Fuel temperature influence on Diesel sprays in reacting conditions

Transcription

1 ILASS-Americas 23rd Annual Conference on Liquid Atomization and Spray Systems, Ventura, CA, May 2011 Fuel temperature influence on Diesel sprays in reacting conditions R. Payri* J.M. García-Oliver J. Manin M. Bardi CMT-Motores Térmicos, Universidad Politécnica de Valencia, Spain Abstract The detailed knowledge of the evaporation-combustion process of the Diesel spray is a key factor for the development of robust injection strategies able to reduce the pollutant emissions and keep or increase the combustion efficiency. In this work the fuel temperature effect on the evaporation and combustion of the spray has been studied thanks to a system able to control the temperature of the injector body installed on a continuous flow, high pressure and high temperature test rig (Nitrogen or air up to 1000 K and 15 MPa). Imaging diagnostics have been employed to visualize the liquid phase penetration in evaporative inert conditions and to have time and spatial description of the combustion event in reacting atmosphere. The results underline a clear influence of the injector body temperature on both conditions, evaporative and, in a lesser degree, reactive; finally the physical models found in the literature have been used to explain the results obtained experimentally.

2 Introduction The importance of the injection process in the global functioning of a direct injection Diesel engine is pivotal. Many works in the last years have highlighted which are the variables that mostly affect the combustion process such as injection pressure, nozzle orifice diameter and the thermo-dynamical conditions inside the chamber at the moment of the injection [1] [2]. The aim of this work is to give a further contribution to this knowledge by studying experimentally and theoretically the effect of the injector coolant temperature on several characteristics defining the injection/combustion processes. The relationship between the temperature of the fuel and the temperature of the injector has been discussed and the results obtained have been compared with what theoretically expected. The tests have been carried out in a continuous flow test rig that allows simulating the thermodynamical conditions in an engine at the time of injection in an optically accessible testing section. The main advantage of this facility is the possibility to create stationary condition inside the chamber and thus to permit a great number of repetitions and to lower in considerable way the shot to shot dispersion of the results. This test rig has been coupled with a purposely designed injector holder that keeps the body of the injector in direct contact with a liquid flowing at a controlled temperature. The injector employed in this work has been designed by Bosch; it is based on second generation Common-Rail systems. The needle is indirectly moved by a solenoid coil and the injector is equipped with a single-hole nozzle coaxially positioned with the injector s body. From the technical drawings given by the manufacturer the value of the outlet orifice diameter is 0.09 mm and its internal geometry is of conical shape with k factor of 1.5 [2]. The energizing time has not been changed all along the test matrix (ET = 2500 µs) and it has been chosen long enough to have a complete stabilization of the vaporization and of the combustion process. The fuel used is n-dodecane: the use of a single-component fuel is justified by the wide amount of the physical and chemical data available related to it gives us the opportunity to try a theoretical evaluation of the experimental results. This document is divided into six sections including the present introduction; the rest of the paper is organized as follows: The experimental facility, the control system and the test matrix. Liquid phase penetration and the experimental methodology applied. Lift-off length and the experimental methodology applied. Results and discussion. Conclusions. Experimental facility The test rig employed is a high temperature and high pressure test chamber where the thermo-dynamic conditions obtained in a Diesel engine at the time of injection can be simulated with a maximum ambient temperature of 1000 K and a maximum pressure of 14 MPa. The testing section has three large optical accesses (120 mm diameter) placed orthogonally in order to have a complete view of the injection event. The facility is basically composed of four parts: gas compressors, gas heaters, test vessel, control system. The gas, stored by volumetric compressor in high pressure reservoirs, passing through electrical resistances is heated above the target test temperature and continuously flows though the test chamber. The control system is a closed loop PID that adjusts both the pressure in the chamber and the power of the heater in order to create in the chamber the conditions selected by the user. The rig can work in open or close circuit in order to test either with air or gas mixtures with different O 2 concentration. The chemical composition of the gas in the chamber is permanently measured by a Horiba system and can be adjusted adding either air or Nitrogen through a reintegration system. 1

