An improved algorithm of image processing technique for film

An improved algorithm of image processing technique for film thickness measurement in a horizontal stratified gas-liquid two-phase flow Hadiyan Yusuf Kuntoro, Akhmad Zidni Hudaya, Okto Dinaryanto, Akmal Irfan Majid, and Deendarlianto Citation: AIP Conference Proceedings 1737, 040010 (2016); doi: 10.1063/1.4949298 View online: http://dx.doi.org/10.1063/1.4949298 View Table of Contents: http://scitation.aip.org/content/aip/proceeding/aipcp/1737?ver=pdfcov Published by the AIP Publishing Articles you may be interested in Fluctuation characteristics of gas‐liquid two‐phase slug flow in horizontal pipeline AIP Conf. Proc. 1207, 162 (2010); 10.1063/1.3366360 Gas‐Liquid Two‐Phase Flow Measurement with Dual‐Plane ERT System in Vertical Pipes AIP Conf. Proc. 914, 734 (2007); 10.1063/1.2747506 Analysis on Dynamic Differential Pressures of Multi‐Loop Flowmeter for the Measurement of Gas‐Liquid Two‐ Phase Flow AIP Conf. Proc. 914, 683 (2007); 10.1063/1.2747499 Study of the Gas‐Liquid Two‐Phase Flow Measuring Method Based on the V‐Cone Flow Meter AIP Conf. Proc. 914, 208 (2007); 10.1063/1.2747432 Gas-liquid two-phase flow across a bank of micropillars Phys. Fluids 19, 043302 (2007); 10.1063/1.2722424

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions IP: 180.245.129.238 On: Wed, 15 Jun 2016 01:47:49

An Improved Algorithm of Image Processing Technique for Film Thickness Measurement in a Horizontal Stratified Gasliquid Two-phase Flow Hadiyan Yusuf Kuntoro1, 2, a), Akhmad Zidni Hudaya2, Okto Dinaryanto2, Akmal Irfan Majid1, 2, Deendarlianto1, 2, b) 1 2

Center for Energy Studies, Gadjah Mada University, Sekip K-1A Kampus UGM, Yogyakarta 55281, Indonesia. Department of Mechanical and Industrial Engineering, Faculty of Engineering, Gadjah Mada University, Jalan Grafika 2, Yogyakarta 55281, Indonesia. a)

Corresponding author: [email protected] b) [email protected]

Abstract. Due to the importance of the two-phase flow researches for the industrial safety analysis, many researchers developed various methods and techniques to study the two-phase flow phenomena on the industrial cases, such as in the chemical, petroleum and nuclear industries cases. One of the developing methods and techniques is image processing technique. This technique is widely used in the two-phase flow researches due to the non-intrusive capability to process a lot of visualization data which are contain many complexities. Moreover, this technique allows to capture direct-visual information data of the flow which are difficult to be captured by other methods and techniques. The main objective of this paper is to present an improved algorithm of image processing technique from the preceding algorithm for the stratified flow cases. The present algorithm can measure the film thickness (hL) of stratified flow as well as the geometrical properties of the interfacial waves with lower processing time and random-access memory (RAM) usage than the preceding algorithm. Also, the measurement results are aimed to develop a high quality database of stratified flow which is scanty. In the present work, the measurement results had a satisfactory agreement with the previous works. Keywords: Gas-liquid two-phase flow, Stratified flow pattern, Film thickness, Image processing technique.

INTRODUCTION Two-phase flow phenomena are frequently encountered in the transportation of gas and liquid in the pipeline systems, which are mainly an important part of the chemical, petroleum and nuclear industries. During the transportation, the interaction between the gas and liquid interfaces can lead to a number of flow patterns, for instance slug [1], plug [2], bubbly [3], annular [4] and stratified flows [5]. Some of them have potentially hazardous effects and can cause the damage to the pipeline systems [6]. For that reason, the study on the characteristics of the flow patterns plays an important role in the industrial safety analysis, which has a strong relation to the safety issues of human and environment. One of the parameters that are often used to study the characteristics of the flow patterns is the film thickness (hL). On the study of slug flow, Kadri et al. [7], Woods et al. [8] and Ujang et al. [9] used the film thickness data to explain the mechanisms of slug initiation in a horizontal pipe. Meanwhile, for the cases of stratified flow, the film thickness data were used by Kuntoro et al. [5], Pitton et al. [10] and Birvalski et al. [11] to describe the characteristics of stratified flow in a horizontal pipe. In this regard, the improvement of methods and techniques to measure the film thickness of the flow in a horizontal pipe is inevitable due to the urge to obtain the data with high spatial and temporal resolutions. The data can then be used to validate the models of the flow.

