Professional Documents
Culture Documents
Paper 16
Paper 16
2
Publication Date: April 25, 2023
DOI:10.14738/aivp.112.14527.
Leiva, C. A., Borjas, J., Acuña, C., & Luukkanen, S. (2023). Applied Online Bubble Size Distribution Measurement in a Pilot Flotation
Cell Based on Image Analysis. European Journal of Applied Sciences, Vol - 11(2). 645-662.
José Borjas
Departamento de Ingeniería Química,
Universidad Católica del Norte, Antofagasta, Chile
Claudio Acuña
Department of Chemical and Environmental Engineering,
Universidad Técnica Federico Santa María, Valparaíso, Chile
Saija Luukkanen
Oulu Mining School, University of Oulu, Oulu, Finland
ABSTRACT
The distribution of bubble size in the pulp is a parameter directly related to the
flotation kinetics, but its measurement is complex to determine due to the presence
of particles and cluster of bubbles. The existing equipment for the measurement of
bubble size (McGill and UCT), which operate manually and batch, requires
specialized operators in image analysis. On the other hand, the McGill technique has
not been directly validated with bubble swarms and only 10% of the sampled
bubbles are analyzed. These aspects have limited the technology transfer and
sustainability in the measurement of bubble size. To solve the problems presented,
a device based on the McGill technique was designed and implemented.
Furthermore, algorithms were implemented to increase the statistical significance
of the measurement of bubbles per image. The validation consisted of a comparison
of the degree of detection using the software manually and automatically
(undetected remaining bubbles). As a result, it is possible to predict the bubble size
distributions with an error of less than 5% and derivations close to 0.1 [mm] in the
determination of D32, using an average of 100 images. In conclusion, the new device
and algorithms improve the accuracy of BSD measurements, helping to optimize the
process, predict, control flotation kinetics, and be used as a troubleshooting tool.
The new device and algorithms improve the accuracy of BSD measurements,
helping to optimize the process, predict, control flotation kinetics, and be used as a
troubleshooting tool.
INTRODUCTION
The mineral concentration process is carried out, to a large extent, by froth flotation. Flotation
consists in the separation of the valuable mineral particles contained in solid-liquid dispersion
(pulp) by means of their selective adhesion to air bubbles, which float and concentrate in the
form of froth. The efficiency of this process depends on the physicochemical aspect in particle-
bubble adhesion [1], characteristics of the circuit [2] and the hydrodynamics of the equipment
in the collection area (dispersion of gas in the pulp) [3], which includes gas superficial velocity
(Jg), gas volume fraction (Eg), size distribution (BSD) and bubble stability, being the bubble size
distribution the parameter that directly affects the kinetics of the process [4]. Nevertheless,
only few techniques have been developed to determine this parameter, in which they can
operate both on a laboratory scale and in industrial plants. Furthermore, studies about the
validation of these techniques are scarce in the literature [5].
The measurement of bubble size is one of the most complex aspects to determine due to the
presence of heterogeneous mixtures in the cells (water, dark particles, single bubbles and
swarms of bubbles) [6] and it has been studied by various researchers [7, 8], reaching a great
level of sophistication in the development of non-invasive techniques [9-11] including:
ultrasound [12], radiography and tomography imaging, nuclear magnetic resonance imaging
(NMR) and velocimetric techniques (Laser Doppler Anemometer (LDA), Particle Image
Velocimetry (PIV), [13]). However, these techniques are substantively complex to be
implemented at an industrial level, although there are some methods that use image analysis
[14, 15]. Only two methods have been used at this level and these are the following: an imaging
sample tracking technique [16] and a capillary technique [17]. The analysis of the bubble size
distribution has been extensively studied [18-26], being a basis for achieving the most accurate
data collection possible.
The current bubble size measurement is built on models based on one-way flow or “drift flux”
analysis. Likewise, the model has been extended to be adapted for bubble swarms, thus relating,
the distribution of bubble size (db) with the superficial gas velocity (Jg), superficial liquid
velocity (Ji) and hold-up gas (Eg), being contrasted with photographic measurements. Several
study mechanisms have been identified on the bubble size distribution, among which the McGill
Bubble Viewer device is the most widely used technique to determine the bubble size
distribution [27-37].
