Stat-of-the-Art Review: Combustion

Introduction

Throughout industry, Computational Fluid Dynamics (CFD) demonstrates its power and effectiveness. It is used in a wide range of design applications, whenever fluid flow, heat transfer, combustion or many other complicated processes are significantly influencing the product. The objective of this document is intended to be a survey on CFD developments and validations in combustion and heat transfer, the second thematic area (TA2) of QNET-CFD network. Started in May 2000, this review has been finally focused on the gas turbine combustor and internal combustion engine, reflecting the industrial application challenges presented by the network partners within TA2.

The review of the paper begins with a short section (section 2) to underline the importance and relevance of CFD combustion and heat transfer for the industry.

Section 3 is a scientific and technical state-of-the-art of the gas turbine combustor modelling that deals with chemical mechanisms, turbulence and/or chemistry interaction, modelling of two phase flows as well as with soot modelling. This is completed by a succinct review (section 4) of the experimental and theoretical works devoted to the instability of the flow types important for aero-engine combustors.

Section 5, focuses on CFD and combustion applications related with automotive engine developments. There one can find some issues of engine processes modelling, turbulence modelling and combustion process modelling.

Next section (section 6), surveys the progress in experimental instrumentations that are essential in CFD validation and proof of the sense of making afford to improve them.

Last part summarises the CFD application in combustion, mentions advantages and disadvantages of the computer simulations and methods that can be used to get results in this field, like RANS and LES.

Importance of CFD for Combustion and Heat Transfer to Industrial Sector

Combustion is of great importance in modern society, as a means of supplying energy for everyday use, for powering vehicles, and for industrial processes. Different aspects of combustion are present in many practical processes, including energy conversion, material processing, propulsion, waste incineration, pollution and fire. Since combustion systems are extremely costly to build and test, CFD can be a valuable tool to model the performance of different designs, perform parametric studies and virtual prototype each design before actual fabrication.

Combustion problems generally involve complicated geometries, complex physics, heat transfer, and fluid flow. This multidisciplinary aspect is reflected in the wide range of research interests in this field, the objectives of which aim at:

• system efficiencies improvement,

• pollutant emissions and fuel consumption reduction to meet the future restrictive legislation,

• design cost and development time reduction,

• disposal of hazardous wastes, prevention fires in buildings and space habitats, and processing of materials.

To achieve these goals, in-depth understanding of the dynamics of the turbulence-chemistry interactions that occur in the combustor, is required. Modelling can play a significant role to identify and optimise the control parameters of the combustion systems.

During the past 30 years, substantial advances have been made in combustion and heat transfer modelling, but only recently they have spread to over many industrial sectors. In fact, CFD developments in this area have reached a state where industrial use has been feasible technically and economically.

Efficient numerical algorithms and mesh generation, associated with parallel computing technology, now permit more and more rapid and practical CFD calculations.

In automotive applications (both spark ignition and compression ignition engines), last decade work was mainly dedicated to fuel consumption reduction and to satisfy tougher emission regulations. Many carmakers are now convinced of the great potential of combustion modelling in term of time and cost effectiveness. This leads them to use CFD calculations from program planning to design and optimisation.

This was made possible thanks to CFD codes (commercial or public domain), which are now widely available for the development of engine technology. Specialised codes for variable geometry combustion chambers due to piston and valves motion in Internal Combustion Engines (ICE) have been developed too (cf. §5).

Scientific and Technical State of the Art of Gas Turbine Combustor Modelling

During the last decades engine efficiency has been improved by increasing the pressure and temperature of the thermodynamic cycle. The improved cycle efficiency reduces specific CO2 emissions, which was declared as an important pollutant in the United Nations conference in Kyoto. At the same time, combustor design has been subject to increasingly stringent environmental legislation as defined by the ICAO. The increased pressure and temperature levels lead to improve combustor design, to keep NOx emissions down to a low level, since these emissions increase exponentially with peak temperature levels in the combustor (Warnatz J. et al., 2001, Bowman C.T, 1993).

Combustor design has historically been dependent on extensive test rig programs, which are time consuming and expensive and offer no internal data at full power conditions. Validated Computational Fluid Dynamics (CFD) offers means to save cost and time through reduction of tests rig, and to move towards new design concepts outside of the empirical knowledge basis.

The prediction of reactive two-phase flow in gas-turbine combustor requires comprehensive models for all relevant physical phenomena involved to ensure accurate results from the simulations (Kim W.W et al., 1998) such as the fuel preparation process, the turbulent fluid flow (Peters N., 2002), reaction kinetics (Warnatz J. et al. 2001), and the radiative heat transfer. Beside the radiation of the gaseous combustion products CO2 and H2O a significant amount of heat loss is due to radiation of soot particles (Bockhorn H., 1994).

