The impact of water slugs on wet steam flowmeters

98

I. Owen et al - The impact of water slugs on wet steam flowmeters

The impact of water slugs on wet steam flowmeters I. O W E N , I. B. HUSSEIN and A. M . A M I N I

The damage that can be inflicted on a flowmeter that has been installed into a steam distribution system without having a water separator upstream of it to offer protection is considered. The situation examined is one where a body of liquid accumulates behind a closed valve in a piping system carrying a wet or condensing gas, then the valve is suddenly opened to expose the slug of liquid to the high pressure gas. The motion of the slug under these circumstances has been analysed and compared with slug velocities measured in a 50 mm pipe. The pressures generated when slugs impact upon three different sizes of orifice plate have been measured and correlated in terms of the system variables. These pressures are seen to be considerable; for example a 2.16 m slug propelled by a source pressure of 10 bar (1 MPa) will generate a pressure of about 300 bar when impacting on an orifice plate with an area ratio of 0.250. When the orifice plate is preceded by a water separator, the impact pressures fall to about one tenth of the corresponding case with no separator. Keywords: wet steam flowmeters, water slugs, water separator, orifice plate

Introduction The measurement of flow in a wet steam distribution system, or in any other wet gas installation, has potential difficulties because of the presence of condensate. From the viewpoint of accuracy, the condensate may cause problems because it can interfere with the functioning of the instrument and also because its own flow rate and properties have to be included in the total flow measurement. These problems are widely recognized and are usually minimized by ensuring the pipework is well drained whilst the measured flow rate is corrected using estimates of the wetness fraction. Another, and more dramatic consequence of the condensate being present, is the damage inflicted by a slug of liquid as it impacts onto a flowmeter. It is this aspect of wet steam metering that is being addressed in this paper. It is good practice to install a steam separator upstream of a flowmeter, not only for protection, but also to reduce the water fraction of the metered fluid and hence to increase accuracy. There are many instances, however, where this practice is not followed and it is common for steam flowmeters to show evidence of water impaction. Consider the worst situation where condensate has collected behind a closed valve in a steam distribution system. If the valve is opened rapidly, the collected water will be propelled along the pipeline as an isolated slug, which has the potential to cause damage to a flowmeter (or any

The authors are in the Department of Mechanical Engineering, University of Liverpool, PO Box 147, LiverpoolL69 3BX, UK 0955-5986/91/020098-07

other obstruction). It is recognized that steam valves are normally opened slowly in practice, but it is useful to consider this worst case to determine the possible extent of the problem. In this paper the dynamics of an accelerating slug have been analysed to show the velocities that can be attained; these have been compared with measured slug velocities. To illustrate the magnitude of forces that can be produced, measurements of the pressure generated as the slug impacts upon an orifice plate are presented. The level of protection offered by water separators is demonstrated by comparing these impact pressures with the corresponding pressures measured at an orifice plate that is preceded by a separator.

Isolated slug flow Fluid transients in pipelines can generate significant pressure surges and the resulting forces are capable of causing considerable damage. Water hammer is probably the best known and most widely reported phenomenon in this respect (see, for example, Sharp1). The case under consideration in this paper is similar to that of two-phase slug flow, but is different in that it concerns a single slug of liquid being propelled along a pipeline by high pressure gas. Much of the work published on two-phase flow is related to the nuclear power and the oil recovery industries where the flow is continuous and, over a reasonable timescale, is steady. Two-phase slug flow has been investigated in detail by, amongst others, Dukler with various coauthors: for example, Dukler & Hubbard 2, Taitel & Dukler 3 and Dukler et al. 4 The force generated by a

© 1991 Butterworth-Heinemann Ltd

Flow Meas. Instrum. Vol 2 April 1991

99

transient slug as it emerged from the end of a pipeline to impact upon a target was measured by Sakaguchi et

al. 5

slug becomes

pl~7-=

Pea 1

2

ZJ

Theoretical analysis of slug dynamics Consider an idealized slug in a pipe, as shown in Figure 1. Behind the slug is a high pressure expanding gas, which is forcing the slug along the pipe, thereby Compressing the low pressure gas in front. Also resisting the motion of the slug will be friction between it and the pipe wall. It is appreciated that as the slug moves through the pipe, it will shed liquid behind it. However, the area of the slug upon which the gas is acting will still be that of the pipe cross-section, and the total mass of liquid being accelerated is that of the whole slug. Some of these practical aspects will be discussed later in the paper in the light of experimental data. Returning to Figure 1, the pressure, P,, of the expansion wave upstream of the slug is given by