Proceedings of the 3rd AUN/SEED-NET Regional Conference on Energy Engineering and the 7th International Conference on Thermofluids (RCEnE/THERMOFLUID 2015) AIP Conf. Proc. 1737, 040010-1–040010-14; doi: 10.1063/1.4949298 Published by AIP Publishing. 978-0-7354-1391-7/$30.00

040010-1 Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions IP: 180.245.129.238 On: Wed, 15 Jun 2016 01:47:49

One of the flow patterns in a horizontal pipe is stratified flow. It is characterized by the liquid film flowing along the bottom of the pipe whereas the gas flowing co-currently in the top with a gas-liquid interface separating the fluids. In the field of modeling, the models of stratified flow are often used as a starting point to predict the transitions of the flow patterns [12]. Thus, a comprehensive study about stratified flow can help to understand the two-phase flow behavior inside a horizontal pipe as well as the relation to the transitions of the flow patterns. Some of the researchers who struggled with the transition models of two-phase flow are Soleimani and Hanratty [13], Kadri et al. [14], AlSafran et al. [15], Zhang et al. [16] and Sanchis et al. [17]. On the previous work, Kuntoro et al. [5] had developed an algorithm of image processing technique to study on the local characteristics of stratified flow on the basis of the film thickness data. Image processing technique was employed in that work due to the advantage of the non-intrusive capability to process a lot of visualization data which contain many complexities. Moreover, this technique had been used by many researchers in the field of two-phase flow, e.g. Ozbayoglu and Yuksel [18], Montoya et al. [19], do Amaral et al. [20], de Castro et al. [21] and Vallée et al. [22]. This is because image processing technique provides the direct-visual information data of the flow which are difficult to be provided by other methods and techniques. Although the algorithm of Kuntoro et al. [5] was able to measure the film thickness of stratified flow and the measurement results had a good agreement with the other works, the algorithm had high processing time and randomaccess memory (RAM) usage when it was executed in the PC. The algorithm [5] will not be effective and efficient in time and RAM when it is used to process the high spatial and temporal resolutions data. In addition, the algorithm [5] had not been equipped with the ability to measure the geometrical properties of the flow which are very important to the validation of CFD models. Therefore, the algorithm [5] must be improved in order to face the mentioned challenges. In the present paper, an improved algorithm with lower processing time and RAM usage of image processing technique to measure the film thickness of stratified flow is proposed. The algorithm had also been designed to be able to measure the geometrical properties of the interfacial waves of stratified flow, such as wavelength and wave amplitude. The measurement results are presented and compared to other works. Moreover, the measurement results are intended to develop a high quality database for further studies of stratified flow.

EXPERIMENTAL SETUP Figure 1 shows the schematic diagram of the present experimental facility. The experiments of stratified gas-liquid two-phase flow were conducted in the Fluid Mechanics Laboratory, Department of Mechanical and Industrial Engineering, Faculty of Engineering, Gadjah Mada University. A horizontal pipe with 26 mm inner diameter (D), 9.5 m total length and made of a transparent acrylic resin was used to observe the flow phenomena. The test fluids used were water for the liquid and air for the gas. There were 24 combinations of the superficial gas and liquid velocities, varying between 1.02 and 3.77 m/s for the superficial gas velocities (JG) and between 0.016 and 0.092 m/s for the liquid superficial velocities (JL). In order to ensure the gas and liquid flows from the mixer section became a stratified flow, a flat plate was used in the mixer section to separate the inlets between the gas and liquid flows. In the visualization test section, a high-speed video camera at a rate of 120 fps (frames per second) and 640 x 480 pixels resolution was used to record stratified flow for 30 s at each combination. The visualization test section was located 200 x pipe inner diameter (D) away from the inlet pipe to ensure that the recorded videos were the fullydeveloped stratified flow. The length of the pipe of the visualization test section was 1 m. The video data were taken after the flow reached a steady state. A correction box was employed to eliminate the refraction effect of the acrylic pipe. The LED lamp was used as the light source. The detail explanation about the experimental facility and the procedures to conduct the experiments had been presented by Kuntoro et al. [5] and Kuntoro [23]. To measure the processing time and random-access memory (RAM) usage performances between the present and preceding [5] algorithms, a PC with specification: CPU Intel®CoreTM i3-3110M, VGA AMD Radeon HD 7670M/2GB DDR3 and RAM 4GB DDR3 was used to execute both algorithms.

