Droplet Size Measurement with Linear Charge Coupled Device Camera

Transcription

1 Journal of NUCLEAR SCIENCE and TECHNOLOGY, 27[4], pp. 295~306 (April 1990). 295 Droplet Size Measurement with Linear Charge Coupled Device Camera Ken-ichi HAYASHIt, Power Reactor and Nuclear Fuel Development Corp.* Toshio KARAKAMA S. G. Engineering Co., Ltd.** Received July 7, 1989 Revised January 31, 1990 Automatic and remote-controlled measuring devices are required to detect liquid leakage from structures in reactor containment vessels. A new device has been developed to measure a droplet diameter based on the distance between two brilliant spots formed on a droplet surface by the reflection and refraction of a collimated light beam. The device is composed of a collimated beam source, a linear CCD (Charge Coupled Device) camera and a data processor. The measurement error is less than 1% in volume for a 5 mm diameter BK7 ball lens falling within a 200 mm horizontal width at a distance of m. This device is applicable to the measurement of droplet volume within a 200 mm diameter horizontal area with 10% accuracy. KEYWORDS: droplets, particle size, measuring method, charge coupled devices, cameras, leakage, collimated beam, light scattering, geometrical optics, ray tracing, reflected light, refracted light, accuracy, errors I. INTRODUCTION Safe reactor operation is becoming more and more important. Automatic and remotecontrolled measuring devices are required to detect liquid leakage from structures, pipes, sealing parts and others contained in a reactor containment vessel. Direct liquid leakages monitoring and observation are impossible for operators because of high radioactivity. Liquid transparency makes it difficult for them to detect liquid leakage with a TV camera monitor. Droplet detection with humidity sensors is impossible because liquid does not diffuse like gas. If a leakage is found, quantitative measurement is required to know the amount. Sampling with a saucer and weighing with a chemical balance is difficult due to uncontinuous measurement. Various devices have been applied to the measurement of droplet diameters, but only the subject device have been successful. Some reports are available to measure the spray mist and rain drop size distribution. Konig et al.(1) measured 10 to 300 mm droplet diameters with an accuracy of 2% based on forward-scattering light intensity distribution using a linear CCD (Charge Coupled Device)-array. Hess & Li(2) measured the side-scattering light intensity of 0.1 to 9 mm diameter droplets. Knollenberg(3) obtained cloud and precipitation particle size distribution by an optical array using a shadowgraph technique. Analog measurement is essential for light intensity measuring methods based on a light * Akasaka, Minato-ku, Tokyo 107. ** Hamamatsu-cho, Minato-ku, Tokyo 105. resent address : Plant & Engineering t P Division, Sumitomo Heavy Ind., Ltd., Yato-cho, Tanashishi

2 296 J. Nucl, Sci. Technol., scattering theory. However, complex and expensive devices are necessary for obtaining high accuracy. The shadowgraph technique requires edge detection, but this makes highly accurate measurement difficult. These methods are not suitable for remote measurement because the detector must be placed near the sampling area. When monodispersed transparent droplets are illuminated by a collimated ray, two brilliant spots appear on the droplet surface. A detector using a linear CCD sensor is being developed to measure the distance between the two spots. The diameter and the volume of each droplet are obtainable through a calculation based on geometrical optics. For the calculation, only two peak points are required to obtain a highly accurate diameter and volume because the peak positions do not shift irrespective of beam intensity and camera gain. Droplets are remotely measurable by this method without sampling. The method can also be used to measure the spray mist and rain drop size distribution. The basic theory and the ray tracing analysis are described in Chap. II, and experimental results obtained by measuring a ball lens and water droplets in Chap. III. II. RAY TRACING ANALYSIS 1. Basic Theory A collimated ray incident on a droplet is partially reflected on the droplet surface and partially refracted through the droplet emerging from the surface. Since a small droplet can be approximated to a sphere, only the rays which satisfy specific conditions are observed as two brilliant spots on the droplet surface when viewed at a scattering angle 0, as shown in Fig. 1. Angles t1 and t2 for the reflected light and the refracted light are respectively : t1=(p-t)/2, (1) t2=(t+2p)/2. ( 2 ) Letting n denote the relative refractive index, the following relationship exists between t2 and p: ( 3 ) Two brilliant spots are reformed on the sensor of a camera when the camera receives scattered rays. Representing the magnification of a camera lens as m and the actual droplet radius as R, the radius r of the droplet image on the sensor is mr=r. (4) Using the distance l between two brilliant spots on the sensor, the following relations are obtained from Fig. 1: Hence, Fig. 1 Scattering from sphere l=r1+r2, ( 5 ) r1=r sin t1 ( 6 ) r2 =r sin t2. (7) ( 8 ) 2