The bubble detection has been enhanced with algorithms that allow discriminating the bubbles
from swarms, differentiate irregular bubbles, and isolate bubbles in areas of irregular lighting.
As detection algorithms, it was proposed to use the quotient between the convex area and the
projected area of the isolated segment and a variable threshold to separate bubbles into
swarms and illumination zones. To validate the algorithms and the continuous measurement
system, the device was implemented in a pilot cell with a controlled bubble size distribution
generation system. Furthermore, to control the bubble size distribution, a porous injector with
controlled axial liquid flow and froth concentration was used.
For the Bubble Viewer case, the volume of the device is designed to capture 100 images per fill.
This number of images allows a representative distribution in terms of bubble population. The
software discriminates irregular bubbles and swarms of bubbles under the criterion of form
factor, which is based on the sphericity of the bubble to classify it as a single bubble.
Additionally, the BBV is inclined to allow bubbles to rise towards the front section of the
viewing chamber. Studies [27] have as a result that an inclination of 45° causes a deformation
in the bubbles affecting their spherical geometry, while an angle of 5° is sufficient to allow the
bubbles to be near the front of the device and thus, can be studied without major effects on its
geometry. Lastly, a 15° tilt is recommended, which enables to fulfill the main objective of
bringing the bubbles closer to the front section of the device and thus, reducing bubble swarms
within the viewing chamber.
Necessary volume, over the focal area, to obtain at least 100 photographs = 10.14 [cm3/s] * 100
[s] = 1014 [cm3]
Additionally, a volume correction factor was considered due to the expansion of the incoming
air, approximately 20%. According to this, the volume over the focal area is = 1014 x 1.2 [cm3]
= 1217 [cm3].
In relation to the above and considering a 6 × 5 [cm] image, the following interior
measurements were chosen to complete the necessary volume: Height = 20 [cm], Width = 12
[cm] and depth = 11 [cm], as seen in Figure 1.
Figure 1. Viewing Chamber Design. Note: The angle used in the viewing chamber is given by the
research carried out.
The system is completed with a metal support, with an inclination of 15° that supports the
viewing Chamber, the rear lighting system and the image capturing camera, allowing the device
to be moved and handled in a simple way in flotation cells. The Figure 2 shown, the installation
of device in flotation cell and the conventional injector bubbler type.
Device Protections:
The exterior changes were made to preserve the device against any environmental conditions
that could undermine the device:
• To protect the viewing chamber from the heat emitted by light, a heat diffusing glass is
placed between the lighting system and the viewing chamber.
• To protect the viewing chamber from external light, the side windows of the chamber
are covered with PVC sheets.
• To protect the viewing chamber from possible scratches, cork sheets are inserted
between the support and the chamber.
• Since the chamber is sealed, a viewing window cleaning system composed of magnets
has been implemented.
Operating Procedure
To keep the chamber sealed, the bottom valve of the sampling tube must be closed, allowing
the tube to be immersed in the flotation cell. The viewing chamber must be filled with clear
liquid by opening the top fill and vacuum valves. Once it is full, it must be completely sealed by
closing the top valves.
Additionally, the device must be located on the flotation machine so that the sampling tube is
approximately 50 cm below the surface.
Once the device is disposed, the digital camera settings must be as follows:
• Shutter speed between 1/1000 [s] and 1/2000 [s]
• Continuous shooting
• Focus to obtain a sharp image of a maximum of 6x5 [cm2].
To determine the magnification factor of the image, a first image must be obtained with a
transparent ruler located in the focal plane. Then, remove the ruler and open the valve of the
sampling tube to allow the entry of the bubbles, obtaining images once the entry of the bubbles
into the chamber is stable.
To avoid coalescence in the device, it is possible to add froth agent in the viewing chamber,
maintaining the concentration of the solution.
characterized by geometric parameters. Triple selection criteria (bubble size range, bubble
curvature or shape factor range and background level intensity) was used. For example,
selecting isolated circular objects from an image as shown in Figure 3 (b); local threshold +
isolated bubbles in Figure 3 (c) and watershed local threshold + isolated bubbles in Figure 3
(d).
