Intermittent multijet sprays for improving mixture preparation in HCCI engines

Transcription

1 Intermittent multijet sprays for improving mixture preparation in HCCI engines Miguel R. Oliveira Panão 1,*, António L. N. Moreira 1, Diamantino F. G. Durão 2 1: IN+, Center for Innovation, Technology and Policy Research, Instituto Superior Técnico, Technical University of Lisbon, Portugal 2: Lusíada University, Lisboa, Portugal * correspondent author: mpanao@em.ist.utl.pt Abstract In this work, the characteristics of droplets produced by a multijet impingement atomization process are measured with a Phase-Doppler Interferometer and statistically described using finite mixtures of weighted probability density functions (pdf). Through this statistical tool, drop size and axial velocity distributions are involved in the physical interpretation of the flow, instead of limiting it to first, and second order distribution moments. Each group of droplets with similar size characteristics has been modeled by lognormal distributions, and normal distributions relatively to drop axial velocity. The analysis based on finite mixtures identified three groups of droplets with similar size characteristics, although the group with smaller sizes is negligibly represented in the mixture. Also, the lognormal standard deviation in all groups is well correlated with the corresponding geometric mean diameter. In terms of axial velocity, mainly one distribution has been identified with a relatively constant standard deviation, and a characteristic velocity dependent on the duty cycle associated with the spray intermittent condition. Furthermore, droplets characteristics are correlated with heat transfer rate for several operating conditions that maintain the surface temperature in steady-state at 125ºC. The effect of the time between consecutive injections is analyzed. Concerning the use of multijet impingement sprays in fuel injection systems, results evidence the importance of an interaction between thin-liquid film heat transfer and droplets axial velocity to enhance heat transfer and promote evaporation. This would decrease the amount of fuel deposited on interposed surfaces, and improve mixture preparation in HCCI engines. 1. Introduction In HCCI engines, the reduction of emissions and fuel consumption requires an accurate knowledge and control of the fuel/air mixture in order to obtain lean conditions, and ensure a high thermal efficiency (Maurya and Agarwal, 2011). When Port-Fuel Injection systems are used, the homogeneous charge is prepared in the intake manifold, where the fuel spray issued from the injector can impinge onto the port or valve surface depending on its position and angle. Namely, the quality of the fuel/air mixture preparation will depend on the interaction between the spray and interposed surfaces. Considering that such impact occurs onto the intake vale, due to the hydrodynamic impact mechanisms, spray droplets eventually deposit on its surface and form a liquid film, but if the engine has already overcome the cold-start period, heat transfer becomes an important agent in the fuel/air mixture preparation. Thus, the characterization of the impinging spray structure, secondary atomization and heat transfer are important features for the modeling of this two-phase flow. Moreover, Alt et al. (2008) stated that the best method for controlling the incylinder mixture is to use the exhaust gas energy by mixing with fresh gas through variable valve timing and durations. However, this would require an individual cylinder adaptation that is not possible. Therefore, compensation using fuel injection has been suggested through the splitting of an injection cycle into multiple consecutive ones. This introduces the effect of the spray intermittency on the spray/wall heat transfer process that forms the fuel/air mixture, which is part of the argument explored in this work