3 Vol. 27, No. 4 (Apr. 1990) 297 and then, ( 9 ) in this way. When N droplets are counted in a given period, the volume of liquid V is Figure 2 shows the values 1/(sin t1+sin t2) of Eq. ( 9 ) calculated with various angles t using Eqs. ( 1 ), ( 2 ) and ( 3 ) under n= The diameter of each droplet is determined (10) where Ri is the radius of each droplet given by Eq. ( 9 ). Fig. 2 1/ (sin t1+sin t2) 2. Analysis Using Ray Tracing Algorithm Analysis was carried out to estimate the size and intensity of each brilliant spot. Calculation model was simplified to be used for a light beam from a ball lens. The light beam was divided into some ray elements, and each ray orbit was calculated using a ray tracing algorithm to clarify whether the ray had reached the effective camera iris or not. An optical design program, ODEPAC (Okudaira Saiteki System Laboratory) was used for the ray tracing algorithm to obtain ray traces and spot diagrams. Analysis was performed using the calculation model shown in Fig. 3, varying the values of the distance L between the droplet center and the sensor surface, scattering angle t, droplet diameter D and refractive index n. A thin single lens with a large refractive index was assumed to be an ideal camera lens because of its minimum aberration. Calculations were carried out using the refractive index of several ball lenses to simulate water droplets. The collimated rays of dht width at the point (2) of each water droplet are refracted and scattered from the point (3) and transmitted through the camera lens, finally to reach the position yr on the plane (8) (the CCD sensor surface). The droplet center plane (1) is regarded as an iris, and the pattern of rays incident on the camera lens is obtained on this plane. The ray widths dht and dhr incident on the camera lens are roughly calculated, and divided into 20 to 40 squares in a grid, and then detailed calculations are performed using the ray tracing algorithm. The relative intensities of the spots are given by the product of the number of rays and the transmittance or reflectance. The calculation results were confirmed by the relation : YT= myt4, yr =myr4 (11) where yt4,ur4,yt and yr are the spot diagram gravity centers of the plane (4) and (8). The droplet diameter is obtained by re- 3

4 298 J. Nucl. Sci. Technol., Fig. 3 Calculation model placing l in Eq. ( 9 ) with l=-yt+yr (12) By using the focal length f of the camera lens and the magnification m, the distance L is approximated geometrically L=f(1+m)2/m. (13) 3. Results ( 1 ) Size of Spot Diagram Under focused conditions, small spot diagrams with diameters from 1 to 1.5 mm were obtained on the sensor surface using the nonaberration lens. The pixel size is 5x5 mm even when the sensor has 4096 pixels. Spot diagram calculation shows that a brilliant spot is sufficiently smaller than a pixel in size. ( 2) Light Intensity of Spots Sufficient light intensity is required for detecting spots clearly although their positions do not shift by intensity. The scattered light intensity is independent of the distance L when the F value of the lens and the magnification in are constant. Refracted light intensity is ten times higher than reflected light intensity in s-polarization. Refracted light intensity becomes higher as scattering angle becomes smaller, and reflected light intensity reaches the maximum value at 30-. The light intensity of each spot t= is proportional to the square of droplet diameter. ( 3 ) Measurement Accuracy of Diameter Droplet diameters thus calculated are independent of parameters including droplet diameters, scattering angles or distances L, except for defocuses. When dripping position is deviated from the focal plane, measurement error will increase in accordance with lens magnification. Error decreases in linear proportion to the distance L, when defocus is constant. Droplet diameters can be determined with an error of less than 0.1% (0.3 v/o) by measuring the distance between the centers of the spot diagrams. II EXPERIMENTAL RESULTS 1. Measuring Device As shown in Fig. 4, the measuring device is composed of a collimated beam source, a linear CCD camera and a data processor. The specification of this device is shown in Table 1. ( 1 ) Collimated Beam Source A 15 mw linearly polarized (TEM00 mode, s-polarization) He-Ne laser beam was expanded 2.5 times by a beam expander and was spread to form a sector-like shape by a 3 mm diameter rod lens. The beam was then collimated into a parallel beam 200 mm wide through a sin - glet cylindrical lens used as a projection lens. 4