IMAGE PROCESSING TECHNIQUE After the stratified flow phenomena had been recorded in the format of video data, the recorded videos were then transferred to a PC where the data were visualized and processed. Next, the recorded videos were extracted into sequential individual RGB images by using VirtualDub, a video extraction software. Each individual RGB image was


then converted into a grayscale image by using MATLAB. After that, there were sequential individual grayscale images for further processes. Figure 2 shows the flowchart comparison between the present and preceding [5] algorithms of image processing technique for the stratified flow cases. The present and preceding [5] algorithms of image processing technique were written in MATLAB. In the preceding algorithm [5], the grayscale images were converted into the binary images by means of filtering and segmenting processes. The filtering process included image complement, zero filtering, and average filtering processes. Meanwhile, the segmenting process included image contrast enhancement and thresholding processes. Later, the processed images were converted into the binary images. Before gathering the film thickness data, the binary images were processed by using morphological image process where the skeletonize method was used. The detail explanation about it had been presented by Kuntoro et al. [5].

FIGURE 1. The schematic diagram of the experimental facility.

On the other hand, in the present algorithm, the processes did not include segmenting process and conversion process into the binary images. For stratified flow at JG = 1.88 m/s and JL = 0.077 m/s, the example results of each process of the present algorithm are shown in Fig. 3 and explained as follows: The film thickness data were obtained after the filtering process (see Fig. 2). The present algorithm used the grayscale images (Fig. 3.b) as the source data for detecting the film thickness of the flow and did not use the binary images. In the present work, the filtering process (see Figs. 3.c – 3.e) included image complement, image subtraction, and average filtering processes. The image complement process was the process where the values of the grayscale images were reversed into the complement values (see Fig. 3.c). The image subtraction process was carried out by subtracting the complement images (Fig. 3.c) with a grayscale background image that had been complemented (see Fig. 4.b). In the present work, a grayscale background image was made by combining a full of air grayscale background image and a full of the water grayscale background image (top area: full of air; bottom area: full of water; see Fig. 4.a). This combining method to create a grayscale background image (Fig. 4.a) was also performed by Vallée et al. [22] in a case of CCFL. For the average filtering process, a 4x4 neighbor pixel size was used to remove the pixel values which were unrepresentative of their surroundings by replacing each pixel value in an image with an average value of its neighbor. The results of image subtraction and average filtering processes are shown in Figs. 3.d and 3.e, respectively. Meanwhile, the grayscale background images of stratified flow at JG = 1.88 m/s and JL = 0.077 m/s are shown in Fig. 4. For the film thickness detection, the principle was by detecting the positions of the maximum gray values in each pixel column of the images (see Figs. 3.e and 3.f). There were 640 pixel columns in the present data. After the positions of each maximum gray value had been recorded, the film thickness data were measured by calculating the distance in y-coordinate between the base position and the position of the maximum gray value in each pixel column of an image (see Figs. 3.e and 3.f). Figure 3.f shows the film thickness detection result which is shown by the red line, whereas Fig. 3.g shows the plot of the film thickness detection result in the RGB image. In the present work, it was found that


the pixel-to-mm (millimeter) ratios were 0.323 mm/pixel for the width (x-coordinate) and 0.351 mm/pixel for the height (y-coordinate). These ratios were used to convert the pixel dimensions into the real dimensions (for example: mm). To measure the local film thickness data, four reference pixel columns were chosen as the samples with the distance 100 pixels between each other. There were 100th, 200th, 300th and 400th pixel columns. The measurement results of the local film thickness data on a chosen pixel column were shown in Figs. 8 – 10 as the time-series plots.

START

The flow line of the preceding algorithm [5] The flow line of

Recording videos Extracting RGB images Converting RGB images into grayscale images Filtering process of grayscale images Segmenting process of grayscale images Converting grayscale images into binary images Gathering the film thickness data, as well as the geometrical properties data of the

Gathering the film thickness data

END FIGURE 2. The flowchart comparison of the image processing technique between the present algorithm and the algorithm of Kuntoro et al. [5].

Furthermore, the present algorithm had been designed to detect the peaks and valleys of the gas-liquid interfaces, so the algorithm was capable to calculate the wavelength and wave amplitude. Figure 5 shows the peaks and valleys detection result of the interfacial waves of stratified flow at J G = 1.88 m/s and JL = 0.077 m/s. The principle of the peaks and valleys detection was by determining the positions of the red line (film thickness data) which had a zero gradient. On the basis of x-coordinate, a peak had a positive gradient at xp – 1 coordinate and a negative gradient at xp + 1 coordinate, with xp coordinate was the position of a peak in x-coordinate. On the contrary, a valley had a negative gradient at xv – 1 coordinate and a positive gradient at xv + 1 coordinate, with xv coordinate was the position of a valley in x-coordinate. The wavelengths of the interfacial waves were calculated by measuring the distance in xcoordinate between the peaks or the valleys (see Fig. 5), whilst the wave amplitudes of the interfacial waves were calculated by measuring the distance in y-coordinate between the peaks and valleys (see Fig. 5). To avoid the same detection of a peak or a valley in a grayscale image, the sequential grayscale images were chosen with a specific interlude depending on the average wave velocity of stratified flow. For example, if the stratified flow data had an interlude 60 images, the sequential images that would be processed were 1st, 61st, 121st images and so on.