3 Figure 1: Global view of the chamber of the test rig. The temperature of the injector body has been controlled employing a special injector holder that maintains the injector in direct contact with a coolant flowing at constant temperature: the temperature of the liquid is controlled through a PID system able to feed the liquid at temperatures ranging from 15 to 80 C. The injection frequency set during the test is 0.3 Hz. Experimental methodology The effect of the temperature of the injector body has been studied observing its influence on two aspects of the injection-combustion processes: the liquid phase penetration and the lift-off length. With the aim of focusing the attention on this aspect the same test matrix has been repeated changing the temperature of the liquid flowing in the injector holder T f ; the test conditions employed in the tests are detailed in table Table 1: Test matrix. Liquid phase penetration through Mie scattering The liquid phase penetration or liquid length (LL) is a measurement commonly performed to characterize the mixing process of the Diesel spray [3]; a simple definition of the liquid length can be given as the distance that the injected fuel has to penetrate until its complete evaporation. Several empirical and physical models have been developed to predict the LL and its dependence on the injection parameters. In the several studies on liquid phase penetration available in the literature the dependence of the liquid length upon different parameters such as ambient temperature, ambient density, injection pressure, nozzle diameter and fuel type can be found [3], [4], [5] and [6]. To predict the variation of the liquid length in the past also many researchers have developed scaling laws basing on physical hypothesis as the model developed by Desantes [7] and the one proposed by Siebers [8]: both of them are based on the hypothesis of mixing-limited vaporization and with relatively easy calculation demonstrated to predict in accurate way the experimental results. The aforesaid scaling laws are interesting for the purpose of this work because they permit to have references to compare with the tendencies experimentally observed. More details about the models will be given in the next sections 1 The injector temperature T f = 343 K has been used only in liquid length tests 2 The ambient density T g = 7.6 kg/m 3 been used only in liquid length tests 3 The 0% O 2 concentration has been employed in liquid length test and 15% in the lift-off length test 2

4 The technique chosen to visualize the liquid- phase penetration is Mie scattering with lateral illumination. A sketch of the optical set-up used in this test is depicted in Figure 2; A high speed CMOS camera (Photron Fastcam) fitted with a Sigma 70 mm lens has been employed to acquire the images of the injection events. The region of the spray has been illuminated by a continuous Xenon arc lamp and the light scattered by the liquid droplets collected by the camera. Figure 2: Scheme of the lay-out employed for liquid length measurement. The image processing applied to the images for the calculation of liquid length has been described by Siebers in [8]. This method consists in subtracting the background and taking threshold of 3% the maximum digital level observed in the core of the spray. This method has demonstrated to be sensible and robust to different optical settings and experimental conditions. Stationary lift off length through OH chemiluminescence The flame lift-off length (LOL) of Diesel fuel sprays is usually defined as the distance from the injector to the reaction zone stabilization after the time of auto-ignition [4], [10]. This characteristic of the flame is known to have a significant effect on Diesel combustion [11]. In this study the effect of the injection parameters together with the effect of the temperature of the coolant liquid on the lift-off length have been observed. The chemiluminescence is a chemical phenomenon consisting in a light emission due to the decay of a molecule from an exited state (OH * ) to a lower energy level (OH). OH chemiluminescence has a well-defined spectrum that permits identifying the emitting molecule [9]. Dec et al [12] demonstrated that the OH radicals develop in a very thin layer covering the surface of the jet and adding that in Diesel-like conditions this layer is so thin and the OH residence time so short that it could almost be considered as the jet stoichiometric surface. The spectrum of the light emitted by OH * decay has its most important peak at 306 nm: in this measurement to use a 310±5 nm interferometric filter gives a double advantage: first it permits visualizing the limits of the combustion area in an accurate way; second it excludes nearly the totality of the soot incandescence radiation, that at higher wavelengths are normally order of magnitudes more intense than OH emissions. In these measurements, an Andor ICCD camera (istar) fitted with an 100 mm focal length UV lens and a 310±5 µm filter has been employed. The camera intensifier has been set to its maximum gain value and 0.1 ms gate time. The camera trigger has been synchronized at 3 ms after the start of the injector energizing (corresponding roughly to 2.7 ms after the start of injection) in order to observe the stabilized liftoff length. The scheme of the optical layout is represented in Figure 3 To determine the position of the LOL in a coherent way for the different test conditions, the methodology described by Siebers in [10] has been followed. This approach calculates the digital level threshold scaling the intensity observed in the reacting area and, considering the two characteristics lobes upstream the flame, it averages the distances found for each one: in the measurements performed, this method has demonstrated to be robust for the different test conditions and also to different changes in the setting of the camera. The image is divided in two parts along the injector axis. For each half-image the maximum digital level of each x coordinate is calculated (see figure 4.b); the curves obtained are characterized by a sharp rise followed by a nearly flat zone. The threshold is obtained dividing by two the digital level reached in this part normally called leveling value; finally the LOL for each image is obtained as the mean of the LOL obtained in the two halfs of the image. 3