2 Panão et al. (2012a) have advanced an argument for the importance of controlling the liquid film deposition on the impinging surface in order to have an efficient heat transfer using intermittent multijet impingement sprays, and eventually promote fuel evaporation. Multijet impingement sprays have already been advanced as an adequate atomization strategy for producing small droplets for gasoline and Diesel sprays (Durst, 2010). However, different atomization strategies produce different spray patterns, and the formation of the liquid film depends on the hydrodynamic characteristics of droplets before impact. Furthermore, these depend on the fluid thermophysical properties, the surface material and topography, but essentially on the drop size and velocity, namely its axial component, before impaction on the surface. Therefore, from a local perspective, this paper addresses the relation between droplets characteristics in multijet impingement sprays, liquid film formation and heat transfer. Namely, it is suggested that this atomization strategy can potentially improve port-fuel injection systems for HCCI. In order to support it, a multijet impingement spray produced from the impact of 3 jets is locally characterized in terms of size and velocity of its droplets and the results are compared with those previously reported on heat transfer for this spray configuration (Panão et al., 2012a). 2. Experimental setup and methodology The spray issues from a multijet impingement atomizer with N j = 3 impinging jets making an impact angle (2θ) between them of 90 and equally spaced in terms of azimuthal angle (ψ =120º). The jets are considered to have the same diameter d j = 400 µm as the hole in the atomizer. The liquid supply line works in a closed circuit and a pressure regulation valve allows controlling its value in the monorail coupled with the injector. This pressure has been set to the same value used in the heat transfer experiments previously reported (Panão et al., 2012a), corresponding to 1.6 bar. For this pressure differential at the atomizer exit, a volumetric flow rate of 3.33 ml/s is obtained with the injector valve fully open, and it varies depending on the duty cycle (DC 0%, 100%, see Fig. 1) as v DC = DC v 100%. The jet velocity is 8.8m/s. The liquid is methanol with density (ρ) of kg/m 3, dynamic viscosity (µ) of kg/(m/s) and surface tension (γ) of mn/m at 30 C. In fact, a recent report on The future of natural gas from the MIT Energy Initiative indicated methanol as the liquid fuel that is most efficiently and inexpensively produced from natural gas, thus, an alternative to gasoline, considering that such modification can be made to the engine at a modest cost (MIT Energy Initiative, 2011). Fig. 1 Illustration of the spray injection cycle and description of the experimental conditions considered for studying the effect of the spray intermittency on heat transfer. DC (%) f inj (Hz) Δt!!"# (ms)

3 Heat transfer experiments previously reported considered the effect of the non-injection time between consecutive injections Δt!"# as it is illustrated in Fig. 1 (Panão et al., 2012a). These noninjection times have been set to 10, 20 and 40 ms while keeping a constant duty cycle (DC, see definition in Fig. 1, where f inj is the injection frequency and Δt inj the duration of injection). Thus, a decrease of this period means that the amount of mass injected in the largest cycle Δt!"# = 40ms is actually being distributed into multiple injections (see illustration in the lower-left part of Fig. 1). The heat transfer results reported in Panão et al. (2012a) will be used in this work in order to correlate with the spray droplet s characteristics. These (size and axial velocity of droplets) have been measured with a Phase-Doppler Interferometer collecting the light dispersed by droplets at a scattering angle of 74, which corresponds to the Brewster angle for methanol, with the purpose of minimizing the reflected light with a parallel polarization. The beam spacing is 60 mm and both transmitting and receiving lens have a 500mm focal length, thus, the maximum measurable diameter is 350 µm. The data on the size and axial velocity of droplets in this work is statistically described by discrete probability density distributions (pdf). Usually, these distributions are further used in a statistical analysis, which is based on its moments, e.g. mean quantities or standard deviation. However, it has been argued in a previous work that such approach, although correct, gives only a partial and limited knowledge of the measured distribution (Panão et al., 2012b). On the other hand, for the development of numerical models, it is an advantage to describe such pdf distributions by mathematical functions (e.g. Lognormal) or empirical models (e.g. Nukyiama-Tanasawa, Rosin- Ramler, Weibull, for a comprehensive review see Babinsky and Sojka, 2002). However, probability distributions describing sprays are often multimodal, or presented with heterogeneities. Thus, when the actual distribution is approximated to a unimodal mathematical function, with its characteristic parameters, and the latter are used to evaluate geometric, operating or environmental parametric effects on the spray formation and development, the analysis becomes limited. In order to overcome these limitations, in this work, discrete probability distributions characterizing the spray have been mathematically modeled using the known statistical tool of finite mixtures of probability density functions. This tool consists in identifying the number K of groups of droplets with similar characteristics and empirically fit the discrete probability distribution to a linear combination of weighted probability density functions as! f!"#,! =!!! w! f x β! (1) where w i is the weight and β! are the characteristic parameters of the mathematical probability density function. In this work, for each group of droplets, the size has been modeled by a Lognormal pdf β! = μ!,!, σ!,!, and drop axial velocity by a Normal pdf β! = μ!,!, σ!,!. The fitting procedure followed a Bayesian approach previously applied to sprays by Panão and Radu (2012) and Panão et al. (2012b). Therefore, instead of using characteristic means values, which are limited for describing multimodal drop size and velocity distributions, the parameters of the finite mixture are used instead. The same formulation for a finite mixture given in eq. (1) can also be used with cumulative distribution functions (F mix or F i ). In terms of flow analysis, this ultimately means that the entire distribution is being considered, and not just its moments. The best finite mixture is that which maximizes the Marginal Likelihood (ML) or the minimum K above which ML stabilizes (Panão and Radu, 2012). The resulting finite mixtures obtained have been subjected to a chisquared test with a 99.9% of confidence interval and the hypothesis that the mathematical curve fits data has been accepted in all cases