The accepted bubbles are shown in Figure 4 (a), where an equivalent diameter is calculated for
each bubble. The results of these diameters are classified in size ranges and their frequency of
distribution is plotted in a histogram, Figure 4 (b).
The image analysis system is based on a triple selection criterion (bubble size range, bubble
curvature or shape factor range and background level intensity) adjusted by the specialist. This
allows the identification of individual and spherical bubbles from the image, while excluding
the remaining objects.
The exposed problems show that image analysis can be a difficult technique to implement in
bubble size measurement. To address some of these difficulties, an improvement in the image
analysis technique has been developed, which incorporates an algorithm to compensate for the
unevenness in the backlight (bubble contour boundary detection) and an algorithm to separate
and analyze cluster bubbles (Watershed).
The new criteria include de convex area over object area, which allows better discrimination of
cluster and single irregular shape bubbles than shape factor (this is the most important
contribution). For example, the shape factor cannot discriminate between cluster and single
irregular bubbles in the next real bubbles and constructed objects, Figure 5. However, if the
convex area-over-area criterion is applied, single irregular bubbles are identified, and the
remaining objects are discriminated.
The shape factor (SF) is a parameter commonly used to discriminate a single ellipsoidal object
from a cluster. This parameter, which is a measure of the circularity of an object, is defined as
P2
SF = Ec.1
4 A
Where A and P are the segment's area and perimeter (bubble), a shape factor of one indicates
a circle. As the value approaches zero, it indicates an increasingly elongated polygon.
Photo frame data (an average of 100 images per test, 11750 bubbles)
• Image size: 1131×1491 [pixels].
• Frame height = 50 [mm].
• Approximate ratio: 30 [pixels/mm]
Figure 6 shows the behavior of the bubble size distribution as a function of superficial gas
velocity for a micropore of 100 [µm] and a frother concentration of 10 [ppm]. D32 and D10
increase with higher Jg.
Figure 7 shows the behavior of the bubble size distribution as a function of superficial gas
velocity for a micropore of 10 [µm] and a frother concentration of 10 [ppm]. D32 and D10
increase with higher Jg. A higher frequency of smaller bubbles can be seen than in Figure
8.
Figure 8 shows the behavior of the bubble size distribution as a function of superficial gas
velocity for a micropore of 5 [µm] and a frother concentration of 10 [ppm]. D32 and D10
increase with higher Jg. A higher frequency of smaller bubbles can be seen than in Figure 8
and Figure 9.
Comparisons of D32 for Same Froth Concentration and Different Micropore Sizes
Figure 9 allows establishing a relationship between the size of the bubbles and the variation in
the superficial gas velocity in tests with different micropore sizes.
Figure 9. Relationship between D32 and Jg, in the Variation of Micropore Size.
1
The Sauter mean diameter varies like D32 ⁄4 as a function of Jg [4] if the whole surface is
saturated with bubbles, but in the case, on larger spargers, there is little effect (reduced
coalescence).
Comparisons of D32 and D10 for the Same Froth Concentration and Different Gas Surface
Velocity
Figure 10 allows relating the variation in the size of the bubbles with the variation in the
micropore size for tests with different superficial gas velocity.
Figure 11 shows how D32 increases when a larger micropore size is used and also increases
when the superficial gas flow is greater.
Figure 11. Relationship between D32 and Froth Concentration, Variation of Jg.
Figure 12. Relationship between D32 and Froth Concentration, Micropore Size Variation.
Figure 11 and Figure 12 shows how D32 decreases when a higher frother concentration is
used at different superficial gas velocity and micropore size.
Technique Validation
Selection of Tests for Validation:
The ratio of the minor over the major axis obtained by the software, allows to determine the
relation or aspect ratio in the size of the bubbles. This allows obtaining a criterion to choose
tests for validation. According to this criterion, two types of sample were studied: (A) small and
spherical bubbles (1 mm) and another sample (B) of large and irregular shaped bubbles (3
mm).
