Measurements of Droplets Spatial Distribution in Spray by Combining Focus and Defocus Images

Transcription

1 Measurements of Droplets Spatial Distribution in Spray by Combining Focus and Defocus Images Kentaro HAASHI 1*, Mitsuhisa ICHIANAGI 2, Koichi HISHIDA 3 1: Dept. of System Design Engineering, Keio University, okohama, Japan 2: Dept. of Engineering and Applied Sciences, Sophia University, Tokyo, Japan 3: Dept. of System Design Engineering, Keio University, okohama, Japan * correspondent author:, hayashi@tfe.sd.keio.ac.jp Abstract: The measurement accuracy of droplets position, size and velocity in a spray flow was improved by combining interferometric laser imaging for droplet sizing (ILIDS) with glare-points technique (GPT). For the combination of ILIDS and GPT, we have developed the advancing optical measurement system, called as Double-planes Particle Imaging (DPI). DPI has an advantage to simultaneously capture both focus images (glare-points) and defocus images (interferometric fringes) by using only one imaging device. For the determination of droplets position, the center position of glare-points was adopted as the actual droplet position. For the calculation of droplets sizing, we need to multiply the spatial frequency of fringe pattern with the fringe length. For reducing the ambiguity of the fringe length, the estimation technique for the fringe length was proposed, which was calculated from the vertical distances between the focus and defocus images. The relationship between fringe length and vertical distance was estimated by the ray tracing method. The velocity was calculated by PTV with the aid of droplet information by ILIDS. The measurement accuracy of droplets sizing for the DPI system was validated by evaluating the diameter of monodispersed droplets. Compared with ILIDS and DPI, the bias error was extremely improved. Furthermore, DPI was applied to the actual spray flow, and the spatial distributions of droplets position, size and velocity were measured. These results indicate that the histogram of droplet diameter was similar to the chi-squared distribution, because the probability of large size droplets was reduced by improving the measurement accuracy of droplets sizing especially in the droplet size area of more than 15 µm. 1. Introduction A liquid spray is widely used for many industrial fields, such as fuel injections in engines, inhalers in medical applications, ink-jet printers and so forth. For further enhancement of industrial devices using sprays, the size and velocity distribution of droplets in a spray are required, because this information enables us to understand the unsteady behavior of droplets and surrounding flow structure. Various optical measurement techniques have developed to investigate the flow structure of sprays. One of the most common measurement techniques is Phase Doppler Anemometer (PDA). This technique gives the accurate data of droplet size and velocity by extracting information from the scattered light. However, the PDA is a pointed measurement technique, so that it is difficult to obtain droplets spatial distributions in unsteady spray flows. Interferometric laser imaging for droplet sizing (ILIDS) is a promising technique to provide the spatial distribution of droplets position, size and velocity simultaneously. This technique captures the interference patterns generated from the reflected and refracted lights in the forward scattering regions (Glover et al. 1995), because the angular frequency of the interference pattern or the number of fringe depends on the size of particles (Hesselbacher et al. 1991). Maeda et al. (2, 22) and Kawaguchi et al. (22) developed the optical system to compress interference images, which can dramatically increase the droplets number density per an image, in order to solve the problem to overlap fringe - 1 -