The average wave velocity was calculated by dividing the distance between two reference pixel columns with the measured time-lag of the wave detection (see Eq. 1). The cross-correlation method was applied to calculate the timelag of the wave detection between two reference pixel columns.

V

s t

(1)

Where:

V = the average wave velocity;

s = the distance between two reference pixel columns;

t = the time-lag of the wave detection between two reference pixel columns. In the present paper, the measured wavelengths and wave amplitudes are stated in the term average (see Eqs. 2 and 3). The measurement results of the average wavelengths and wave amplitudes are shown in Fig. 12 in the term normalized with the pipe inner diameter (see Eqs. 4 and 5). Also, the relation between the average wavelength and wave amplitude can be stated in the term wave aspect ratio (see Eq. 6).

O1 A1

¦O

i

(2)

n

¦A

i

(3)

n

O D

Normalized wavelength Normalized wave amplitude Wave aspect ratio

O A

(4)

A D

(5) (6)

Where:

O = the average wavelength; Oi = the individual wavelength; n = the total number of detected wave; A = the average wave amplitude; Ai = the individual wave amplitude; D = the pipe inner diameter; Subscript 1 = method 1. The average wavelength can also be calculated by dividing the average wave velocity with the dominant wave frequency (see Eq. 7). The power spectral density (PSD) method was used to calculate the dominant wave frequency. On the other hand, according to Johnson et al. [24], the average wave amplitude can be calculated by using Eq. 8. The comparison results between the calculation using the equations (method 2) and the measurement of the visual data using the present algorithm (method 1) are shown in Fig 11.

O2 A2

V f

(7)

2 2V hL

(8)


Where:

f

= the dominant wave frequency;

V hL = the standard deviation of the time-series data of the film thickness; Subscript 2 = method 2.

Gas (a)

Liquid cm

(b)

(d)

The nth pixel column The position of the maximum gray value

Image filtering process

(c)

The base position

(e)

The position of the maximum gray value (f)

The base position

(g) FIGURE 3. The image processing results of stratified flow at JG = 1.88 m/s and JL = 0.077 m/s by using the present algorithm: (a) RGB image; (b) Grayscale image; (c) Image complement result; (d) Image subtraction result; (e) Image after average filtering; (f) Film thickness detection result, which is shown by the red line; (g) Film thickness detection result is plotted in the RGB image.


(a)

(b) FIGURE 4. The grayscale background images of stratified flow at JG = 1.88 m/s and JL = 0.077 m/s: (a) Grayscale background image before image complement process; (b) Grayscale background image after image complement process.

Y (mm)

Wavelength Wave amplitude

X (mm) FIGURE 5. The peaks and valleys detection result of the interfacial waves of stratified flow at JG = 1.88 m/s and JL = 0.077 m/s (Red circle: the peak of the wave; blue triangle: the valley of the wave; unit is in mm).

Even though there is a possibility for the interfacial curvature to be concave or convex [25], for the cases of stratified air-water two-phase flow, the interfacial curvature can be assumed to be plane curvature [26]. The effect of concave or convex interfacial curvature in stratified air-water two-phase flow can be neglected [27]. In the present work, the plane curvature assumption was used. The positions of the gas and liquid interfaces were based on the positions of the maximum gray values in each pixel column. Researchers who used a plane curvature assumption on the study of stratified two-phase flow were Taitel and Dukler [28] and Hall and Hewitt [29]. To measure the processing time performance between the present and preceding [5] algorithms, tic and toc functions of MATLAB were used. Meanwhile, to measure the RAM usage performance between the present and preceding [5] algorithms, memory function of MATLAB was used.