A – 10 [µm] – 1.0 [cm/s] – 10 [ppm].
B – 100 [µm] – 1.0 [cm/s] – 10 [ppm].
In both tests, about 90% of the objects or bubbles detected have an Aspect Ratio greater than
0.8 and less than 1.
Software Validation:
The validation of the technique was carried out by processing the images by the same software
in which, manually, known color values were highlighted and assigned to the uncounted
bubbles during the automatic processing.
The bubbles not automatically recognized were highlighted with numbered colors according to
the RGB color model in Microsoft Power Point, differentiating those that were not counted due
to being in a swarm or if they were not recognized by the software when isolated. It is shown
that the images in the left column correspond to the ones automatically analyzed by the
software, while the images in the right column are presented with bubbles not detected
automatically marked with the respective color, according to the proposed criteria and
highlighted with a red circle.
The results obtained for the validation by processing the modified images according to the
explained criteria, are reflected in accumulated distributions in Figures 13 and 14:
The data obtained show a variation, in the D10 and D32 statistics, of 5.5% and 4.4%
respectively.
The data obtained show a variation, in the D10 and D32 statistics, of 3.6% and 1.3%
respectively.
In this way, the error obtained in the accumulated frequency, between the automatic analysis
and the manual analysis performed for the validation, is not very significant (close to 0.1 mm)
considering that of the total of the analyzed bubbles, the 70% was detected automatically in
both operating conditions the remaining 30% is not significant for the analysis. This can help
reduce the number of images captured. Likewise, the error obtained for the statistics reached
is negligible due to the uniformity in bubble sizes and aspect ratio, for which about 90% of the
counted bubbles show a value close to 1.
The total of bubbles counted manually, for both tests, reveals that more than 80% correspond
to bubbles in swarms and about 20% correspond to unrecognized isolated bubbles.
Nevertheless, due to the large number of items counted automatically (about to 70% in both
cases), it does not have a great impact on the determination of the bubble size distribution.
The image processing software was validated, considering that the results obtained through
the exposed criteria present little significant deviations when determining the Bubble Size
Distribution. Therefore, the incorporation of the software to the device solves the problems
exposed regarding the analysis of the images by granting automatic data processing.
The importance in the diagnosis and control of the Bubble Size Distribution parameter, to
improve the recovery of concentrated material in flotation systems, shows a great potential for
the market and for the device.
It is suggested to perform some work at an industrial level with the device. This would affirm
that its main characteristics make it suitable for this type of work. In addition, this would allow
the development of a sampling procedure for the determination of bubble size distribution in
industrial flotation machines.
It is proposed for future work to fully automate the device through the implementation of
automatic valves, corrections in the rear lighting system and improvements in the capture and
quality images. This would allow to have an online sensor that allows taking adequate criteria
to optimize performance in industrial flotation machines.
Author Contributions:
All authors. conceived, designed, and performed the experiments; all the authors analyzed the
data; C.A.L. wrote the paper.
Funding:
This research received no external funding.
Acknowledgments:
The authors wish to acknowledge the support provided by Universidad Católica del Norte and
to the student José Borjas and David Díaz his contributions.
Conflicts of Interest:
The authors declare no conflict of interest.
Data Availability:
All raw data remains the property of the University that allowed this study. The input data used
to support the endings of this study could be available from the correspondent author’s email
with appropriate justification.
References
1. Rao, S.R., 2003. “Surface Chemistry of Froth Flotation.” Kluwer Academic/Plenum Publishers; 2nd Rev
Edition
2. Loveday, B.K., Brouckaert, C.J.,1995. “An analysis of flotation circuit design principles.” The Chemical
Engineering Journal, 59: 15-21.
3. Nesset, Jan E., Hernandez-Aguilar, J.R., Acuña. C.A., Gomez, C.O., Finch J.A., 2005. “Some gas dispersion
characteristics of mechanical flotation machines.” Published by Department of Mining, Metals and
Materials Engineering, McGill University, Canada.