4 3. Results and Discussion 3.1. Spray characteristics The set of results presented corresponds to a multijet impingement atomizer with three impinging jets (N j = 3). Firstly, one considers the characteristic parameters (w i, µ i, σ i ) of the finite mixture that best describes the distribution of drop sizes for the experimental conditions included in the table of Fig. 1. In all the spray intermittency conditions, the best finite mixture describing the discrete probability distribution is made of three groups of droplets (K = 3). An example of the discrete and its finite mixture of cumulative drop size distributions is given in Fig. 2. Fig. 2 Example of the discrete cumulative size distribution and the corresponding finite mixture of K = 3. In Fig. 3, the characteristic mean diameter µ i,d of every group of droplets identified is plotted against the standard deviation σ i,d, and the size of symbols is proportional to the weight w i that each cluster described by f i (x µ i,d, σ i,d ) has within the finite mixture f mix. Therefore, the first group with a lognormal distribution centered on small droplets, widely polydispersed, is the least representative, while the remaining two clusters have similar weights. There are two noteworthy observations in these results. The first is the fact that µ i,d is well correlated with σ i,d = g(µ i,d ), representing a simplification by reducing the need of two parameters to one when describing the spray characteristics, f i (x µ i,d, g(µ i,d )). The second observation is that changing the spray intermittency does not seem to significantly affect the correlation between µ i,d and σ i,d for the experimental conditions considered. Fig. 3 Correlation between the characteristic mean diameter µ i,d and the standard deviation σ i,d

5 Relatively to spray impaction, the most important velocity component is the axial one, perpendicular to the wall (U). Thus, our analysis also includes discrete probability distributions of droplets axial velocity. In this case, Normal distributions describe the experimentally obtained discrete pdf with reasonable accuracy and the best mixture is obtained for K = 2, therefore, two clusters were identified (Fig. 4). However, the cluster with the lowest characteristic velocity values and higher RMS has a negligible overall weight. Therefore, the main focus is given to the second cluster (µ 2,U, σ 2,U ). Fig. 4 Correlation between the characteristic mean diameter µ i,u and the standard deviation σ i,u of drop axial velocity. The standard deviation σ 2,U varies only 1.8% between cases, and the variability of µ 2,U is also moderate, about 6.2%. In fact, the maximum variation between operating conditions is approximately 47% of average standard deviation of all operating conditions, indicating that the effect of the spray intermittency of the spray characteristics is relatively low. Nonetheless, if this component is correlated with the duty cycle (DC), one observes a systematic pattern where the axial velocity distribution shifts to lower characteristic values for every non-injection time condition (see Fig. 5). Namely, for lower DC s, the effect of injection splitting is slightly more pronounced. Fig. 5 Variation of the second group of the axial velocity distribution (µ 2,U ) with the duty cycle (DC). Relatively to droplet size, given the presence of three groups of droplets, a weighted characteristic - 5 -

