MODELING, IDENTIFICATION AND CONTROL, 2005, VOL. 26, NO. 2, 95-109 Riser slugging—a mathematical model and the practical consequences S. L SAGATUN® SPE Keywords: Muliphase, slug, riser ‘This anticle presents @ novel approach to estimate severe riser slug build up time ard ‘consequently the slug period. The slug model is based on a simplified mechanical model. This information har nubrequently been weed to ilunrate the effects of the traditional actions to prevent severe riser slugging. New field data from an offshore floating production platform and large scale experimental data are included. ‘The ‘experimental set up is described in detuil. The estimate on the slug built up time provided by the simplified mosel matches data from de experiments, the full scale data and data in relevant references. Published in SPE Journal of Production and Facilities August 2003, SPE 87355. 1. Introduction ‘This article contains a method for estimating the build up time for riser slugging and consequently the slug period. The developed model is subsequently used to explain the cffcct of traditional meusures uved to prevent riser slugging. The method presented i also useful to identify new methods and explain how and why these methods work, Riser slugging, also called severe slugging is « phenomena which is characterized as a low rate (celative to the dimension of the riser) phenomena. Severe slugging may therefore occur during startup and tail production of a field. Severe riser shigging is also experienced during startup of flowlines after temporary shutdowns of wells. when the wells are routed cone by one to the flowline. ‘This problem is frequently experienced on floating production platforms (FPO) with satellite wells tied to the FPO though marine risers in a lazy S or a similar configurations. Figure | contains a plot of a time scrics of severe riser slugs occurring during startup of one flowline to the Troll C floating production platform operating west of Bergen Norway. Severe riser slugging may cause damage to the first stage separator imernals due to the impulse energy in the liquid slug. Reduced efficieney of the separation in the first stage separator and fluctuations in the rest of the oil-water treatment plant are other undesired effects from large severe riser slugs. Slugging may also cause flaring which is not environmentally benign and also costly. This is particular ‘uu in areas where a COs gas tax regime is enforced. This article uses a simplified mechanistic model to predict severe slug build up ‘The use of simplified mechanistic models to model two phase flow is a well proven technique. Reference Taitel & Barnea (1990) contains a simplified mechanistic two phase model for slug flow. Various correlations are used to calculate the frietion components. A transient simulator is then developed in Taitel & Barnea (1990) based on extended mechanistic model with the moznenturn equations. Neither of these contain explicit expressions for the slug build up time, Reference Scbinidt er al, (1980) also °S.1 Sagatun is with Norsk Hydro Operation and Production, ‘Troll Petroleum Technology in Bergen. Norway. E-mail: svein.varsagatun@hydro.com. "This paper ws originally presented as SPE 87355 in Journal of SPE Production of Facilites, Vol. 19, No. 3, pp. 168-175, August 20041. Sagatun. i Figure 1. Example of severe riser slugs during starup of « fowline to the Troll C FPO. The top plot contains the pressure before the platform chokes as 2 function of time. The bottom plot contains the same time series in the frequency domain, contains a simplified mechanical model based on a set of differential equations. The model uses correlation data to model liquid fallback and hold up. Estimates for the slug build up time is then achieved by integrating the equations with respect to time, thus no explicit expression is used to predict the build up time. In a recent article Tengesdal cet al, (2003) a steady state mechanistic model is developed to predict the effect of riser base gas injection to prevent severe slugging. Since the relerence contains a steady state model no slug build up time estimate is provided. The simplified two phase fiow model in Taitel & Barea (1990) is the starting point for this article. LL. Alternative slug control mechanisms Several methods are used to avoid or reduce the effect of riser based slugging. A common way is to use riser based gas lift. The gas injected at the bottom of the riser increases the superficial gas velocity and reduce the pressure at the riser bottom, thus the flow is forced outside the slug regime. Increasing the pressure in the line by reducing the ‘opening of the platform choke is also a common measure. This may for some well the dowiside that the increased well head pressure reduces Ue production. AM oer method is to install a slug catcher. A slug catcher is @ large drum with volume large ‘enough to contain the liquid and gas from one slug. The gas and liquid is then transferred separately to the first stage separator. This method is rarely used on offshore installations due to the extra cost related to the space and weight of the slug catcher drum. A method which uses the principle of a slug catcher together with active control is the $3 slug suppression system Dill-Quip (2002). This system uses a small drum which acts like mini separator which crudely separates the gas and the liquid, The pressure and the liquid! level is automatically controlled and the gas and liquid flow separately to the first stage separator in a controlled manner. Active suppression of severe riser slugs by automatic control of the platform production valves is a relatively new approach.Riser slugging—a meshematical model and the practical consequences 97 Feedback may be tuken from the choke position and the riser bottom pressure. One of the fist references addressing the issue of automatic control of terrain slugging is Hedne & Linga (1990). Recent developments related to automatic control of slugging in wells Jansen ef al. (1999) and pipelines (Courbot, 1996; Bjune, 2001; Havre et al., 2000 and Storksas er al., 2001) are releted, but not similar phenomena. References Fard & Godhavn (2001); Skofteland & Godhavn (2003) and Godhavn er al. (2003) contains more experimental results from these experiments and a full scale implementation on the Heidrun tension leg platform located in the North Sea. These references also contain experimental results from different control structures including cascade type controllers with feedback from other states ie. feedback from riser top pressure, manifold pressure and density measured at the riser top. ‘The mathematical model developed in this article is useful to explain why and how the traditional methods for slug suppression work. Included in this article are both field ddata and results from large scale experiments. The next section will analyze the severe phenomena with a pure mechanistic approach. ‘The final result from this section is an estimate of the slug build up time and consequently the slug period. Section 3 contains the analysis of why and how riser based gas lift and adjustment of the platform based production choke works, while section 4 and 5 present the experimental setup and the experimental results. Section 6 and 7 contain a field description and field data 2. A mechanistic slug model ‘The aim of this section is to establish » slytient expression fir the severe shig period. Reference aitel & Barnes (1990) is used as a starting point for this analysis. A pure mechanistic dynamic model of the slug in the pipe will be derived using the following assumptions Schmidt et al. (1980) AL ‘The input flux rates of gas and liquid are constant, AZ The pressure Lefure the valve is consta A3 The effect of gas bubbles in the slugs are neglected. ‘A4 The liquid hold up in the pipe is constant AS The process is isothermal. ‘The starting point of this analysis is when the shug front is at the valve, see Figure 2. The liquid mass in the pipe is Where is is the superficial velocity of the liquid defined as, % ws= A where v; is liquid volumetric flow rate, a is the void fraction, py is the density of the liquid phase, A is the pipe cross sectional area and zx; and / are defined in Figure 2. Henee, rit, becomes: rie = Apes Q Expressing the liquid mass as a function of x and z yields: m= prAlx +2) + 1~ at) eB)98 S.f. Sagatun Figure 2. Left plot: The riser-pipeline during the slug formation. The shaded area represents liquid ‘and the white area is gas. Right plo: ‘The riscr-pipeline at «= 7) ‘The corresponding equation for the gas in the pipe becomes: me = poVer+ [arse (4) o My ‘ me= Vi Wesapor Fe rer | Aueseett 6 where the initial pipe pressure per and initial yas volume Voy are found from the following expressions: pnt pusle ed) o Vea= alla) am ‘The subscripts GO represents gas velocity and density at standard conditions. ze(x) is the clevation in the pipe of the tal of the initial slug. zr(4,) may be approximated to x; sin B for a relatively straight pipe with negative inclination fl. Me is the molecular weight, R is the universal gas constant and T represents the temperature. The pressure p, is the top riser pressure, Hence, ric, becomes: tig = Aucspen C3) Combining (1) and (A) recat in the fallewing equation: 1 raat tust— (2d) o whereas, combining (4-6) ids, for the gas Taitel & Barnes (1990): Px + (—xsin f)| ~al—x)| P+ @—x:sin f| = espe at af te »| a [2 re np] Mapigtorrot (10) Inserting (9) into (10) and leting 2 =A results in the following polynomial in time 1. at tnte=0 apRiser slugging—a mathematical model and the practical consequences. 