9th Asia-Pacific Conference on Combustion, Gyeongju Hilton, Gyeongju, Korea 19-22 May 2013
Effect of grid generated turbulence on the atomization of a liquid jet of acetone 1
Agisilaos Kourmatzis , Assaad R. Masri 1
1
Clean Combustion Research Group, Aerospace, Mechanical & Mechatronic Engineering The University of Sydney, New South Wales, 2006, Australia
Abstract This paper uses a newly developed wind tunnel, fitted with grids to generate high levels of turbulence, in order to investigate the effect of velocity fluctuations on the atomization of a liquid jet of acetone. The mean gas velocity at the liquid jet location has been kept roughly constant such that the nominal break-up regime is not affected by an increase in mean shear at the liquid-air interface. Microscopic high-speed shadowgraphy is applied to resolve the evolution of the droplets during the atomization process. Results indicate a significant effect of velocity fluctuations on the primary atomization of the liquid jet when increasing the turbulence intensity from approximately 7% to 45%. Further processing of the images will be conducted in the future in order to obtain a more quantitative statistical analysis of this process.
1 Introduction The preparation of a spray for use in combustion applications generally involves two liquid break-up stages, referred to as primary and secondary atomization. The latter stage involves the break-up of droplets into further smaller droplets and this can occur in a variety of ways depending on the droplet Weber Wed and Ohnesorge numbers Ohd. Low droplet Weber numbers will generally only yield minor droplet deformation or oscillatory deformation, while higher Weber numbers (Wed>11) will result in break-up [1-2]. The full definition of these numbers is widely available in the literature [1-3]. The focus of this paper is on primary atomization, which involves the break-up of a liquid jet or sheet into fragments. The same dimensionless parameters nominally control this process although defined with respect to the bulk liquid properties with the liquid jet Reynolds number Rej , the liquid jet Weber number Wej and the jet Ohnesorge number Ohj given by (1)-(3) respectively [3].
Re j
l u d l
(1)
We j
l u 2d l
(2)
Oh j
l l l d
(3)
At high liquid jet velocities and thus Reynolds numbers, significant shear may develop at the liquid jet/air interface to move the break-up regime from a laminar one to an air assisted one [3-4]. Further increase in the liquid jet velocity will ultimately lead to a turbulent break-up regime, where the turbulence generated in the liquid core acting in combination with air assistance at the liquid-air interface will further fragment the jet [1,5]. Prediction of liquid jet break-up becomes more complex when the surrounding air is not stagnant. Such a situation is particularly relevant in practical spray combustors which involve the delivery of a liquid jet into a turbulent gas flow. Industrial atomization systems such as air-blast and effervescent atomizers, both very commonly used in gas turbines, utilize non-stagnant air to accelerate the atomization process [6-8]. Other relevant non-dimensional parameters now include the surrounding air Reynolds number Rea as well as factors such as the ratio of liquid jet to co-flowing air velocity [6]. The primary atomization characteristics of air assisted atomization are fairly well documented [1,6] and the impact of the surrounding air is also understood [1,6,8,9]. However, there is a severe lack of simple systematic studies investigating the effect of gas phase turbulence intensity on primary and secondary atomization. Kolmogorov [10] and Hinze [11] have studied the effects of turbulence on secondary atomization and suggested that any droplet larger than the Kolmogorov scale will exhibit turbulence assisted atomization. Kourmatzis and Masri [12] have recently studied the effects of turbulence on secondary atomization in a simple pipe geometry [12]. Of particular importance is to experimentally investigate the effect of turbulent fluctuations on atomization, whilst keeping mean flow conditions constant. This ensures that the liquid jet is in the same nominal break-up regime in all cases with only a superposition of velocity fluctuations contributing to atomization. In this paper we suggest an experiment that may be utilized to investigate the effect of turbulence on primary atomization (as well as secondary atomization, although this aspect is not examined here). Three grids are used to generate different levels of turbulence subjected to laminar liquid jet. A base case is also studied where no grid is introduced in the flow. Laser Doppler anemometry is used to characterize the flow behavior of the wind tunnel, while microscopic high speed shadowgraphy is employed to visualize the liquid jet evolution as a function of turbulence intensity.
2 Experimental Methods Corresponding author. Tel: +61 2 9351 2835 E-mail address:
[email protected]
A general layout of the experiment is shown in Figure 1 and
consists of a small wind tunnel (54x54mm) which is fed air via a fan resulting in a mean velocity at the exit plane of approximately 17m/s. A 27 gauge syringe with an inner diameter of 210µm is mounted in the tunnel such that the tip of the syringe is just behind the turbulence grid or the exit of the wind tunnel. The arrangement is chosen such that the coflowing air flows around the ejected liquid jet and imparts velocity fluctuations onto the jet.
fixed such that the gaseous Reynolds number is of the order of 50,000.
