Experimental study on the flow pattern and pressure gradient of air-water two-phase flow in a horizontal circular mini-channel ＊

Sudarja , Aqli Haq Deendarlianto , Indarto , Adhika Widyaparaga

1. Departement of Mechanical and Industrial Engineering, Gadjah Mada University, Yogyakarta, Indonesia

2. Centre for Energy Studies, Gadjah Mada University, Sekip K-1A Kampus UGM, Yogyakarta, Indonesia

3. Departement of Mechanical Engineering, Universitas Muhammadiyah Yogyakarta, Yogyakarta, Indonesia

Abstract: Experimental studies on the flow pattern and the pressure gradient of gas-liquid co-current two-phase flow in a mini-channel were conducted. The test section was a transparent circular channel of 1.6 mm inner diameter. The working fluids were air and water.The superficial velocities of gas and liquid were in the range of 0.025-66.300 m/s and 0.033-4.935 m/s, respectively. In the present work, the flow pattern and the pressure gradient data were obtained by analyzing the flow images captured by a high-speed camera, and by using the pressure transducer, respectively. As a result, it was found that (1) the obtained flow patterns were bubbly, plug,slug-annular, annular, and churn flows, (2) new experimental correlations on the bubble and plug lengths were proposed, whereas the lengths are the function of the homogeneous void fraction, (3) both gas and liquid superficial velocities affect proportionally to the pressure gradient, whereas it increases with the increase both of JG , JL . In addition, the obtained flow patterns are in a good agreement with that of the available flow pattern maps in the open literatures, such as, Triplett et al. (1999) and Chung and Kawaji (2004).

Key words: Two-phase flow, mini channel, flow pattern, pressure gradient, plug length

Introduction

Nowadays, mini and micro channels have been used in many engineering applications, such as compact heat exchangers, microelectronic cooling systems,research nuclear reactors, chemical processing, and small-sized refrigeration systems[1]. Moreover, Zhao and Bi[2]noted that those channels can be used in the field of the cooling of high-density multi-chip modules in supercomputers, high-powered X-ray and other diagnostic devices, high-flux heat exchangers in aerospace systems, and cryogenic cooling systems in satellites. Next, it can be found also in the micropower generation, and fuel cells[3].

Fukano and Kariyasaki[4]conducted an experimental investigation on the characteristics of twophase flow in capillary tubes with the inner pipe diameters of 1 mm, 2.4 mm and 4.9 mm. The attention was paid on the flow pattern, void fraction, and pressure drop. The static pressure and pressure difference were measured by using the pressure transducer.The void fraction was measured by the constant current method. They noticed that the flow direction did not affect on the flow pattern, for Dc≤ 4.9 mm,herecD is defined as inner pipe diameter. In the pipe with a smaller inner diameter, the pressure gradient is much larger than those estimated by Chisholm’s equation.

Mishima and Hibiki[5]reported the characteristics of gas-liquid two-phase flow in the small diameter tubes with the inner diameters of 1-4 mm. They proposed a new Chisholm parameter ( C) as a function of inner diameter, in which,is applicable to both horizontal and vertical capillary tube. Furthermore Triplett et al.[1,6]performed an experimental study on the small size circular channel (1.10 mm, 1.45 mm of the inner diameter) and semi-triangular channel (with hydraulic diameter of 1.09 mm, 1.49 mm). The work focused on the flow pattern, void fraction, and pressure drop. They reported that the observed flow patterns were bubbly, slug, slug-annular, annular, and churn. They also reported that based on previous studies, significant differences between small size and large size channels in term of two-phase flow patterns,two-phase pressure drop, as well as boiling heat transfer and critical heat flux were found. The predominance of surface tension in small size channel reduces significantly the slip velocity, and renders the independency of the flow characteristics against channel orientation in terms of gravity.

Serizawa et al.[7]carried out the experimental studies by using air-water and steam-water as working fluids. The inner pipe diameters were 20 μm, 25 μm and 100 μm. The observed flow patterns were dispersed bubbly flow, gas slug flow, liquid ring, liquid lump, annular, frothy or wispy annular, rivulet, liquid droplets and special type of flow. In terms of the void fraction, their results were in good agreement with Armand correlation. They concluded that =α 0.833β. Here α, β are measured void fraction and homogeneous void fraction, respectively.