5 temperature is related with the temperature of the injected fuel and if the injected fuel is hotter its evaporation is faster. In order to have a value of the LL representative for each test condition, it has been defined a mean value of LL, LL m that is the average of the values measured in the interval from 2800 to 3200 µs after start of injection (ASOI). In Figure 5 the time window employed for analyzed for the LL averaging is represented. Figure 3: Experimental lay-out employed for the visualization of the OH radical. a. Averaging window b. Levelling value Figure 4: Sample of the image processing. The image recorded by the camera (a) and the intensity profiles of the two half images. Figure 5: Liquid length vs. time for three different values of Tf. Pr = 150 MPa, T g = 900 K, ρ g = 22.8 kg/m 3. In Figure 6, 7 and 8 the effect of the coolant temperature is shown by plotting the LLm as function of ambient density, ambient temperature and injection pressure for the three different Tf. Results and discussion Liquid phase penetration The time resolved LL obtained for the three different T f have been plotted in the same axes in Figure 5 for one sample test condition ( p r = 150 MPa, T g = 900 K and ρ g = 15.2 kg/m 3 ). The LL curve can be divided in two parts: the first characterized by fast penetration rate, where the three cases do not present any revealing difference, and the second part where the penetration ceases and stabilizes to a nearly constant value; in this second part of the injection event it appears clear as the temperature of the cooler influences in a substantial way the liquid length penetration in particular as the higher temperature of the cooler causes the liquid length to decrease. This fact appears logical, if we consider that the cooler Figure 6: Effect of the Tf variation over LLm at different ambient densities for p r = 150 MPa and T g = 900 K. 4

6 Figure 7: Effect of the Tf variation over LLm at different chamber temperatures for ρ g = 22.8 kg/m 3 and p r = 150 MPa. that each test condition has on the result as well as the possible interaction between the effects. This analysis gives also the statistical significance of the tendencies calculated in other words the level of credibility of the information. In the diagram presented in figure 9, the results of the statistical analysis are shown as the qualitative influence of the test conditions on liquid length. This table shows the effects that the different parameters of the experiment have on the liquid length together with the blue line representing the p-value 5% limit; summarizing the bars that are above that limit are considered as statistical evidence and above the limit of the experimental uncertainties. Figure 8: Effect of the Tf variation over LLm for different injection pressures at ρ g = 22.8 kg/m3 and T g = 900 K. From a first look at the results two important conclusions can be laid: first, the effect of the temperature of the injector coolant on the LLm is evident: in fact the same tendency is observed in each test condition; moreover, its influence on the liquid length is substantial regardless of the test conditions, showing from the case Tf = 15 C to the case Tf = 70 C differences up to 13 % on LL value have been observed. To go deeper in the understanding of the experimental results obtained, taking advantage of the factorial design of the experiment a statistical study has been performed with statistical data analysis software: Statgraphics. This software allows pointing out the dependencies of the results obtained upon the experimental condition. Using the statistical theories related to the factorial experimental design described in [13], Statgraphics analyze the data evaluating the effect Figure 9: Normalized effects diagram obtained from Statgraphics. Blue indicates a negative influence and gray a positive one. This diagram shows that the ambient density has dominant effect on the LL but there also the ambient temperature and in a slightly lower way the injection pressure and coolant temperature have an important impact on the liquid penetration. In the same table are shown also the interactions between the different parameters: for example the interaction between ambient density and rail pressure (ρ g and p r ) is a measure of how much the effect of the ambient density changes, changing the injection pressure. Observing the interactions it is important to note that T f appears coupled only with gas density with a really slight effect and close to the significance threshold. The reason of the positive interaction between gas density and coolant temperature might be related with the characteristics of the rig: increasing the gas density the velocities of the flowing ambient gas in the chamber decrease and a thicker boundary layer near the injector tip is formed. 5