99 where snp on“ sinp De panafinp(@e-o+20-1}- 1 : eta fang (tanta) aman 2 +5] : fue Saving (1) ith eset tne eng eh es in neti fe sao tne fy Keung ea we 8 pre (1) a= tsa 13) b= msba—ucsba results in the following analytical expression for 7; T= [mish + uessba® Vetusba — web assuming 7; is positive and real. ‘The slug period is the sum of the liquid build up time 7, and the time necessary to transport the slug into the separator denoted 7>. According to Schmit et al. (1980) To is experimentally determined to be in the range 0.57) ~ Ti, thus using 7; as an estimate of the slug period is conservative for slug control design. Notice that (14) can not be used as a slug criteria. Equation (14) can only be used to estimate the slug build up time when the flow is in severe slugging regime 2.1. Verification of the estimate of the slug period ‘The liquid build up time prediction given by (14) has been tested on two different ‘cases reported in the literature, full scale date logged on the Troll C floating production Uuit and on our experimental set up. the experimental setup and the full scale case reported in Schmidt et al. (1980) are used as external references. Table 1 presents the results from the tests. The estimate is consistently conservative in the sense that the liquid build up time is estimated to be faster than the actual values. Effects like frietion and the assumption that the liquid hold up is constant over time will both increase the liguid build up time. ‘Table |. Comparison between estimated and measured Ty Reference actual 7) in s predicted 7 in s (14) Schmidt of of, experimental stip. 50 38 Schmidt etal, field data ~ 600 24 Tiller experimental setup 206 nz “Troll C field data ~ 900-1000 867 ‘Table 1: Verification of the prediction of 7). The following paraineters are used in the calculation: the superficial liquid velocity us the superficial gas velocity at standard condition uc, the inclination of the inflow pipe /}. the void fraction 2, acceleration of gravity g, liquid and gas densities p and eas the gas property fraction Kr Me the pressure atthe riser top p, the riser height A, the pipe length / and initial iquid positions x ant100 SL Sagatun 3. Traditional actions to suppress severe riser sluggi ‘The traditional actions usually employed offshore to suppress severe riser slugging are either by adjusting the choke opening of the platform production choke or using riser based gas lift. The latter action increases the superficial gas velocity uesy and reduce the pressure at the riser base. An increase of the platform choke results in a lower riser and production line pressure. This will result in an increased liquid rate from wells with little for no differential pressure over the subsca chokes. A decrease in the platform choke results in a higher riser pressure, thus p, increases. If possible, te riser slugging can also be suppressed by routing more liquid in to the production pipe connected 10 the riser. ill inerease the superficial liquid velocity, thus moving the flow away from the severe slug regime. 3.1. Adjustment of the platform choke Choking action on the platform choke results « higher line pressure and consequently an increase in the p,. Using (12 and 13), we notice that an inerease in the p, pressure results in an increase of ¢ and a decrease of the by parameter. We observe by inspection of (14) that both these changes reduces the value of the minimum positive real T,, thus, the slug period reduces and likewise the amount of liquid in the slug, Hence, increased Jine pressure improves the line conditions with respect to severe riser slugging. The increased riser and production line pressure does not necessarily affect the production. This in i accordance with the findings in Schmidt e¢ uf, (1980). Puuxuction wells with high gas liquid ratio have usually a well head pressure much higher than the required line pressure, thus there is a large differential pressure over the subsea production choke. 3.2. Riser based gas lift Adding riser basod gas lift increases the superficial gas velocity denoted iso. In the same manner as above we can analyze this effect. Using (12 and 13), we notice that an increase in the superficial gas velocity tug results in an reduction of the absolute value of the term (usb — eau) in (14). Thus the minimum positive value of 7; is reduced. Addition of gas at the riser base will also increase the pressure p, at the riser top, which also contributes to the reduction of slug periods, see the paragraph above. ‘The effect of riser based gas lift is also investigated in Tengesdal et al. (2003). QA Increasing the liquid flow rate Routing more wells or increasing the production rate from existing wells are also beneficial when severe riser slugging is experienced. Increasing the liquid rate by ‘opening the platform choke is also possible for some field. An increase of the platform choke opening will reduce the riser and line pressure. This will result in an increase in the liquid rate for wells with little or no differential pressure over the