Kawahara et al.[8]carried out an experimentalwork on 100 m of inner diameter of circular channel made of fused silica. The working fluids were de-ionized water and nitrogen. They reported that the observed flow patterns were intermittent and semiannular flows, while bubbly and churn flow patterns were absent. In terms of void fraction, all of their data was far below the homogeneous line model. They explained that, even at low gas flow rates, the slip ratio in micro channels is higher than those of the channels with the hydraulic diameter bigger than 1 mm.The two phase flow multiplier data were in good agreement with the Lochart-Martinelli’s separated flow model, but were over-predicted against the homogeneous flow model.

Similar to that of Kawahara et al.[8], Chung and Kawaji[9]also conducted the experimental works on the micro channel two-phase flow with the inner pipe diameters of 530 μm, 250 μm, 100 μm and 50 μm.The working fluids were nitrogen gas and water. They reported that the identified flow patterns for the flow channels with the inner diameters of 530 μm, 250 μm were bubbly, slug, slug-annular, annular, and churn.Meanwhile, for the channel with the inner pipe diameters of 100 μm, 50 μm, the observed flow pattern was only slug. The absence of other flow patterns in smaller channel diameters was caused by higher surface tension and viscous effects on the liquid flow. In this case, the inertia force has inadequate force to change the slug flow to churn neither annular flow even both of superficial velocities were increased. In terms of pressure gradient, they reported that for the flow channel with inner diameters larger than 250 μm, the two-phase pressure gradient is affected by mass flux, as found in mini channels.

The measurement of the slug length and the trailing bubble velocity was done by Xia et al.[10]by using EKTAPRO 1000 high speed motion analyzer and the optical probe.They stated that the correlation between the length of liquid slug ahead of that bubble and the trailing bubble velocity is derived from the experimental data. They then proposed a model for the slug length distribution at any designated locations along the pipe based on this correlation. The predicted results are in agreement with the experimental data.

Saisorn and Wongwises[11]utilized a fused silica channel with the inner pipe diameter of 0.53 mm as test pipe in their study. The identified flow patterns were slug, throat annular, churn, and annular rivulet.Next, the frictional pressure gradient is affected by mass flux as well as flow pattern. Sur and Liu[12]conducted the experimental study with the inner diameters of 100 μm, 180 μm and 324 μm of fused silica. They used air and water as the working fluids.They identified four distinctive flow regimes, namely,bubbly, slug, ring, and annular flow. Concerning the frictional pressure drop, they concluded that the flow pattern-based model provides a more accurate result in comparing with the homogeneous and separated two-phase flow models. Recently, Barreto et al.[13]also analyzed the pressure drop and void fraction in a circular pipe gas-liquid two-phase flow with an inner diameter of 1.2 mm. They concluded also that the predicted pressure drop was affected by the flow pattern and the liquid flow regime.

Beside the above mention studies, there are some researchers who paid special attention to the effect of surface tension to the two-phase flow patterns as well as pressure drop, such as Liu and Gao[14]and Xia and Chai[15]. Liu and Gao[14]performed an experimental study on the effect of surfactant on two-phase flow patterns in 1.6 mm inner diameter pipe. The pairs of working fluids used in their study were air-water and air-water/sodium dodecyl benzoyl sulfate (SDBS)mixture. They reported that the flow pattern transitions in vertical small tube, for air-SDBS mixture, tend to occur at lower flow velocities compared with those of the gas-water flow. This was mainly caused by the fact that surfactant decreased the surface tension of liquid working fluid.

Xia and Chai[15]used air-water and air-100 ppm sodium dodecyl sulphate aqueous solution as working fluids in their investigation on the influence of surfactant on the two-phase flow regime and pressure drop in upward incline pipes. From the study, they concluded that the pressure gradient reduced significantly when the surfactant was added in liquid, for annular and slug flow patterns. Furthermore, they also found that particularly in the annular flow regime, the pressure gradient gradually became free of the upward angle inclination effect, and was only dependent on the two-phase flow properties.

Fig. 1 Schematic diagram of experimental apparatus

From the above facts, it is possible to be noticed that the channel size and fluid properties significantly influence the flow patterns and pressure gradient.However, there are some differences on the obtainedflow patterns and the formulation of the pressure gradient among those studies, although the used channel size and fluid properties were almost the same.For example, when some available maps were compared among each others, the discrepancies of the transition line was still observed[16]. For the above reason,further investigation on the mini channel gasliquid two-phase flow is needed.