Pe

[1

Ue] 2Y/(Y-1)

y-1

Pea

2

~e

(1)

where Pea refers to the stagnation pressure upstream of the slug, 7 the ratio of specific heats, U~ the speed of the pressure wave and ae the speed of sound upstream of the slug. Downstream of the slug the pressure, Pc, of the compression wave is given by

Pc

[1 +

Pco

Uc] 2r/(r-1)

7-1 2

aTc

(2)

where Pco refers to the downstream stagnation pressure, Uc the speed of the compression wave and ac the downstream speed of sound. The wall friction force, Fw, is given in terms of the wall shear stress, r, as

Fw = rhrD

(3)

where / is the length of the slug and D the pipe diameter. The wall shear stress is in turn related to the slug velocity, Us, by 1 2 r = TpUsG

(4)

where p is the liquid density and Cf is the skin friction coefficient calculated using the Blasius equation: G = 0.079(Re)-°25

(5)

where the Reynolds number is based on the slug velocity and the pipe diameter. The expansion wave, the slug and the compression wave will all be moving with the same velocity, say U, and therefore the equation of motion for the

2pU2G I} D

Equation (6) can be integrated numerically to obtain a graphical relation of velocity with time. By curve-fitting this relationship and then integrating it, distance can be found as function of time. Consider now the impact of the slug with an orifice plate, as shown in Figure 2. If the slug was assumed to behave as a solid and to remain intact during the impact, the impact pressure would be the same as that calculated for water hammer. If, on the other hand it is assumed that the flow passes easily through the orifice, as it would through an orifice plate flowmeter, then the pressure drop across the orifice might be expected to follow the typical orifice plate flowmeter equation. In practice the situation is somewhere between these two. The slug does not remain as a solid on impact, but extrudes through the orifice so that not all the inlet momentum flux is given up. Neither does the slug pass easily through the orifice, but slows down significantly on impact so that the actual velocity with which it approaches the orifice after the initial impulse, and the velocity with which it then flows through the orifice, are indeterminate. In this case, therefore, it is not possible to produce an analytical expression for the impact pressure or force and if a practical expression is desired, then an empirical approach using dimensional analysis must be used. Collecting the variables shown in Figure 2, the following relationship suggests itself F

pU2A

-function { ( ~ o ) ( / ) }

r

Experimental investigation The situation being considered is one where a steam valve has been opened rapidly. Upstream of the valve is a pressure source of infinite volume; downstream is an isolated slug of water being propelled into an empty pipe. How this was modelled experimentally is shown in Figure 3. Air with pressures up to 10 bar was used as the Slug length, l "

"

,reom

Downstream pressure, Pc

pressure~Pe Local sound speed,oe

(7)

This expression will be considered later in the paper in the light of experimental data.

Slug length /

U

(6)

pi~eea~

~

Orificeplate (target)

" ~

[~Force on Plate, F=

Local sound speed, oc

Slug velocity, U

Figure 1 Isolated slug in pipe

Slug velocity,L.' .

Figure 2 Slug impacting on orifice plate

\

Area of orifice A0

1O0

I. Owen et al - The impact of water slugs on wet steam flowmeters

Compressed air receiver

I l

9

)

!

