Experimental Technique to Measure Droplets Diameter Distribution in Spray

Transcription

1 Experimental Technique to Measure Droplets Diameter Distribution in Spray. Tomassi, T. Pagliaroli, A. Di Marco, G. Aloisio, R. Camussi, E. Giulietti, A. erretti 3. Industrial and Mechanical Eng. Dept. (DIMI) - University Roma TRE, Rome - ITALY. TER-ENE-IMP - ENEA Casaccia, S.M. Galeria, Rome - ITALY 3. Kempro International S.r.l. - Ariccia, Rome - ITALY. Introduction In many industrial and fossil energy processes, concentration, size and velocity distributions of solid or liquid particles suspended in any clear fluid play an important role. Therefore, there is a need for obtaining fundamental non-intrusive measurements which include the above-mentioned parameters [-3]. The objectives of this study are to develop a digital image technique potentially capable of sizing two-phase (solid-liquid) particles of arbitrary shape, size and spatial distributions and to understand the effects of the slurry composition (solid concentration) and flow condition (injection pressure) variations on the diameter distribution. Therefore, a test plant, able to atomize the slurry at various injection conditions has been designed and manufactured. The test plant has been mounted in the Thermo-luid Dynamics and Aerodynamics Section of the Mechanical and Industrial Engineering Department (DIMI) of the Roma TRE University. The droplet-sizing technique works by acquiring scattered signals of the spray using a highresolution camera [4,5]. The scattered images are obtained by illuminating a section of the spray cone with a short pulse laser sheet. The short pulse from the laser freezes the motion, allowing visualization of drop size and shape. Images from the digital camera are transferred to a computer and high-speed particle sizing homemade software (RaM low: Reactive and Multiphase low) analyses the images obtained in order to build up the diameter distribution and the morphology characterization [6].. Image processing The general procedure of the method used in this study for the analysis of droplets images involves quite different image processing tasks [7,8]. The main tasks are: extraction of the green component from the RGB image, image pre-processing using morphological operation, suppression of noise, thresholding, edge detection and as final step, (not yet implemented) image quality analysis of the droplets. The first step is the extraction of green component from the RGB image,, and then a preprocessing operation that eliminates the non-spherical particles. The image noise is suppressed smoothing the image with a Gaussian convolution filter. Gaussian filtering is mathematically expressed as a convolution integral of the raster image and a kernel ( K ), that in discrete form becomes: = i + k, j+ w K + k, + w k = w=, K = 6 4 X-,

2 3st Meeting on Combustion In order to distinguish droplets side scattering from background within the image an optimal threshold is applied. The applied threshold is set at a green level between the two peaks corresponding to the droplet and image background. In particular the optimal threshold is chosen applying the Otsu s method [7]. In this method the threshold level k is the level that maximizes the difference between ω(k) and μ(k), where those values are calculated with: k L p i ω ( k) = p i ; ( ) μ k = i= i= k + Where p i is the probability of a pixel having green level i and L is the number of green scales, so that L- is the largest [7]. The edges are detected applying a first-order Prewitt derivative operator (ig. ). The operator is applied in both x- and y-direction. The two resulting images are squared and summed, finally the square root of the sum is computed [8] and another threshold is applied to extract an annular area. The threshold is mathematically expressed by [9]: = m n m i= n j=, bin, =, < ( inside the droplet) ( outside the droplet) Then the contour of the annular area is calculated with a box count algorithm. The droplets diameter is calculated operating a circular fitting on the contour. a) b) c) d) e) f) ig. Edge detection: a) Green layer D-intensity curve of the droplet; b) Green layer Dintensity curve after the edge detection filtering; c) Droplet image; d) Green layer intensity profile of the droplet at x=6 pix; e) Green layer intensity profile at x=6 pix after the edge detection filtering; f) Comparison of the original and filtered intensity curve. X-,

3 Italian Section of the Combustion Institute The software also calculates statistic quantities such as mean, standard deviation, skewness, kurtosis, Sauter-mean diameters (SMD), maximum diameter. The diameter distribution is also expressed in terms of Probability Density unction (PD) and Cumulative Density unction (CD). urthermore it approximates the volume CD with the Rosin-Rammler distribution [,]. 3. Experimental Set-up The experimental setup employed in the present study includes a test chamber, a pressurized biphasic fluid supply system, an injection assembly, a liquid collection and suction system and a visualization, acquisition and analysis system. A water and carbonate mixture with variable density and viscosity is used as biphasic fluid. The biphasic fluid is first pressurized by a screw pump and then pumped into the pressure swirl atomizer. The. kw screw pump could elaborate biphasic fluids with high solid concentration, up to 6%, and high molecular viscosity. urthermore, a pressure reducer valve with a 8-6 bar range is mounted between the screw pump and the atomizer. The pressure swirl atomizer is shown in figure together with its main component. In particular, the figure.c shows the two swirl generators used in the experimental tests. a) b) c) d) ig. Pressure Swirl Atomizer: a) Nozzle; b) Two Swirl Chamber; c) Swirl generators d) assembled atomizer. The test chamber is the clear Plexiglas box showed in figure 3. It is a box of m side with panels.5 m thick. The two-phase fluid sprayed into the test chamber is collected into a reservoir by the drain at the end of the inclined flat plate (see fig. 3). inally the hydraulic circuit is closed by a tube which connects the reservoir to the screw pump. Spray support low direction Slide & ½ Drain ig. 3 Plexiglass Cube. X-, 3