The objective of the present study is to investigate the basic fundamental data on the gas-liquid of two-phase flow in a mini-channel. A large experimental data set of flow pattern and pressure gradient were obtained to achieve the goal. The obtained experimental data are also dedicated also for further commercial CFD codes validation. In the present paper, the flow pattern map is displayed also in the form of phasic Weber number to reveal the effect of inertia and surface tension in the mini channel.Moreover, the effect of both Reynolds number and flow patterns to the pressure gradient is also discussed in detail.

1. Experimental apparatus and procedure

Figure 1 shows a schematic diagram of the experimental apparatus used in the present study. It consists of a test section, an air compressor, a pneumatic pump, gas and liquid flow meters, pressure sensors, a camera, a data acquisition system, and a personal computer. The test section was a transparent circular pipe made of glass with the inner diameter of 1.6 mm,400 mm length. The test section was equipped with an optical correction box in order to avoid the effect of tube curvature as conducted also by Kawahara et al.[8].

In the present experimental study, the air was supplied by air compressor, whereas it was equipped with a water trap and an air dryer to ensure free water content in the compressed air. The stainless steel pressure vessel, as a pneumatic pump, was provided to feed (instead of pumping, to avoid pulsation by pump)the liquid. The water flow meters from Omega and TOKYO KEISO with the accuracy of ±5 %, ±3 %,respectively, were used. To measure the differential pressure between two locations, a differential pressure transducer from Validyne with an accuracy of ±0.25 %was utilized. The differential pressure transducers were connected to a T-junctions fitted on the upstream and downstream of the test section. Meanwhile, to measure the static pressure, a static pressure sensor from Copal Electronic was used. It was mounted on the test section upstream.

A high-speed camera, Nikon J4, with a recording speed up to 1200 fps and a resolution of 640×480pixel was used to capture the flow images. A data acquisition system from National Instrument was applied to convert the analog to digital data. The working fluids were air and water. The experiments were conducted in the range of gas and liquid superficial velocities of 0.025-66.300 m/s and 0.033-4.935 m/s,respectively as shown clearly in a test matrix (Table 1).

Table 1 Test matrix

In the experiments, the water was injected by a pneumatic pump (stainless steel pressure vessel), then flowed to the mixer, through the flow meter. Air was compressed from the air compressor to the mixer through a regulator, water trap, and a gas flow meter.In the mixer, water and air were mixed, and then the two-phase fluids mixture flowed to test section, and finally returned back to the storage tank and separated.The visualization data were captured by using a highspeed camera. The video images from the high-speed camera were then displayed in the form of slow motion images for the detail observation. Hence, the flow regime map can be plotted.

2. Results and discussion

2.1 Flow pattern and flow pattern map

The flow pattern was obtained by capturing the flow images in the test section by using a high-speed video camera. The morphology of the observed flow patterns and the time series of the void fraction are shown in Table 2. As has been previously reported by Sudarja et al.[17], the observed flow patterns are bubbly, plug, slug-annular, annular, and churn flow.Meanwhile, the stratified flow is absent in the present study due to the dominance of surface tension force.Here the gravitational force is less than surface tension force, therefore, stratified flow does not occurred.

Bubbly flow occurs at very low0.3 m/s), and moderate to highAs shown in Table 2(a), the bubbly flow pattern is characterized by the occurrence of air bubbles with the diameters are smaller than or the same as the inner pipe diameter, in a continuous liquid flow. Here the air and water superficial velocities were JG=0.041m/s, JL= 0.879 m/s , respectively, as an example. From the image, it is shown that most of the form of air bubbles are non-uniform either shape or size. Most of them are non-spherical and the sizes are smaller than the inner pipe diameter. The time variation of void fraction of this flow pattern is in the form of sharp curve with the height disperse from zero up to little under one, and sometimes blunt curve is appeared. This means that in the bubbly flow, the flow consist of liquid, bubbles in various size, and occasionally accompanied by a plug, since the height of curve indicates the height of bubble, while the thickness of the curve shows the axial bubble length.

