MID TERM REPORT CONTRACT N : G4RD-CT PROJECT N : GRD ACRONYM: MUSCLES

Transcription

1 ` MID TERM REPORT CONTRACT N : G4RD-CT PROJECT N : GRD ACRONYM: MUSCLES TITLE: Modelling of UnSteady Combustion in Low Emission Systems PROJECT CO-ORDINATOR: H. T. Brocklehurst, Rolls-Royce plc PARTNERS: Organisation Name Rolls-Royce plc MTU Aero Engines GmbH Fiat Avio S.P.A Snecma Moteurs Turbomeca SA Office National d'etudes et de Recherches Aeronautiques Centre National de la Recherche Scientifique (CORIA) Centre National de la Recherche Scientifique (EM2C) Instituto Superior Tecnico, Lisbon Universita degli Studi di Genova (DIMSET) Centre National de la Recherche Scientifique (LEMTA) Universitat Karlsruhe (ITS) Universitat Karlsruhe (EBI) University of Cambridge (Engineering dept) Loughborough University Dipartimento di Ingegneria Chimica - Universita di Napoli "Federico II" Université de Rouen Institut National des Sciences Appliques de Rouen Country UK D I F F F F F P I F D D UK UK I F F REPORTING PERIOD: FROM 1 June 2002 TO 1 June 2004 PROJECT START DATE: 1 June 2002 DURATION : 42 months Date of issue of this report: October 2004 Project funded by the European Community under the Competitive and Sustainable Growth Programme ( )

2 1.0 TABLE OF CONTENTS 2.0 EXECUTIVE SUMMARY OBJECTIVES AND STRATEGIC ASPECTS SCIENTIFIC AND TECHNICAL ASSESSMENT 4 Work Package 2 Prediction of Unsteady Reacting Flow and Validation 4 Task 2.1 Prediction of Lean Blowout in Liquid Fuelled Combustion 4 Subtask Investigation of the Non-Stationary Two-Phase Flow at Lean Blowout... 4 Subtask Prediction of Lean Blowout in Liquid Fuelled Combustion.. 6 Subtask Numerical Simulation of Industrial Fuel Injection for Lean Stoichiometry.. 8 Task 2.2 Radiation-Vaporisation Interaction.. 8 Task 2.3 Investigation of Unsteadiness for Swirl Premixers.. 10 Subtask Experimental Study on Swirling Premixing Devices. 10 Subtask Numerical Simulation of LPP Swirl Premixers.. 12 Work Package 3 Advanced Diagnostics in Two-Phase Flow Field Task 3.1 Development of Advanced Optical Diagnostics.. 15 Subtask Light Scattering Predictions 15 Subtask Development of Optical Techniques for Simultaneous Characterisation of the Two Phases on Laboratory Experiments.. 16 Task 3.2 Validation and Comparison of Measurement Techniques Performances. 19 Task 3.3 Database Construction for Numerical Models Validation 19 Task 3.4 Fluctuation Analysis of Swirled Kerosene Flame 19 Subtask Design and Manufacture of Swirled Kerosene Flame Test Configuration. 19 Subtask PDPA and Laser Light Scattering Measurements.. 20 Work Package 4 Analysis of Pressure/Acoustic Waves.. 21 Task 4.1 Pressure Waves in Combustion Chambers Task 4.2 Combustor Flow Field Sensitivity to Acoustic Oscillations 22 Task 4.3 Influence of Acoustic Waves on Spray Vaporisation and Combustion 24 Subtask Influence of Acoustic Waves on Spray Vaporisation and Combustion (Numerical Aspects) Subtask Influence of Acoustic Waves on Spray Vaporisation and Combustion (Experimental Part). 27 Subtask Industrial Validation. 29 Task 4.4 Fluctuations Analysis of Single (Multiple) Jet in Under/Super Critical Regimes LIST OF DELIVERABLES PLANS FOR DISSEMINATION AND USE MANAGEMENT AND COORDINATION ASPECTS CONCLUSIONS.. 50

3 2.0 EXECUTIVE SUMMARY All low emission combustor designs are inherently prone to suffer from unstable combustion. In practice it is often this issue that limits the NOx levels attainable in practical gas turbine designs. Hence the understanding and prevention of acoustic oscillation in gas turbine combustors is fundamentally linked to the ability to deliver low NOx and hence a cleaner environment. The project consists of a series of coupled experiments at leading research centres in Europe, where local experts have devised the experiments in detail and developed existing measurement techniques to study the time-varying properties of the flow under investigation. Studies to date outside this program have concentrated on either eliminating problems with specific hardware or in describing a pre-supposed mechanism for coupling the flow pressure variations to the fluctuating heat release rate. This project is different, in that no presumption of the modes of oscillation will be made, and so methods developed should be equally applicable to understanding oscillation issues at all frequencies of interest. At present most models of combustion instability are based on both linearised theory and also rely on the provision of a "flame transfer function". While these are undoubtedly useful, and can be run rapidly, they do suffer from two major flaws - all the physics is wrapped up in the flame transfer function, and thus it is difficult to extrapolate to new geometries and conditions, and the linearised theory, while predicting unstable frequencies well, does not give information about the pressure levels which may be reached. In the current project both these deficiencies are addressed, and a new model of the process will be delivered which can be applied to all gas turbine combustors using partners own CFD codes. Progress to date (the mid term point) has been very good and well aligned with the planned activity. The programme is achieving its stated objectives and delivering useful results that will benefit both the consortium and the wider research community. Initial results show exciting progress in the understanding of unsteady phenomena and it is anticipated that the final period of the project will deliver further significant insights, as the initial period of the project was the dedicated to the design and commissioning of new rigs and measurement equipment. The major milestones achieved by the programme are described below. Prediction of Unsteady Reacting Flow and Validation There have been developments in the joint PDF combustion model, which was extended for highly diluted combustion and has been applied to the simulation of lean blow-out of a combustion system. Initial calculations were unsuccessful, but after refinement of the JPDF tables, lean blow-out was successfully predicted. A detailed measurement program of a fuel injector has shown that the number of droplets generated is strongly influenced by the presence of oscillations, although there are only slight fluctuations in size. The combustion oscillations affect the spray characteristics at other positions within the spray field similarly. A high number of droplets lead to stronger reaction in the combustor, causing air density decreases, which in turn leads to higher air velocities. Advanced Diagnostics in Two-Phase Flow Field There has been good progress in developing state-of-the-art measurements techniques, with their associated data analysis software. This has been facilitated by the close collaboration of the partners concerned and will deliver the facility to measure droplet temperature, including internal temperature profiles, and local vapor concentration simultaneously. Thus far, these techniques have been applied to mono-disperse ethanol and acetone droplets successfully, whilst also establishing any limitations of the techniques being developed. The next period will focus on the extension of the techniques to poly-disperse sprays, thus offering more flexible measurement tools. Analysis of Pressure /Acoustic Waves The first ever DNS of evaporating sprays has been completed and delivered as a resource to test models developed elsewhere in the project. It shows different response of the droplets to turbulence dependent on the droplet size. Measurements of an axial fuel injector under a range of acoustic excitation (50H 1500HZ) has shown that different areas of the fuel injector respond differently to the excitation, although this is the subject of continuing analysis. 3.0 OBJECTIVES AND STRATEGIC ASPECTS Currently, although there have been low emissions technologies developed over the past 10 years through Commission funding, the successful application of these technologies in aircraft engines has been limited. One significant barrier to this take-up is that in nearly all cases low NOx gas turbine combustors have been much more affected by unsteady combustion than is the case for conventional designs. Although pressure oscillations are always present in combustors, the opportunities for amplification are much greater, principally because the sensitivity of the heat release rate to equivalence ratio is large in devices operating near to lean extinction. Industry had developed some semi-empirical tools for describing the feedback process, but these were generally limited to one mode of instability and required empirical input not available at the design stage. Although these tools had been relatively useful in the past, in designing or operating around areas of instability, these were usually observed only at low power or start up and were of nuisance

