulence mixing that are neglected by the regular existing zonal models. Comparisons with measurements show that our code is able to correlate the auto ignition timing and to prove the effects of inhomogeneities and of mixing.

Abstract

In recent years detailed kinetic models have been developed and validated for both auto ignition and laminar combustion processes. The range of validity of these models is limited by the range of physical parameters covered by the available experimental data. Often, a re-optimization of the models is necessary. Optimization of reaction models has been automated in the past for example for the development of the GRI-Mech. In this work a similar procedure has been applied to optimize a gasoline fuel reference model as developed by Curran and Westbrook for HCCI engine cases. The HCCI-process was simulated using a compressed homogeneous reactor. Computations have been carried out on a PC in a graphical user interface environment. Experimental data were collected at Lund Institute of Technology on a Volvo engine with variable compression ratio and port injection mixes of iso-octane/n-heptane. Results show that, after optimization, HCCI experiments can be predicted for the given conditions.

Abstract

The aim of this work is to generate a semi-detailed mechanism for the oxidation of n-heptane and iso-octane mixture with 241 species and 1905 reactions and to optimize it for the different engine operating conditions to cover the full range of temperature. A step wise efficient lumping strategy for different reaction types combine with necessity analysis has been used for the simplification and reduction of mechanism in order to make it faster and less complex in terms of both number of reactions and species. Its validation has been done against the shock tube experimental data from Fieweger et al. for both low and high temperature regions, lean and rich fuel conditions, a range of octane numbers and pressures from 13 bar to 40 bar. In addition, the mechanism was tested using a zero dimensional Homogenous Charge Compression Ignition engine model against experiments under a range of different octane numbers and initial temperatures.

Abstract

The Quasi Steady State Assumption (QSSA) is employed to reduce chemical mechanisms within automatically recognized reactive zones. In order to identify the distinct reactive zones within a chemical process, a domain splitting method using a machine learning clustering technique is used in a pre-processing step. The present reduction procedure carries out the reduction within each domain separately according to a steady state selection parameter, the level of importance (LOI), valid for the corresponding domain. This implies that a clustered domain has the same set of species in steady state throughout the domain. The transitions between the domains are treated according to traditional domain splitting methods. It is found that the method enables a stronger reduction than traditional methods, as only the true non-steady state species for the given local conditions needs treated in detail. The method is applied to calculations of ethene diffusion flamelets.

Abstract

We present a method for developing dynamically reduced chemical mechanisms based on the quasi steady state assumption and a domain splitting procedure. This procedure carries out the reduction within each domain separately according to a steady state selection parameter valid for the corresponding domain. The domains are defined according to a clustering algorithm. The method is applied to calculations of ethene diffusion flamelets.