5 Vol. 27, No. 4 (Apr. 1990) 299 Fig. 4 Schematic view of device Table 1 Specification of device ( 2 ) Linear CCD Camera A linear CCD camera with 4096 pixels (NEC SC-4096B) and a camera with 2048 pixels (NEC SC-2048NB) were used. Experiment was also conducted with a camera having a linear PCD (Plasma Coupled Device) sensor (Hamamatsu S Q) attached to the camera head of the linear CCD camera. The camera lens was a combination of a zoom lens (SMC Pentax A Zoom 70~210 mm F4) and a rear converter (Pentax A1.4X-S). ( 3 ) Data Processor Light intensity data obtained from each pixel of the CCD sensor is processed by a personal computer (NEC PC-9801VM2) after being digitized into 8 bits by a camera controller (NEC SC-4NB). Data from the PCD sensor processed after being digitized into 16 bits by the control unit (C S) and the data aquisition unit (C2890). The gain of the operational amplifier, used in the data acquisition unit, was adjusted to provide a full-scale output at 0.5 V. Output obtained after A/D conversion was at a level approximately 3 orders greater than that from the CCD camera. Though noise level also increased, total S/N ratio improved nearly tenfold. One thousand and five hundreds of data samples are recorded on an 8 MB RAM disk with the maximum data sampling intervals of 100 ms for 2.5 min. For further data storage, a hard disk with a cassette streamer is available and the RAM disk is expandable. Assembly language was used in pixel data transfer while Basic or Basic Compiler was used in other programs. The software can perform : (1) Length calibration by reference scale input (2) Continuous recording of 1,500 data samples (3) Data display to a CRT, printer and plotter (4) Brilliant spot peak searching (5) Correction of scattering angles in accordance with camera view angles (6) Calculation of droplet diameters, volumes and the total liquid quantity. For geometrical correction, the optical scale of BK7 was placed at the focus of the camera to read the scale data. The precision of the device depends on the measurement accuracy of the distance between two brilliant spots. In order to precisely determine the brilliant spot peak positions, a smoothing and differentiation technique by Savitzky & Golay(4) was used. Application of this technique was extended to the range of real number to obtain a high resolution of 1/10 pixel pitch. 5

6 300 I. Nucl. Sci. Technol., 2. Measurement of Droplet Dripping from Nozzle ( 1 ) Measurement with Linear CCD Camera The sizes of water droplets dripping from a nozzle were measured by a linear CCD camera with 4096 pixels and a collimated beam. The droplets were collected in a saucer disposed under the nozzle and weighed by a chemical balance (sensitivity 0.1 mg). Silicon oil was put on the saucer to avoid splashing and vaporization of the water droplets. Measurements were conducted varying the dripping height at a dripping rate of about 60 droplets per minute with droplets of about 4.1 or 4.8 mm diameters. The droplet diameters calculated from the weight of the collected droplets and the number of droplets were and mm, respectively. The results show the droplet vibration caused by surface tension(5). The following equations are obtained by regression analysis : (14) (15) where D represents the droplet diameter (mm) and h the dripping height (mm). The first terms in each of these equations show the mean diameters unaffected by vibration. The diameters agree well with values calculated from the weight with a small error of 2% or less in diameter and 5% or less in volume. The second terms are factors attributable to vibration. The ratio of the second term to the first term increases when the droplet diameter becomes large. The measurement involves an error of 8% in diameter (25% or more in volume). ( 2 ) Photographic Measurement In order to verify that the error is mostly caused by deformation due to vibration, droplets from three types of nozzles were recorded with an apparatus incorporating a Sugawara stroboscope SVS-1 and a single lens reflex camera. The measurement was conducted using a photosensor disposed immediately below the nozzle and activating the stroboscope for 4 ms with various delay times to vary the dripping height. This photographic measurement showed clear droplet deformation as shown in Fig. 5. Regression analysis was performed with the Fig. 5 Maximum horizontal droplet diameter 6