4 value. Now, however, unsteady combustion is observed potentially at any condition, over a wide range of frequencies, and the magnitude of the fluctuating pressure is such that severe damage to the combustor can occur very rapidly if left to operate in the unsteady mode for more than a few minutes. This poses a severe hazard for the potential flight use of low emissions technology. Clearly, operators, equipment suppliers and customers will not be prepared to accept an environmentally friendly engine which in perceived to be less safe. CFD methods are now becoming available which adequately describe time-evolving combustion flows. It is therefore appropriate to consider how best to describe the unsteady heat release, and how that may be correlated with the fluctuating pressure. In this project, the fundamentals are studied without prescribing the ways in which coupling may occur. Thus, we should develop enough understanding to completely describe the process and hence to model it without empirical input. The object of this project is to build up the understanding of the mechanisms by which pressure waves can influence heat release rates. We will use this understanding to develop theoretical, fundamental models of these processes suitable for inclusion in partners' CFD codes, and to validate these models by comparison with other data generated in the programme. Thus, designs can be assessed and developed before expensive testing, and demonstrated to be free of damaging oscillations. The overall deliverables are methodology reports describing modelling processes, novel measurement techniques for instantaneous spray evaluation, and routines that can be used in partners' CFD codes. This process will enable the takeup of low emission technologies throughout the aerospace sector of the EU, and thus remove a significant barrier to the improvement of air quality. Not only is there potential for improvements in the environment, but the capability to design oscillation free combustion systems gives EU industry a competitive advantage over the US, and ensures the continuing growth of our aerospace industries. The project is well placed to deliver its goals and work is progressing as planned. 4.0 SCIENTIFIC AND TECHNICAL ASSESSMENT WORK PACKAGE 2 - PREDICTION OF UNSTEADY REACTING FLOW AND VALIDATION Task 2.1 Prediction of Lean Blowout in Liquid Fuelled Combustion Subtask Investigation of the non-stationary two-phase flow at lean blowout (ITS Karlsruhe) Work Completed The objective of subtask is the experimental investigation of the unsteady two-phase flow in the presence of combustion oscillations. The experimental data will be used by other subtasks within work-package 2 to verify and validate numerical methods for predicting two-phase flows under unsteady flow situations. For the present subtask, an atomizer was selected which has been studied at the Institut für Thermische Strömungsmaschinen previously. It is a prefilmer design, consisting of a hollow cone pressure atomizer located at the centre. The cone angle of the spray is 80. The air is fed into the nozzle through 2 passages, both equipped with swirling vanes. The droplets produced by the hollow cone pressure atomizer are deposited on the pre-filming lip leading to a build-up of liquid film. This film is driven by the primary airflow to the atomizing edge, where the final break up into droplet takes place. The secondary airflow is contra-rotating to the primary one, and assists the atomization of the liquid sheet. The swirl number of the primary air is 0.45 and that of the secondary air Two different designs of the atomizer were used for the experiments; design 1 of the nozzle features a round diffuser geometry, design 2 has a sharp edge. The atomizer was mounted into a combustor with a cylindrical cross section. The inner diameter of the combustor is 100 mm. For optical access, the first section of the combustor consists either of a quartz glass tube or a metallic housing with plane windows (LDA, PDA). The characteristics of the atomizer have been studied under non-reacting conditions and documented in deliverable report D2.1. Also, the investigation of the reacting two-phase flow under stable conditions has been reported in deliverable report D2.2. The investigations up to month 12 revealed that under oscillating flow conditions with reaction unsteady flow patterns are present which strongly disturb the atomization process of the nozzle. The objective of this

5 task therefore was modified in accordance with the other partners at the expert group meeting in June 2003 (Rouen). Thus, it is focused on what causes these flow instabilities and how they interact with the atomization process. For the deliverable report D2.3, the reacting two-phase flow under oscillating condition has been investigated. Figure 1 Phase Discriminated Distribution of Number of Particles per Phase Angle Figure 1 shows the number of detected particles per phase angle. It can be seen that the number does strongly vary. The maximum number of detected droplets at each position (A-C in the main spray, D-F on the spray axis) is up to 50 times higher as the minimum number of droplets. Therefore, the atomization process in terms of the number of generated droplets is strongly influenced by the presence of oscillations. The combustion oscillations affect the spray characteristics at other positions within the spray field similarly. A high number of droplets leads to stronger reaction in the combustor. As consequence, the air density decreases, this in turn leads to higher air velocities. Figure 2 shows the droplet velocity over the phase angle. It is obvious that the mean axial velocities e.g. at position A (x=14mm, r=30mm) oscillate between 40 m/s and 65 m/s. The maximum velocities are reached when the maximum number of particles is detected. Because of the high velocities at the first measurement plane (Position A ), it can be concluded that the velocities at the prefilmer must be high enough to produce a fine spray. Even if the minimum velocity is reached, the shear forces on the liquid fuel film are high enough to produce small droplets. Thus, oscillations of the mean axial velocity have almost no influence on the mean droplet size. Figure 3 shows the arithmetic mean diameter D 10 as function of the phase angle. As expected, the mean droplet size at position A does vary just slightly between 13 µm and 18 µm. Figure 2 Phase-Locked Mean Axial Velocity Fluctuations

6 Figure 3 Phase-Locked droplet Diameter Fluctuation The investigations showed, that under oscillating flow conditions with reaction unsteady flow pattern are present which strongly influence the atomization process of the nozzle. Therefore, extensive PDA measurements downstream of the prefilming nozzle have been performed under oscillating combustion condition. The phase-locked PDA data revealed that there are only slight fluctuations of the droplet size. However, the number of droplets produced as well as the droplet velocity is significantly affected by combustion oscillations. Subtask Prediction of lean blow out in liquid fuelled combustion) (Karlsruhe University Engler-Bunte-Institute) Work Completed The aims are investigations of the influence of turbulent mixing and reaction on lean blow-out characteristics. The 2-domain-1-step kinetic model was extended for highly diluted combustion. It has been shown that reaction pathways for fuel-consumption- and CO-oxidation-domain can be decoupled and that the thermo-chemical state of mixtures can be determined by one single reaction progress variable. Comparing the 2-domain-1-step kinetic to a detailed chemical reaction mechanism shows that laminar flame speeds are reproduced satisfactorily (Figure 4) and that the 2-domain-1-step kinetic is capable of calculating concentration profiles and temperature profiles (Figure 5). Using one single reaction progress variable makes it possible to implement this kinetic into presumed PDF models without making further assumptions concerning the statistical independence of multiple progress variables. Figure 4 Comparison of laminar flame speeds calculated using detailed chemical reaction mechanism and 2- domain-1-step kinetic for different preheating temperatures at 1bar.

7 Figure 5 Temperature and concentration profiles as a function of the flame position for a fuel equivalence ratio of 1.0 and a preheating temperature of 498K at 4bar. For testing the JPDF reaction model the matrix burner test case was selected, which was developed for systematically analysing stationary highly turbulent premixed free jet flames. This first test case has shown that the JPDF-model applying the 2-domain-1-step kinetic is capable of reproducing flow field measurements of the selected turbulent flameconfigurations with good accuracy (Figure 6). Figure 6 Comparison of the JPDF-Calculation (Simulation) with experimental data for temperature CH4- and CO-concentration. As a test case for LBO-calculations a co-swirling airblast atomizer application has been chosen, for which 3D LDVmeasurements of turbulent flow and mixing fields are available within the Engler-Bunte-Institute. The computational grid was generated from geometry data of the combustor, including the airblast atomizer nozzle and boundary conditions were derived from experimental settings. Once the turbulent flow field was calculated the computation of the mixing field using the JPDF model, which is shown in Figure 7 for a Schmidt-No. (Sc) = 0.2. Figure 7 Comparison of the JPDF-Calculation (Simulation Sc=0.2) with experimental data for mixture fraction.