RESULTS AND DISCUSSION Figures 6 and 7 show the comparisons between the present and preceding [5] algorithms on the processing time and random-access memory (RAM) usage performances when both algorithms were used to process the stratified flow data. As shown in Figs. 6 and 7, there are the data at JG = 1.02 m/s, JL = 0.063 m/s; JG = 1.02 m/s, JL = 0.077 m/s; JG = 1.88 m/s, JL = 0.031 m/s; JG = 1.88 m/s, JL = 0.047 m/s; JG = 1.88 m/s, JL = 0.063 m/s; JG = 2.83 m/s, JL = 0.031 m/s; and JG = 3.77 m/s, JL = 0.031 m/s. Each data contain 3598 RGB images with 640 x 480 pixels resolution. As seen in Fig. 6, the present algorithm has the faster processing time performances than the preceding algorithm [5] when both algorithms were used to process the given data. The average processing time for the stratified flow data using the present algorithm is 5 minutes 58 seconds, while the average processing time of the stratified flow data using the preceding algorithm is 7 minutes 52 seconds. It means that the average processing time using the present algorithm is 1 minute 54 seconds faster than using the preceding algorithm. The comparison of the RAM usage performances between the present and preceding algorithms is shown in Fig. 7. It can be noted that the present algorithm uses fewer random-access memory (RAM) than the preceding algorithm when both algorithms processed the same stratified flow data. The average RAM usage of the present algorithm is 1.1 gigabytes, while the average RAM usage of the preceding algorithm is 3.6 gigabytes. It means that the preceding algorithm uses 3.27 times more RAM than the present algorithm. The present algorithm uses 2.5 gigabytes fewer RAM than the preceding algorithm. Although the present algorithm had been equipped with the capability to measure the geometrical properties of the interfacial waves, the present algorithm still has a lower processing time of the stratified flow data and fewer RAM


Minutes

usage than the preceding algorithm [5] as shown in Figs 6 and 7. It is because the script of the present algorithm had been improved and managed to be able to be executed in the effective and efficient allocations of processing time and RAM usage. The improvements included the simplification of the looping process, the modification of the saving process and the deletion of the unnecessary dummy data in the looping process. By implementing these improvements, the present algorithm has better performance in processing time and RAM usage than the preceding algorithm. Figures 8 – 10 show the film thickness’ measurement results of stratified flow at 3 combinations of JG and JL. There were 3 distinguishable sub-flow patterns of stratified flow which were observed in the present work. There were stratified smooth (Fig. 8), stratified ripple (Fig. 9) and stratified roll (Fig. 10) sub-flow patterns. To determine the characteristics of each sub flow pattern, the measurement results were plotted in the time-series and PDF plots (see (c) and (d) of Figs. 8 – 10). Also, to amplify the interfacial wave pattern differences between each sub-flow pattern, the film thickness measurement results were plotted again in a graph (see (b) of Figs. 8 – 10). The film thickness data were expressed in the dimensionless film thickness (hL/D) by normalizing the data with the pipe inner diameter (D). 9 8 7 6 5 4 3 2 1 0

The present algorithm The preceding algorithm [5]

JG =

1.02

1.02

1.88

1.88

1.88

2.83

3.77

(m/s)

JL =

0.063

0.077

0.031

0.047

0.063

0.031

0.031

(m/s)

FIGURE 6. The processing time performances between the present and preceding [5] algorithms.

4

The present algorithm

Gigabytes (GB)

3.5 3

The preceding algorithm [5]

2.5 2 1.5 1 0.5 0

JG =

1.02

1.02

1.88

1.88

1.88

2.83

3.77

(m/s)

JL =

0.063

0.077

0.031

0.047

0.063

0.031

0.031

(m/s)

FIGURE 7. The RAM usage performances between the present and preceding [5] algorithms.

As shown in Fig. 8, the stratified smooth sub-flow pattern is characterized by the flat interface which has no significant fluctuation of film thickness in the time-series plot. As a result, the probability density function (PDF) plot shows a gathered dominant distribution in the narrow range of hL/D. This indicates that in the stratified smooth sub flow pattern there is no significant fluctuation of pressure inside the pipe. Figure 9 shows the characteristics of the stratified ripple sub-flow pattern. This sub-flow pattern is characterized by the occurrence of the periodic interfacial waves or the ripple waves with the wavelengths and wave amplitudes that are relatively uniform. In this sub-flow pattern, the distance of a peak to the front valley and the distance of a peak to the rear valley are relatively similar (see Fig. 9.b). In contrast with the PDF plot in Fig. 8.d, the PDF plot in Fig. 9.d


shows a widened dominant distribution where the values of hL/D are distributed in a wider range. As seen in Fig. 9.c, the periodic fluctuation of the film thickness indicates the periodic fluctuation of pressure inside the pipe. The characteristics of the stratified roll sub-flow pattern are depicted in Fig. 10. This sub-flow pattern is characterized by the occurrence of the irregular interfacial waves or the roll waves. The shape of a roll wave is like “the whale-head shape” which has longer distance of a peak to the rear valley than a peak to the front valley (see Fig. 10.b). In this sub-flow pattern, the PDF plot has a widened dominant distribution (see Fig. 10.d). This characteristic is similar with the PDF plot in the stratified ripple sub-flow pattern (Fig. 9.d). Hence, many researchers [28, 30 and 31] grouped the stratified ripple and roll sub-flow patterns as the stratified wavy sub-flow pattern. As seen in Fig. 10.c, by observing the fluctuations of the film thickness, the irregular fluctuations with the sudden increases of pressure occur inside the pipe.