7 Vol. 27, No. 4 (Apr. 1990) 301 correlations of the dripping height and the maximum horizontal droplet diameter obtained from the photograph. The following vibration frequency o of a falling droplet is obtained by dimenssional analysis(5) : Table 2 Water droplet deformation frequency (16) where r : Density of water (kg/mm3) Surface tension of s: water (dyn/mm) D0: Diameter of water droplet (mm). As shown in Table 2, the value of the vibration frequency o obtained through this analysis agreed well with the value of Fig. 5 obtained from the photographs. Regression analysis using the similar Eqs. (14) and (15) was made to determine the variation coefficients of diameter, i, e. the ratios of the second terms to the first terms. The results are shown in Fig. 6. The larger the droplet diameter, the greater the droplet deformation ratios. The results of the measurements obtained from the photographs agree well with those obtained from the linear CCD camera. Fig. 6 Deviation rate of water droplet diameter 3. Random Dripping Droplets The volume of water randomly dripped in different sizes of droplets was measured and compared with that determined by weight measurement. Unlike the dripping from a nozzle, the dripping position and the droplet size varied randomly. The testing apparatus shown in Fig. 7 has a test vessel provided with an oil bath for heating the whole vessel. Preheated water at a temperature of about 95dc is dripped. The dripped water volume was measured with varying the amount of supplied preheated water. The dripped water was collected in a saucer for weighing with a chemical balance (sensitivity 0.01 mg). Photograph 1 shows the dripping of water from the test vessel. The droplet diameter distribution measured with a linear CCD camera showed the highest value at about 3 mm. Tests were conducted controlling the volume of supplied water as shown in Table 3, but vaporization reduced falling water. The temperatures at the bottom of the testing vessel and the temperatures of the test chamber were set at 90 and 80dc, respectively. 7

8 302 J. Nucl. Sci. Technol., Fig. 7 Flow sheet of testing facility Photo. 1 Dripping from test vessel Table 3 Comparison of measured droplet volumes 8

9 Vol. 27, No. 4 (Apr. 1990) 303 Tests with 200 to 2,000 cms/h of preheated water decreased dripped water quantity to less than 1/3 of the volume of preheated water supplied for vaporization. Droplets at a pixel data sampling interval of 100 ms were often measured simultaneously. Assuming Poisson distribution, it was confirmed that simultaneous measurement can be reduced with shortening interval to 10 ms. This is also effective for eliminating the adverse effect of mist formed in the measurement area. As shown in Table 3, the value measured by a linear CCD camera agreed well with the value determined by weight measurement with an error of about 10%. IV. DISCUSSION 1. Measurement Accuracy ( 1 ) Perfect Spheres Since a small droplet can be approximated to a sphere, measurement error is mainly caused by deformation of a large droplet. Falling ball lenses which are considered as perfect spheres were measured to simulate droplets. Table 4 shows the testing conditions. 1± mm diameter artificial sapphire, 1.5, 2, 3, 5±0.006 mm diameter BK7, and 5±0.005 mm diameter artificial quartz balls were used. Table 4 Testing conditions The results showed that measurement accuracy is independent of parameters including scattering angle t, distance L or refractive index n. And the small ball lens of 1 mm diameter was measured with an error of less than 1% using a camera with magnification. Figure 8 shows waveforms obtained with various numbers of sensor pixels. Ball lens diameters were measured with these pixels and calculated as described in Chap. II. Figure 9 shows the results. Measurement was repeated 10 times with different numbers of sensor pixels. Although the difference of averaged values was small, data from fewer pixels showed larger deviation. The analysis of Chap. II showed that measurement accuracy was not significantly affected by the diameter of the object. However, since the numbers of sensor pixels are limited, a smaller diameter causes a larger error or deviation. ( 2 ) Deformation of Droplet Maximum horizontal diameters that showed the maximum horizontal length of the droplets were measured photographically to examine the deformation of droplets which have a vertically symmetrical axis. Brilliant spots are formed just when a plane with the maximum horizontal diameter meets the optical axis. As shown in Fig. 6, results obtained by photographic measurement agree well with those obtained by the new device. Some experiments were performed for droplets with a diagonal axis dripping from a diagonal nozzle tip. A lower degree of deformation was observed by the new device. Droplets from the testing vessel are mixture of those with different falling height, diameter and defocus. Average diameters show a smaller deformation than that of each droplet. The highest value of the histogram was 3 mm in diameter. It showed a smaller deformation than that shown in Fig. 6. 9