7 In this case the heat transfer between the gases and the injector decreases and thus, the coolant temperature will have greater importance on the temperature of the injected fuel. Despite this slight interaction the effect of the coolant temperature appears to be independent upon the specific test condition suggesting a strong link between coolant temperature and fuel temperature. A further evaluation of the results about liquid length has been performed comparing the influence of T f experimentally observe with the influence predicted by the model developed by Desantess [7] and mentioned before. The model derived by Desantes assumes the hypothesis of mixing-limited evaporation of the spray: in other word the heat exchange between entrained air and fuel droplets is considered to be infinite and the fuel evaporation is solved as an energy balance between fuel and entrained air; in this way liquid penetration is calculated as the location at which the energy supplied by the entrained air is enough to vaporize all the fuel. From a mass and energy balance, this location corresponds to a specific fuel mass fraction named Y f,evap. The form in which the scaling law has been obtained is reported below: LL = Kd 0 ρ f 1 ρ Y a f, evap Where K is a constant of the spray, d o is the orifice diameter and ρ f and ρ are the fuel and air density. a The last term in equation (1), as stated before takes in account the energy balance needed for the fuel evaporation. Y f,evap can also be written as function of variation of the specific enthalpy of the fuel and of the air. This variation is from the startt condition (T D and T a ) to the equilibrium temperature of evaporation (T evap ). 1 hf = 1+ Y h, f, evap a ( TD, Tevap ) ( T T ) a evap In equation 2, it can be appreciated that the temperature of the fuel appears in the law. To draw a comparison between the experimental and theoretical tendencies, the temperature of the fuel has been considered to coincide with the coolant temperature 4. Moreover to 4 (1) ( 2) This assumption appears reasonable considering that the injection frequency is relatively low during the tests (0,3 Hz) and the temperature of the exclude all the other deviation between the model and experimental results the model has been calibrated on the experimental datum of liquid length of T f = 70ºC for each test condition and only the tendency regarding the fuel temperature has been observed. In figure 10 a comparison of the model and the experimental results is presented. For the results related to T f = 70ºC the results of the model and experimental overlap as long as the model has been calibrated on each of these datum. In the case T f = 70ºC the relation within experimental and prediction can be seen. This comparison put in evidence a good agreement between the prediction and the experimental observation. However it is important to point out, that the model normally under-predicts the influence of T f. The cause of this deviation is unclear. Figure 10: Influence of the coolant temperature on liquid length. Comparison between experimental results and theoretical predictions; pr = 150 MPa, Tg = 900 K. One possible reason can be the fact that the model does not take in account the variation of the physical property of the fuel in the liquid phase such as cinematic viscosity and surface tension parameters that can affect the spray atomization [1]. Lift-off length In figure 11 and 12 the results for LOL obtained in all the conditions tested are depicted. To ease the understanding of the resultss the experimental standard injector body, that has been monitored during the tests, was always less than 5 degrees higher than the coolant liquid. 6

8 deviation has been plotted together with the mean value: the error bars in the figures indicate ± σ. Although the values of LOL appears to be affected by the temperature of the coolant liquid this difference in some case is very low and in most of the cases is of the same magnitude as the standard deviation. Figure 11: Results of LOL obtained for ambient density of 15.2 kg/m 3. prove the existence of the interactions observed, but at least put in evidence one possible way to improve the empirical models. In this case the temperature of the coolant appears to have a very low effect, slightly above the confidence limit. However, also if the effect is very low, this tendency can be explained through the non-premixed gas-jet theory developed by Peters [11]. Peters reviewed much of the experimental gas-jet flame lift-off investigations and analytically derived a scaling law for lift-off length that is in agreement with the experimental trends observed for gas-jets. The scaling law was derived observing that the lift-off length stabilizes in a thin region very close to the stoichiometric contour, and making the hypothesis that the Damköhler number is of order of one at the stabilization location, that means that the local flame reaction rate is comparable to the local mixing rate. The scaling law proposed by Peters is reported below: U 0Z st D LOL = (3) 2 s ( Z ) L st Where U 0 is the theoretical velocity at the exit of the orifice, Z st the stoichiometricc fuel mixture fraction, D is the thermal diffusivity and s L (Z st ) is the laminar flame speed at stoichiometric fuel-air mixture. Figure 12: Results of LOL obtained for ambient density of 22.8 kg/m 3. As in the previous case to have a better understanding of which is the effect on the LOL of the different test conditions a statistical analysis has been carried out with Statgraphics. As in the case of the liquid length the effects of ambient density, ambient temperature and rail pressure are far above the confidence limit. Observing this diagram, has to be noted the importance of the interactions between experimental parameters in the statistical analysis. This is particularly important when we consider that the empirical models normally shown in the literature consider the effects separately excluding the interactions. It has also to be said that the database of this experiment might not be enough to Figure 13: Normalized effects diagram obtained for the LOL from Statgraphics. Blue indicates a negative influence and gray a positive one. s L (Z st ) for hydrocarbons has an analytic expression usually of the form: α T 2 s = L ( Z st ) sl0 298K p 1atm β (4) 7