Gas Liquid (a)

1

flat interface

gathered dominant distribution

0.8

PDF

hL / D (-)

(b) 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

no significant fluctuation

0.6 0.4 0.2 0

0

1

2

3

Time (s) (c)

4

5

0

0.2

0.4

0.6

hL / D (-) (d)

0.8

1

FIGURE 8. The measurement results of stratified flow at JG = 1.88 m/s and JL = 0.016 m/s: (a) Film thickness detection result; (b) Film thickness plot (unit is in mm); (c) Time-series plot; (d) Probability density function (PDF) plot.


Gas Liquid (a)

1

0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

ripple wave

0.8

PDF

hL / D (-)

(b)

periodic fluctuation

widened dominant distribution

0.6 0.4 0.2 0

0

1

2

3

Time (s) (c)

4

5

0

0.2

0.4

0.6

hL / D (-) (d)

0.8

1



Gas Liquid (a)

1

sudden increase

roll wave

0.8

PDF

hL / D (-)

(b) 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

irregular fluctuation with the sudden increase of the film thickness

widened dominant distribution

0.6 0.4 0.2 0

0

1

2

3

4

Time (s) (c)

5

0

0.2

0.4

0.6

hL / D (-) (d)

0.8

1


Figure 11 shows the comparison results between methods 1 and 2 to determine the average wavelength and wave amplitude which are stated in the term normalized wavelength and wave amplitude. The difference of the results is ±3% for the normalized wavelength (Fig. 11.a) and ±10% for the normalized wave amplitude (Fig. 11.b). It can be inferred that methods 1 and 2 are capable to determine the average wavelength and wave amplitude with a good agreement of the results between each other. 2

0.1

+ 3% - 3%

1.6

+ 10%

0.08

A1 / D (-)

λ1 / D (-)

1.8

1.4

0.06

- 10%

0.04 0.02

1.2

0 1.2

1.4

1.6

1.8

2

λ2 / D (-) (a)

0

0.02

0.04

0.06

0.08

0.1

A2 / D (-) (b)

FIGURE 11. The comparison results between methods 1 and 2 on the: (a) normalized wavelength; (b) normalized wave amplitude.

Figure 12 shows the effects of the relative superficial velocity (JG/JL) on the normalized wavelength (Fig. 12.a), normalized wave amplitude (Fig. 12.b) and wave aspect ratio (Fig. 12.c). All data points in Fig. 12 were obtained by using method 1 (see Eqs. 2 and 3). In Figs. 12.a and 12.b, as the relative superficial velocity increases, the normalized


wavelength and wave amplitude decreases. The wavelength becomes shorter and the wave amplitude becomes lower when the relative superficial velocity increases. At the high relative superficial velocity, the superficial gas velocity becomes dominant resulting the higher disturbance and creating waves to the gas-liquid interface. The higher the disturbance, the shorter the wavelength and the lower the wave amplitude of the interfacial waves. In the concept of wave aspect ratio (Fig. 12.c), as the relative superficial velocity increases, the wave aspect ratio increases as well. In the present paper, the data presented in Fig. 12 fulfill the linear trendlines as stated in Eq. 9 for the normalized wavelength, Eq. 10 for the normalized wave amplitude and Eq. 11 for the wave aspect ratio on the relation with the relative superficial velocity.

O1

§J 0.0018¨¨ G © JL §J 0.0004¨¨ G © JL

D

A1 D

O1 A1

· ¸¸ 1.6551 ¹ · ¸¸ 0.0815 ¹ §J · 0.1405¨¨ G ¸¸ 19.474 © JL ¹

2

r2

0.3476

(9)

;

r2

0.7241

(10)

;

r2

0.9245

(11)

0.1

(λ1/D) = -0.0018(JG/JL) + 1.6551 r² = 0.3476

0.08

A1 / D (-)

1.8

λ1 / D (-)

;

1.6 1.4

0.06 0.04

(A1/D) = -0.0004(JG/JL) + 0.0815 r² = 0.7241

0.02

1.2

0 0

20

40

60

80

100

120

140

0

20

40

JG / JL (-) (a)

60

80

100

120

140

JG / JL (-) (b)

40

(λ1/A1) = 0.1405(JG/JL) + 19.474 r² = 0.9245

λ 1/ A1 (-)

35 30 25 20 15 0

20

40

60

80

100

120

140

JG / JL (-) (c) FIGURE 12. The effects of the relative superficial velocity (JG/JL) on the: (a) normalized wavelength (λ1/D); (b) normalized wave amplitude (A1/D); (c) wave aspect ratio (λ1/A1). The data λ and A were obtained by using method 1.

