4.2 Description of the experiments
4.2.1 Image processing
With the objective of facilitating image analysis, the images are subjected to a process of binarization. As a result, we obtain new images that are easier to analyze, in which only bubbles can be visualized. Firstly, the background is sub-tracted from the actual image (Fig.4.4-a), in order to remove the illumination heterogeneities that could have been produced by the lighting system. Addi-tionally, the image contrast is also changed to achieve a more homogenized im-age (Fig.4.4-b). Due to the capillary curvature, shadows appear at the capillary upper and bottom parts. Several image filters are used to soften them (Fig.
4.4-40 4 Experimental setup
c). Finally, a segmentation tool allows us to choose which range of colors will appear as black, leaving the remaining colors as white (Fig.4.4-d).
(a)
(b)
(c)
(d)
Figure 4.4: Image processing protocol: (a) actual image, (b) background subtraction and changes in contrast, (c) image filtering, (d) image segmen-tation.
Analysis of the videos permits one to identify the flow pattern and classify each experiment into bubble, slug, churn or annular flow regime. This task can be susceptible to the subjectivity of the researcher, especially in the two former flow regimes. Both bubble and slug flow patterns are defined as proposed by Dukler et al. [22]. Under this definition, the transition between bubble and slug flow patterns is considered to take place when the bubble diameter reaches the value of the capillary diameter (see Subsection3.2.1).
Due to the operative limitation of the gas and liquid flow meters, most of the experiments recorded belong to bubble or slug flow patterns. Since gas flow me-ters with higher operational limits are required for an accurate analysis of churn and annular flow regimes, such analysis were not performed in this study.
4.2 Description of the experiments 41
Figure 4.5:Bubble being detected by a rectangular AOI with a length of 1 pixel.
In addition, measurement of bubble generation frequency, bubble velocity, as well as bubble, unit cell and liquid slug lengths are carried out by means of the standard image software.
The software enables the user to active an Area of Interest (AOI) on the im-age, which can have any shape and be placed anywhere. We use a rectangular AOI with a length of 1 pixel, the minimum allowed by the software. Black ob-jects are detected when entering the AOI (see Fig.4.5) and the software automat-ically records it on the datafile with a 1. Fig.4.6shows an example of several bubbles detected during 100 ms (400 frames). The generation frequency, f , can then be calculated by knowing the total number of tracked objects, the recording speed and the number of frames by means of Eq.4.1. The larger the number of frames, the more accurate the results will be.
f = # objects · # f ps
# f rames (4.1)
The maximum generation frequency registered in our experiments is roughly 1200 bubbles per second (see AppendixB), almost four times smaller than the recording speed. In each case, frequency aliasing is thus avoided, hence this process is correctly resolved by the image acquisition system. 1000 Frames per video results in being an appropriate number to accurately determine the fre-quency. A macro was programmed to define the procedure from the original images to the final data output. This macro automatically performs the entire process of image processing, enhancement and bubble detection.
The analysis of the recorded images also allows the measurement of the gas velocity. The gas is contained within the bubbles, and thus it is acceptable to assume that the gas velocity is equal to the bubble velocity, which can be mea-sured with the standard image software. To this end, the image must be cali-brated, what we do through the capillary diameter. A tracking option permits
42 4 Experimental setup
0 0.5 1 1.5 2
0 20 40 60 80 100
Black object detection
t (ms)
Figure 4.6:Bubble detection by using the standard image software during 100ms (400 frames). USG=0.242m/s and USL=0.212m/s.
0 2 4 6 8 10 12 14
0 20 40 60 80 100
X coordinate of the bubble geometric center (mm)
t (ms)
Figure 4.7: Bubbles tracking and gas velocity estimation by using the standard image software during 100ms (400 frames). USG=0.242m/s and USL=0.212 m/s.
4.2 Description of the experiments 43
Element Uncertainty
Manometer < 0.1 bars Pressure controller < 0.01 bars Gas flow meter < 0.5 ml/min Liquid flow meter < 0.5 ml/min Liquid pump < 1 ml/min Image processing < 0.07 mmm
Table 4.5: Uncertainties from the experimental apparatus and the image processing.
user to track all black objects, reporting the x coordinate of their geometrical center. In bubbles longer than the capillary diameter, no movement is observed in the y-axis. In smaller almost spherical bubbles this is the case only just after the detachment until they reach the centerline. Therefore, we assume that the velocity has its main component in the x direction. Fig.4.7shows the tracks of several bubbles as an example. Once the bubble velocity is known, we can then estimate the mean void fraction with Eq.2.11.
Finally, the bubble, unit cell and liquid slug lengths are measured directly from the images as defined in Fig.2.2. The gas/liquid interface is often blurred and overshadowed, particularly in those cases where the bubble velocity is high (on the order of 1 m/s) and thus, some uncertainties appear. An error of ap-proximately ±0.07 mm has been estimated when measuring the bubble interface, which corresponds to an error of two pixels.