8 Next, the 2-domain-1-step kinetic scheme was improved to cover the preheating temperature of 373K of the reacting test case selected, a co-swirling airblast atomizer application. This new kinetic scheme was validated by means of laminar premixed flame calculations and homogeneous reactor calculations and then implemented into the JPDF model. The results of the CFD calculations have shown that the calculated JPDF table was not capable of providing data adequate for the near blow out case. Therefore the PrePDF software has been improved for generating adaptive JPDF tables with higher refinement. Using this it was possible to simulate lean-blow out by raising the thermal load of the combustion system (see Figure 8), even if the combustion system was considered to be adiabatic. Figure 8 CFD-calculation of stable combustion (P=29kW) and blown-out case (P=70kW) 3.0 Subtask Numerical Simulation of Industrial Fuel Injector for Lean Stoichiometry (MTU) Work Completed Technical activity not started yet. Task 2.2 Radiation - Vaporisation Interaction IST Lisbon. 4.0 Work Completed The work proposed by IST within the MUSCLES project is mainly experimental and involves the design and manufacture of a test rig, related initial tests and a detailed set of measurements of droplet vaporisation rates for different liquid fuels. An experimental test rig was designed and manufactured and IST (deliverable 2.10). A literature survey was provided with the objective of gathering existing data regarding previous experimental work on droplet vaporisation. This survey includes the analysis on the type of liquid fuels used, test rig geometry identification, experimental test conditions (temperature, pressure, initial droplet diameter etc.) amongst others. The experimental results are to be compared with the results already selected from the literature survey (deliverable 2.11). The experimental apparatus was tested and calibrated. Improvements were introduced and some preliminary tests were performed concerning the initial droplet diameter, flow and temperature characterisation. Initially, the test rig was tested and the temperature distribution inside the cylindrical tube was measured at maximum heat output (T w = 973 K). The main objectives were to check the experimental apparatus itself and test for possible asymmetries in temperature distributions. The results showed that the mean temperature in the central region of the reactor could be considered to be stable and within measurement errors. The next step included the characterisation of the nitrogen flow inside the reactor and the calibration of the N 2 flow meter. The mean axial velocity in the reactor s centerline was measured with a hot-wire anemometer. The initial drop diameter, D 0, was also measured using water and three needles of 0.5, 0.8 and 1.2 mm in diameter at ambient temperature. Measurements were also performed to assess droplet residence times. Figure 9 compares the drop residence time for isothermal conditions and at maximum heat load, set point of 973 K, for a nitrogen flow rate of 17.5 l/min (292 ml/s). More detailed results (other operating conditions) and a complete description of the experimental set-up (diagrams and detailed views) as well as of the experimental techniques is provided in deliverable 2.12.

9 Following the preliminary tests for drop initial diameter measurements using water further experiments were carried out using n-heptane (needle diameter = 0.7 mm). The N 2 flow rates and the temperature range were kept constant (nheptane = 50 ml/h, N 2 = 17.5 l.min). Sample results are presented in Figure 10. To improve the future results, which are highly dependent on the quality of the optical system and light source, the test rig was moved to another laboratory in order to allow the use of a more powerful laser source. 450 Residence Time (ms) WITH N2 T = 300 K T = 973 K Axial Location (mm) Figure 9 Drop residence time (water) in the presence of N2 Figure 10 Drop diameter in the presence of N2 For further tests with n-heptane a droplet generator was used, instead of the initially used needle injection system. The initial drop size was determined as a function of the liquid (water) flow rate for different frequencies of the drop generator with a pinhole of 100 µm and the results are presented in Table 1. Experiments with n-heptane were performed with a 50 µm pinhole for different heat fluxes in the presence of CO 2 for the operating conditions shown in Table 2. Sample results are presented in Figure 11 and Figure 12, which show the contribution of the radiative term, respectively, for the vaporisation of n-heptane drops. Table 3 summarises the contribution of the radiative term for these experimental conditions (deliverable 2.12 and status of deliverable 2.14). Based on these results the vaporization constant was determined, as shown in Table 4. Figure 11 Drop diameter evolution for n-heptane for different radiative fluxes as a function of time

10 Figure 12 Drop diameter evolution for n-heptane for different convective conditions as a function of time Frequency Flow rate (ml/h) Table 1 Intial Drop Diameter (in µm) for the droplet generator (pinhole = 100 µm) Temperature (K) U (m/s) n-heptane (ml/h) Frequency generator Pinhole (µm) Off Off Off Off 50 Table 2 Operating Conditions T w (K) U (m/s) Q r /Q T (%) Table 3 Radiative term contribution for the different experimental conditions T w (K) U (m/s) K (m 2 /s) T d (ms) Table 4 Vaporisation Constant (K) for the different experimental conditions Following discussions with the project coordinator and MUSCLES partners the IST deliverables have been amended from the original contract, and are shown in the latest version of the Technical Annex and section 5.0. Due to delays introduced by the need to build new experimental set-up as described in section 5.0 the proposed work is running approximately 4 months late. No further delays are forseen and the next deliverables are expected to be on time. A reference database is being built which includes the Table of Contents from several scientific journals in this field. It is continuously updated and is already available at the MUSCLES project website ( Task 2.3 Investigation of Unsteadiness for Swirl Premixers SubTask Experimental study on swirling premixing devices

11 University of Genoa & Avio Group Work Completed During the first year the assembly of the test rig was achieved. The air fan, the electrical air preheater, and the control instrumentation have been installed and commissioned. The presence of possible velocity fluctuations produced within the test-rig itself was investigated and this demonstrated that the test-rig was fully operational. The expected functional parameters, in particular the airflow rate and preheating temperature, were achieved. In months the research activity concentrated on the set-up of and testing of LDA, PDA and PIV systems and defining the plan of measurements. In order to investigate the airflow within the premixing duct and the flame tube without the presence of the spray, the airflow needs to be seeded and the flow field is measured with LDA (PDA). For the characterisation of the spray the use of both PDA and PIV systems was planned. PDA was used to characterise the nozzle spray alone in terms of spatial distribution of droplets and their characteristic diameters and the PIV system can be used to determine the spray velocity field. For unsteady measurements an air-cooled pressure transducer was utilised as trigger signal. In the period from month 19 to month 24 the research activity concentrated on the experimental investigation of the Avio LPP injector, in order to understand effect of the spray on the unsteady airflow. First of all the fuel spray was characterised to determine the diameters of the droplets and the shape of the spray as generated by the nozzle. After that some traverses were used to determine the air velocity field and to detect the frequencies of the velocity fluctuations. With an air temperature of 293 K a characteristic frequency of 200 Hz was found; this frequency shifted to 270 Hz when the air temperature was increased up to 451 K. These preliminary measurements demonstrated that a pressure transducer was suitable to generate a trigger signal for the phase-averaged acquisitions. The in phase measurements used to detect characteristic fluctuations of the velocity in the air flow field, which have strong effect on the droplet formation, showed that the diameter of the generated droplets fluctuates at the same frequency (see deliverable D2.17 for details). For understanding the effects of the fuel injection on the air flow field, a PDA system performing simultaneous measurements of the air flow and fuel spray was utilised. After the acquisition it was possible to separate the information related to the airflow (seeded with 4 microns droplets) from the information related to the fuel spray (characterised by 30 microns droplets) by filtering the data on the basis of the droplet diameters. Then comparing the airflow field obtained in such a way with the airflow field measured without fuel injection it was possible to determine the effect of the fuel injection. The Particle Image Velocimetry was used for determining the spatial distribution and the velocity field of the fuel droplets in the meridional plane and in the frontal plane. These measurements demonstrate the periodic characteristics of the velocity field and show that the injected fuel impinges on the premixing duct walls, generating a film of fuel that at the end of the premixing duct produces droplets characterised by high values of the diameters. Figure 13 PDA Measurements for the Characterisation of the Fuel Spray