To achieve low emissions, there has been a concerted effort to develop lean, premixed (LP) combustion technology in which none of the current aero-engine gas turbine manufacturers have existing design or service experience. This is compounded by the inability to take internal measurements at high power conditions due to the hostile conditions, which make it difficult to establish design rules quickly and reliably.

Computational Fluid Dynamics (CFD) offers a route by which these problems can be alleviated, since it can give detailed internal information across the range of operating conditions. Nevertheless, this is only true if the results obtained by CFD are accurate and reliable, as incorrect results could actually hinder the progress.

For several physical sub-processes of combustion, the modelling has currently reached a status, which satisfactorily can describe the relevant parameters’ limits for dedicated burner configurations. These are for example the heat release, the lean blow out or formation of NOx. In addition, for other parameters like CO or UHC emissions, the relevant physical phenomena can be described by numerical models but only with rather demanding computational effort.

Nevertheless, the CFD codes currently available to aero-engine gas turbine manufacturers fall short of the ideal, both in terms of range and accuracy of the models, particularly for lean, premixed flames. Premixed combustion modelling requires a greater knowledge of the chemical behaviour of the fuel, in contrast to diffusion flames where combustion is mainly controlled by the mixing of fuel and air. These types of combustion processes are controlled by the fuel spray formation and liquid fuel evaporation behaviour. Modelling the break-up of liquid fuel sheets and jets is another area that is little understood, and thus cannot be modelled currently with any confidence, thus affecting the accuracy of emissions calculations. Another area of interest is the field of soot modelling, which is relevant to the conventional and rich-quench-lean (RQL) combustors with a fuel rich primary combustion zone, as these are proposed as pilots to the LP modules. Soot is important not only as pollutant, but also because thermal radiation from the soot particles is the primary source of heat transfer to the combustor walls in the primary zone and hence impacts durability.

As mentioned above, there are four areas of special interest to gas turbine combustion, chemical mechanisms, turbulence/chemistry interaction, two-phase flow and soot modelling that status still, is not fully understood. Improvements and validation of modelling approaches of these phenomena not only improve predictive capabilities but also would help to optimise existing technologies because not all the information necessary for design optimisation can be collected by rig experiments due to hostile conditions and limited accessibility to the flame especially at high pressure conditions. In this case, the use of numerical models is the only way to get the necessary information to derive rules for design optimisation from regions inside the combustor, which are usually not accessible.

Another field of interest that should also be mentioned, but is not subject to further discussion here, are combustion instabilities (Fureby C., 2000). In particular, for lean premixed configurations the thermo-acoustic behaviour of combustion chambers is often subject to demanding effort during the development phase. Although a lot of scientific work on this topic has already been carried out, the phenomena cannot yet be modelled in a satisfactory manner due to the highly non-linear nature of the relevant physics. © ERCOFTAC 2004 Chemical Mechanisms

Established approaches to generate chemical schemes with respect to pollutant chemistry including NOx and soot usually formulate a kinetic mechanism comprising approximately 1000 elementary reactions. The compilation of detailed reaction mechanisms is well advanced and the techniques have been demonstrated and validated for several single component fuels. There are mechanisms for some specific model fuels like n-heptane or n-dodecane that are validated by comparison to experimental data of burning velocities, ignition delay times and also by distributions of the concentrations of main species in the flow field of the flame.

It is obvious that detailed kinetic schemes are not applicable to combustion modelling in 3-D CFD due to the enormous computation effort. The second problem is the fact that the liquid fuels usually used as Diesel fuel or aviation Kerosene are ill-defined mixtures of a huge number of different chemical species, mostly aliphatic and aromatic components that vary in composition and thermo physical- properties from batch to batch.

In order to overcome this problem, reduced schemes have to be used for dedicated model fuels that approximate the evaporation behaviour, ignition delay, flame speed and pollutant formation within a defined range of accuracy. These model fuels consist of a small number of pure hydrocarbons with well-established kinetic behaviour. © ERCOFTAC 2004 Turbulence/Chemistry Interaction

It is obvious that many combustion-related issues of technological interest to the gas-turbine industry depend upon the interaction of thermo- chemistry and aerodynamics, which is poorly accounted for in current combustion models by simply ignoring either the mixing or chemical reaction. There are several models of varying complexity and computational cost to model the interaction.