CONCLUSION An improved algorithm of image processing technique to measure the film thickness data of stratified gas-liquid two-phase flow has been presented. The stratified flow data from the experiments of Kuntoro et al. [5] in a 26 mm i.d.


horizontal pipe were used for the input data of the image processing algorithm. The present algorithm shows lower processing time and RAM usage performances than the preceding algorithm [5] when both algorithms were used to process the same stratified flow data. The present algorithm is 1 minute 54 seconds faster on the processing time of the stratified flow data and 2.5 gigabytes fewer on the RAM usage than the preceding algorithm [5]. In the present paper, the image processing results from the present algorithm are presented and explained on the basis of the film thickness data. The results are in good agreement with the previous works. Furthermore, the measurement results of the geometrical properties of the interfacial waves are presented and explained in the relation with the relative superficial velocity (JG/JL). As the relative superficial velocity increases, the normalized wavelength and wave amplitude decreases, while the wave aspect ratio increases. The data points of the geometrical properties’ measurement results fulfill the linear trendlines as written in Eqs. 9, 10 and 11 on the relation with the relative superficial velocity.

ACKNOWLEDGMENT The authors would like to thank Yosephus Ardean, Sugiyanto and Arif Widyatama for the experimental raw data.

REFERENCES 1. 2. 3.

4. 5.

6. 7. 8. 9. 10. 11. 12. 13.

E. Al-Safran, “Investigation and prediction of slug frequency in gas/liquid horizontal pipe flow,” in Journal of Petroleum Science and Engineering 69, 143-155 (2009). A. I. Majid, O. Dinaryanto, B. Hartarto, Deendarlianto, Indarto, “The application of the image processing technique to analyze the flow structure of the horizontal air-water plug two-phase flow,” in Proceedings of the 10th International Forum on Strategic Technology (IFOST 2015), Bali, Indonesia, 3 – 5 June 2015. H. Y. Kuntoro, M. Banowski, Deendarlianto, “A study of three dimensional bubble velocities at co-current gas-liquid vertical upward bubbly flows,” in Proceedings of the 1st International Conference on Engineering Technology and Industrial Application (ICETIA 2014), Universitas Muhammadiyah Surakarta, Surakarta, Indonesia, 4 December 2014. A. Setyawan, Indarto, Deendarlianto, “The effect of the fluid properties on the wave velocity and wave frequency of gas–liquid annular two-phase flow in a horizontal pipe,” in Experimental Thermal and Fluid Science 71, 25-41 (2016). (Available online 20 October 2015). H. Y. Kuntoro, A. Z. Hudaya, O. Dinaryanto, Deendarlianto, Indarto, “Experimental study of the interfacial waves in horizontal stratified gas-liquid two-phase flows by using the developed image processing technique,” in Proceedings of the 10th International Forum on Strategic Technology (IFOST 2015), Bali, Indonesia, 3 – 5 June 2015. M. Abdulkadir, V. Hernandez-Perez, S. Lo, I. S. Lowndes, B. J. Azzopardi, “Comparison of experimental and Computational Fluid Dynamics (CFD) studies of slug flow in a vertical riser,” in Experimental Thermal and Fluid Science 68, 468-483 (2015). U. Kadri, R. F. Mudde, R. V. A. Oliemans, “Influence of the operation pressure on slug length in near horizontal gas–liquid pipe flow,” in International Journal of Multiphase Flow 36, 423-431 (2010). B. D. Woods, Z. Fan, T. J. Hanratty, “Frequency and development of slugs in a horizontal pipe at large liquid flows,” in International Journal of Multiphase Flow 32, 902-925 (2006). P. M. Ujang, C. J. Lawrence, C. P. Hale, G. F. Hewitt, “Slug initiation and evolution in two-phase horizontal flow,” in International Journal of Multiphase Flow 32, 527-552 (2006). E. Pitton, P. Ciandri, M. Margarone, P. Andreussi, “An experimental study of stratified–dispersed flow in horizontal pipes,” in International Journal of Multiphase Flow 67, 92-103 (2014). M. Birvalski, M. J. Tummers, R. Delfos, R. A. W. M. Henkes, “PIV measurements of waves and turbulence in stratified horizontal two-phase pipe flow,” in International Journal of Multiphase Flow 62, 161-173 (2014). D. Gorelik and N. Brauner, “The interface configuration in two-phase stratified pipe flows,” in International Journal of Multiphase Flow 25, 977-1007 (1999). A. Soleimani and T. J. Hanratty, “Critical liquid flows for the transition from the pseudo-slug and stratified patterns to slug flow,” in International Journal of Multiphase Flow 29, 51-67 (2003).