6 ! mean diameter is considered and defined as μ!,! =!!! w! μ!,!, and when plotted against DC, unlike droplets axial velocity, it becomes evident the absence of correlation (see Fig. 6). Even so, the magnitude of the measured value is within the range expected for multijet impingement sprays formed from turbulent liquid sheets, since the hydrodynamics of the atomization process has been shown to be similar regardless the number of impinging jets (Panão et al., 2012c). For N j = 2 jets, Anderson et al. (1993) have provided an empirical correlation for the arithmetic mean diameter as d!" = d! We! f θ!!.!"#, where We j is the Weber number of the jet (= ρu j 2 d j /σ) and f θ = 1 cos θ! sin!! θ. Fig. 6 includes the value obtained for the geometrical configuration of the atomizers used in this work, confirming the similarity in the order of magnitude. Fig. 6 Weighted characteristic mean diameter of droplets (µ K,D ) as a function of DC. The dashed line corresponds to the estimation of the arithmetic mean diameter for sprays produced by turbulent sheets formed from the impact of two jets according to Anderson et al. (1993) Local influence of spray characteristics on liquid deposition and heat transfer Spray impingement heat transfer events depend on the deposition of the liquid injected on the heated surface and the ability to control it. In previously reported results, if spray impaction is unavoidable, the best heat transfer efficiency for fuel vaporization purposes is achieved if a thin liquid film remains present on the heated surface (Panão et al., 2012a). Consequently, the highest heat transfer efficiencies were obtained with short time intervals between consecutive injection cycles (10 ms). In terms of predicting the outcome of spray impact, the criteria in the model of Bai et al. (2002) are used to discern between the basic hydrodynamic mechanisms: i) Deposition: a. stick (We d 2) and b. spread (20 < We d We c = 1320 La d ) which contribute to the formation of a liquid film; ii) Secondary Atomization: c. rebound (2 < We d 20) or d. splash (We d > 1320 La d ), where We d is the Weber number of droplets, We d = ρu 2 d D d /σ, and La d is the Laplace number (La d = ρσd d /µ 2 f, where µ f is the fluid dynamic viscosity)

7 These criteria have been applied to drop size and velocity data, and the percentage of drops in each mechanism has been estimated. Fig. 7 shows the effect of the duty cycle (DC) on the efficiency of heat transfer process (below) and the percentage of droplets in the spray expected to deposit on the impinging surface (above). The heat transfer efficiency is given by η =!! "!!!!!!!"!!!" (2) where q! is the heat flux dissipated at the wall, m! " = ρ! v DC the mass flow rate, c p and h fg are the fuel specific heat and latent heat of vaporization, respectively, and ΔT bf is the subcooling degree. According to the criteria of Bai et al. (2002), slightly more than half of the spray droplets are expected to deposit on the surface, and a trend for slightly increasing this fraction or percentage is observed as DC approaches the condition of a continuous spray (DC = 100%). These values are within the typically measured in the intake port of Port-Fuel Injection (PFI) systems (Senda et al., 1999). Furthermore, considering the differences between the multijet impingement atomization strategy in this work and a pressure atomizer from a pintle-type injector tipically used in PFI engines, despite the larger drop sizes and the lower values of axial velocity, parameters such as the particle response time τ! =!!!!!"!!"# are similar. Therefore, this supports that multijet impingement sprays are an alternative atomization strategy for PFI systems, with the advantage of requiring lower pressure differential at the injector nozzle, which would decrease the pumping power of the overall injection system and corresponding losses. Fig. 7 Effect of the duty cycle (DC) on the efficiency and expected deposited fraction of droplets for non-injection times Δt!"# of 10, 20 and 40 ms. The evolution of deposition percentage with DC, observed in Fig. 7, is coherent with the observed fact that a higher DC implies a larger mass flux injected, leading to thicker liquid films and jeopardizing heat transfer (Panão et al., 2012a). However, while it is evident that a shorter time between consecutive injections improves the heat transfer efficiency, no such trend is observed relatively to the fraction of droplets expected to deposit on the surface. On the other hand, the behavioral pattern of the axial velocity with DC, shown in Fig. 5, is similar to the monotonic decrease observed for the heat transfer efficiency in Fig. 7 (bottom). This would be consistent with the fact that impinging droplets are likely to pierce the liquid film, allowing that cooler liquid reaches the surface more efficiently, particularly in the cases of thinner liquid films, thus, enhancing heat transfer. These results suggest that not only the axial velocity might be a determinant parameter - 7 -