The state of the combustion process in the reaction zone can be represented by one or more progress variables Libby P.A and Williams F.A, 1994, Poinsot T. andVeynante D., 2002, Veynante D. and Vervisch L., 2002. The use of a single progress variable is suitable for joint presumed PDF (Probability Density Function) approaches for the distribution of the fuel fraction and the temperature. For these models, the PDF is represented by its mean value and variance where the form of the PDF has a given shape. Alternatively, methods exist to solve additional equations for the transport of the PDF by using more progress variables and advanced tabulation methods to model the combustion (Pope S.B, 1990, Saxena V. and Pope S.B, 1999). However, even if tabulated reduced chemical systems are used, the calculation of the chemistry sub- models is still the most CPU intensive. Computationally less expensive models exist like presumed PDF approaches or flamelet models (Bray K.N.C and Peters N., 1994) that can be regarded as a compromise between numerical effort and accuracy.

The most simple approach, the usage of eddy dissipation concepts (Spalding D.B, 1971) in combination with a global chemistry scheme (usually not more than three chemical steps), is currently still widely used in particular for complex 3-dimensional geometries usually connected to industrial devices and configurations due to its robustness and small computation effort. © ERCOFTAC 2004 Modelling of Two-Phase Flows

A critical factor in establishing combustor performance is a good understanding of the fuel atomisation and distribution (Williams F.A, 1990). Modelling of this process is difficult as atomizer performance is a function of liquid film thickness, fuel viscosity, air to fuel flow ratio, and other nozzle characteristics that are known in principle but not in a quantitative manner. Single nozzle or single sector tests do not offer an alternative, as nozzle-to-nozzle interaction effects are not accounted for.

The atomisation of fuel inside the combustor of jet engines is based on the concept of air blast atomisation, which employs the kinetic energy of an air stream to shatter the fuel sheet into ligaments and then droplets. In order to minimize emissions, especially NOx and soot, a homogeneous mixture of the fuel with the combustion air should be achieved, since local inhomogeneities lead to temperature peaks that result in enhanced NOx formation. Until now, evaluation of the performance of different injector configurations has been possible only by experiments, using both exhaust gas analysis and detailed optical measurements in the region close to the nozzle. The results from these experiments lead to an optimisations of the injector geometry.

Numerical simulations could be used to optimise the injector geometry, but the models would need to accurately capture the atomisation, which is strongly dependent on the liquid fuel flow inside the fuel nozzle and the coupling between the liquid and gas phases. Currently, there are no suitable models that describe these processes in realistic injector geometries. In order to simulate the atomisation process accurately, there is a great need for models to describe the earliest stage of the fuel preparation process, the formation of the liquid fuel film in the injector. Existing models usually are based on empirical correlations, which obviously are problematic to be extrapolated out of the range of applicability. In particular, those approaches are often derived from the measurements at atmospheric pressure and extrapolation to higher pressures (up to 5 MPa) is sometimes more than doubtful.

Regarding the evaporation of the fuel droplets, existing evaporation models have the drawback of using a single component fuel to describe the heat and mass exchange from the liquid to the gaseous phase. As outlined above, kerosene or diesel are a multi-component fuel, and this will affect the distribution of the gaseous fuel species and hence the pollutant formation. Secondly, liquid properties have a strong effect on the atomisation process but also on the evaporation behaviour, so data need to be available for liquid viscosity, thermal conductivity, surface tension, heat capacity, enthalpy of phase change, vapour pressure and diffusion coefficients and also vapor phase properties for pressures up to 5 MPa.

Numerical spray combustion models have been proposed over the past two decades. The comprehensive reviews of spray and combustion models can be found in the literature (Chigger 1976, Law 1982, Faeth 1983, Crowe 1991, Sirignano 1993, Oefelein and Young 1996). In general, these models can be classified into three types: 1) two-fluid models which treat the particles as a continuum, (2) separated flow Lagrangian trajectory models which track individual particles in the gas field, and (3) direct numerical simulations. © ERCOFTAC 2004 Soot Modelling

The formation and emission of soot particles from aero-engines is of concern from an environmental perspective and remains of predominant importance in the context of radiative heat transfer through its impact on combustor life and the emission of pollutants. Beside the prediction of heat release rates and the concentration of the main combustion products there is a need to predict emission data. In particular, soot modelling has been an area in which there has been a great deal of interest in recent years. Nevertheless, current smoke production and burnout calculations fail to predict combustor exit smoke sometimes by considerable margins, due to problems related to the complexity of the fuel and the effects of pressure, heat loss and turbulence.