The plug flow pattern, as shown in Table 2(b),occurs at a relatively low JG. Here, the air and water slugs flow intermittently. An air plug is a long bubble or called elongated bubble. The diameter of air plugs are approximately smaller or the same as the inner pipe diameter, as also indicated in the time variation of void fraction shown clearly in Table 2(b). The magnitude of the void fraction are also intermittently 0 (represent liquid alone) and 1.0 (represent plug with diameter same as its channel diameter). The length of air plug depends on the JG, here, the higher the JGthe longer the air plug. The dependency of air plug length on JGwas also reported by Fukano and Kariyasaki[4], Saisorn and Wongwises[11]. In line withthem, Triplett et al.[1]explicitly reported that the parameter changes (increasing JGor decreasing JL)lead to a higher void fraction and implies to longer air plug or shorter liquid slug.

Table 2 Sample of flow patterns observed in pipe with inner diameter of 1.6 mm

In order to confirm the location of bubble and plug originated, detailed observations were taken at the fluid mixer (Fig. 2). At low JG, the formation of bubbles or plugs mostly takes place in the mixing area(in the convergent region before leave the mixer). At low JG, gas is restrained by liquid and it forms an air bubble, prior to entering the mixing chamber. After the bubble size large enough, it pierces the existing liquid inside mixing chamber in the form of a large air plug.Furthermore, the air plug passes though the mixer outlet side. At the outlet side, air plug transforms into bubbles when JLis high (Fig. 2(a)), while at low JLit remains in a plug form (Fig. 2(b)).

When the direction of the working fluidsentry is changed (the air from the top and the liquid from the left side), the observed flow patterns were exactly the same, meanwhile, the mechanism of the gasentry in the mixing chamber is different. Gas enters the mixing chamber from the top and directly pushed by liquid to the convergent side, then, form bubbles or plugs,depends on the JG, JL.

Fig. 2 (Color online) Mechanism of air bubbles and plugs formation in the mixer

The slug-annular flow pattern is the transition from the slug flow to annular flow. The gas penetrates the liquid bridges that separate gas plugs and produce liquid ring. In some cases, the liquid rings are thicker than those in other position and the liquid neck was also observed, as shown in Table 2(c). The void fraction in this flow condition is illustrated by the periodically downward pulse line. The slug-annular being observed in the present study is in agreement with that of other studies. In addition, some researchers defined this flow pattern with the different terminologies, such as Saisorn and Wongwises[11],Serizawa et al.[7], Kawahara et al.[8], Sur and Liu[12],Triplett et al.[1]as throat-annular flow,liquid ring flow,gas core with a deformed interface, ring flow, slugannular flow, respectively.

Annular flow is characterized by the gas core flows in the center of the channel surrounded by liquid at the channel wall as liquid film, as shown in Table 2(d). Experimental observations indicate that the film thickness depends on the liquid superficial velocity,whereas, the higher JLthe thicker liquid film. The liquid films in the present experiment are sometimes in the form of the wavy, similar to those observed by Triplett et al.[1]. This condition is shown also in the time variation of void fraction by the ripple curve,with low amplitude and high frequency. In general,annular flow occurred under high JG, low JL.

Churn flow is characterized by the appearance of flow disruptions, and flow tends to be an irregular form as shown in Table 2(e). This condition is also illustrated by the rapid fluctuation of time variation of void fraction with extreme magnitude. It was also occurred at high both JG, JL. Churn flow was also observed in some previous studies in circular minichannel, such as Mishima and Hibiki[5], Triplett et al.[1], Zhao and Bi[2], Chen et al.[18], and Pehlivan et al.[16].

Fig. 3 (Color online) Two-phase flow pattern map for air-water flow through a 1.6 mm diameter channel

To present the flow pattern data, it is needed to classify the observed flow patterns and then plot them as a flow pattern map. The parameters used in the present study are gas superficial velocity ( JG) as an abscissa and liquid superficial velocity ( JL) as an ordinate, and both are in logarithmic scales as shown in Fig. 3. In the figure, the boundaries of the flow patterns are indicated by the bold lines.

The flow pattern map of the present study is compared with the available maps, such as those from Triplett et al.[1], Chung and Kawaji[9], Saisorn and Wongwises[11], and Sur and Liu[12], as depicted in Figs.4(a)-4(d). As shown in Fig. 4(a), it is noticed that the present data are in good agreement partially with those obtained by Triplett et al.[1]. The region of some flow patterns (bubbly, slug, and churn) of their study also corresponds well to bubbly, plug, and churn of present study, respectively.