4. Finch, J.A. y Dobby, G.S., 1990. Column Flotation, Pergamon Press: Elmsford, New York, pp. 23-24
5. Gómez, C.O, Acuna C.A., Girgin E.H., Finch J.A., 2006. “Bubble distributions in laboratory flotation
machines measurement and evolution with increasing gas velocity”. International Mineral Processing
Congress.
6. Gómez, C.O., Finch, J.A., 2002. Gas dispersion measurements in flotation machines. CIM Bulletin, Vol. 95
(1066), pp. 73-78.
7. Grau, R.A., Heiskanen, K., 2003. Gas dispersion measurements in a flotation cell. Minerals Engineering,
Vol. 16, pp.1081-1089.
8. Hernandez-Aguilar, J. R.; Coleman, R. G.; Gomez, C. O. and Finch, J. A. 2004. A comparison between
capillary and imaging techniques for sizing bubbles in flotation systems. Minerals Engineering. Volume
17, Issue 1, Pages 53-61. https://doi.org/10.1016/j.mineng.2003.09.011.
9. X. Bu, S. Zhou, M. Sun, M. Alheshibri, M. S. Khan, G. Xie, et al., Exploring the relationships between gas
dispersion parameters and differential pressure fluctuations in a column flotation, ACS Omega 2021 Vol.
6 Issue 34 Pages 21900-21908
10. Letzel, H. M.; Schouten, J. C.; Krishna, R.; van den Bleek, C. M. Characterization of Regime and Regime
Transitions in Bubble Columns by Chaos Analysis of Pressure Signals. Chem. Eng. Sci. 1997, 52,
4447−4459.
11. Kwon, H. W.; Han, J. H.; Kang, Y.; Sang, D. K. Bubble properties and pressure fluctuations in bubble
columns. Korean J. Chem. Eng. 1994, 11, 204−210.
12. Boyd and Varley, 1998 J.W.R. Boyd and J. Varley, Sound measurement as a means of gas-bubble sizing in
aerated agitated tanks, A.I.Ch.E. Journal 44, 1731–1739.
13. Deen, N. G., Hjertager, B. H., Solberg, T., 2000. Comparison of PIV and LDA measurement methods applied
to gas-liquid flow in a bubble column, 10th Int. Symp. on Appl. of Laser Techniques to Fluid Mech., Lisbon,
Portugal.
14. Mondaca, S.; Leiva, C.; Acuña, C. and Serey, E. 2020. “Flow Enhancement of Mineral Pastes to Increase
Water Recovery in Tailings: A Matlab-based imaging processing”, Scientific Programming, vol. 2020, 9
pages, 2020. https://doi.org/10.1155/2020/5607242.
15. Molina, M.; Acuña, C. and Leiva, C. 2020. “Characterization of filamentous flocs to predict sedimentation
parameters using image analysis”, Journal of Sensors, febrero 2020,
https://doi.org/10.1155/2020/5248509.
16. Jameson, G.J., Allum, P., 1984. “A survey of bubble sizes in industrial flotation cells.” Reporte a AMIRA'Ltd
(Australian Mineral Industries Research Association Limited.
17. Tucker, J P. Deglon, D A. Franzidis, J -P. Harris, M C., y O'Connor, C T., 1994. “Evaluation of a direct method
of bubble size distribution measurement in a laboratory batch flotation cell.” Minerals Engineering, 7:
667-680.
18. Chen, F., Gomez, C.O., and Finch, J.A., 2001 “Technical note: bubble size measurement in flotation
machines.” Minerals Engineering, 14: 427-432.
19. De Lasa, H., Lee, S.L.P., Bergougnou, M.A., 1984 “Bubble measurement in three-phase fluidized beds using
a U-shaped optical fiber.” Canadian Journal of Chemical Engineering, 62: 165-169.
20. Cathignol, D., Chapelon, J.Y., Newhouse, V.L., Shankar, M., 1988. “Bubble sizing with a high spatial
resolution.” Ultrason. Symp. Proc., 2: 933-936.