14. U. Kadri, R. F. Mudde, R. V. A. Oliemans, M. Bonizzi, P. Andreussi, “Prediction of the transition from stratified to slug flow or roll-waves in gas–liquid horizontal pipes,” in International Journal of Multiphase Flow 35, 1001-1010 (2009). 15. E. Al-Safran, C. Kora, C. Sarica, “Prediction of slug liquid holdup in high viscosity liquid and gas two-phase flow in horizontal pipes,” in Journal of Petroleum Science and Engineering 133, 566-575 (2015). 16. H. Q. Zhang, Q. Wang, C. Sarica, J. P. Brill, “A unified mechanistic model for slug liquid holdup and transition between slug and dispersed bubble flows,” in International Journal of Multiphase Flow 29, 97107 (2003). 17. A. Sanchis, G. W. Johnson, A. Jensen, “The formation of hydrodynamic slugs by the interaction of waves in gas–liquid two-phase pipe flow,” in International Journal of Multiphase Flow 37, 358-368 (2011). 18. A. M. Ozbayoglu and H. E. Yuksel, “Analysis of gas–liquid behavior in eccentric horizontal annuli with image processing and artificial intelligence techniques,” in Journal of Petroleum Science and Engineering 81, 31-40 (2012). 19. G. A. Montoya, Deendarlianto, D. Lucas, T. Höhne, C. Vallée, “Image-processing-based study of the interfacial behavior of the countercurrent gas-liquid two-phase flow in a hot leg of a PWR,” in Science and Technology of Nuclear Installations 2012, Article ID 209542 (2012). 20. C. E. F. do Amaral, R. F. Alves, M. J. da Silva, L. V. R. Arruda, L. Dorini, R. E. M. Morales, D. R. Pipa, “Image processing techniques for high-speed videometry in horizontal two-phase slug flows,” in Flow Measurement and Instrumentation 33, 257-264 (2013). 21. M. S. de Castro, C. C. Pereira, J. N. dos Santos, O. M. H. Rodriguez, “Geometrical and kinematic properties of interfacial waves in stratified oil–water flow in inclined pipe,” in Experimental Thermal and Fluid Science 37, 171-178 (2012). 22. C. Vallée, T. Nariai, T. Futatsugi, A. Tomiyama, D. Lucas, M. Murase, “Experimental characterisation of the interfacial structure during counter-current flow limitation in a model of the hot leg of a PWR,” in Science and Technology of Nuclear Installations 2012, Article ID 298452 (2012). 23. H. Y. Kuntoro, “The implementation of image processing technique to determine the interfacial behavior of gas-liquid wavy two-phase flow in a horizontal pipe,” Bachelor’s Thesis (Skripsi), Gadjah Mada University, 2013. 24. G. W. Johnson, A. F. Bertelsen, J. Nossen, “An experimental investigation of roll waves in high pressure two-phase inclined pipe flows,” in International Journal of Multiphase Flow 35, 924-932 (2009). 25. A. Ullman and N. Brauner, “Closure relations for two-fluid models for two-phase stratified smooth and stratified wavy flows,” in International Journal of Multiphase Flow 32, 82-105 (2006). 26. N. Brauner, J. Rovinsky, D. M. Maron, “Determination of the interface curvature in stratified two-phase systems by energy considerations,” in International Journal of Multiphase Flow 22, 1167-1185 (1996). 27. T. S. Ng, C. J. Lawrence, G. F. Hewitt, “Interface shapes for two-phase laminar stratified flow in a circular pipe,” in International Journal of Multiphase Flow 27, 1301-1311 (2001). 28. Y. Taitel and A. E. Dukler, “A model for predicting flow regime transitions in horizontal and near horizontal gas-liquid flow,” in AIChE Journal 22, 47-55 (1976). 29. A. R. W. Hall and G. F. Hewitt, “Application of two-fluid analysis to laminar stratified oil-water flows,” in International Journal of Multiphase Flow 19, 711-717 (1993). 30. J. M. Mandhane, G. A. Gregory, K. Aziz, “A flow pattern map for gas—liquid flow in horizontal pipes,” in International Journal of Multiphase Flow 1, 537-553 (1974). 31. J. Weisman, D. Duncan, J. Gibson, T. Crawford, “Effects of fluid properties and pipe diameter on two-phase flow patterns in horizontal lines,” in International Journal of Multiphase Flow 5, 437-462 (1979).