In comparing with that of Chung and Kawaji[9],Fig. 4(b), it is found that, in general, they are in quite good agreement, although the working fluids are different. The regions of bubbly, plug, and churn of the present study corresponds well with those of obtained by them. Next, Saisorn and Wongwises[11]carried-out the experiments by using air and water as working fluids in a fused silica circular channel. The inner pipe diameter was 0.53 mm. The observed flow pattern map obtained from their works are compared with the present result, as shown in Fig. 4(c). Close observation from figure, it is revealed that the slug and churn flow regions of their data correspond well to that of the present study.

Fig. 4 (Color online) Comparisons of the observed flow patterns with the transition lines by other authors

Figure 4(d) indicates the comparison between the flow pattern map obtained from the present study and Sur and Liu[12]. The inner pipe diameters used in their study were 0.324 mm, 0.18 mm of fused silica pipe.The figure shows that the transition of patterns of bubbly, slug, ring, and annular flows obtained in their experiment are fairly agree with those of the present study. Meanwhile, the churn flow pattern was not observed in their experiment. The reason is not understood exactly, therefore, further investigation on this regime is needed in order to clarify the occurrence of churn flow in mini or micro channels.

From the above fact, it is noticed that the inner pipe diameter, the physical properties of the fluids,and the pipe material may affect to the observed flow pattern, since those parameters directly influence to the competition of inertia, viscous, and surface tension forces. From Fig. 4, it is noted that, when the inner pipe diameter is decreased, there are two implications.

First, the transition of slug-annular and annular to churn flow pattern are shifted up or toward higher JL,as confirmed in Figs. 4(a)-4(c). This phenomenon indicates that in the smaller inner pipe diameter the churn flow is formed in higher JL.At high JG, low JL, the flow patterns observed are slug-annular or annular. When JLis increased up to a certain value,the flow pattern changes to be churn flow since the inertia force overcomes the surface tension force. As a result, the liquid film is disrupted. However, in the smaller inner pipe, the inertia force generated by such value of JLis not enough yet to disrupt the liquid film. The consequence of this phenomenon is that to change slug-annular flow and annular flow to churn flow need higher JL.

Second, the transition of slug-annular to annular shifts to the left or towards the lower JG, as shown in Figs. 4(a)-4(d). This means that the annular flows obtained by other studies are formed easier than that in the present study. This is probably caused by the predominance of surface tension in a smaller inner pipe diameter used by them, therefore, the thin liquid film strongly stick on the wall surface and gas-liquid interface becomes straight and establish an annular flow.

However, it also can be emphasized that the discrepancies of the transition lines which caused by the channel diameter change, will be less when they are in the same size category refer to classification from Kandlikar and Grande[19]. They classified the channel to: conventional channel ( dh ＞ 3 m), mini channel (0.003 m ＞ dh ＞ 200 m), micro channel(200 m ＞ dh ＞1 0 m), transition channel (10 m ＞dh ＞ 0.1m), and molecular nano-channel ( dh＜0.1 m). Also, air can be used instead of nitrogen without significant discrepancies as confirmed by Fig.4(b). This may be caused by the fact that in near sea level, air mostly contains nitrogen.

In respect of the flow pattern map, in which surface tension is taken into consideration, Deendarlianto et al.[20]proposed the phase Weber number as a system parameter to elaborate the effect of surface tension and inertia on the flow pattern map. Here,Weber number is defined as the ratio of inertia force to force caused by the surface tension, which defined as follows

where We is Weber number, ρ is density, J is superficial velocity,cD is the inner pipe diameter of the channel, andLσ is surface tension of the liquid.The subscript x describes the phases of fluids.

Fig. 5 (Color online) Comparison of flow transition to the others, based on Weber number

Figure 5 shows the comparison between the available maps in open literature with that of present work. Close observation of the figure indicates that the present map fits well to that of Akbar et al.[21],although in terms of boundary lines, there are some discrepancies among each other. This is due to the difference definition of some flow patterns. As an example, churn flow in the present study include all disrupted or churning flows, while they limit it only the dispersed or frothy ones. Area of bubbly and plug flows in this result correspond well with bubble and plug/slugflow of Akbar et al.[21], even though they do not clearly split those two flow patterns each other,and both flow regimes laid in the surface tension dominated region. Slug-annular flow of this study correlated well with transition zone of Akbar et al.[21],which is the transition from surface tension to inertia dominated region.