21. Grau, R.A., Heiskanen, K., 2002. “Visual technique for measuring bubble size in flotation machines.”
Minerals Engineering, 15: 507-513.
22. Hernandez-Aguilar, J.R., Gomez, C.O., Finch, J.A., 2002. “A technique for the direct measurement of bubble
size distributions in industrial flotation cells.” Proc. of the 34th Annual Meeting of the Canadian Mineral
Processors, ed. J.E. Nesset: 389-402.
23. Meernik, P.R., Yuen, M.C., 1988a “Optical method for determining bubble size distributions - Part I.
Theory.” Journal of Fluids Engineering-Transactions of the ASME, 110: 325-331.
24. Meernik, P.R., Yuen, M.C., 1988b. “Optical method for determining bubble size distributions - Part II.
Application to bubble size measurement in a three-phase fluidized bed.” Journal of Fluids Engineering-
Transactions of the ASME, 110: 332-338.
25. Randall, E.W., Goodall, C.M., Fairlamb, P.M., Dold, P.L., O’Connor, C.T., 1989. “A method for measuring the
sizes of bubbles in two- and three-phase systems.” Journal of Physics E: Scientific Instruments, 22, 827-
833.
26. Acuña, C. and Leiva, C. (2020) Sensor device and system for in-line measurement of bubble-size
distribution in flotation cells PCT WO2020132768: World Intellectual Property Organization
27. Hernandez-Aguilar, J.R., Finch, J.A., 2005. “Validation of bubble sizes obtained with incoherent imaging on
a sloped viewing window.” Chemical Engineering Science 60, pp. 14
28. Acuña, C. A.; and Finch, J. A. 2010. Tracking velocity of multiple bubbles in a swarm. International Journal
of Mineral Processing, 94(3-4), 147–158. https://doi:10.1016/j.minpro.2010.02.001.
29. Redlinger-Pohn, J.; Grabner, M.; Zauner, P.; Radl, S. 2016. Separation of cellulose fibres from pulp
suspension by froth flotation fractionation. Separation and Purification Technology, volume 169, pages
304-313. https://doi.org/10.1016/j.seppur.2016.06.004
30. Schäfer, J.; Schmitt, P.; Hlawitschka, M.; Bart, H. 2019. Measuring Particle Size Distributions in Multiphase
Flows Using a Convolutional Neural Network. Chemie Ingenieur Technik, volume 19, Issue 11, pages
1688-1695. https://doi.org/10.1002/cite.201900099
31. Jamshidi, N.; Mostoufi, N. Investigating bubble dynamics in a bubble column containing shear thinning
liquid using a dual-tip probe, Experimental Thermal and Fluid Science, Volume 94, 2018, Pages 34-48.
https://doi.org/10.1016/j.expthermflusci.2018.01.034.
32. Leiva, C., Acuña, C. and Castillo, D. Development and Validation of an Online Analyzer for Particle Size
Distribution in Conveyor Belts. Minerals, mayo 2021.
33. Besagni, G., Brazzale, P., Fiocca, A. and Inzoli, F. Estimation of bubble size distributions and shapes in two-
phase bubble column using image analysis and optical probes. Flow Measurement and Instrumentation.
December 2016.
34. Mesa, D., Quintanilla, P., and Reyes, F. Bubble Analyser—An open-source software for bubble size
measurement using image analysis. Minerals Engineering, April 2022.
35. Schneider, C. A., Rasband, W. S., and Eliceiri, K. W. NIH Image to Image J: 25 years of image
analysis. Nature methods, june 2012.
36. Tunwal, M., Mulchrone, K. F., and Meere, P. A. Image based particle shape analysis toolbox
(IPSAT). Computers & Geosciences, February 2020.
37. Heilbronner, R. and Barrett, S. Image analysis in earth sciences: Microstructures and textures of earth
materials (Vol. 129). Springer Science & Business Media. 2014.
38. Leiva, C. and Acuña, C. 2021, Dispositivo sensor y sistema para la medición en línea de la distribución del
tamaño de burbujas en celdas de flotación, Chile Granted Patent, CL2018003886.