12 Figure 14 PIV Measurements for the Characterisation of the Fuel Flow Field The configuration of a premixing duct with two radial swirlers was defined for the tests. It was the duct originally designed for LOPOCOTEP small engines combustor. The configuration was already tested during the project ICLEAC but only with single-phase flow conditions. In the Project MUSCLES tests with liquid fuel have been performed. The liquid fuel was chosen in order to have the same behaviour of kerosene at real working conditions of the duct in the combustor. In the last 6 months new blades for the radial swirlers were defined. The original configuration had co-rotating swirlers. The new one has contra-rotating swirlers with the same swirl number. CFD analyses were performed in order to assess and optimise the flow field and the evaporation rate of the new configuration. This will be manufactured and tested in the next future and its behaviour will be compared with the old one. SubTask Numerical Simulation of LPP swirl premixers University Of Genoa Work Completed 1. Fuel Spray Dynamics Modelisation There has been development, implementation, and calibration of a sophisticated, organic set of self-complementing two-phase flow models suitable to increase the fuel-spray behaviour prediction capabilities of solver NastComb. More in detail, the NastComb's previously available TAB ("Taylor Analogy Break-up") spray-model, has been complemented by a new model named RT-KH ("Rayleigh Taylor - Kelvin Helmholtz"), rooted on instability analyses performed at the inter-phase between the liquid and the gas. Kelvin Helmholtz instabilities are considered as responsible for the droplets primary break-up whereas the Rayleigh Taylor instability governs, together with the former, the secondary break-up. The two models, introduced into NastComb, have been compared in connection with a few reference test situations in order to have guidance toward the respective calibrations. Figure 15 presents an example of comparisons among the prediction capabilities of the different models. TAB Model SMR= 43 micron Droplet N = 6000 RTKH model SMR= 25 micron Droplet N = Figure 15 Comparisons of the Results of Different Models 2. LPP-System Parametric Optimisation Parametric application of NastComb to several geometrical and functional configurations of the Avio-Group LPP swirl premixer, in order to achieve its optimisation in terms of: complete flow stability, adequate levels of fuel

13 prevaporisation and air-fuel premixing in the outlet section, combined with an outlet swirl number of order 0.5. The optimisation strategy, involving many successive fully 3D flow-field predictions, both in single- and in two-phase flow situations, has delivered the final configuration that was then manufactured and has undergone extensive experimental testing. This activity has produced the Deliverable D 2.19 transmitted, as required, at Month Premixed-Combustion Modelling Experimental Validation In order to help validate the time-dependent, fully-reactive predictions a laboratory combustion-test was performed at UNIGE-DIMSET, wherein a rapid-mix combustion system, made up of a radial swirl-premixer followed by a cylindrical combustor, has undergone detailed experimental testing. The cross comparisons between the experimental temperature distribution within the cylindrical combustion chamber and NastComb predictions have turned out at all positive (Figure 16). The tests are proceeding in order to cross compare predictions with the corresponding experimental data in specific connection with the flame process stability as well as the emissions. C B A C Radial position (Section A) Experimental data NastComb C C Radial position (Section B) 1000 Radial position (Section C) Experimental data NastComb Experimental data NastComb Figure 16 Comparison of Measurements and Experimental Data 4. Unsteady Two-Phase Numerical/Experimental Comparisons in the LPP System A first important outcome of all the numerical investigations has been the clear evidence that no steady conditions could be achieved even for significant parametric variations both in the geometrical and the functional parameters of the premixer, pointing out an intrinsic fluid dynamic instability of the swirling flow, of periodically "snaking" behaviour, showing frequencies ranging from 180 to 260 Hz, of the same order as the experimentally measured. Interesting is also the observation, coming from the numerical analyses performed in parametric mode, that the said instability showed a marked increase in its intensity depending on the length the discharge chamber as well as of the increase of the mass flow in the inner swirler. All the detailed time-dependent results of the numerical investigations performed in unreactive mode (enlarged scale model) are presented in the required Deliverable D2.20, due at month 24, will be delivered before the mid-term meeting. Figure 17, Figure 18 and Figure 19 show the locations of the measuring traverses, the predicted unsteady trajectories of the droplets, as well as the numerical- experimental cross comparisons with reference to the averaged radial distributions of the droplets diameters. In Figure 19, the accuracy of the predictions is remarkably good, which has resulted from the imposition of droplets elastic rebounds from the premixer walls in correspondence of the points of their impingements. No such precision could be obtained by excluding wall rebounds and imposing the condition of droplets adherence to the walls, with corresponding formation of a liquid boundary layer.

14 Figure 17 Locations of measuring traverses Figure 18 Unsteady Droplets Trajectories with Wall Rebounds D32 [um] D32 [um] Traverse 1 Experimental NastComb prediction radial distance [mm] Traverse 2 Experimental NastComb prediction radial distance [mm] D32 [um] D32 [um] Traverse 3 Experimental NastComb prediction radial distance [mm] Traverse 4 Experimental NastComb prediction radial distance [mm] Figure 19 Averaged radial distributions of the droplets diameters: 5. Unsteady, fully reactive, numerical predictions within the real-scale prototype Recently, a preliminary unsteady numerical investigation has been performed, in fully reactive conditions, within and downstream of the real-scale LPP prototype, at operating pressures, with kerosene fuel. The reactive scheme adopted was an extended partial oxidation mechanism (EPOM, 12 species, 28 reactions), but, very recently, implementation in NastComb of an advanced, detailed-chemistry mechanism (ADCM, 68 species, 260 reactions), keeping fully unsteady interaction with turbulence and radiation, has been successfully performed. Contrary to expectations, the intensity of pulsations appeared as somehow decreasing, for the same air mass flow, entering the reactive conditions with respect to the unreactive one. As a first outcome, it can thus be stated that the basic fluid-dynamic instability of the swirling jet is still present but it does not link with the chemical kinetic process so that no significant thermo-acoustical waves (humming) are produced. Of course, these results are preliminary and new evidence, possibly coming from the adoption of ADCM mechanism, could point toward different conclusions. In Figure 20 a time-dependent picture (one out of a sequence of about 500) of the temperature distribution in the overall combustion system (real scale, real pressure) is given.

15 Figure 20 Preliminary reactive calculation (time-dependent temperature distribution) To be noticed as a potentially dangerous situation, the tendency of the flame front to protrude back into the premixer duct, with risk of inducing both unsustainable wall thermal stresses as well as metal surface damage. Flame flash-back risk appears also present. 6. Next six months activities According to work-plan, after the above positive cross comparisons in unreactive mode, during the next six months the theoretical/numerical investigations will proceed with a progressive emphasis into the fully time-dependent reactive calculations within the real scale Avio-Group prototype. WORK PACKAGE 3 - ADVANCED DIAGNOSTICS IN TWO-PHASE FLOW FIELD. TASK 3.1 Development of Advanced Optical Diagnostics Subtask Light Scattering Prediction (CORIA) For the 12 first months of the programme the tasks have been to develop Lorenz-Mie codes and Geometrical optics code to predict the interaction between the light and a spherical object. Two main kinds of interactions have been studied: Prediction of the internal field to be able to predict the fluorescence (experiment of Nancy group) Prediction of the far field to be able to predict the rainbow and forward scattering (experiment of Toulouse group). A particular effort has been devoted to the user-friendly aspect of such codes. The codes corresponding to task 1 (homogeneous particle) and task 2 (multi-layered particles) have been delivered. Codes for the homogeneous particle and for multi-layered spheres have been delivered to the partners developing fluorescence and rainbow techniques. Furthermore, to answer a demand from the Nancy group, a GLMT code for a LDV configuration (two interfering beams at different angles) has been generated and delivered. This is an extra code beyond the promised deliverables. However because of the generic nature of the work done it is possible to provide other specific configurations that partners may require. The figure below shows the user-friendly input screen for one of the codes and a typical output plot of the intensity distribution. Other examples are shown on the project web site. This completes all four planned deliverables for the first year on time, which were needed by the experimental partners to conduct their test programme. Figure 21 Example of an input screen and of an internal field. More recent activity was mainly dedicated to the description of fluorescence emission from micro-droplets. The fine description of the droplet illumination have been carried of during the 12 firsts months of the program. In particular,