In the region of low JG, where the bubbly and plug flow patterns observed, sometimes the flow pattern is not-single regime, but multiple flow which consist of plug and bubbly flows as indicated in Figs.6(a), 6(b). Figure 6(a) shows the appearance probability of bubble, water slug, and gas plug of flow with JG= 0.116 m/s and various JL. This figure shows that the bigger JLimplicates the increase of percentage of water slug, decreasing of void fraction or decreasing of plug number, increasing the percentage of bubbles. From that figure, it is defined that flows with liquid superficial velocities of 1.493 m/s,3.425 m/s are bubbly, since the bubble is dominant,while the rest are plugs.

Fig. 6 (Color online) The percentages of bubbles and plugs emergence

Figure 6(b) denotes the appearance probability of bubble, water slug, and gas plug of flow with JL= 0.149 m/s and various JG. From the figure, it is seen the effect of JGto the appearance of air plugs and bubbles. Water slug and air plug dominates the flow. Meanwhile, number of bubbles fluctuates and tend to increase, even though the number is still far below the plug. Therefore, at JL= 0.149 m/s , all of the flow patterns are plug flow which accompanied by various number of bubbles; even we change the JGfrom low to medium 0.025-0.423 m/s.

Fig. 7 (Color online) Frequency of appearance

2.2 Frequency, velocity, and length of plug/ bubble

Fig. 8 (Color online) Plug and bubble velocities

Figure 7 shows the appearance frequency of plug,bubble and the combination of them. The frequency was obtained by counting them manually from the video images. As shown in Fig. 7(a), for plug flow,under a constant JG, plug frequency increases as the JLincreases. Meanwhile, for bubbly flow Fig. 7(b),the bubble appearance does not increase proportionally with the increase of JL. From Fig. 7(c), it is seen that under constant JG, the plug frequency increases with the increase of JLup to JL=0.879 m/s, in which the bubbles begin to appear.Upon entering the bubble flow, the frequency of bubble/slug fluctuates. This is because the bubbles turned into dispersed bubbles and partly suffered coalescence, causing fluctuation of bubble counting.Figure 8 shows the plug and bubble velocities(UG) obtained from the present experimental study.Here, the velocities were obtained from the cross correlation analysis of two correspond time variation of void fraction data, which is by getting the time lag between the movement of the bubble or plug from one reference point to another one. In Fig. 8(a), those velocities were correlated to the total volumetric flux,J =( JG+ JL). From the figure, it is shown that UGincreases with the increase of J linearly. The data then compared with the a vailable experimental correlations proposed by Nicklin et al.[22], Fukano and Kariyasaki[4]. The empirical equation proposed by Nicklin et al.[22], is as follows

where UGis bubble velocity, J =( JG+ JL) is the total volumetric flux, g is acceleration of gravity, and D is the inner pipe diameter. Since it has low accuracy in small pipe (no relative velocity due to buoyancy force), Fukano and Kariyasaki[4]modified Eq. (2) as follows

where Csis a constant and approximately equal to 0.2.

The comparison between the present data against Eq. (3) is shown in Fig. 8(b). It is seen from that figure,the present work is in good agreement with both Nicklin et al.[22], and Fukano and Kariyasaki[4].Next, Fig. 8(c) illustrates the distribution of constant ofsC obtained in the present work. From the figure,it is noted that Csbecomes zero when the J is close to zero. This means physically that a liquid slug moves as a solid body and the velocity equals to a large bubble. In the region of high J ( J ＞ 1 m/s), the Csare slightly higher than zero. It means that the bubble velocity can slightly higher than liquid.Meanwhile, in low J ( J ＞ 1 m/s) area, some data of Csare negative. This means that the liquid moves faster than gas bubble, because large gas bubbles deflected toward the tube top as claimed by Fukano and Kariyasaki[4]. The average value of Csof this study is 0.233. This value is then used for calculating the UGas remarked in Eq. (3) and plotted against the experimental Us, as shown in Fig. 9.