4 3st Meeting on Combustion The analysis of the droplet distributions from the spray nozzle involves the use of a system consisting of a high-resolution digital camera, laser, mirrors, optics, data analysis software, and computer. The camera, a Canon EOS 35 D, has a resolution of over 8 megapixels. The laser has a beam length of 53 nm (green) and a maximum frequency of Hz. The laser can also pulse twice during a single camera frame to produce a double image to be used in determining the velocity field. A series of two 45 mirrors elevates the laser beam to the spray axis height. A thin light sheet of high intensity is generated combining a cylindrical and a spherical lens. The light sheet intersects a selected planar section of the nozzle spray. The scattered light of the spray droplets is detected by the camera positioned on the upper side of the box at.5 m, about a half the height of the box, the minimum allowable distance from the spray section. A general schematic of the experimental set-up is shown in igure 4. Digital Camera Spray Nozzle Lenses and mirrors Test Box Laser ig. 4 General schematic of the experimental Set-up. The focusing distance is obtained focusing the image on a target aligned with the spray axis and then calculating the meter/pixel ratio (scale factor). or the optical set-up used in the present investigation, the resolving power of the objective lens and the camera characteristics limited the interrogation area and the droplet sizing. As a matter of fact the size of the ield of View is.4 m x.5 m and the scale factor is μm/pix. 4. Results The droplet size distributions are measured for two atomizer configurations, three injection pressures and two densities ( test cases). The atomizer configurations differ in the swirl generator length (7 mm and mm). The complete test matrix is reported in table. igure 5 shows a comparison of the PDs obtained from analyzing a sequence of about 6 images (about droplets analyzed) for each test case. The variation of the SMD value for the mm swirl generator is shown in igure 6. As would be expected igure 6 reveals a reduction of the SMD as the injection pressure increases []. X-, 4

5 Tab. luid properties Italian Section of the Combustion Institute Atomizer (Swirl generator 7 mm) Atomizer (Swirl generator mm) ρ=6 kg/m 3.5 MPa MPa.5 MPa.5 MPa MPa.5 MPa ρ=5 kg/m 3.5 MPa MPa.5MPa.5 MPa MPa.5 MPa Test Matrix.. ρ=5 kg/m 3-7 mm. ρ=5 kg/m 3 - mm ρ=6 kg/m 3-7 mm. ρ=6 kg/m 3 - mm ig. 5 Diameter PDs obtained for the cases listed in the test matrix. Black lines.5 MPa; Blue lines MPa; Red lines.5 MPa. Dot lines individuate the SMD positions SMD (µm) Pressure ( 5 Pa) ig. 6 The SMD as a function of pressure for the mm swirl generator: Black line ρ=5 kg/m 3, Blue line ρ=6kg/m 3. X-, 5

6 3st Meeting on Combustion Less intuitive is the SMD variation with the fluid density. As reported in [], no reliable experimental data exist to test the SMD variation with the liquid density, but all the evidence suggests that the effect of this property is quite small, hence the SMD decrement with an augmented density in igure 6 is probably due to the effect of the surface tension. 5. Conclusions In the present work a digital image technique has been implemented to carry out measurements of droplets size and shape for different fluid properties and flow conditions. The technique was based on the analysis of scattered images, obtained illuminating a cone section of the spray with a sheet of light and acquired with a high-resolution digital camera. The software analysis consisted in a series of subroutines which elaborate the raw image applying a series filters, thresholding and tools in order to detect the in-focus particles and determine the diameter of the droplets. Subsequently a statistical analysis have been performed. The results here described have to be considered as preliminary. They have to be deepened through software improvements (more sophisticated detection tools and velocity field computation), software validation and further experiments needed to broaden the statistics and confirm the results. 6. Acknowledgements The authors gratefully acknowledge Eng. Jean-Sébastien Dunand, student of the INSA Lyon, for his collaboration during the program implementation and the experiments. 7. References. Lecuona, A., Sosa, P.A., Rodriguez, P.A., Zequeira, R.I.: Meas. Sci. Technol.,, 5-6 ().. Kadambi, J.R., Martin, W.T., Amirthaganesh, S., Wernet M.P.: Powder Technology,, 5-59 (998). 3. Nasr G.G., Yule A.J., Bendig L.: Industrial Sprays and Atomization Design, Analysis and Applications. Springer (). 4. Putirti A.D., Everest Jr D., Atreya A.: Simultaneous Measurement of Drop Size and Velocity in Large-Scale Sprinkler lows Using Particle Traking and Laser-Induced luorescence. NIST GCR 3-85, NIST report, (3). 5. Crowe C., Sommerfeld M., Tsuji Y.: Multiphase lows with Droplet and Particles. CRC Press, Boca Raton, LISA (998). 6. Blaisot, J.B., Yon, J.: Experiments in luids, 39, (5). 7. McAndrew A.: Introduction to Digital Image Processing with Matlab. Thomson Course Technology (4). 8. Jähne B.: Digital Image Processing. Springer-Verlag (995). 9. Pagliaroli T. et al: AIVELA 6 Conference, Rome, Italy, November (6).. Lefebvre A.H.: Gas Turbine Combustion. Taylor & rancis, Bristol (983).. Lefebvre A.H.: Atomization and sprays. Taylor & rancis, Bristol (989). X-, 6