subsea production choke. These wells are typically wells with low GOR and high produetivity index. A decrease in the well head pressure with less than a bar may result in substantial increase in the liquid rate with a negligible increase in the superficial gas rates. If the well head pressume is lowered substantial the increase in the superficial gas rate (due to expansion of free gas) may be significant. The increase in the gas rate is beneficial with respect toRiser slugeing—a mathematical model and the practical consequences 101 severe slugging since the superficial gas rate will increase. An increase of the liquid flow resulis in an increases of the superticial Uquid veloc. Equation (14) reveals (not trivially) that a 2 aus ‘Thus, the volume of each slug and consequently the consequences of each slug will be reduced. The effect on the liquid rate is increased with increased PI. A typical Troll C ‘well has a PI larger > 2000 Sm'Vdlbar, thus, opening the platform choke is very efficient «with respect to reducing the riser slugs on production lines dominated by wells with low GOR. 4. The experimental set ‘The experiments were carried out at SINTEF Multiphase laboratory in Trondheim, Norway. The experimental setup and the operations of the facility are carried out according to internal safety and quality procedures according to Norwegian rules and regulations. The test facility is a 231 m, 3 in closed loop with a vertical riser for circulation of oil and gas. ‘The SF: gas and the Exxsol D80 test oil is used in the experiments. The first 100 m have a — 0.1 deg declination, then 180 deg horizontal Uctum. The diameter of the U-tum is 3.5 m, corresponding to a length of 70 diameters. ‘fice the U-turn, the pipe is declined — 0.7 dag for about 100 m, and finally = 15 m vertical riser The riser ends in a double bend, where the flow is directed downwards into. ‘an 8 in vertical drop-leg. The vertical 8 inch drop leg ends in a gas-liquid separator, ‘where the gas is drawn into a de-mister to remove droplets and then into the compressor. ‘The oil is drained to the horizontal separator and recycled through the oil pump. The separated phases are fed to the singlephase velocity measurement stations and routed through the correct flow meters by manually operated valves. At the inlet section the gas is mixed with the oil through a 45 deg downward inclined pipe. The oil and gas then passes through a7 m long flexible (eubber) pipe section, and on to the = 011 deg section. The scientific instrumentation is located along the entire loop, but mostly along, the last 100 meter section and in the riser. Here the flow regime, pressure gradient, absolute pressures, and hold-up are measured. This pipe section is declined approxi- mately — 0.7 deg downwards. The hold-up is measured by means of seven single-energy natrow-beam gamma densitometers distributed along the pipe, with the two Tast ones in the riser. Mainly, acid proof 316L steel pipes are used, but in this project, two steel sections were replaced with PVC pipes for visualization purposes. The design pressure is 10 bara, and in the current experiments the nominal system preccure varied from 2.2 to 3.0 bara. ‘The loop is equipped with basie process and scientific measurement instrumentation. ‘The instrumentation consists of single-phase flow rate meters, temperature sensors, differential and gauge pressure sensors, and gamma densitometers. The flow rate meters fre mounted on single-phase flow lines upstream the mixing point, and the signals from the oi and water rate meters are converted to velocity in a 69 mt pipe to direetly measure the superficial velocity in the test section, All superficial velocities are calculated using the absolute loop pressure, which is defined at the 165 m location. The temperature sensors, pressure cells and gamma densitometers are distributed along the test section of the loop. Vortex meters are used to measure the gas volumetric flow rate,102 SL Sagatun 0.9—is followed with r<0.25—after a blow out ‘The pulsating flow regime is established when a semi steady state situatio Will Feas ~ Finan > 0.25—. A bubble flow is defined as a situation when h vari than +0.125—. All experimental results are taken from Fard & Godhavn (2001). lt be observed from the owmaps that the experiments confirm the analysis carried out in ‘2 previous chapter. Thus, increasing the riser pressure is efficient with respect to slug reduction, Figure 5 and Figure 6 contain the results from one experiment. The superficial ‘208 velocity eso and liquid velocity ues are kept constant to 0.18 and 0.22.ms-! during, the experiment. Figure 5 is a plot of the riser top and riser bottom pressure. Notice that the flow is in the severe slugging regime before the choke is actuated. The measured slug period is 210s with a measured build up time of. 195s. The estimated build up time using (14) is 112. Figure 6 plots the corresponding choke valve opening and the parameter h. It is easily observed by looking on the parameter fr that the riser is in severe slugging. before the valve is actuate. Remark, one should notice that the experiments were carried out with the same liguid ‘and gas rate before and after the pressure was inereased. This is an idealized situation which may be difficult to implement for some fields. An increase in the riser pressure will consequently inerease the production Tine pressure and result in a subsequent reduction in the production. This will be the case for production lines with wells with litle differential pressure over the subsea production choke. v,, ms") Uline") Left plot the flow mp for a 100% open valve. Right plot: the flow mp for a valve with 159% stroke104 Sf Sagatun 2 (ba70) 6 2s 3% 36 o Time {rin Figure 3. Top plot he pressure at me top of ine riser upstream me choke. Bottom plot: the pressure atthe bottom of the ris c up Tame fin} igure 6. Top plet: the non-dimensional parameter plotet versus time. Bottom plot: the valve ‘opening.Riser slugging—a mathematical model and the practical consequences 109 6. Field data ‘The field data are taken from ‘roll C a floating production platform (FPO) operating in the North Sea west of Bergen Norway. The Troll C is operating on approximately 350 m water depth. The field lay out consist of a number of well clusters (four wells on ‘each cluster) tied to the FPO with subsea pipelines and flexible marine risers in a lazy S riser configuration, see Figure 7. The pipelines and the marine risers all have a ID of 10 in The length of the subsea pipelines are between 2000 to 10000 m and the marine riser are approximately 720 m. The wells are controlled with a subsea proxiuction choke ‘on the well head and a platform production choke on the top of the riser. The riser water ‘cut (WC) is in the range between 10-70% and the riser gas oil ratio (GOR) is in the range of 58-250—. The density of the oil and gas are approximately 893 kgm and (0.861 hyn at stundatd wuunlitions, acts siner Hay dhe capability Of riser xan injection at the riser bottom. Typical pressure at the riser top is between 15 and 35 bara, Measure- ‘ments available are pressure and temperature before and after the topside production ‘choke and the same measurements before and after the subsea chokes. The gas oil and. ‘water rates are estimated using an online state estimator using the measurements and test separator data as input. The uncertainties of the volumetric rates of the three phaes are for the cases presented in this article assessed to be within + 10%. The volumetric rates, of the riser gas lift is consistently under predicted and the error of the gas lift rates are estimated from comparisons on the test separator 10 be ~ 10-50%. There ure m0 measurements at the bottom of the risers, a a Figure 7. The Troll C riser configuration,106 SL Sagann ‘Two cases are recorded in this article. The first case demonstrate the effect of increasing the riser gas lift rate on the slug frequency. The other case shows the effect ‘of adding, more liquid to the riser by increasing the opening of the top side choke, 6.1. Case I—Riser based gas lift Figure 8 illustrates nicely the effect of increasing the superficial gas velocity in the riser. The riser gas lift rate is increased from 0 to 3000 Sm'k-', The peried of the slugs is reduced from approximately 1450 s to 200s. The variance of the pressure on the riser top is reduced from 4.2 to 1.4 bur” or a reduction of 67%. The variance of the pressure may be used as a metric f the energy in the slugs and may be useful to quantify the f wear and tear of process equipment from fatiguc cased by the slugs. It is also beneficial for the separator efficiency to have an even liquid Flow into the separator to reduce the liquid-gas shear forces and fluctuations of the separator level. ‘The period of the slug is a measure on the slug size, thus it is beneficial to reduce the slug period as much as possible. The total gas rate before the rise gas lift was added was estimated to approximately 6500 Smith ' and the liquid rate was approximately 80 Sim'h before the riser gas was added. A significant increase of the liquid production was observed after the gas lift was added. ‘The increase of the liquid rate is mostly due to that the pressure in the production line is reduced. Consequently, 2 new three branched well with little differential pressure over the subsea choke produced significantly more to this production fine, ‘The water cut was approximately 52%. ‘The predicted slug build up time (using (14)) before the increase of liquid rate was 877 s while an estimated build up time from looking on a detail from Figure 8 see Figure 9 is approximately 1100s. Figure 8. Top plot: the volumetsic rate [Sm] ofthe riser gase Tit. Bottom plot: the pressure before the platform choke on the top of the riser.Riser slugging—a mathematical modet and the practical consequences 107 T=1100 [s] 500 os sso Figure 9, Detail from Fig. # estimating the og build up time 7 we 10, The effect of increase liquid rate (inercased superficial iguid velocity urs) on severe slugging. Top plot: the volumetcic liquid rate before and after the subsea choke of one well is increased, Bottom plot the pressure upstream the top side choke on the top of the riser.108 ST Sagatun 62. Case 2—Inereasing the liquid flow rate Figure 10 shows the efllect of increasing the superficial liquid velocity on severe slugging, The riser gas lift rate and the top side choke position are both kept constant. ‘The liquid rate is increased by opening a subsea production choke. ‘The pronounced liquid slugs with a period of 1200 s are reduced both in pressure magnitude and period. “The variance of the pressure on the riser top is reduced from 7.7 to 1.9 bar or a reduction of 75%. The total gas rate was estimated to approximately 3900 Sm'A' and the water cut was 22%. The liquid rate went from 81 Snrh-' to 107 Sah '. The predicted slug build up time (using (14)) before the inerease of liquid rate was 867 s ‘while an estimated build up time by looking on Figure 10 is between 900 0 1000s. 7. Conclusion This article presents a novel approach to estimate slug build up time and conse- quently a slug period. This estimate has been successfully compared with several sets of field data and a lange scale experiment. The model is used to explain why and how the traditional means of reducing severe slugging works. The inerease in liquid rate is much more efficient than increasing the volumetric riser gas lift rate. The former will automatically follow the latter on the Troll field where sn reduction in line automatically leads to a higher liquid production. In terms of 14. an | ah [gh urs! Buesn Experimental results as well as observation from a producing field is used to illustrate the model. The method presented is useful to identify new methods and explain how and why methods for severe slug suppression work. The estimated slug period may also be useful in initial settings of control gains an time constants for automatic slug control ystems ‘Notice that the method in this article does not consist a slug criteria. The method will ‘only work when the flow is in the severe slugging regime. Acknowledgements ‘The author is gratefull to Hydro and the ‘roll field licence partners for allowing this article to be published. ‘The author will like to thank Dr. M. Fard (Hydro), Dr. .M. Godhavn (Statoil) and Dr. J.R. Sagli (Statoil) with their cooperation with the large scale experiments, References BrUNE, B, (2001). Stabiliza Oil-Gas Magazine, (1). ‘CouKsor, A. (1996). Prevention of severe slugging in the dunbar 16" multiphase pipeline. in O7C (Houston, TX), May 1996. Duus-Quir, (2002), Brochure: The s3 slug suppression system, tech. rep.. Dril-Quip Ine, Houston a of slugging flowlines with active flowline control, Scandinavian 1X, Pago, M. & Gooniavn, J. 2001). Slug control experiments at tiller 2002—ah report k-0080792, tech. rep., Norsk Hydro Exploration and Production—Research Center Rergen, NS020 Bergen, Norway. Govuavy, J., Fan, M. & Fucts, P. (2003). New slug control strategies —tuning rules and experimental results, Joumal of Process Contsol August 2003, No. 5, Vel. 15, pp. 597-557.Riser slugging—a mathematical model and the practical consequences 109 Havke, K., SroxNEs, K. & StRAY, H, (2000), Taming slug flow in pipelines, ARB Review, 4/2000, Pp. 55-63, Hone, P. & Lincs, H. (1990). Suppression of terrain slugging with automatic and manual riser choking, in ASME Winter Annual Meeting printed in Advances in Gas-Liguid Flow (Dallas, TX), ASME, Nov 1990. Jansen, B., DALSMO, M.,. NOKLERERG, L., HAVRE, K., KRISTIANSEN, V. & Lewrayen, P, (1999). ‘Automatic control of unstable gas wells, SPE, SPE 56832, Oct 1999. Sonapr. Z.. But. J. & Daue-Beocs, H. (1980). Experimental study of severe slugging in a two-phase-flow pipeline-riser system—SPE. SPE Journal. 20. Skortetann, G. & Gopnavn, J. (2003). Suppression of slugs in multiphase flow ines by active use of topside choke”Field experience and experimental results, in Proceedings of ‘Multiphase'03 (San Remo, Italy), p. na, June 2003. SrowKAAS, E, SKOGESTAD, 8. & ALSTAD, V. (2001). Stabilizing of desired flow regimes in pipelines, in AIChe Annual Meeting, (Reno, NV). Nov 2001 TaTTEL, Y. & GARNEA, . (1990). Simplified transient simulation of two phase now using ‘quasi-equilibrium momentum balances. in. J. Multiphase Flow. 2303), Tare, Y. & BARNEA, D. (1990). Two-phase slug flow, Advances in heat transfer, 20. Tuncaspat, J, Towson, L. & Sarica, C. (2003). A design approach for self-iftirg method for climinate severe slugging in offshore production systems—spe 84227, in SPE ann. tech, confr. and exehib., (Denver, Colorado).