Quick-opening valve Filling point / ~Polythenesheet i~,~J~ betweenflanges

50 mm diameter steel pipe ! Orifice plate

/

I$m Figure 3 Slug impact rig

driving gas. The volume of the air reservoir was sufficiently large that the drop in pressure as the air was expelled was less than 3%. For the majority of the tests the air and the water slug were expelled through a 50 mm diameter steel pipe. To enable the rapid expulsion of the air, a quarter-turn butterfly valve was located between the reservoir and the pipe. The water slug was held between the closed butterfly valve and a thin polythene sheet sandwiched between flanges downstream of the valve. The water was poured into the reservoir so formed through a filling port. By using two different lengths of pipe in which to hold the slug (0.99 and 2.16 m), it was possible to obtain three different lengths of slug (0.99, 2.16 and 3.15 m) whilst maintaining the overall length of the rig constant at 13.0 m. To propel the slug down the pipe, the valve was opened quickly by hand so that the sudden rise in pressure behind the water forced it through the polythene sheet. To confirm the consistency of the procedure, a number of valve openings were made and no discernible difference in the slug velocity was detected. To measure the slug velocity at the end of the pipe, two conductivity probes 0.215 m apart were used, as shown in Figure 4. The leading edge of the slug was detected as it passed the probes and this was recorded on a transient recorder to give the time interval between the slug passing the two positions. To measure the impact at the orifice plate, a piezoelectric pressure transducer was located close to its upstream face (Figure 4). The output from the charge amplifier was also connected to the transient recorder. Three different orifice to pipe area ratios were used: 0.25, 0.375 and 0.5. For each slug length, different air reservoir pressures were used to propel the slug down the pipeline. The velocity measurements were made with the orifice plate in position, and with it removed so the water discharged freely from the end of the pipe. It was observed that for the tests with the 2.16 and 3.15 m Transient recorder "

f

~ I

I0 0 Io

Charge amplifier

I

slugs, the trace recorded from the conductivity probes showed a step-change as the front of the slug passed by. For the 0.99 m slug, however, the trace recorded from the conductivity probe was extremely erratic, indicating that the slug had broken up. That this had happened was confirmed by the trace from the pressure transducer, which showed a much lower impact pressure. These observations led to a supplementary experiment being made, where a 1.24 m slug was propelled along a 28 mm bore pipe of much longer length relative to the 50 mm pipe. With the initial pipe length of 12.5 m (length to diameter ratio of 446), the slug was propelled along the pipe at four different air pressures and the velocity probes were used to indicate whether the slug had disintegrated. The tests were repeated another six times with the pipe length being shortened on each occasion until, at a length of 3.55 m, the slug was seen to remain intact. Installation of water separator upstream of orifice plate

To measure the effectiveness of protecting the orifice plate with a water separator, a Spirax Sarco 50 mm type 1808 separator was installed in the pipeline 2 m before the orifice plate. Three different situations were studied: the first was where the separator was empty and the slug of water was propelled through it; the second was where the separator was flooded with water and the slug propelled through it; the third was where the separator was flooded with water and only the expanding gas was propelled through it. These three situations are representative of what can be found in practice depending upon whether the separator is drained correctly or not. How the water will collect in an undrained separator is shown in Figure 5.

Baffle plate

1/ Separator flooded to level of inlet/outlet

\ Conductivity probe

Pressure transducer

Figure 4 Velocity and pressure measurement

Orifice plate

Drain plugged Figure 5 Flooded water separator

Flow Meas. Instrum.

Vol 2 April 1991

101 70J

For each of the three situations the slug length was kept at 2.16 m and the tank pressure varied up to 6 bar. The orifice area ratio used in this series of tests was 0.5.

Pipe diameter : 50 mm []

[]

6C

l = 2.16m x = 10.9m

Results and discussion Slug velocities

/ D

E]o.,~

[]11

50

Figure 6 shows the calculated velocity of the 2.16 and 3.15 m slugs as they travel along the pipe. The initial pressure upstream of the slug corresponds to that of the air receiver whilst the initial downstream pressure was taken to be atmospheric (all pressures shown are gauge pressures). After the initial rapid acceleration, the rate of increase of velocity reduces as the pressure of the expansion wave reduces and that of the compression wave increases. It can be seen how the heavier slug accelerates more slowly, as do both slugs for lower driving pressures. How the calculated velocities compare with the experimental data is shown in Figure 7. For each of the slug lengths in this figure, there is also a different length between the starting position of the leading edge of the slug and the position where the velocity is measured; this is because the overall length of the pipe was constant. It can be seen that there is no distinguishable difference between the slug velocities measured with and without the orifice plate in place. This suggests that the air being compressed between the orifice plate and the slug has little effect on the velocity of the slug as it approaches the plate. Results are presented for only two of the three slug lengths because, as stated earlier, it was observed that the shorter slug disintegrated before it reached the end of the pipe. To investigate this further, a suplementary experiment was made with a smaller diameter (28 mm) pipe, the results of which are shown in Figure 8. The approximate distance that the slug

Z"

E 40

g

2O

x

Distance between slug and velocity probes Theory <> Free end

l0

D 0

I

I

2

4

Orifice (25ram diameter)

I

I

I

6 8 Tank pressure (bar)