Soot formation and oxidation is a complex process involving both chemical and physical steps, not all of which are well understood (Wagner H.G, 1979, Haynes B.S and Wagner H.G, 1981, Homann K.H, 1984, Bockhorn H., 1994). Even if these complex paths could be described completely, the computational cost would be prohibitive for combustor calculations and thus simplified models, which capture the overall behaviour, must be developed. A number of such models have been proposed and found to perform adequately for the geometries and conditions for which they have been developed. However, their application to conditions outside the initial domain of interest has lead to varying degrees of failure. In particular, for combustion chambers there are problems in extrapolating to high-pressure conditions and, notably, in representing the large amount of oxidation occurring. Thus, there is a need for the development of models that address these limitations if reliable predictions are to be achieved. Experimental approaches do not offer a viable alternative to numerical modelling, given the hostile environment inside the combustor and the problems in obtaining suitable access to the combustor within an engine.

Special focus has to be drawn on the fact that there are intermittent processes associated with the interaction between the gas and particulate phases, notable in soot growth and oxidation, and that parameters such as particle size distributions have to be taken into account. © ERCOFTAC 2004 Unstable Flow Regimes Underlying Combustion Processes in Aero-engine Combustors

The CFD modelling of the gas turbine combustor involves complex physics including heat transfer, turbulence, chemical reaction mechanism and turbulence/chemistry interactions. Whatever complex models of these phenomena are applied, the accuracy of the numerical prediction of combustion process is strongly affected by the quality of the underlying aerodynamics modelling. It seems that understanding of combustor aerodynamics has a fundamental importance for the quality of the overall combustor modelling. There are several flow types that could be considered as gas turbine combustor underlying flow regimes. We would like to concentrate here on aerodynamics of flow types like free jets, variable density jets, swirling jets and annular jets that are far from being fully understood. All these flow types are substantially affected by flow instability and large scale vortex structures growing as a result of the flow instability. Knowledge of the large-scale vortex structures is lost in Reynolds averaging and therefore the RANS modelling of unstable flow types can lead to quite significant errors of predictions. These errors not only affect the turbulence structure which is crucial for chemical reaction modelling but may also cause some significant discrepancies in mean velocity flow field.

At the end of eighties Sreenivasan K.R et al. (1989), Kyle D. & Sreenivasan K.R (1993), Riva R.et al. (1990), and Monkewitz P.Aet al. (1990) verified experimentally that low density jets underwent an absolute instability when the density ratio of the jet and surroundings was low enough. When the flow is absolutely unstable then according to linear stability theory (Landau L., 1959) a small disturbance imposed on a flowfield grows infinitely in time and in real flows this growth is limited only by non-linear interactions of large amplitude flow oscillations. It means that absolute instability may lead to very intense oscillations with length scales comparable to mean flow ones and such a flow type becomes even more difficult for RANS modelling.

Some experimental works performed by Lehman et al. (1997) on a swirl nozzle with two concentric air streams that is used in aero-engine combustor showed that a strong coherent velocity fluctuations of the first azimuthal mode existed close to the nozzle. This experimental outcome stimulated theoretical work of Michalke A. (1999) who studied numerically single ring jet with backflow by applying spatiotemporal linear stability theory. These investigations showed that both backflow and swirl promoted absolute instability of the first azimuthal mode while axi-symmetric mode was convectively unstable for all flow parameters. A very interesting experimental work on swirling jets was earlier carried out by Panda J. and McLaughlin D.K (1994) who observed that for some flow configurations a swirl could diminish jet-spreading rate but for some circumstances the spreading rate was increased by the swirl action. Theoretical study of Lim D.W and Redekopp L.G (1998) showed that when the vorticity generated by swirl motion is concentrated near the jet axis then the swirl renders the flow more absolutely unstable even for axi-symmetric mode. On the contrary, when the vorticity is distributed uniformly across the entire jet the swirl renders the flow more convectively unstable.

The flow configurations, which are important for combustion applications, include certainly combustor geometries where instability processes have minor influence on the structure of the mean velocity field as well as on turbulence and in this case fairly good predictions may be achieved with RANS modelling. In these physical situations the progress in modelling does not depend on aerodynamics because the chemical reaction, heat exchange and soot formation modelling are then crucial.

However, one could expect that for certain configurations instability could govern flow aerodynamics especially when the absolute instability regime appears that can be promoted i.e. by swirl action. In these cases, a significant progress in combustor modelling can be attained only by improvement of aerodynamics modelling. It seems that a key issue in such problems is LES application in aero-engine combustor modelling.

This short review of the experimental and theoretical works devoted to the instability of the flow types important for aero-engine combustors shows that aerodynamics of the flow in combustor is far from being fully understood and it creates the need of further extensive investigations. © ERCOFTAC 2004 Aerodynamics and Combustion in Automotive Engine

Due to the increasing environmental concerns, there has been growing requirement to develop more efficient methodology for cleaner engine design in less time and lower cost. This is particularly true in the vehicle industry where there are strong competitive pressures to bring new vehicle to market in shorter time to respond quickly to costumers need.