A ‘coarse’ turbulence grid with cylindrical bars at a square pitch D of 15mm in addition to a ‘fine’ grid with a square pitch D of 10mm was utilized to generate isotropic decaying turbulence. A fractal grid was also used in order to significantly increase the turbulence intensity as demonstrated by [13] and results from these grids are compared to the case with no grid. The repeating fractal grid structure is depicted in Fig. 2 where D was chosen as the representative length-scale. The wind tunnel was characterized by seeding the flow with silicon-oxide powder which contained a nominal powder size of less than one micron such that the seed responds to the gas phase instantaneously. Seed velocities were determined using a commercial laser Doppler anemometry (LDA) system (TSI Model FSA 3500/4000). For brevity, full information on the geometric layout and associated errors is not provided here but is available in the literature [14].
Figure 1. Experimental lay-out showing wind tunnel, liquid delivery method and imaging technique. The liquid jet and generated droplets were visualized using microscopic high speed imaging in a shadowgraph arrangement as shown in Fig. 1. A high speed (10kHz) Nd-YAG laser was used as the light source, which was diffused through a series of opal glass diffusing optics. The CMOS high speed camera was used in conjunction with a Questar QM-100 long distance microscope to give an optical resolution in these experiments of 3.4 µm with a field of view of 2.6mm at the imaging plane.
3 Results The results to be displayed are all for a constant liquid jet velocity equal to 1.6m/s, where acetone was used as the test liquid due to its low surface tension (.025N/m) and well defined properties. The liquid Ohnesorge and Reynolds numbers are 0.005 and 842 respectively placing the liquid in a laminar Rayleigh break-up regime [3]. The tunnel velocity is
Figure 2. Fractal grid utilized for generation of turbulence
3.1 Wind tunnel characterization Of main importance in these experiments is to ensure a constant mean velocity at the exit plane centerline where the jet is located, such that the mean boundary condition is as similar as possible yielding the same nominal break-up regime. Figure 3(a) shows that this has been achieved for the case of no grid, for the coarse grid and the fine grid. Due to the geometry of the fractal grid it is not possible to achieve the same mean axial velocity at the centerline. This is attributed to the degree of blockage that exists with that grid. However at x/D=5 it can be seen that all mean velocities decay to a value that varies between 14 and 17 m/s. The abrupt increase and decrease in velocity in the coarse and fine grid cases at x/D=0.5 occurs because of the presence of the bars which change the local mean velocity periodically. The turbulence intensity at the exit plane increases from 7%-19%-20%-45% for the no grid, coarse, fine and fractal cases respectively. At x/D=5 the centerline rms trend remains the same however now with the fractal turbulence intensity dropping to approximately 16%. The radial fluctuation results of Fig. 3(c) are a further indicator of how likely the jet is to undergo random radial displacements. It is clear that for the case of no grid the normalized fluctuation is less that 10% while for the coarse and fine grids the intensity is at 20% and at almost 60% for the fractal geometry. The characterization results indicate that any difference in the primary break-up behavior of the liquid jet will be attributed solely to random turbulent fluctuations which impart energy onto the liquid jet surface. This is in contrast to how further break-up is achieved in conventional experiments or atomizers which simply involves an increase in the jet velocity.
3.2 Jet break-up imaging Figure 4 shows representative sequences of liquid jet break-up for four cases where no grid, coarse, fine and fractal grids were used as shown in Fig. 3. In all cases the lens is focused on to the break-up region. A threshold has been applied to all images such that the background is fixed at a constant 1000 counts. This is made even easier due to the shadowgraph technique which results in zero counts whenever a large droplet crosses the imaging plane. The images of Fig. 4(a) show a typical break-up pattern of the laminar acetone jet in the wind tunnel without any grid. The effect of the co-flowing air is to
bring the break-up location much closer to the liquid jet exit plane which is a known result. The turbulence intensity present in the tunnel in this instance is not enough to notably change the break-up structure from a conventional Rayleigh breakup `dripping’ mode break-up. The droplets that issue from the jet are both deformed and spherical and thin ligaments are seen to form which will occasionally result in smaller droplets. Figure 4(b) shows the effect of the coarse grid. These frames indicate how the turbulent energy present in the grid is imparting a slight radial sinusoidal break-up instability, as opposed to Fig. 4(a) where the break-up of the jet is vertically straight. This radial instability is quite apparent in the fine grid case of 4(c) which shows the radial jet fluctuation occupying more of the field of view with more jet thinning and smaller droplets forming. Further deformed droplets are also present in this case.
x/D=0.5
atomization process. Figure 5 shows sequences of images in time at x/D=5. The fine grid case of Fig. 5(b) indicates a larger population of deformed droplets and this is confirmed through observation of further frames not shown here. The droplets of Fig. 5(b) also seem better dispersed when compared to Fig. 5(a) which is the case of the coarse grid. While the turbulence intensity is very similar between the fine and coarse cases, the absolute axial and radial velocity rms is greater for the fine grid case. Figure 5(c) shows the cases for the fractal grid. Via observation of these frames and further ones not shown here this case exhibited the most significant dispersion. However, droplets generally seemed larger and more spherical when compared to the fine and coarse grid variants. The precise reason for this is not yet known and requires further investigation. What is clear from these images is that turbulence has a very pronounced effect on the radial fluctuation of the jet at the exit plane therefore affecting the degree of radial dispersion further downstream.