2 pattern images in dense particle areas. In addition, they measured the droplets diameter and velocity simultaneously by combining ILIDS with Particle Tracking Velocimetry (PTV). Recently, several researchers pointed out that ILIDS has two problems about the measurement accuracy of droplets position and sizing. Hardalupas et al. (21) noted that the droplet center discrepancy is found between focus and defocus planes, which was made clear by using the geometrical optical method. The center discrepancy is several times larger than droplet diameter, which causes the calculation errors of droplets velocity and mass flux. Our previous work (Shigeta et al. 21, 212) presented that ILIDS has the error of droplet size detection, because of the ambiguity of the fringe length in interference patterns. To solve the above two problems, Shigeta et al. (212) developed a new measurement technique called DPI (Double-plane Particle Imaging). DPI combined ILIDS with glare-points technique (GPT), which simultaneously captures focus images (glare-points) and defocus images (interferometric fringes) by using only one imaging device. The errors of droplets position and sizing were corrected by using the droplets diameter and the center position obtained from GPT, respectively. This technique has a powerful tool to overcome the problems of ILIDS. However, the measurement accuracy of droplets sizing of GPT is lower, because the distance between glare-points is proportional to droplets diameter, and its accuracy dominates the spatial resolution of the imaging device such as a CCD camera. The objective of this work is to develop the new analytical algorithm of interference patterns from the DPI system for the estimation of droplets position, sizing and velocity. For the determination of droplets position, the center position of glare-points was adopted as the actual droplet position. For the calculation of droplets sizing, we need to multiply the spatial frequency of fringe pattern with the fringe length. For reducing the ambiguity of the fringe length, the estimation technique for the fringe length was proposed which was calculated from the vertical distances between the focus and defocus images. The relationship between fringe length and vertical distance was estimated by the ray tracing method. The velocity was calculated by PTV with the aid of droplet information by ILIDS. The measurement accuracy of droplets sizing was confirmed to measure a monodispersed droplets diameter by ILIDS and DPI. Furthermore, DPI was applied to the actual spray flow, and the spatial distributions of droplets position, diameter and velocity were measured at several points in an air atomizing spray. 2. Improvement of ILIDS 2.1. Principles of ILIDS and GPT When light illuminates a spherical transparent droplet, the intensity of the reflected and the first order refracted light from the droplet are stronger than that of higher order refracted light in the forward region for the scattering angle between around 3 to 8 degrees. In the focal plane, two bright spots, which are derived from these two lights and known as glare points, are observed (van de Hulst and Wang 1991). On the other hand, in the defocus plane, these two lights interfere with each other, and a parallel interferometric fringe is observed as shown in Fig. 1. A measurement technique using focus images is called Glare-Points Technique (GPT), and a technique using defocus images is called as Interferometric Laser Imaging for Droplet Sizing (ILIDS). In GPT, a diameter of spherical transparent droplet, D, is calculated from the following - 2 -

3 equation, which was derived from the geometrical optics, D= 2GP E, where E cos θ msin θ m 1 2mcos θ = + +, (1) where GP is the distance of glare-points, m is the relative refractive index of the liquid droplet, and θ is the scattering angle. Equation (1) is available when qualifying the following condition. πd x = p p > 2, (2) λ where d p is the particle diameter, λ is the laser wavelength. The planar positions of droplets are defined as center positions between these two bright dots, and the planar velocity-vectors are determined by calculating the deviation of center positions of glare-points in a short time. The advantage of this measurement system is that the mid-points of the glare-points correspond closely to the actual droplets center positions, and their discrepancies were estimated to be only.1d (Hardalupas et al. 21). However, GPT has the difficulty to detect the diameter of droplets of the order of 1 µm, because the distances between glare-points are narrow to need the sub-pixel evaluation in images. On the other hand, ILIDS measures the diameter of a spherical transparent droplet from the pattern frequency or number within the interferogram. The relationship between the droplets diameter and frequency of interferometric fringe is derived from the phase difference between reflected and the first order refracted light originated from one droplet. The droplet diameter, D, is obtained by the following equation, which is derived from the geometric optics, 2λN D= E, (3) α where N is the number of fringe, λ is the wavelength of the laser light source, m is the relativerefractive index of the liquid droplet, θ and α are scattering and collecting angles, Laser sheet 1 Droplet θ α Collecting lens Defocus plane Focus plane Interferometric fringe Glare-Points Fig. 1. Principle of glare-points and interferometric fringe pattern