Fig. 9 (Color online) Experimental Vs predicted UG using proposed Cs ( Cs=0.233)

Besides the aforementioned equation, Fukano and Kariyasaki[4]also recommend the use of simple correlation, as follows

where Ckis a constant, Ck= 1.09 for D=4.9 mm, Ck= 1.17 for D = 2.4 mm , and Ck=1.21 for D =1 mm. Figure 10 shows the comparison between the experimental data obtained from the present study and Eq. (4) withkC equals to 1.21. As shown in Fig. 10, the present data is in good agreement with those of Fukano and Kariyasaki[4].

Fig. 10 (Color online) Bubble and plug velocities of present study compared with the predicted ones using, Eq. (4),with Ck=1.21

Fig. 11 (Color online) Comparison between experimental result of the bubbles/plugs normalized length of the present study to the predicted normalized length of slugs obtained by the correlation by Qian and Lawal[23]

The average of bubble and plug length can be obtained by multiplying the velocity to the time,whereas the time is found from the void fraction data.The void fraction data consist of liquid slug or liquid alone (ε =0) (ε =0) and bubble or plug(ε ≠0) . We have compared the data of bubbles/plugs length of the present work with those obtained by the correlation by Qian and Lawal[23]. It is apparent that there is a difference (Fig. 11). Several assumptions and test conditions may have caused the discrepancy.

Fig. 12 (Color online) Influences of JG , JL to the pressure gradient

First, Qian and Lawal[23]employed a CFD simulation where their test section consisted of a mixing zone with a length of 6 d and a reaction zone with a length of 60 d. For the present study, the observation point was 120 d from the mixer outlet,while our total pipe length was 250 d. This means that our observation is clearly in more downstream and in the fully developed zone compared with their study.Second, Qian and Lawal[23]assume that the slug formation was in the T-junction micro-channel reactor where the cross sectional dimension is the same as the reactor dimension, while in the present study the bubbles/plugs are formed in the convergent zone of the mixer. Third, They assumed that the gas-phase is in form of uniform slugs, while in the present study, it is in form of bubbles, plugs, and sometimes very long plugs. The flow pattern occasionally resembles multiple flow, as shown in Fig. 6.

Thus, the dependency between bubble or plug length and the homogeneous void fraction can be expressed by the non dimensional empirical equation with a mean absolute error (MAE) of 19.55%, as follows

2.3 Pressure gradient

Figure 12 shows the pressure gradient data,( Δ P / ΔZ )exp., obtained from the present experimental study. The data are arranged as a function of JG, JL.Close observation of the figure reveals that both gas and liquid superficial velocities proportionally affect the pressure gradient, whereas the pressure gradient increases with the increase both of JG, JL.

Figure 13 shows the relationship between the two-phase flow Reynolds number and the pressure gradient. The data are also arranged according to correspond flow pattern. Here the two-phase flow Reynolds number is defined as

where G is the total mass flux,TPμ is homogeneous mixture viscosity. From the figure, it is obvious that the pressure gradient increases with the increase of Reynolds number, applicable to all the flow patterns. However, the flow pattern also plays an important role on the pressure gradient. Event, under the same Reynolds numbers, the pressure gradients are different for other flow patterns. For example, at the same Reynolds numbers, the pressure gradient for bubbly is higher than that for slug-annular flow.Nevertheless, at the different Reynolds numbers, the pressure gradients can be same when the flow pattern changes to the other one.

Fig. 13 (Color online) Influences of Re and flow pattern to the pressure gradient

Furthermore, all the pressure gradient data are compared to the theoretical predictions from several established two-phase pressure drop models, which consist of the homogeneous model[24-29]and separated models[30-34,5]as also done by other authors, such as Kawahara et al.[8], Sur and Liu[12], and Qu and Mudawar[35]. The comparison between experimental data and the homogeneous models are presented in Fig.14. Based on Fig. 14, it is noticeable that the homogeneous models fail to predict the experimental data,even if the viscosityHμ calculation was changed according to the proposed experimental correlations.This discrepancy may be caused by the fact that most of the flows are less homogeneous, as also observed by Kawahara et al.[8], Sur and Liu[12]. In addition, it should be noted that the homogeneous model was developed from the assumption that there is no velocity slip between the liquid and gas phases.