10 304 J. Nucl. Sci. Technol., ( 4 ) Smoothing and Differentiation of Data The measurement value of the ball lens is 4.995±0.016 mm using a camera with 1024 pixels, and the error is 0.016/4.995 x 100=0.32 (%). (18) When the ball lens is measured with a 45dc scattering angle, the distance between the spots is mm. This distance corresponds to 22 pixels using a camera with 1024 pixels. As the peak position has an error of ± 1/2 pixel, the measurement error is 1/2x 2/22x100=4.5 (%). (19) Fig. 8 Light scattered from ball lens ( 3 ) Defocus One of the most important parameters affecting the measurement accuracy is defocus as described in Chap. II. For example, to achieve a measurement error below 0.33% (1% in volume), defocus at the distance L=1.75 (m) must be smaller than : x O = 4.5 (mm), (17) since the distance from the principal point of the lens is 1,375.5 mm. Offset limit will be looser for a longer distance L. The difference between Eqs. (18) and (19) is due to the peak positions. In Eq. (18), high accuracy is obtained by a smoothing and differentiation technique(4) expanded to the real number. ( 5 ) Measurable Diameters In the analysis described in Chap. II, each spot diagram is smaller than 1 pixel and the light intensity is proportional to the square of a droplet diameter. In the experiments, however, brilliant spots spread to 10 pixels in the case of 4096 pixels as shown in Fig. 8. And the peak value of the spots is proportional to the diameter as shown in Fig. 10. Differences between the analysis and the experiments show that the spread was caused by lens aberration and defocus. The smaller droplets scatter the smaller light intensity, and the smaller droplets decrease the numbers of pixels between two brilliant spots. The minimum diameter is 1 mm according to the specification of the device. To measure smaller droplets, the device must have a higher magnification, narrower measuring area and higher light intensity. 2. Application of Measuring Device ( 1 ) Liquid Leakage Detector As mentioned previously, the device can measure dripped water volumes with an error of about 10%. Each droplet falling across collimated rays scatters light based on the principles of geometrical optics. This is why liquid leakage is detectable by a charge-accumulating CCD sensor with a negligibly small counting loss. An error of 10% is practically 10

11 Vol. 27, No. 4 (Apr. 1990) 305 Fig. 9 Diameters measured with various numbers of sensor pixels permissible for a liquid leakage detector. The beam source had a large deviation in the light intensity profile due to a large aberration caused by the combination of a rod lens and a cylindrical lens. Two other types of beam sources were prepared to reduce aberration. An anamorphic prism theoretically free from aberration was used. Achromatic cylindrical doublet lenses with long focal lengths were also used. They produced satisfactory light intensity profiles and beam divergence. Although no proof data is available, radiation resistance problems will be solved by completely shielding the collimated beam source and the linear CCD camera or by installing laser equipment outside the reactor container, forming an optical guide with radiation-resistant plane-polarized light transmitting optical fibers and mirrors. Installation of the laser Fig. 10 Light intensity peak values 11

12 306 J. Nucl. Sci. Technol., equipment and camera outside the reactor vessel will be easier for maintenance. A device within 200 mm diameter measuring area requires a scanning equipment to monitor a larger area. The equipment must be provided with devices which travels in parallel to the beam, oscillates or zoomes with ±1 mm alignment accuracy. High-speed and real-time data processing requires to accelerate the speed up to 4,096 data per 10 ms. It is advisable to use an algorithm for high-speed selection and transfer of droplet data to memories. A DSP (Digital Signal Processor) or a parallel processing unit is usable. 3. Application to Other Purposes High precision droplet diameter measurement makes this device applicable to various purposes including spray nozzle performance evaluation and fine liquid particle diameter measurement. Collimated beam source light intensity must be increased in accordance with a particle size. As the collimated beam source is obtained by enlarging He-Ne laser beam, the light intensity profiles show Gaussian distribution and intensity at each edge is about half of that at the center. In smaller droplet measurement, a flatter intensity profile is needed for the efficient use of laser power. Droplet diameter distribution must be practically made to evaluate coincidence probability for simultaneous measurement of many droplets. Correction is required to eliminate deformation due to aerodynamic resistance. V. CONCLUSION How collimated rays are scattered by liquid droplets was analyzed by a ray tracing algorithm. It proved that the volume of a ball lens is measurable with a high precision of about 0.3% through measurement of the distance between two brilliant spots on the ball lens surface. Tests using a ball lens of 5± mm diameter in a 200 mm horizontal width showed a measurement error of less than 1% in volume. Tests using droplets within 200 mm diameter measuring area showed an error of about 10% in volume. ACKNOWLEDGMENT We are greatly obliged to Mr. Y. Hayamizu for the calculations made by the ray tracing method in Chap. II. We would also like to express our gratitude to some members of Fuji Electric Co., Ltd. who supplied us with most of the data given in Chap. III. REFERENCES (1) KON1G, G., ANDERS, K., FROHN, A.: J. Aerosol Sci., 17[2], 157 (1986). (2) HESS, C. F., Li, F. : Optical technique to characterize heavy rain, AI AA 24th Aerospace Science Mtg., (1986). (3) KNOLLENBERG, R.G. : J. Appl. Meteor., 9, 86 (1970). (4) SA VITZKY, A., GOLAY, M. J. E. : Anal. Chem., 36[8], 1627 (1964). (5) KATUTO, Y. : "Dennetu Gairon" (General Considerations on Heat Transfer), (in Japanese), 286 (1987), Yokendo. 12