x/D=5
a
a
b
c b
d Figure 4. Typical time sequences of liquid jet break-up for the case of no grid (a) for the case of a coarse grid (b) for the case of a fine grid(c) and for the case of the fractal grid (d) at the exit plane of the tunnel.
c Figure 3. Mean axial velocity (a), axial turbulence intensity (b) and normalized radial fluctuation (c) at x/D=0.5 (left) and x/D=5 (right)
It is worth noting that while the turbulence did result in accelerated break-up it can also lead to situations where colliding droplets simply coalesce together forming somewhat larger droplets. Such events are present in Fig. 4, and Fig. 6 shows a coalescence event for the fractal grid case.
The fractal grid shows a very similar pattern to the fine grid case even though its turbulence intensity is significantly higher. This may be because the mean axial velocity for the fractal case is slightly lower as seen in Fig. 3(a) resulting in lower shear at the liquid-air interface. However, given that this is the case and the jet is still exhibiting such a significant fluctuation indicates how efficient turbulence is at accelerating the primary
It is uncertain if turbulence in this particular scenario acts to increase or decrease the probability of droplet coalescence. While the conditions required for coalescence have been investigated previously [15], the statistical probability of such a phenomenon occurring in a well defined turbulent flow has not been rigorously studied. This is a possibly avenue of further research which would have very important implications in
spray combustion.
4 Conclusions and future work An experiment has been constructed which involves imparting turbulence onto a liquid jet in order to observe the changes that occur in the atomization process. While keeping the mean flow constant and increasing the rms it is clear that fluctuations move the break-up regime from a conventional ‘dripping’ mode to a sinusoidal break-up which results in the presence of further deformed droplets and smaller ligaments at the primary breakup location. These smaller ligaments eventually result in smaller droplets downstream. Droplet collisions and coalescence have also been observed in all of the turbulent scenarios. Future work must include a statistical analysis of the images such as computing the two dimensional probability density functions while also acquiring a statistical picture of how the number of deformed droplets correlates with the degree of turbulence intensity in the flow. The work is of great importance in spray combustion applications as the effect of turbulence on atomization shall be isolated.
a
b
c Figure 5. Typical sequences of droplet dispersion for the case of a coarse grid (a) a fine grid(b) and a fractal grid (c) at x/D=5.
Figure 6. Coalescence of droplets occurring in the fractal generated turbulence case.
5 Acknowledgments The authors thank Mr. Oliver Smith for constructing and designing portions of the experimental rig. Mr. Phuong X.
Pham and Dr. Vinayaka Prasad are thanked for assistance with the experiments and for fruitful discussions. The authors are funded by the Australian Research Council.
References [1] Faeth, G., Hsiang, L., and Wu, P. International Journal of Multiphase Flow 21 (1995) 99-127. [2] Guildenbecher, D.R., Lopez-Rivera, C., and Sojka, P.E., Experiments in Fluids 46 (2009) 371-402. [3] Lefebvre, A.H., Atomization and Sprays, Hemisphere Publishing Corporation, 1989. [4] Dumouchel, C., Experiments in Fluids 45(2008) 371-422. [5] Sallam, K.A., Dai, Z., and Faeth, G.M., International Journal of Multiphase Flow 28 (2002) 427-449. [6] Engelbert,C., Hardalupas, Y., and Whitelaw, J., Proceedings Of the Royal Society A 451 (1995) 189-229. [7] Sovani, S.D., Sojka, P.E., and Lefebvre, A.H., Progress in Energy and Combustion Science 27 (2001) 483-521. [8] Konstantinov, D., Marsh, R., Bowen, P., and Crayford, A., Atomization and Sprays 20 (2010) 525-552. [9] Wu, P.K. and Faeth, G.M., Atomization and Sprays 3 (1993) 265-289. [10] Kolmogorov, A.N., Doklady Akademii Nauk SSSR 66 (1949) 825-828. [11] Hinze, J.O., American Institute of Chemical Engineering, 1 (1955) 289-289. [12] Kourmatzis, A., and Masri, A.R., Turbulent Secondary Atomization of non-evaporating dilute spray jets, 12th Triennial ICLASS conference, Heidelberg, Germany, 2012. [13] Vassilicos, J.C., and Seoud, R.E., Physics of Fluids, 19 (2007) 105108 (1-11). [14] Oloughlin, W., and Masri, A.R., Combustion and Flame, 158 (2011), 1577-1590. [15] Orme, M., Progress in Energy and Combustion Science, 23 (1997), 65-79.