9 Where T and p are the temperature and the pressure of the air-fuel mixture and s L0 is the laminar flame speed at reference thermo-dynamic conditions (T = 298 K and p = 1 atm). The exponents α and β are dependent on the hydrocarbon properties. However the variation of these coefficients is limited, and as long they are not available in the case of n-dodecane, following the approach used by Siebers in [10] the following coefficients have been employed: α = 2,1 and β = - 0,36. This approach for the estimation of the LOL has been employed in the past to the Diesel spray and despite the simplifying hypothesis of the model when applied to liquid sprays injected under evaporative conditions; it has demonstrated satisfactory results predictions in line with the tendency observed during the experiments. Substituting equation (2) in equation 1 and considering the dependence upon temperature of the thermal diffusivity of a gas ( T 0.5 ) the dependence on the temperature on LOL is obtained: 3.7 LOL T (5) Considering that variations in the coolant temperature cause equivalent variation in the fuel, and considering the adiabatic mixing it seems logical that also the stoichiometric temperature of the mixture will be affected. Calculations performed following these hypothesis have been performed to calculate the variation of the stoichiometric mixture temperature due to the variation in the fuel temperature, and thus the variation of LOL predicted by the scaling law. These calculations have shown that a variation of fuel temperature of 30 C can cause variations in LOL of 1-1.5% that is of the same order of magnitude of what has been observed experimentally. However, this means only to show one possible path to explain the phenomenon observed, but more detailed studies on the real temperature of the fuel at the nozzle outlet and its relation with the temperature of the coolant have to be performed to elaborate a consistent theory. Conclusions A study on the influence of the temperature of the injector coolant on the liquid phase penetration and lift off length has been carried out in a novel high temperature high pressure test rig. The study has been carried out with a mono-orifice 90 µm diameter nozzle injector under several ambient conditions typical of Diesel engines. The analysis of the data has put in evidence: A substantial influence of the coolant temperature on the liquid length, (variation of LL up to 13%) A statistical analysis remarked that the effect of the coolant temperature is observed independently of the test conditions. This fact evidences a strong link between the temperature of the coolant and the fuel temperature at the orifice outlet. The influence of T f on LL has been compared to the influence predicted by a model making the assumption of equality between the T f and the temperature at the orifice outlet. This comparison has pointed a good agreement between the experimental results and the model; however the model seems to underestimate the effect of T f probably because in the model the effect of fuel temperature on the atomization process is not considered. The temperature of the injector coolant showed to affect in a lower way the LOL and variations. These influence can be explained using the scaling low for the gas-flame jet developed by Peters A statistical analysis on the results of LOL pointed out significant interactions between the different test conditions in particular between the ambient temperature and the ambient density. These relations have to be investigated deeper to be included in the models and improve their accuracy. Acknowledgements This research has been funded in the frame of the project FLEXIFUEL reference TRA from Ministerio de Ciencia e Innovación. The injectors are part of the ECN international project. References [1] Payri, R., Salvador, F.J., Gimeno, J., Bracho, G., The effect of temperature and pressure on thermodynamic properties of diesel and biodiesel fuels FUEL 2011, [2] Payri, R., Salvador, F.J., Gimeno. J., De la Morena, J., Influence of injector technology on injection and combustion development Part 1: Hydraulic characterization a, Appl. Energy (2011) [3] Payri, R., Salvador, F.J., Gimeno, J., Zapata, L.D, Diesel nozzle geometry influence on spray liquid- 8

10 phase fuel penetration in evaporative conditions, Fuel 87 (2008) [4] Desantes, J.M., Pastor, J.V., Payri, R., Pastor J.M., Experimental characterization of internal nozzle flow and Diesel spray behavior. Part II evaporative conditions, Atomization and Sprays 2005, [5] Myong K, Arai M, Suzuki H, Senda J, Fujimoto H. Vaporization Characteristics and Liquid-Phase Penetration for Multi-Component Fuels (SAE ( 2004) [6] Siebers D.L., Liquid-Phase Fuel Penetration in Diesel Sprays, SAE (1998) [7] Desantes, J.M., Pastor, J.V., Garcia, J.M., Pastor J.M., A 1-D model for the description of mixingcontrolled reacting Diesel spray,, Combustion and flame 156 (1), , 2009 [8] Siebers, D.L., Scaling liquid-phase fuel penetration in Diesel sprays based on mixing-limited vaporization SAE [9] Gaydon, A.G., the spectroscopy flames, 2nd ed; FRS Chapman & Hall; London [10] Siebers. D.L., and Higgins B.S. Flame Lift-Off on Direct-Injection Diesel Sprays Under Quiescent Conditions, SAE , [11] Peters, N., Turbulent Combustion, Cambridge University press, Cambridge UK 2000 [12] Dec, J.E.; Coy, E.B., Radical imaging in a direct Diesel engine and structure of the early diffusion flame. SAE paper , 1996 [13] George E.P., Box, J., Stuart Hunter, William G. Hunter Statistics for experimenters,, John Wiley & Sons