However, in reality, the real flow condition in the present study is far from that assumption. It is clearly seen from the video images and the void fraction data shown in Sudarja et al.[17]. This condition is also found in Kawahara et al.[8], Sur and Liu[12].

Fig. 14 (Color online) Predicted vs experimental pressure gradients using homogeneous flow model

On the other hand, the separated flow model assumes that the gas and liquid phases flow separately with each phase traveling with a different velocity.Figure 15 shows the the predicted pressure gradients were calculated using the empirical correlations from Lockhart and Martinelli[30], Bankoff[31], Gronnerud[32],Chisholm[33], Friedel[34]and Mishima and Hibiki[5].

Lockhart and Martinelli[30]correlated the twophase frictional pressure gradient by using the twophase multiplier,and as defined as follows

where f, G, x, D and ρ are friction factor,mass flow rate, quality, channel diameter and density,respectively. Next, the Chisholm and Laird[36]proposed a simple experimental correlation and is widely used to calculate the friction multiplier, and is defined as follows

In Eq. (9), C is Chisholm’s parameter and ranges from 5 to 20, depending on the flow condition, while X is Lockhart-Martinelli’s parameter, which can be calculated by Eq. (10)

Furthermore, Mishima and Hibiki[5]modified the Chisholm’s parameter as follows

It is noticed from Fig. 15 that predicted data using Lockhart-Martinelli and Mishima-Hibiki’s correlations are relatively close to the present experiment data, even though there are some over predicted data points, while the other ones have large discrepancies.It is noted that the available correlations are valid only for the condition where the experiment was conducted[37].

Fig. 15 (Color online) Predicted vs experimental pressure gradients using separated flow model

The Lockhart-Martinelli’s correlation was developed from their experiment data using 1.5-25.8 mm pipes diameter, adiabatic mode, and the working fluids were air-liquids (benzene, kerosene, water, and various oil). Mishima-Hibiki’s model was derived from the study on 1.05-4.08 mm pipes diameter,adiabatic mode, and the working fluids were air-water.Thus both these studies incorporate minichannel pipe dimensions. Meanwhile, the other four correlations employed con- ventional macro-channels, which is the most likely cause of discrepancy. Friedel[34], for example, employed pipes with diameter of larger than 4 mm.

To find the most appropriate C for the present experimental pressure gradient data, Eqs. (7)-(9) were used, and it was best found being C =9.58. This value of C is then used to calculate the two-phase flow multiplier of all experimental data and plotted it against the Lockhart-Martinelli parameter, as shown in Fig. 16(a). Figure 16(b) presents the relationship between experimental against the predicted pressure gradient using the proposed C value of 9.58.

Fig. 16(a) (Color online) Relationship between Martinelli parameter ( X) and two-phase multiplier for proposed Chisholm parameter (C =9.58)

Fig. 16(b) (Color online) Predicted vs experimental pressure gradients using separated flow model with proposed C, C =9.58

3. Conclusion and remarks

An experimental study on the flow pattern and pressure gradient of air-water two-phase flow in a 1.6 mm inner diameter of circular channel was carried out.The superficial gas and liquid velocities were varied in the range of 0.025-66.300 m/s and 0.033-4.935 m/s,respectively. The results are summarized as follows:

(1) Five basic flow patterns were observed,namely, the bubbly flow, plug flow, slug-annular flow,annular flow, and churn flow.The configuration of the transition line between one flow patterns to another in the flow pattern map of the present study is in high similarity with that of other maps previously being reported, such as Triplett et al.[1], and Chung and Kawaji[9], which use mini-channel, but in fairly agreement with those of Sur and Liu[12], which use micro-channel.

(2) Both gas and liquid superficial velocities proportionally affect to the pressure gradient. The pressure gradient increases with the increase of JG,JL. The experimental pressure gradient data were also compared with the some available separated flow model correlations. The available correlations are over predicts to the present experimental data due to the effect of surface tension was not considered in their correlation. Therefore, we proposed a new Chisholm parameter ( C), which is C =9.58, apply well for all flow patterns within ±35% deviation.

Acknowledgement

The authors would like to express their appreciation to Mr. Hasan Imaduddin and Faris Mahendra Putra for their contribution in preparing the experiment rig and collecting data of this study.