For this purpose, new engine technologies are becoming more rapidly implemented such as gasoline direct injection, turbo-charging, exhaust gas re-circulation and treatment. These recent advances would have been impossible without the modelling progress.

Early efforts were made on the closed part of engine cycle, i.e., the compression/ combustion/ expansion phase. These models have evolved from cycle calculations in the 1950s to simple components matching models in the 1960s, full thermodynamics models during the late 1970s and multi-zone and multidimensional combustion models in the 1980s. An overview of engine system modelling is performed by Chow A. et al (1998). Over the past two decades, aided by specially developed process oriented tools, many efforts have been devoted to the development of 3D modelling of turbulence, heat transfer, liquid fuel sprays, mixing, combustion and pollutant formation in reciprocating engine,

Various codes have been used for model development and applications in engine and vehicle industry. Developed by Los-Alamos Laboratory, the well-known KIVA codes family is widely used due the availability of the code sources. However, its capability for solving real engine geometries is limited. For this reason, commercial codes, such as FLUENT, VECTIS, FIRE and STAR CD, are frequently used in automotive engineering because of their superior meshing generation, pre- and post- processing, added to their availability of user support.

While these advances do not replace experimental development entirely, they do allow design engineers to explore options early in the design phase at a much lower cost than if the developments were just experimental. © ERCOFTAC 2004 Issue of Engine Processes Model

The internal combustion engine represents a challenging fluid mechanics problem to model for many reasons:

• In the combustion chamber, gas flow is compressible with large density variation, turbulent, cyclic and unsteady.

• The combustion characteristics are greatly influenced by the details of the fuel preparation process and the distribution of fuel in the engine, which is, in turn, controlled by the in-cylinder mechanics.

• The geometry of combustion chambers is complex and has moving walls (pistons and valves).

Because of the competition between carmakers, another challenge consist of, not only to produce a sufficiently accurate mathematical description of the physics, but also a modelling that meets the designer needs at an acceptable time delay and cost. © ERCOFTAC 2004 Turbulence Modelling

Because the air motion impacts directly on the combustion process and then on heat transfer, fuel sprays, mixing and combustion, simulation has to be very careful in the choice of turbulence models as well as convective differencing schemes. These are essential for the accuracy and reliability of simulation results.

To approach the effects of turbulence, several variants of Reynolds Averaged Navier Stokes (RANS) models have been largely implemented into engine combustion CFD Codes, Gossman A.D (1999). The well-known k-ε models with a wall function have been widely used to approach the in cylinder, despite its deficiency to predict the compression phase.

In these models, only the mean flow is calculated, showing a coherent large-scale tumbling/ swirling structures. Haworth D.C et al, 1996 estimate that computational meshes typically used for RANS modelling of practical in-cylinder configurations (with 105 to 106 computational elements with second order or higher spatial discretization) are sufficient to capture 80-90% of flow 's kinetics energy.

As the turbulence in the combustion chamber covers several order of magnitude in time and space, from those determined by the size of cylinder bore to those determined by the viscosity. These averaged models are unable to capture accurately the impact of unsteady fuel/air mixing on the combustion process or cycle-to-cycle, Haworth D.C et al (1996, 1999).

In this case, only Large Eddy Simulation (LES) combined with a suitable wall function, is appropriate to predict turbulent, mixing and combustion cyclic fluctuations. In LES, the effects of the large scale eddies are resolved using the time dependent Navier-Stokes equations. The effects of the smaller sub grid eddies, are modelled.

After a period at the academic stage, this numerical technique is now recognized as a high potential methodology for industrial simulation of turbulent in cylinder reacting flow. Direct calculation of transient phenomena such as combustion instabilities, more flow details available for the combustion modelling added to greater accuracy could explain the reasons of such interest.

Verzicco et al. (1998) described a large-eddy simulation in complex geometries using boundary body forces, as an alternative numerical method or simulating complex flow of industrial interest. This technique is based on the use of body forces, which allow the assignment of boundary conditions independently of the grid. The main advantage of this procedure is that the above forces can be prescribed on a simple cylindrical mesh so that all the advantages and the efficiency of solving the Navier-Stokes equations in simple constant-metric coordinates are preserved when dealing with complex geometries. A motored axisymmetric piston-cylinder assembly with available experimental data was used (Morse A.P, Whitelaw J.H and Yianneskis M., 1978). The selected flow was particularly suitable for testing the numerical procedure since it has complex geometries with moving boundaries and its Reynolds number is high enough to require a turbulence model.

Nevertheless, application to reacting flows in reciprocating engine has not yet been accomplished; only encouraging tentative and approaches have been reported in open literature, Huilai Zhang et al (2000) and Haworth D.C (1999).