8 for heat transfer (Arcoumanis and Chang, 1993), but more so the relation between it and the liquid film thickness. Using the heat transfer data with that of the spray characteristics, a correlation is sought. However, the purpose is not to provide a new empirical correlation per se, but physically interpret the value obtained for its constant parameters. In order to isolate the effects of drop size and velocity, two dimensionless numbers have been used, the Laplace La d = Re d 2 /We d ~ g(d) and Capillary Ca d = We d /Re d ~ g(u) numbers. The Nusselt number is defined as Nu d = h w µ K,D /k f, with h w as the heat transfer coefficient, µ K,D the weighted characteristic mean diameter and k f the liquid thermal conductivity. In Fig. 8, it is observed how heat transfer is higher for the most efficient operating conditions Δt!"# = 10ms, while increasing the time between consecutive injection cycles leads to lower heat transfer rates. This evidences the importance of the ability to control the presence of a thin liquid film for fuel vaporization. Fig. 8 Variation of Nu as a function of the duty cycle (DC) for different non-injection times between consecutive cycles. The La d and Ca d numbers have also used weighted characteristic mean parameters (µ K,D and µ K,U ) in their calculation. The empirical correlation has the typical form of Nu! = ala!! Ca!! (3) where a, b and c are the correlation constants. Although heat transfer correlations usually use the Prandtl number (Pr), relating the kinematic and thermal diffusivities, the fluid is the same in every experiment, as well as the environmental conditions. Therefore, the effect of Pr would be constant and, thus, included in constant a. Using a direct least square method, b and c are estimated to be negative (see Fig. 8-left), suggesting that increasing the size, since La d >> 1 b < 0, heat transfer is jeopardized, while increasing the axial velocity of impinging droplets, since Ca d < 1 c < 0, heat transfer is improved. Usually, there is a positive correlation between droplet size and velocity, i.e. larger droplets are less prone to interact with the surrounding environment, thus, are faster than smaller droplets. Therefore, the constant parameters in this correlation indicate a competition between these two effects. However, Fig. 8-right shows how the effect exerted by the dimensionless parameter Ca d, associated with the axial velocity is more influential relatively to drop size associated with La d