16 micro-droplets internal fields created by an arbitrary located focused beam are able to be computed with different software developed during this period and based on a rigorous resolution of Maxwell s equations (GLMT codes) or geometrical optics (GO codes). Two complementary approaches are now under development to take account and take advantage of the emission process. In an homogenous medium, the fluorescing process is isotropic with a relatively smooth spectral distribution. On the contrary, when fluorescing occurs from micro-droplets the emission is directive and the spectral distribution let appear resonance peaks connected with the droplet size, refractive index and shape (Morphologically Dependant Resonances). A ray tracing approach has been used and coupled with the GLMT codes to complete the process, from excitation to detection. This approach allows one to describe the directivity of the fluorescence emission. The codes are now written but their validation is still progressing. To describe and take advantage of MDRs, we need another approach, based on an electromagnetic description of the micro-cavities properties. The possibility of droplet characterization in terms of size and refractive index from a fluorescing spectrum has been formally and numerically demonstrated. We are now working on an experimental set up to measure such spectrum with enough accuracy to detect MDRs and to test our inversion strategy. Subtask Development of Optical Techniques for Simultaneous Characterisation of the Two Phases on Laboratory Experiments (LEMTA and ONERA) Work Completed 1. The full validation of the two colours laser fluorescence, developed in LEMTA in order to perform combusting droplets temperature measurements, has been completed. The tests of the technique in combustion have been realized on a combusting mono-disperse droplet stream for droplet diameters ranging from 106 µm to 223 µm. Variations of the distance parameter, which influences strongly the combustion rate, have also been considered. It has been shown, that the fluorescent tracer (rhodamine B) added in the fuel, ethanol in the present case, must be in a basic solution, in order to overcome the problem of ph variations of the liquid phase during the combustion process. The deliverable report D3.6, dealing with this allows one using the technique to construct a database of the temperature evolutions of combusting droplets. 2. The implementation of the measurement of the internal temperature field of combusting droplets has been completed. The technique is based on an extension of two colours laser-induced fluorescence, which eliminates the influence of the laser excitation volume, tracer concentration and local laser intensity distribution. Both the experimental set up and associated processing software have been developed. A laser probe volume of about 50 µm length and 20 µm in diameter is created to scan the droplets. The periodicity and reproducibility of the phenomenon allows scanning progressively and at different instants successive slices of the droplets. This processing software is linked to a geometric optics model, in order to position properly the measurement points into the droplets. The grid used for the implementation of the geometrical optics model in the droplet can be easily re-used to implement the models developed in CORIA (Rouen), based on Global Lorentz Mie Theory (GLMT). Both experimental results obtained by the geometrical optics and GLMT approaches are in good agreement. An example of temperature map of a combusting droplet is given in Figure 22. The results suggest a non spherical heat transfer scheme and agree well with numerical simulations of internal droplet heat transport by internal fluid circulations, combined with pure conduction. The corresponding deliverable report (D3.7) was provided on time (month 21).

17 Temperature ( C) Figure 22 Temperature map within a combusting droplet (t=9.6 ms after injection, droplet diameter: 216 mm) 3. The optical set-up related to the implementation of the two colours laser-induced fluorescence technique in acetone droplets has been designed and validated; an initial study has provided several spectra of rhodamine B dissolved in the liquid acetone in order to select the adequate spectral bands of detection. The conclusions and recommendations have been delivered in month 12 (deliverable D3.10 and milestone M3.2). Qualification tests for both evaporating and combusting mono-disperse acetone droplets have been successfully carried out. The temperature measurement set-up of acetone droplets (using two colours LIF) has been combined with the PLIF measurement set-up of the concentration of acetone vapour in the vicinity of the droplet stream (developed in cooperation with ONERA). The experiments have started in ONERA (Palaiseau) in January 2004 and the first commissioning tests of the combined set-up are successful. 4. The set-up combining ethanol droplets temperature measurements by two colours LIF and droplet diameter measurements by PDA has been developed and validated. The measurement system uses one excitation source and two detection chains: one for the fluorescence signal, the other for the laser scattering reception. The PDA parameters have been optimised in order to measure a size variation, since the goal is to estimate the evaporation flux by measuring the diameter evolution of mono-disperse droplets. The droplets heating is measured in parallel, which makes it possible to have a quantitative evolution of the droplet enthalpy and also to take into account the variation of the fuel refractive index, used for the PDA measurement, as a function of the fuel temperature. Qualification tests on combusting ethanol droplets have been successfully achieved (deliverable D3.8). 5. At ONERA the same mono-disperse droplet stream is used, with a focused laser beam and CCD cameras positioned at the three preferential diffusion angles. These cameras record the intensity of the scattered signal whose characteristics can be post-processed. An Infrared measurement device will be added to quantify the droplet s surface temperature. This method is based on the measurement of the infrared emission of the droplets in the 8 12 µm wavelength range. As shown in Figure 23, the IR flux Φd emitted by droplets is focused on the detector by an Infrared lens. The analysis of the resulting signal (Figure 23) gives an estimation of the droplet flux. Ambient Monodisperse droplet stream Detector s angle of view Chopper with mirrors Infrared detector Optics Ambient Φamb(λ,Tamb) Φd(λ,Td) Φeb(λ,Teb) Φd refl-amb (λ,tamb) Φeb refl-amb (λ,tamb) Extended blackbody Figure 23 Radiative flux sources present in the experimental configuration and detector signal, as seen on the scope. 6. Measurements with Infrared (IR) and Laser Induced Fluorescence (LIF) techniques

18 Te mp56 era54 tur 52 e [ 50 C] Mean temperature (LIF) Surface temperature (IR) Times [s] Figure 24 Time evolution of surface and mean ethanol droplet temperature Mono-disperse ethanol droplets are injected above the ambient temperature to observe the downstream droplet cooling. The droplet s surface temperature, measured by IR, and its mean volumetric temperature, measured by LIF, are plotted in Figure 24. Both decrease as the distance from the injection point increases. For such a droplet size, the heat transfer from the droplet surface to the droplet inner layers is mainly due to convective motion. The cold area of the droplet head is transported to the centre; the cooling of the inner layers occurs before the surface is affected. The IR measurement focuses on the droplet s side surface, whereas the LIF averages the temperature in the colder inner layers. This explains the difference between the IR and LIF temperatures. The surface temperature is constantly above the mean temperature. This can be explained by the following observations: the droplet cooling is mainly due to the evaporation at the surface the convective heat transfer contributes to the global cooling, but its effects are limited by the hot gas layer surrounding the jet for aerodynamic reasons, the vapour concentration is quickly depleted in the droplet head, which tends to enhance the local evaporation, and the subsequent cooling of this area is more important. 7. Comparison between rainbow and LIF measurements In the following graph (Figure 25) two temperature measurement techniques are compared: rainbow (droplet is assumed to be homogeneous) and LIF, which gives the average volumetric temperature of a droplet. Figure 25 Comparison between LIF and Rainbow Measurement A difference appears clearly during the first phase of the heating, when the thermal gradient is most important. This denotes a high sensitivity of the rainbow technique to the thermal gradient. 8. Droplet with a thermal gradient At the present stage of development, approximate thermal gradients inside a droplet can only be estimated due to the need for improved accuracy in the measurement of droplet temperature, diameter and thermal gradient. The trends of the results from the approximate method (Figure 26) agree with those of several evaporation models. These preliminary results meet our expectations from a qualitative point of view and motivate the present research.