4 Droplet Ray axis Cylindrical lens Droplet center discrepancy Fringe s center Glare-Points fringe Captured image on focus and defocus plane Intensity [-] Actual length [pixel] Focus plane Defocus plane Fig. 2. Problem of droplets position detection by ILIDS. Full length in an image [pixel] Fig. 3. Relationship between actual and full length of fringe pattern. respectively. Droplet planar position is determined as the center of the fringe pattern. The twodimensional velocity is calculated from consequent frames by using the cross-correlation method. In conjunction with the cross-correlation coefficient, the validity of the planar displacement estimation can be validated by comparing the droplet size obtained in the first frame with the size of the candidate droplets inside the predefined interrogation window. Maeda et al. (2) improved the ILIDS technique by developing the optical compressed system in order to overcome the circular interferograms overlapping problem in high particle concentration regions. Hardalupas et al. (21) pointed out that the droplet center discrepancy is found between interferometric fringe and glare-points by the geometric optics method. Figure 2 illustrates the horizontal center discrepancy at two droplet location. Their results indicate that the discrepancy depends on not only horizontal droplet position but also vertical and depth droplets position and degree of defocus. Shigeta et al. (21, 212) noted that ILIDS has the error of droplet size detection. For the calculation of droplets sizing, we need to multiply the spatial frequency of fringe pattern with the fringe length. As shown in Fig. 3, the full length of fringes includes fringes from noise areas, which causes the ambiguity of the fringe length in interference patterns. 2.2 Present system based on DPI In order to overcome the above two problems, a new algorism was developed using the Double-planes Particle Imaging (DPI) system. DPI combined ILIDS with GPT, and simultaneously can capture both focus image (glare-points) and defocus image (interferometric fringe) by only one imaging device. Figure 4 shows the overview of the present algorithm using the DPI system, and the detailed procedure of the algorithm is summarized in chapter 3. The droplets position was corrected by adopting the position data of glare-points image center. The measurement accuracy of droplets sizing of ILIDS was improved by reducing the ambiguity of the fringe length detection, which corrects the number of fringe pattern. The droplets velocity was calculated from the displacement of glare-points image center. A pair of droplets images in subsequent photos was matched with the aid of the interferometric fringe pattern from ILIDS

5 t =t 1 t =t 1 +Δt Position: GPT Diameter: ILIDS To adopt Glare-points center as a droplet position To calculate fringe number N by ray tracing Fig. 4. Overview of present algorithm using Double-planes Particle Imaging (DPI) system. 3. Optical Arrangement and Image Processing 3.1. Configuration of optical system For the combination of ILIDS and GPT, the receiving optical system called as DPI system was developed as shown in Fig. 5(a). The optical glass ((1) in the figure) and two cylindrical lenses ((2): curvature radius = ±51.9 mm) were equipped at the positions between the collecting lens (Nikon Corp., focal length = 15 mm) and the imaging device ((3): CCD camera (Imperx ICL 252 SC, pixels, 3 fps)). As shown in Fig. 5(b), the imaging device was set with sheimpflug arrangement, and the scattering angle and collecting angle were set as 73 degrees and 6.53 degrees, respectively. The total length of this system (the length from optical axis to imaging device) was 412 mm. The distance from optical axis to the optical lens was 219 mm. The distance from the center of laser sheet to collecting lens was 185 mm. The degree of defocus was 11. mm. The refracted and first-order refracted scattered lights from droplets can transmit two ray paths guided by the apertures, as shown in Fig. 5(c). In the top circular aperture (diameter = 3.5 mm), both scattered lights pass through the optical glass and produce two glare-points in a focus plane. On the other hand, in the middle rectangular aperture (3 mm 25 mm), both scattered lights go through two cylindrical lenses and provide interferometric fringe patterns in a defocus plane, whose images are vertically compressed. The outside dimension of the optical glass is equal to that of the cylindrical lenses, which gives that the optical path lengths of focus images are the same as those of defocus images. Thus, focus and defocus images from droplets, illuminated by a laser sheet, can be captured by using only one camera, as shown in Fig. 5(d). In this study, the measurement area was 1 1 mm, and the thickness of laser sheet was approximately 2 mm