9 Fig. 8 Comparison between Nu obtained experimentally and through an empirical correlation (left); and influence of dimensionless numbers Ca d and La d on the Nusselt number (right). An additional point is that faster droplets have greater impact energy and their likely outcome is not to deposit on the surface for vaporization purposes, but eventually trigger secondary atomization mechanisms and generate secondary droplets. In this case, part of the liquid is is in contact with the surrounding air instead of remaining deposited on the surface, enhancing fuel vaporization, and improving mixture preparation. 4. Concluding remarks It has been suggested that injection-splitting strategies are a way of controlling the in-cylinder mixture in HCCI engines. However, this introduces the effect of intermittency in the spray/wall interaction process, in terms of secondary atomization/deposition and heat transfer, which conditions the fuel/air mixture preparation. In this work, a local analysis is made on the characteristics of a multijet impingement spray and the heat transfer that keeps the surface temperature in a steady-state condition at 125ºC. The spray characterization through statistical analysis uses finite mixtures of weighted Lognormal distributions for the size of droplets and normal distributions for their axial velocity component (perpendicular to an impinging surface). The advantage of this more advanced statistical tool is that one is considering the entire distribution in the analysis, instead of its moments alone. Results for the spray characteristics evidence: the presence of three groups of droplets characterizing the spray in every intermittent operating condition. However, the groups with medium and large characteristic sizes are dominant relatively to the group of smaller droplets; droplets size polydispersion expressed by the standard deviation of the log-normal is well!!.!"# correlated with the characteristic size as σ!,! = μ!,! (R 2 = 0.996); droplets axial velocity is mainly constituted by a group around 86% of the jet velocity (8.8m/s), which does not significantly vary with the duty cycle (DC). The characteristics of droplets have been measured for the same operating conditions of previously reported heat transfer experiments, in order to evaluate their correlation, as well as the expected outcome of impact. Results evidence the following: - 9 -

10 spray deposition is within the typical values observed for the usual sprays in PFI engines produced by pintle-type injectors. However, in the case of multijet impingement atomizers, a lower pumping pressure is required; multiple injection pulses with shorter time intervals between consecutive cycles, induce a greater importance to the relation between droplets axial velocity and liquid film thickness for heat transfer enhancement. Consequently, this improves evaporation and reduces fuel deposition; Further research is required to assess multijet impingement sprays performance in a more real HCCI engine environment. Acknowledgements The author would like to acknowledge Fundação para a Ciência e Tecnologia for supporting his research through a postdoc fellowship SFRH/BPD/45170/2008. References Alt M, Kafarnik P, Pritze S (2008) The Gasoline HCCI engine by GM, in Alternative Propulsion Systems for Automobiles II, Cornel Stan, Giovanni Cipolla (Eds.), Expert-Verlag, 137. Anderson W, Ryan S, Santoro R (1993) Spray formation processes of impinging jet injectors. NASA Prop Eng Res C 2: Arcoumanis C, Chang J-C (1993) Heat transfer between a heated plate and an impinging transient diesel spray, Experiments in Fluids, 16: Babinsky E, Sojka PE (2002) Modeling drop size distributions, Prog. Energy and Combustion Science, 28: Bai C, Rusche H, Gosman A (2002) Modeling of gasoline spray impingement, Atomization and Sprays 12:1 27. Durst F (2010) Multi-jet spray generation for small droplet production, 23 rd Annual Conference on Liquid Atomization and Spray Systems, ILASS-Europe, Brno, Czech Republic. Maurya RK, Agarwal AK (2011) Statistical analysis of the cyclic variations of heat release parameters in HCCI combustion of methanol and gasoline, Applied Energy 88: MIT Energy Initiative (2011) The future of natural gas an interdisciplinary MIT study, MIT. Panão MRO, Moreira ALN, Durão DFG (2011) Intelligent thermal management for full electric vehicles, 6 th Dubrovnik Conference on Sustainable Development of Energy Water and Environment Systems. Panão MRO, Correia AM, Moreira ALN (2012a) High-power electronics thermal management with intermittent multijet sprays, Applied Thermal Engineering 37: Panão MRO, Moreira ALN, Durão DFG (2012b) Statistical analysis of spray impact to assess fuel mixture preparation in IC engines, submitted to Fuel Processing Technology. Panão MRO, Moreira ALN, Durão DFG (2012c) Transient analysis of intermittent multijet sprays, Experiments in Fluids, DOI /s Panão MRO, Radu L (2012) Advanced statistics to improve the physical interpretation of atomization processes, submitted to International Journal of Heat and Fluid Flow. Senda J, Ohnishi M, Takahashi T, Fujimoto H, Utsunomiya A, Wakatabe M (1999) Measurement

11 and modeling on wall wetted fuel film profile and mixture preparation in intake port of SI engine, SAE Technical Paper