19 thermal gradient deduced from the LIF temperature difference (in C) time of heating (in s) angular difference angular difference angular difference Figure 26 Thermal gradient deduced from a comparison between rainbow and LIF Subtask Extension in Polydisperse Sprays (LEMTA) Work Completed The goal is to extend the two colours laser induced fluorescence, applied to droplet temperature measurement, already implemented for mono-disperse droplets in linear stream, in the case of polydisperse sprays. In first, the spray test bench has been constructed: mechanical injection is used and droplets around 30-µm diameter are generated, with a spray half angle of about 15. The fuel, ethanol, can be heated before being injected. The fluorescence response in the spray has been measured, but the main difficulty is to take into account the multi-scattering effects of the fluorescence signal, i.e. the wavelength dispersion of the fluorescence emission in one point of the spray, interacting with the surrounding droplets, between the emission point and the detector surface. A third spectral band of detection must be used in order to make an in-situ correction of the fluorescence multi-scattering phenomena. The technique (hardware and correction process) has been developed on a water spray and first results have been obtained on an evaporating ethanol spray. Task 3.2 Validation and Comparison of Measurement Techniques Performances The goal is to compare droplet temperature measurements performed with the use of two colours LIF (LEMTA) and rainbow thermometry (ONERA/DMAE, Toulouse). Common experimental points have been defined and are under investigations in both research centres. For the case of non-burning droplets, the two colours LIF and rainbow thermometry results are in good agreement: the temperature gradient within the droplets is low. For the case of combusting droplets, the non-uniformity of the temperature gradient and subsequently of the refractive index inside the droplet is significant and the rainbow thermometry leads to erroneous measurements in such a case. Limitations and range of use of both techniques will be defined. Task 3.3 Database Construction for Numerical Models Validation The workplan for this substask was defined in cooperation with Turbomeca (milestone M3.4 achieved in month 18). The main outlines were: 1. combusting ethanol droplet stream: simultaneous droplet diameter and temperature measurements, for various inter droplet distance (LEMTA task), with additional gas phase temperature characterisation (ONERA task ) 2. combusting acetone droplet stream: simultaneous droplet diameter and temperature measurements (LEMTA task), with additional gas phase temperature characterisation 3. acetone droplet stream in heated air: simultaneous droplet diameter and temperature measurements, with acetone vapour characterisation (combined ONERA+LEMTA task on the same test bench, in relation with subtask 3.1.2, point 3) All these actions are in progress, and no problems are expected to deliver the database at time (month 36). Task 3.4 Fluctuation Analysis of Swirled Kerosene Flame Subtask Design and Manufacture of Swirled Kerosene Flame Test Configuration (University of Naples & Avio Group) Work Completed The aim of this task was the design of a premixer-atomizer, which could be adapted to an experimental facility

20 equipped with advanced optical diagnostics based on single point measurements. Initially a facility that was already used for studying industrial furnaces was adapted to take a swirl-stabilized combustor configuration. In addition, easy access for the probe as well as optical measurement techniques had to be provided at each location inside the combustor. The combustion chamber was designed in modular concept providing ready exchange of the swirlers and fuel-injection systems by mounting different bottom plane sections with integrated atomizers. The rig had to accept LPP and LP injectors from Avio Group plus a modular generic injector. It is planned to use the same premixing duct for the tests at University of Genoa. For security reasons and to adapt the LPP duct to the capability of Naples test rig scaled, conditions were defined in order to get a working point with the same flow number of the real conditions. In this subtask, UNINAP worked in collaboration with Avio Group on a burner of new generation as a result of a previous contract call LOPOCOTEC. A new configuration of the LPP duct with different height of the blades in order to assure the same conditions of kinetic energy of the gas phase at fuel injection was designed in collaboration with the AVIO Group designers. The original burner is presented in Figure 27. The main modification concerns the height of the air duct blades (red colour on Figure 27). This lead to the final version of the test burner that is now ready and is under construction (Figure 28). The design phase and the detailed drawings of the ducts have been completed. The test article, that is the premixing duct, the bladings and the feeding chamber have been manufactured by AVIO and provided to IRC. Modified blades Modified nozzle place Original burner design (courtesy of Avio Group). Revisedl burner design (courtesy of Avio Group). Figure 27 Original and Revised Burner Designs Figure 28 External view of the final design of the burner (courtesy of Avio Group). Another aspect of the study concerned the choice of the injector series that will be tested on the burner. According to the habits of AVIO Group, it was considered to use the DELAVAN spray nozzles with different spray angle and flow rates. Commercial sets of hollow cone nozzle were characterized in terms of droplet size and concentration distribution function and velocities. Subtask PDPA and Laser Light Scattering Measurements (University of Naples) Work Completed The work described in task will be performed in the LPP duct burner designed in collaboration with AVIO. This burner has now been manufactured and delivered to IRC Naples. Instruments and optical diagnostics are ready to be set

21 up around the burner. This schedule in manufacturing this burner should not influence the task program. The activities have been started in any case at month 13. They were focused on calibration and commissioning of some ancillary devices. The original and the modified configurations will be tested in the same working conditions at IRC Naples that is participating as a sub-contractor of Avio. The tests at IRC Naples will start in September. WORK PACKAGE 4 ANALYSIS OF PRESSURE/ACOUSTIC WAVES Task 4.1 Pressure Waves in Combustion Chambers (Cambridge University) Work Completed The main objective of this task is to investigate experimentally the flame transfer function at very high amplitudes of acoustic excitation and to incorporate the measured experimental results to a nonlinear model. This model will be used to predict the frequencies and amplitudes arising from unsteady combustion in an aero engine. Small-scale spray flame rig A small scale swirl stabilised spray flame (< 5kW) with capability to modulate the air has been successfully designed and tested. Flame transfer measurements in this experimental rig have been measured. The flame transfer function for various frequencies is shown in Figure 29, with details given in Deliverable 4.4. Presently, Phase Doppler anemometry is being setup to measure the droplet size and droplet velocities. Larger-scale rig Initial experiments were performed with an existing half-scale Rolls-Royce industrial RB211 LPP burner by injecting kerosene through the gas fuel bars. A mini combustor placed upstream of the modified burner was used as a pre-heater. It was run lean and this supplied preheated vitiated air for partially vaporising the kerosene. A kerosene flame was successfully established in the burner at high flow rates. Images of the flame were acquired using an ICCD camera to look at the CH * chemiluminescence distribution. Figure 30(a) shows the 2D projection of the flame. An Abel transform was used to convert the 2D projection into a 3D axisymmetric flow field. Figure 30 (b) shows a slice through the 3D flow field. However this is not being pursued as it was found that this form of injecting kerosene from the fuel bars led to a flame which was not premixed due to a lack of oxygen in the vitiated supply. High-amplitude forcing using the existing siren was tested on the same LPP burner with ethylene as the fuel and at ambient inlet temperature. It was possible to achieve forced amplitudes of acoustic velocity at the fuel injection plane as high as 60% of the mean axial velocity at certain frequencies. The amplitude in a typical self-excited oscillation is 20-30% of the mean. Flame response to this unsteady forcing has been investigated by measuring the OH chemiluminscence with a photomultiplier and from this a transfer function deduced. This is normalised by the value at low forcing levels to give the transfer function gain shown in the Figure 31, for three different forcing frequencies. It can be seen that the response is linear even at very high acoustic velocity amplitudes. Further insight into the flame response will be obtained by analysing phased-averaged images of CH * chemiluminescence at various forcing amplitudes. A swirl stabilised spray burner at high flow rate has been designed with a pressure atomiser positioned on the centre line using the RB211 LPP burner as a basis. Figure 32 shows a cross section of the burner and shows that the design is similar to the burners used by other partners. A new fuel delivery system, which is air cooled, has also been designed. The new heater to preheat the inlet air to high temperatures (up to 450 C) is currently being commissioned. A new siren has been designed which will siren to impose large amplitude acoustic velocity perturbations has been redesigned while allowing for the higher temperature inlet air. Currently, the manufacture of the redesigned burner is in progress. Work items for the next six months The tasks for the next six months include the measurement of the droplet size and velocities of the swirl stabilized flame. The nonlinear response of the flame at high flow rate will be measured and incorporated into the nonlinear flame transfer function model.