6 Receiving optical system Droplet Ray axis Droplet Collecting angle (1) Optical glass (1) Scattering angle Aperture + Collecting lens (2) Cylindrical lens (3) CCD with sheimpflug arrangement (2) (a) (3) (3) (1) (b) Focus image (2) (c) (d) Defocus image Fig. 5. (a) Schematic of DPI system. (b) Top view and (c) side view of optical path lines by DPI system. (d) One sample image including glare-points (focus image) and interferometric fringe pattern (defocus image) captured by DPI system Algorithm of image processing The algorithm of image processing of the DPI system is shown in Fig. 6. Firstly, fringe patterns and glare-points are individually detected from an image. Secondly, the center positions are calculated from fringe patterns and glare-points, and its fringe frequencies are obtained from fringe patterns. Thirdly, as shown in Fig. 7, a fringe length, L, is proportional to a vertical distance, s, between a fringe pattern and glare-points (Willert and Gharib 1992). The calibration curve between L and s was estimated by using the ray tracing method, as shown in Fig. 8. The actual fringe length was estimated by calculating the vertical distance from the image processing and using the calibration curve in Fig. 8. Fourthly, the droplets position, size and velocity were calculated. The droplets position was determined as the center position of glare-points, because discrepancies between glare-points and actual droplets positions are less than 5%, compared to the droplet size. For the droplets sizing, the fringe number was calculated by multiplying a fringe length and a fringe pattern frequency, and the droplets size was calculated by substituting the fringe number for equation (3). The droplets velocity was calculated from the displacement of center positions of glare-points which was based on the PTV method with the aid of fringe patterns. Finally, the droplets position and velocity-vectors were transformed into actual coordinate by using the calibration curve between the camera s and actual coordinates

7 - 7 -

8 Fringe pattern Glare -points Ray tracing Center position Fringe frequency Center position Calculate fringe length by estimating depth position Calculate position, droplet size and velocity Coordinate transformation Fig. 6. Flow chart of algorithm of image processing of DPI system. Object plane Aperture Collecting lens Imaging plane s 1 L 1 L 2 s: Vertical image distance [mm] L: Fringe length [mm] Fig. 7. Relationship between fringe length, L, and vertical distance, s, between fringe pattern and glare-points. s (1) x [mm] [mm] y [mm] [mm] (2) (1) z [mm] (3) frin ge GP i [pixel] j [pixel] Fringe length L [mm] Vertical disance s [mm] -5-1 (2) (3) z [mm] Fig. 8. Procedure to obtain calibration curve between fringe length, L, and vertical distance, s

9 (a) (b) Mono disperse generator Droplets Air atomizing nozzle x Optical glass CCD camera Collecting lens (f =15 mm) and aperture Cylindrical lenses Spray y Fig. 9. Experimental set up for (a) validation of droplet sizing and (b) application to actual spray Experimental condition The measurement accuracy of droplets sizing using the DPI system was validated by measuring the diameter of monodispersed droplets. The droplets were generated by using the mono disperse generator as shown in Fig. 9(a), and their true diameter was measured by GPT and evaluated to be 179 µm. In order to clarify the availability of the proposed system, the results from DPI were compared with those of DPI. The DPI system was applied to measure the spatial distributions of droplets position, size and velocity in an air atomizing spray, and compared with ILIDS. The spray was generated by an air atomizing nozzle (H. Ikeuchi & Co., Ltd, BIMJ222), and the water and air pressure was set to be.5 bar and.1 bar, respectively. The spatial distributions of droplets were measured at (, ) = ( mm, 3 mm), ( mm, 4 mm), and ( mm, 5 mm) in Fig. 9(b). 4. Results and Discussion 4.1 Validation of measurement accuracy of droplet sizing For the validation of measurement accuracy of droplets sizing, the droplets with the diameter of 179 µm were measured by using ILIDS and DPI. Figure 1 shows the histograms of droplets diameter from ILIDS and DPI. The result indicates that the peak value of histogram from DPI is significantly close to the true value, compared with ILIDS. Furthermore, the measurement uncertainty of droplets sizing of ILIDS and DPI was evaluated and summarized in Table 1. It was obvious that the bias index of DPI was drastically improved. This was caused by reducing the ambiguity of the fringe length from the area of image noise. 4.2 Application to actual spray The spatial distributions of droplets position, sizing and velocity in an actual spray flow were measured by ILIDS and DPI. Figure 11 shows the histograms of droplets diameter at (x, y) = ( - 9 -