Direct Numerical Simulation applied to a turbulent reacting flow is at its exploratory stage of research. It is limited by computer memory and speed and even the largest computers are unable of computing exact solutions to complex engineering problems. However fundamental problems treated with DNS yield results, which are used to improve the physical sub-models to describe turbulent combustion. © ERCOFTAC 2004 Combustion Process Modelling

Engine combustion models involve a complex, interactive mixture of different physical and chemical processes. Multiphase flows with liquid, gaseous and solid phases (in the case of soot formation) must all be considered.

Fuel and air are introduced into the combustion chamber according to patterns that depend on the type of application. In the case of Diesel engine and GDI, fuel is injected as liquid spray, resulting in a two-phase CFD problem.

The injector mass flow rate and the flow conditions at the nozzle exit is a key issue for a successful simulation of mixture formation, combustion and pollutant formation.

In order to take into account the impact of geometrical details of the transient nature of cavitating injector flow, only a comprehensive multidimensional two-phase flow models are able to provide the relevant information required as input for IC engine spray simulation.

The mathematical approaches used in spray modelling are based on Continuum Eulerian or Discrete Lagrangian. The well-established Lagrangian discrete droplet method proposed originally by Dukowicz J.K (1980), offers a large range of sub-models for break-up, collisions and evaporation, coupled with the turbulent gas field. Gosman A.D and Clerides D.(1997) reviewed diesel spray modelling status, including some important mathematical/ computational aspects. It appeared that current main weakness and uncertainty are in the modelling and measurements of atomisation. To validate the spray modelling, measurements of penetration, droplet sizes and velocities are used rather than the distribution of mixture concentration and temperature. These are determining factors of ignition and combustion but are still difficult to measure especially within the spray.

Over the past two decades, advances have been achieved in the modelling of homogeneous charge and less in stratified spark ignition engine. Weller H.G et al (1994). The "Eddy Break Up"' is a widely used but deficient turbulent premixed combustion model as little explicit account is taken of the flow/flame interaction. "Coherent Flame Model", "Wrinkled Flame Model" and "G-equation" model are considered among the most predictive approaches homogeneous combustion of SI engine combustion. The WFM approach was extended to stratified SI combustion with encouraging results, Gosman A.G (1999).

The challenge in simulating Diesel engine combustion is due to the strong coupling and simultaneity of injection, ignition, and combustion processes. In most time, reduced laminar kinetics' models with simplified interaction with turbulence are developed. © ERCOFTAC 2004 Validation Experiments

Together with more traditional probes, advanced flow visualisation and measurements are playing a very important role in providing accurate data bases for validating CFD combustion codes for automotive engine. Instantaneous, non- intrusive measurements are now available using optical diagnostics instruments, such as LDA, PIV, PLIF, CARS and others.

• Laser Doppler Anemometry (LDA), widely used in the industry for a wide variety of flows and ready-to-use commercial setups that are available. With LDA, biphasic flows can be analysed, and temporal fluctuations can be observed. One major drawback is the size of the domain measured, which is the intersection of two laser beams.

An example of the experimental characterization of the flow in an IC engine by LDA is given by Miles P. et al (2001).

• Particles Image Velocimetry (PIV), is more and more used in the industry. The measurement zone is a wider planar zone comparing to LDA. Modern PIV setups can allow access to all the three components of velocity and it has been used to visualize the in-cylinder flow in several vertical sections. It can be particularly useful to visualize and quantify velocity in sections of the flow in an IC engine.

PIV method has been used by Zhang et al. (1994) to investigate cycle-to-cycle variations in a motored engine and by Devesa et al. (2004) to validate LES calculations in a square piston.

• Planar Laser- Induced Fluorescence (PLIF) used to monitor OH and CH combustion intermediates, and to have fuel/air ratio in the engine's combustion chambers. PLIF has been also used to obtain instantaneous two-dimensional images of OH and CH radicals in a turbulent, swirling, gas/ air flame.

• Coherent Anti- Stokes Raman Spectrometry (CARS) used to obtain major species (CO, CO2, O2 and N2) concentration and temperature measurements. It can be done in rather harsh environments with good geometrical resolution, but in a single point only. The main advantage of this method is its ability of temperature measurement when other methods fail.

• There are many other techniquesm, eg.:

- Laser Sheet Visualisation (LSV) produce image of laser light, scattered by particles in natural gas, oil and coal flames. It provides a simple way to obtain images of combustion flame in burners.

- Interferometric Mie Imaging used to determine the size of the particles in the studied flow.

- Degenerate Four-Wave Mixing (DFWM) can be useful when more established methods like CARS, cannot achieve their goals.

- Polarization Spectroscopy (PS) is experimentally simpler than DFWM, but the underlying physical principles will probably restrict its use to open flames. Windows are indeed causing birefringence problems.

- Laser Induced Incandescence (LII) provides quantitative visualisation of soot volume fraction, through heating by laser light. LII is commercially available.

- Schlieren is a photography technique that allows gas temperature measurements, eg. Reiz R.D and Ruttland (1995). © ERCOFTAC 2004 Summary and Conclusions

The CFD offers a great potential to improve gas combustor design and save cost and time through reduction of very expensive tests rig. On the other hand, complexity of physical and chemical processes involved in the combustion process ranging from aerodynamic instability through turbulence-chemistry interaction to combustion instabilities and thermo-acoustics of combustion chambers presents an extremely difficult task for CFD applications. It seems that constraints of RANS modelling applied to such complex flows limit very much the CFD simulations. Thus a significant progress that can be expected in near future as far as combustion modelling and combustion-turbulence interaction are concerned seems to be related to Large Eddy Simulations. LES modelling of gas turbine combustors that will be introduced to industrial applications in several years will probably bring a qualitative progress in combustors optimisation.

In the same way, in IC engine, Large Eddy Simulation (LES) represents an alternative to the averaged models allowing a direct access to the in-cylinder physical processes. Nevertheless, the need of computations through several engines’ cycles, added to the complexity of the combustion chambers (moving boundaries) are still limiting points in application of LES in automotive engine. © ERCOFTAC 2004 References © ERCOFTAC 2004 Bibliography

Bockhorn H., 1994, Soot Formation in combustion, Springer, Berlin/Heidelberg.

Bowman C.T.,1993, Control of combustion-generated nitrogen oxide emissions: technology driven by regulation, Proc Comb Inst 24:859.

Bray K.N.C., Peters N, 1994, Laminar flamelets in turbulent flames, Turbulent reactive flows

Chow A. and Wyszynski M.L., 1998, "Thermodynamic Modelling of Complete Engine Systems-a Review", MSAE.

Devesa A., Moreau J., Poinsot T., Helie J., 2004, Large Eddy Simulation of Jet/tumble Interaction in a GDI model Engine Flow, technical report, CERFACS, ref. TR/CFD/8.

Dreier T. and Ewart P., 2002, Coherent techniques for measurements with intermediate concentrations, in Applied Combustion Diagnostics, edit. Katharina Kohse-Hoeinghaus and Jay B. Jeffries, Taylor & Francis.

Dukowicz J.K. (1980), A particle-fluid numerical model for liquid sprays", Journal of Computational Physics, Vol.35, pp. 229-253.

Fureby C., 2000, Large Eddy Simulation of Combustion Instabilities in a jet Engine Afterburner, Combust. Sci. and Tech., Vol. 161.

Gosmann A.D., 1999, "State of the art of Multi-Dimensional Modelling of Engine Reacting Flows", Oil and gas sciences and Technology, revue de l'IFP, vol.54

Gosmann A.D., Clerides D., 1997, "Diesel Spray Modeling: a Review", Proc. ILASS-Europe Annual Meeting.

Haworth D.C. and Jansen K. 1996, "LES on unstructured deforming meshes: Towards reciprocating IC engines", Center of Turbulence Research, Proceedings of the Summer Program.

Haworth D.C., 1999, "Large Eddy Simulation of In-cylinder Flows, Oil and gas sciences and Technology, revue de l'IFP, vol.54.

Haynes B.S. Wagner H.G., 1981, Soot Formation. Prog. Energy Combust. Sci. 7:229

Huilai Zhang, Toshio Kobayashi, and Nobuyuki Taniguchi, 2000, Large Eddy Simulation of Motored Engine, Fisita, Seoul.

Kim W.W., Menon S., Mogia H.C., 1998, Large Eddy Siumaltion of Gas Turbine Combustor Flow Fields. Comb. Sci. and Tech., Vol.143

Kyle, D., M., Sreenivasan K.,R., 1993, “The instability and breakdown of a round variable density jet”, J. Fluid Mech. vol. 249, pp. 619 – 664.

Landau L., 1959, Fluid Mechanics, Pergamon.

Libby P.A., Williams F.A., Turbulent Reacting Flows, 1994, Academic Press, New York.

Lim D.W., Redekopp L.G., 1998, “Absolute instability conditions for variable density, swirling jet flows”, European Journal of Mechanics, B/Fluids, Vol.17, No.2.

Michalke A., 1999, “Absolute inviscid instability of a ring jet with back-flow and swirl”, Eur. J. Mech. B/Fluids, Vol.18, No.1, pp. 3-12.

Miles P., Megerle M., Sick V., Richards K., Nagel Z., Reitz R., 2001, Measurement and Modelling of Large-Scale Flow and Turbulence in a High-Speed Direct-Injection Diesel Engine, Proc. Comodia’01.

Monkewitz P.A., Bechert D.W., Barsikow B., Lehman B., 1990, “Self-excited oscillations and mixing in a heated round jet”, J. Fluid Mech., vol. 213.

Morse, A. P., Whitelaw, J. H. & Yianneskis, M. 1978 Turbulent flow measurement by Laser Doppler Anemometry in a motored reciprocating engine.

Panda J., McLaughlin D.K., 1994, “Experiments on the instabilities of a swirling jet”, Phys. Fluids, Vol.6, pp.263-276.

Peters N., 2002,Turbulent combustion, Cambridge Monographs on Mechanics.

Poinsot T., Veynante D., Theoretical and Numerical Combustion, R.T. Edwards Inc. ,2002

Pope S.B., 1990, Computations of Turbulent Combustion: Progress and Challenges

Riva R., Binder G., Favre-Marinet M., 1990, “Jets of Air-helium Mixtures in Air”, Proceedings of 3rd European Turbulence Conference, Advances in Turbulence III, Springer-Verlag

Saxena V., Pope S.B., 1999, PDF Simulations of Turbulent Combustion Incorporating Detailed Chemistry, Combustion and Flame, 117:340-355

Spalding D.B., 1971, Mixing and Chemical Reaction in Steady Confined Turbulent Flames, Thirteenth Symposium (International) on Combustion, pp.649-657, The Combustion Institute, Pittsburgh.

Sreenivasan K.R., Raghu S., Kyle D., 1989, “Absolute instability in variable density round jets”, Experiments in Fluids, Vol.7, pp.309-317.

Veynante D., Vervisch L., 2002, Turbulent Combustion Modelling, Proegress in Energy and Combustion Science, Vol.28.

Wagner H.G., 1979, Soot formation in combustion, Proc Comb Inst 17:3

Warnatz J., Maas U., Dibble R.W., Combustion-Physical and Chemical Fundamentals, Modelling and Simulations, Expeimnets, Pollutant Formation, Springer, 2001.

Weller H. G., Uslu S. Gosman A.D. Maly R;R,1994., Herweg R., and Heel B., Prediction in homogeneous Charge Spark Ignition Engines, Proc. Comodia'94.

Useful Links

Portals, links to links, symposiums:

• Efluids, a generic portal for fluid dynamics.

• ERCOFTAC www Server, European Research Community On Flow, Turbulence And Combustion. Links to links, events, publications, databases.

• CFD Review, news and articles about CFD.

• CFD Resources Online: An online center for CFD. Services, including a comprehensive link section, a discussion forum and a job database.

• George B. Ross’s Turbulence Links. Not just turbulence links, but also combustion, heat transfer, simulation and experimentation links.

• COMODIA 2004, the Sixth International Symposium on Diagnostics and Modeling of Combustion in Internal Combustion Engines. Papers for COMODIA 85, 90, 94, 98, 01.

CFD & experiments

• CERFACS (Centre Européen de Recherche et de Formation Avancée en Calcul Scientifique) – Combustion modelling bibliography database.

• Professor Gosman’s CFD Group, Large Eddy Simulation (LES) and Combustion.

• Cambridge University - Computational Fluid Dynamics (CFD) Many applications in combustion, turbines, engines.

• Hardware benchmarks.

• Combustion Data: Measurements in Turbulent Nonpremixed Flames.

• Combustion Research: Combustion Diagnostics (Paul Sherrer Institute).

• ERCOFTAC Summer Course on Laser spectroscopy in combustion research (pdf documents).

• CYCLONE Fluid Dynamics - Numerical simulation and experimental verification of DI diesel intake port designs.

• A collaborative framework and a library on experimental and computational Turbulent Non-premixed Flames.

Commercial sites for CFD codes:

CD Adapco (StarCD) FLUENT FIRE

Theory and fundamental research, universities:

• Cranfield University - Computational Flows and Combustion, recent publications.

• University of California @ Berkley- Combustion Modeling Laboratory -Recent and Current Research Topics.

• Annual Research briefs in the Stanford University.

• UC Berkeley: Vitiated Coflow Burner. Data and design information on the VCB.

• Institute for advanced Engineering- Engine Research laboratory. © ERCOFTAC 2004 Acknowledgments

The contributions of these people are graciously recognized:

• Stefan Hohmann / MTU Aero Engines, Heat Transfer and Combustion, Germany

• Andrzej Boguslawski, Stanisław Drobniak / Technical University of Częstochowa, Poland.