22 Gain global air/fuel ratio = 29, Bulk velocity = 18.4 m/s) global air/fuel ratio = 20, Bulk velocity = 12 m/s Frequency (Hz) Phase (Degrees) global air/fuel ratio = 29, Bulk velocity = 18.4 m/s global air/fuel ratio = 20, Bulk velocity = 12.1 m/s Frequency (Hz) Figure 29 Transfer function measurements in the swirl stabilised spray flame as a function of frequency. (a) CH Chemiluminescence image (b)abel inverted image Figure 30 A Typical Image of a Swirl Stabilised Kerosene Flame with Pre-Hetaed air Figure 31 Transfer function measurements as a function of normalised acoustic velocity amplitude. Pressure atomiser Quartz tube Figure 32 Schematic of the cross-section of the modified burner. Task 4.2: Combustor Flow Field Sensitivity to Acoustic Oscillations (University of Loughborough)

23 Work Completed An acoustic excitation facility has been assembled and commissioned and the Stage 1 series of measurements are now complete. These measurements investigated the effects of acoustic excitation by loudspeakers over a range (50Hz~1500Hz) of frequencies on the flow field produced by an axial and radial fuel injector. A data acquisition system was developed in which 4 fast response ('Kulite') pressure transducers were used to characterise the acoustic pressure field both downstream of the injector and at the injector exit plane. The response of the velocity field issuing from the injector was measured using hot wire anemometry, which allowed the relative phase and magnitude of the pressure and velocity signals to be determined at the frequency of excitation. Further measurements were also undertaken in which water was introduced through the fuel galleries and a light sheet used to investigate the effects of acoustic excitation on the simulated fuel sheet. An example of the axial fuel injector response to acoustic excitation (50Hz~1500HZ) is presented at the injector exit plane for one of up to 6 radial positions where data was obtained (Figure 33). The acoustic velocity perturbations represent the predicted fluctuations in velocity due to the acoustic pressure fluctuations in the downstream duct, whilst the velocity perturbations associated with the injector flow field turbulence is also indicated. The measured velocity perturbations are indicated by the 'injector response' whilst also shown is a 'quasi-steady' effect. This assumes that the total pressure upstream of the injector is constant and the acoustic pressure fluctuations at the exit plane of the injector effectively alter the pressure drop across the injector and hence the flow through it. In addition to these absolute values the injector response has been expressed in terms of this 'quasi-steady' effect for several locations and injector types (Figure 34). In addition data has been removed where the magnitude of the response is thought to have been influenced by the background turbulence. It can be seen that at low frequencies the injector response appears to reflect this 'quasisteady' behaviour, but not at the higher frequencies so reflecting a low pass filter style of characteristic. However, local maximum and minima also appear to be apparent in the response characteristic whilst different parts of the injector also appear to respond differently. This data is the subject of a continuing analysis. The effect of acoustic excitation on the fuel sheet issuing from both the central (pressure) and outer (air-blast) style of injection systems are also presented (Figure 35 and Figure 36). Each processed image represents an average of approximately 3000 instantaneous images, with each instantaneous image being taken at the same phase of the acoustic cycle. In this way fluctuations due to acoustic excitation could be identified from the more random fluctuations of the fuel sheet that are present with no excitation present. This process was repeated at 10 different phases of the cycle and at 2 different frequencies of excitation, so allowing movies to be obtained of the injector response over an acoustic cycle. As an indication of the response images are presented for both the central and outer fuel injection systems at different phases of the acoustic cycle 'Quasi steady' 1.0 U' rms (m/s) Injector response 0.4 Turbulence Level 0.2 Acoustic Signal Frequency (Hz) Figure 33 Axial injector response (r=35mm)

24 Axial Injector (Outer Stream) Axial Injector (Centreline) Radial Injector (Centreline) 1.0 'quasi steady (qs)' U rms /U qs Frequency Figure 34 Injector response non-dimensionalised w.r.t. the quasi-steady velocity a) Minimum perturbation at 288º b) Maximum perturbation at 180º Figure 35 Axial injector s Main Burner response to excitation (200Hz, 135dB) a) Minimum perturbation at 36º b) Maximum perturbation at 180º Figure 36 Axial injector s Pilot Burner response to excitation (200Hz, 135dB) Task Influence of Acoustic Waves on Spray Vaporisation and Combustion

25 Subtask Influence of Acoustic Waves on Spray Vaporisation and Combustion (Numerical Aspects) (CORIA) Work Completed A Direct Numerical Simulation (DNS) database of a turbulent evaporating spray has been developed to study the impact of local sources of vapour in fuel-air turbulent micro-mixing. Those DNS s are designed so that they can easily be probed for testing and to propose new closures that are needed when modelling unsteady spray combustion. DNS allows for a model-free simulation of the gas phase. The flow is described in an Eulerian context using the Navier- Stokes equations. The spray is composed of a large number of individual droplets convected by the turbulent flow. They are tracked within a Lagrangian frame. The gas phase evolution involves a two-way coupling and additional source terms appear in the density, the momentum and the fuel mass species equations, due to droplet evaporation. Spray evaporation and turbulent mixing were studied in grid turbulence using four (two Lagrangian and two Eulerian) coupled solvers. The first set of two solvers is necessary to generate a time evolving 3D forced turbulence laden with drops. This is done with a spectral solution of a forced turbulence transporting drops. This spectral solution provides a synthetic grid turbulence carrying the discrete phase that is used as inlet condition for the spatially decaying grid turbulence, which is simulated by two additional physical space solvers. The coupling between evaporation and turbulent mixing generates a polydisperse evaporating spray. Three-dimensional simulations have been performed and analyzed through a post-processing tool (Figure 37). Because of the large amount of information contained in the database, it cannot be stored easily. In order to provide a set of DNS results to the partners, additional two-dimensional simulations have been generated and delivered. The partners may use those simulations to determine which information they would like to see collected in the three-dimensional runs. All simulation domains size are with Eulerian nodes and involve about 55,000 droplets. More details may be found in the 12 th month report. Figure 37 Three-dimensional DNS of spray evaporation. Once the DNS database had been developed, we began to work on the modelling of the subgrid mixture fraction Z ~ and its fluctuations in a LES context. Those are the key parameters of many turbulent combustion models for non-premixed or partially premixed flows. If, in the case of purely gaseous flows, the mixture fraction definition is straightforward, it is not the case in multiphase systems where both liquid and gaseous fuel are present. Thus, we first selected an appropriate definition for the mixture fraction that is similar to the one-phase simulation to have an easy coupling with ~ the turbulent combustion models. Then we derived the balance equation for the mixture fraction subgrid variance Zv to evaluate the order of magnitude of each unclosed term. Compared to the classical gaseous balance equation, new unclosed source terms appear due to spray evaporation. These new terms are of the same order of magnitude of the dissipation terms. Thus, it is impossible to neglect them. Several existing models have been implemented and tested, either to determine directly the mixture fraction variance or to close its balance equation. All subroutines concerning these models are available to any interested partner. To determine Z ~ v directly, two methods may be used: the scale similarity model of Cook and Riley that supposes a direct link between the known large scale fluctuations and the subgrid fluctuations through a constant parameter C. Because of C, this model does not give satisfactory results. Although a dynamic determination has been proposed and tested, its accurate determination is a difficult task. The use of a similarity model can be a problem since the theory is based on a

26 direct link between the turbulence and the mixing but it does not take into account the presence of an evaporating dispersed phase. An alternative to the scale similarity model has been proposed by Pierce and Moin with a dynamic determination of the local model coefficient. However, it appears that these dynamics procedure may leads to non-physical values of the coefficient. Moreover, from a general point of view, the model is not able to capture the correct levels of Z ~ v. Because a direct determination of the mixture fraction subgrid fluctuations seems to be an issue, a second alternative is to solve the balance equation of this parameter. Then, several unclosed terms appear and, among them, the subgrid dissipation which is a very important parameter because its determination is crucial (1) to determine Z ~ v and (2) as an input parameter to any flamelet model for the combustion. Three models for the dissipation have been tested: (1) A linear relaxation model (LRM) (2) A subgrid scale equilibrium model with a one-phase closure (3) A Two-phase subgrid scale equilibrium model (TP-SGSEM) After comparisons of the three various models, it appears that the TP-SGSEM closure is the best model to use, even it some difficulties appear when the spray is not entirely in equilibrium with the gas phase. The LRM model gives good results but more specifically when the droplets have a lot of inertia. In the balance equation of Z ~ v, new source terms have to be modelled. The problem is then to express the correlation between the mixture fraction and the evaporation rate. Both data are mainly linked to the droplet dispersion and their inertia. A model has been developed to uncouple this correlation into two known terms linked by a correlation coefficient, which is the unknown of the problem. A dynamic procedure has been proposed for the correlation coefficient and the model gives very good results although it can be improved when the droplet inertia increases. To finish our work concerning the subgrid model for evaporating sprays, we introduced a new parameter: the mixing factor ~ φ. This factor allows us to determine Z ~ ~ ~ ~ v through the following relation: (1 ) ~ Zv = Z Z φ. It appears that the balance equation for φ ~ is much more easy to solve than the one for Z ~ v. Indeed, there are far less unclosed terms. Remaining unclosed terms have been evaluated and models have been tested. The results are very encouraging. Indeed, they are more accurate than any other models or balance equations tested up to now. Another part of our work during the last 6 months has been dedicated to the setup, still in progress, of the numerical code that will be used to work on the spray combustion experiment from EM2C laboratory.

27 Figure 38 Snapshot of the vorticity (a), mixture fraction field (b), the mixture fraction subgrid variance (c), evaporating droplets (d). Left: streamwise, right: spanwise. Subtask Influence of Acoustic Waves on Spray Vaporisation and Combustion (Experimental Part) (EM2C) Work Completed The main objective of this task is to provide validation measurements for numerical simulations performed by INSA/CORIA in Rouen on interactions between spray combustion and acoustic waves. The first year of this project was devoted to the design and building of the experimental setup. First a laminar, conical and gaseous methane / air premixed flame is anchored on a Bunsen burner. Then, in order to seed the methane / air flow with fuel droplets, a spray generator has been placed on the axis of the burner. According to the provider, the mean drop size of the ultrasonic atomizer should be around 45 µm at the atomizer exit, but they are smaller due to liquid vaporisation. For example, for droplets of heptane, the arithmetic diameter is around 4 µm at the burner exit. In the first configuration, a loudspeaker was placed at the bottom of the burner to generate longitudinal plane acoustic waves in the burner. In order to integrate the spray atomizer, the burner was modified: two loudspeakers are placed on both sides of the burner. Following the technical meeting of April 2003, the experimental setup has been modified so that vaporized heptane could be used as gaseous fuel instead of methane. A small amount of liquid heptane flows in the primary airflow. A

28 heating device has been introduced around the main air duct so that liquid heptane present in the airflow is completely vaporized. Finally, the setup can be placed in upward or downward directions thus varying the influence of gravity on the biggest droplets. In the upward direction, we observe that there are a lot of droplets that have a negative velocity, indicating that these droplets are not able to follow the air flow. These droplets fall back on the generator plate and there can be an accumulation of fuel at the burner base. To analyze and understand the behavior of the droplet generator, measurements were performed at the generator exit, using first water, then heptane and decane. Measurements were then performed at the burner exit, in the non-reactive case with phase Doppler anemometry, tomography and particle imaging velocimetry. Results indicated that the vaporisation rate is a limiting parameter, which may prevent us from obtaining large droplets at the burner exit, depending on the liquid fuel used. Finally, a study of the spray flame was realized. Two situations are considered: - a preheated gaseous heptane / air flame seeded with heptane droplets - a non-preheated methane / air flame seeded with heptane droplets In the stationary situation, the results show that there are little effects of the flame itself on the droplet size distribution. The second year of this project is devoted to the study of the effects of low frequency / low amplitude modulations on the spray flame. We first search in literature for works related to this subject. It has been demonstrated that, properly applied and controlled, the presence of acoustic oscillations in two-phase flows results in reductions in the particle lifetime and thus improvement of the combustion process. Following the results obtained on the effect of gravity on the droplet formation, we choose to put the burner in the downward direction. Thus, it was necessary to protect the setup during combustion experiments from the heat of the burned gases and we build a water-cooled sheet of metal placed near the burner exit. Problems concerning the flow homogeneity at the burner exit have been observed. Thus we reduce the distance between the atomizer tip and the burner exit to 180 mm and a metallic restriction has been put below the primary airflow entrance (Figure 39) since it improves mixing between air and droplets. liquid fuel ultrasonic atomizer secondary air primary air + gaseous fuel secondary air primary air + gaseous fuel loudspeaker loudspeaker 180 mm (a) schematic of the setup (b) photograph of the setup Figure 39 Experimental Facility. Because of high vaporisation, only small droplets were detected at the burner exit. Various liquid fuels have been tested in the non-reactive case: heptane, decane and methanol. With decane fuel, we obtain bigger droplets at the burner exit. We also perform gaseous flames of heptane but we could not obtain gaseous flames of methanol and decane because the liquid fuel could not be heated enough to vaporize it with the existing device. The effect of acoustics has been studied by phase Doppler anemometry, in the non-reactive jet and in the spray flame for several fuels. As we cannot have gaseous flames of methanol and decane, we perform measurements for methane flames seeded with methanol or decane droplets. Interpretation of the results is in progress but the first conclusion is that the response of droplets to acoustic excitation depends on the droplet diameter: the biggest droplets are less affected than the smallest ones. Deliverable 4.8, the results on the effects of low frequency / low amplitude modulations on the spray flame, will be delayed for less than six months. This delay is due to:

29 The modifications that were necessary in the device to resolve problems concerning the flow homogeneity at the burner exit Improvement of measurements with phase Doppler anemometry Attempts to obtain bigger droplets at the burner exit Subtask Industrial Validation (SNECMA) Work Completed Work on this sub task has not commenced at this time. Task 4.4 Fluctuations Analysis of Single (Multiple) Jet in Under/Super Critical Regimes (University of Naples) Work Completed As it was planned in the research activities program the experimental system and the diagnostics have been realized or adapted in due time. The milestone marking the first 24 months was the realization of a preliminary measurement campaign. The milestone has been achieved and the results demonstrated the feasibility of the analysis of the jet behaviour in the different test conditions. The grid of measurements conditions has been defined on the ground of the preliminary measurements and of the industrial partner (AVIO Group). Some minor corrections to the experimental set-up are currently being realized to correct some problems detected during the preliminary measurement campaign and to widen the operating conditions of the test rig. The activities that have been completed in the first 24 months of the task research work are described below. 1. Implementation of test rig: An experimental facility was designed to reproduce geometry and operating conditions of the premixing channel of LPP gas turbine engine. The test rig consisted of a fully accessible chamber with a square cross section of 25x25 mm, capable to resist to high pressures, up to 10 MPa, and high temperatures, up to 1000 K. Three side walls of the duct hosted quartz windows for the optical access, while on the fourth side is mounted the injector with the axis normal to the channel one. The injector is a plain nozzle. A 45 taper introduces the liquid flow to the terminal straight section of the nozzle having an L/D ratio equal to 4. The discharge coefficient of the nozzle was preliminary determined and resulted to be The liquid was supplied to the nozzle by means of a nitrogen-pressurized vessel and regulated by a pressure control valve. That system allowed a precise control of the liquid velocity in the field m/s. The gas flow rate and the airflow velocity were regulated by using a variable area diaphragm mounted at the end of the channel. The stability of the air cross flow pattern has been improved with mounting of a high quality pressure reduction valve. The level of air velocity uniformity up to 50 m/s has been checked by means of numerical simulations of the flow through the whole inlet duct and velocity measurements made using both a hot wire anemometer and a Pitot tube probe. 2. Variable injection angle device: A first release of the system was designed and realized. A set of preliminary test demonstrated the possibility to modulate the jet behaviour very effectively by changing the orientation of the jet with respect to the airflow (see for instance the figure below).