10 Probability[-] D = 179[µm] ( validated by GPT) ILIDS Present work Diameter [µm] Fig. 1. Histograms of droplets diameter measured by ILIDS and DPI. Table 1. Measurement uncertainty in 95% confidence level of droplets sizing of ILIDS and DPI ILIDS mm, 3 mm), ( mm, 4 mm), and ( mm, 5 mm). The origin was located at the nozzle exit. It was found that the histogram of droplet diameter form DPI was similar to the chi-squared distribution, because the probability of large size droplets, especially more than 18 µm, was reduced by improving the measurement accuracy of droplets sizing. Figure 12 indicates the spatial distribution of droplets position, sizing and velocity, located at (x, y) = ( mm, 4 mm). The droplets positions from ILIDS corresponded to those from DPI at x-location ranging from x = 1. mm to x = 2. mm. On the other hand, the droplets positions from ILIDS were deviated away from those from DPI at around x = 2.5 mm or 5. mm. Thus, the droplets position error from ILIDS was improved by using the DPI system. DPI Averaged diameter [µm] Bias index [µm] 47 1 Precision index [µm] Measurement uncertainty in 95% confidence level [µm] Conclusions A new measurement technique called as Double-plane Particle Imaging (DPI) was developed by combining ILIDS and GPT, for the improvement of measurement accuracy of droplets position, size and velocity. The system enables to simultaneously capture focus and defocus images by using only one CCD camera. The center position in focus image was determined as the actual droplets position, which causes to improve the droplets position error of ILIDS. The ray tracing method was - 1 -

11 Probability[-] (, ) = ILIDS DPI (,3) (,4) (,5) Diameter [µm] Fig. 11. Histograms of droplet diameter measured by ILIDS and DPI. [mm] [mm] 25 [µm] 1. [m/sec] ILIDS Fig. 12. Spatial distribution of droplets position, diameter and velocity at (x, y) = ( mm, 4 mm). employed to estimate fringe lengths in defocus images, which causes to improve the measurement accuracy of droplets sizing. The DPI system was applied to measure the diameter of the monodispersed droplets and the spatial distribution of droplets in a spray flow. The important conclusions obtained from this work are summarized below. (1) The measurement accuracy of droplets sizing of DPI was validated by measuring the diameter of monodispersed droplets. The bias index of DPI was significantly reduced, compared with ILIDS. This was caused by reducing the ambiguity of the fringe length from the area of image noise and improving the measurement accuracy of droplets sizing. (2) The DPI system was applied to measure the spatial distributions of droplets position, size and velocity in a spray flow, and compared with ILIDS. The histogram of droplet diameter form DPI was similar to the chi-squared distribution, because the probability of large size DPI

12 droplets was reduced by improving the measurement accuracy of droplets sizing. Furthermore, the droplets position error from ILIDS was improved. Acknowledgements This work was subsidized by Grant-in-Aid for Scientific Research (S) (No ) of Japan Society for the Promotion of Science. References Glover AR, Skippon SM, Boyle BD (1995) Interferometric laser imaging for droplet sizing method for droplet-size measurement in sparse spray systems. Appl Optics 34: Hardalupas, Sahu S, Taylor AMKP, arogoulidis K (21) Simultaneous planar measurement of droplet velocity and size with gas phase velocities in a spray by combined ILIDS and PIV techniques. Exp Fluids 49: Hesselbacher KH, Anders K, Frohn A (1991) Experimental investigation of Gaussian beam effects on the accuracy of a droplet sizing method. Appl Optics 3: Kawaguchi T, Akasaka, Maeda M (22) Size measurements of droplets and bubbles by advanced interferomentric laser imaging technique. Meas Sci Technol 13: Maeda M, Akasaka, Kawaguchi T (22) Improvements of the interferometric technique for simultaneous measurement of droplet size and velocity vector field and its application to a transient spray. Exp Fluids 33: Maeda M, Kawaguchi T, Hishida K (2) Novel interferometric measurement of size and velocity distributions of spherical particles in fluid flows. Meas Sci Thecnol 11:L13-L18. Shigeta, Ichiyanagi M, Hishida K (21) Practical improvement of spatial droplet distribution by interferometric laser imaging technique. 15th Int Symp on Applications of Laser Techniques to Fluid Mechanics, Lisbon, Portugal, 21. Shigeta, Hayashi K, Ichiyanagi M, Hishida K (212) Measurement of droplet size, velocity and spatial distribution of mass flux in spray by combining focus and defocus imaging technique. Trans JSME B 78, No.788:867-88, (in Japanese). van de Hulst HC, Wang RT (1991) Glare points. Appl Optics 3, No.33: Willert CE, Gharib M (1992) Three-dimensional particle imaging with a single camera. Exp Fluids 12: