CONDENSATE AT "T'S" - CONCLUSION EXPERIMENTS VERIFY PREDICTIONS OF CONDENSATE MOVEMENTS

Jan M. H. Fortuin, Peter J. Hamersma, Jaap Hart
University of Amsterdam
Amsterdam

Harry J. Smit, W. P. Baan
NV Nederlandse Gasunie
Groningen, The Netherlands

Experiments carried out in laboratories at the University of Amsterdam and NV Nederlandse Gasunie have verified a model for predicting the route preference of condensate at T junctions in gas-transportation pipelines.

The double-stream model, whose development was the subject of Part 1 of this series (OGJ, Jan. 21, p. 37), can be applied to sharp-edged and radiused, regular, and reduced T and Y junctions.

The model explains condensate behavior at a junction. From this information, precautions can be taken to install gas-liquid separators at those pipeline sections where condensate accumulates.

Such accumulations may cause problems in gas-fired appliances or damage delicate instruments in gas control and compressor stations.

Verification of the model was derived from experimental results from literature and from experiments carried out at the University of Amsterdam with air-water flow in a horizontal tube with a horizontal regular T (ID: 0.051/0.051 m; 2/2 in.).

Moreover, a comparison has been made between results obtained with the double-stream model and those from high-pressure experiments carried out at the research department of NV Nederlandse Gasunie with natural-gas condensate in a horizontal reduced T (ID: 0.075/0.050 m; 3/2 in.).

The agreement between measured and calculated results was good.

FLOW LOOP, EXPERIMENTS

A schematic representation of the air-water flow loop is shown in Fig. 1.

The inlet and run consisted of a straight, horizontal copper tube (No. 12 on the flowloop, Fig. 1) with an internal diameter of 51 mm and a length of about 17 m.

The branch, with the same diameter as the run, was also made of copper and had a length of about 5 m. Two different T junctions (13) were used, a sharp-edged regular T junction (R/D1 = 0, D1 = D2 = D3; Fig. 1 of Part 1) made of Perspex and a radiused regular T junction (R/D1 = 0.1, D1 = D2 = D3; Fig. 1 of Part 1) made of glass.

This junction was used because in practice it was recognized that forged T fittings are not sharp-edged but radiused.

The deviation from the horizontal plane of the tubing was less than 0.050.

A total number of 12 pressure taps with internal diameters of 1 mm were located at several places in the pipeline system at the top of the tube to measure the pressure gradient over the gas phase with a precision micro manometer having an accuracy of 0.4-4.0 Pa.

An essential part of the test rig was provided by the water-ring compressor (1). This type of compressor allows control of the air temperature and permits a relative humidity of the air of more than 90%. This is essential in investigations of gas-liquid flow with low liquid-holdup values because a significant evaporation of the liquid in the pipe may occur if the relative humidity is insufficient.

The air flow rate was controlled by one of the two rotameters (3). Before entering the test section, the air passed a sieve plate (4) to eliminate effects of bends.

Water was supplied from a vessel (5) and placed in a raised position so that the liquid was injected into the system by means of gravity. This was done to provide very stable liquid flow rates during the measurements.

The liquid flow rate was controlled by one of the five rotameters (6), of which only three are drawn in Fig. 1, together covering a range of superficial liquid velocities of 5.10 5 < UL/(m/sec) < 5.1 0-2 . The liquid was injected through an injection point (7) at the bottom of the pipe.

At the end of the branch a gas-liquid separator (8) was mounted and provided with a gas rotameter (9) to measure the gas branch mass-intake fraction. The flow rate of the liquid after the T was metered at the end of the run (10) by weighing a timed efflux of water.

At the end of the run the air was released to atmosphere (11).

Valves on the two outlet tubes (14, 15) were used to control the division of the gas flow in the junction so that approximately 10-20 lambda G values were set and lambda G or lambda L varied from 0 to 1.

Measurements were carried out under atmospheric pressure conditions (P = 0.1 MPa) at superficial air velocities ranging from about 7 to 20 m/sec, superficial water velocities ranging from 0.1 10-3 to 30 . 10-3 m/sec, and temperatures in the range of 15 C. to 25 C.

EXPERIMENTAL VERIFICATION

In subsequent sections results calculated with the double-stream model are compared with experimental results obtained at the University of Amsterdam. The double-stream model was also tested against experiments carried out under operational conditions (natural gas-condensate flow at a pressure of 30 bar) at NV Nederlandse Gasunie.

Fig. 2 reveals the strong interdependence between the branch liquid-mass intake fraction (lambda L) and the junction pressure difference DELTA P2 3.

A comparison has been made between experimentally determined lambda L values as a function of experimentally determined DELTA P2 3 values (markers) and calculated lambda L values as a function of calculated DELTA P2 3 values (solid curves).

The calculated lambda L values for the regular T junctions were determined after having solved the set of four extended Bernoulli equations without elimination of the junction pressure difference. The equations of Gardel6 were used for the determination of k1 2 and k1 3 (and therefore of lambda 0, Equation 5). (References and equations are in Part 1.)

From Fig. 2 it can be found that for the smallest inlet liquid-mass flow rate (0.6 g/sec), a change of DELTA P2 3 only from -10 Pa to +10 Pa results in a "flip-flop" change of lambda L from 0 (all of the liquid flowing into the run) to 1 (all of the liquid flowing into the branch).

In view of the present analysis, it is obvious that the junction pressure difference (DELTA P2 3) is a driving force for liquid route preference at junctions and that models for the two-phase flow split in which this is not incorporated are incomplete.

In Fig. 3, lambda L values have been plotted as a function of lambda G values. Solid lines were obtained with the doublestream model for regular junctions (Equation 8) and markers refer to experimental results measured at the University of Amsterdam on the regular sharp-edged Perspex T junction.

For the lines A, B, and C, the liquid-film flow in the inlet is laminar (ReL < 2000 and BETA L = 1.54); for the line D the liquid film flow in the inlet is turbulent (ReL 2000 and BETA L = i).

It can be seen that good agreement is found between experimentally determined and calculated values concerning the liquid route preference. Note that for the smallest inlet liquid-mass flow rate (Line A), an increase of lambda G only from 0.05 to 0.15 results in a sudden increase of lambda L from 0 to 1.

This flip-flop effect particularly occurs in gas-transportation pipelines where the average kinetic energy per unit volume of the gas is much larger than that of the condensate which is normally present in small amounts (EL < < 0.01).

A very important experimental and theoretical finding is that equal flow split of gas and liquid (lambda L = lambda G) occurs if K = 1.0; that is, when in the inlet the average kinetic energies per unit of volume of gas in the gas phase and liquid in the liquid phase are equal (Line C, Fig. 3). It seems clear that this situation is exceptional.

The experiments carried out with the radiused glass T junction suggested that the liquid route preference was only slightly influenced compared with the sharp-edged junction.

Both from the experiments and from the model, it was found that the branch liquid-mass intake fraction (lambda L) was affected in such a manner that for the radiused junction a slight shift towards the line equal flow Split (lambda L - lambda G) was obtained.

In Fig. 4, a comparison has been made between results obtained with the present model and experimental results from NV Nederlandse Gasunie (Oranje1). These experiments refer to natural gas-condensate-flow in a horizontal dividing reduced T junction (D1 = D2 = 0.075 m, D3 = 0.050 m) at a system pressure of 30 bar.

As can be seen, the agreement between the calculated and experimental results is reasonable, taking into account that the data were not well tabulated and that the physical properties of the gas and the condensate had to be estimated.

IMPLICATIONS

An illustrative calculated example of severe maldistribution of liquid that might occur in a manifold consisting of a horizontal main tube of diameter DR and four horizontal side tubes of equal diameter DB has been given in Fig. 5.

Fig. 5a reveals that, if the four side tubes receive 15 wt % each of the inlet mass flow rate of the gas, only 9.4 wt % of the inlet mass flow rate of the liquid is extracted into the first branch.

The other three side tubes receive dry gas and the rest of the liquid (91 wt %) flows straight on into the main tube.

If, however, the third side tube receives 20 wt % of the inlet gas mass flow rate instead of 15 wt %, the situation is completely different (Fig. 5b): According to the model, the first side tube receives 9.4 wt % of the inlet liquid mass flow rate, the second tube "dry" gas, the third tube 60 wt % of the liquid, and the fourth tube "dry" gas again.

The remaining liquid flows straight on into the main tube.

Thus, for two-phase manifold flow only small variations of the gas branch intake may lead to large variations of the liquid branch mass intake. This result shows that liquid route preference is a very delicate phenomenon and that it may have a large impact on many practical situations.

SUMMARY

The work reported in this two-part series has shown the remarkable phenomenon of liquid route preference in T junctions following in a straightforward manner from the extended Bernoulli equation, applied both to the gas and liquid streams in the junction.

The procedure for the calculation of lambda L values with the double-stream model requires no complex numerical methods. Therefore, the model is easily accessible for design and operation engineers.

Implementation of the model into a computer might be helpful to reveal the effects of flow parameters or junction geometry on the liquid route preference. In that way more insight in the phenomenon can be gained.

A calculation procedure is shown in the accompanying box. For more detailed information see Fortuin.

Although in theory the problem of liquid route preference can be solved, in practice it is very sensitive to subtle changes in flow. Therefore, it is necessary to extend the present work to gain more insight into the effect of small differences of physical quantities and of junction geometry and slope of the main pipe and branch on the two-phase flow split.

Copyright 1991 Oil & Gas Journal. All Rights Reserved.