10

12

Figure 7 Comparison between calculated velocities and experimental data travelled before disintegrating is shown for different air pressures. How these distances relate to the overall velocity-time histories calculated for the slug is shown in Figure 9. The curves suggest that the slug approaches a terminal velocity asymptotically and that the slug will in fact break up before it reaches the asymptotic region. This data is for one slug length and

Pipe diameter = 28mm Slug length = 1.24m (440) 80

Pipe diameter : 50mm Tank pressure

60

~

O

0

10.3 bar

12

( / = 2.16m)

0

F, m

E

o

40

2O

~

10 (l= 3.15m)

0

8

0

I

t

I

2

4

6

I

I

I

8 I0 12 Distance along pipe (m)

I

I

14

16

Figure 6 Slug velocity as it travels along pipe

2.0 18

I

I

2.5

5.0

I

I

.3.5 4.0 Tank pressure (bar)

I

4.5

Figure 8 Approximate distance at which the slug breaks up

5.0

102

I. Owen et al - The impact of water slugs on wet steam flowmeters

Impact of slug on orifice plate

80 Pipe diameter = 28 mm Slug length = 1.24m Tonk pressure (bar) 4.83

I

60

I i

q~n

40

~4, -• /"

2.76

/ Terminal

Breaking Approximate distance tr0vel)ed (m)

20

0

4,14 5.45

I

I

0.2

0.4

I

(I) 20.5

(1') 12.5

(2) 19.5

(2') 12.2

(5) 17.5

(5') 11.5

(4) 16.0

(4') 9.6

I

I

0.6 0.8 I0 Time from slug release (s)

I

I

1,2

1.4

1.6

Figure 9 Slug velocity in the 28 mm diameter pipe

the observation just made cannot be applied universally. However, it is useful to have insight into the practical aspects of the slug dynamics since one might worry needlessly about the impact damage from a slug that has already disintegrated in the pipeline. The main difference between the slug being considered in the present work and more normal twophase slug flow is that in the latter the slug is 'riding' over a layer of liquid; picking up liquid at the front and shedding it behind. In the present case the slug is only shedding from the rear. TaiteP* proposed that the rate of shedding from the rear of the slug, X, is approximated by X = 0.2ApUR

(8)

where A is the pipe cross-section, p is the liquid density, U is the slug velocity and R is the liquid volume fraction in the slug. In the present case the slug will initially be all water and R will be unity. However, Barnea & Brauner 7 suggest that R can quickly fall towards 0.5 in a moving slug. The original mass of a slug of length _/is Apl and therefore for an average slug velocity of U, the time taken for the slug to disintegrate is I/0.1 U. If an average velocity for the slug between its release and its break-up of 30 m s -1 is taken from Figure 9 for the slug length of 1.24 m, then according to this relation the time for the slug to disintegrate is about 0.4 s, which is comparable with the time shown in the figure. Regardless of the simplistic nature of this calculation, it nevertheless demonstrates that a single slug being propelled along an empty pipeline has only a limited time before it disintegrates into a bubbly mixture.

Figure 10 shows a typical pressure trace for a 2.16 m slug being propelled along the pipeline from a tank pressure of 4.8 bar and impacting upon the 0.5 area ratio orifice plate. The pressure rises quickly to its peak within about 5 ms and then falls again; the pressure is not maintained at its maximum level for any period of time as would be expected for water hammer, but decays relatively slowly as the slug extrudes through the orifice. Figure 11 shows the impact pressures generated by the two different lengths of slug as they impacted upon orifice plates of different area ratios and with different velocities. Comparing Figure 11a with 11b, it can be seen that the shorter slug has higher impact pressures for a given air pressure; this is because, as shown earlier in Figure 7, the shorter slug has the higher velocities because of its lower inertia and because it has travelled a little further. The pressures generated are considerable. Figure 12 shows how the impact pressure consistently amplifies the original tank pressure for a given orifice diameter and slug length. Orifice plate flowmeters are required to have sharp edges for them to comply to the relevant standard (e.g. BS 1042 (1964) 9) The orifice plate used most frequently in the present tests, and for setting up and developing procedures, was made from mild steel, had a 25 mm orifice, and was 3 mm thick. After about 20 impacts, the plate had deformed very substantially and it was therefore frequently replaced thereafter. The potential for damage to orifice plate flowmeters is clearly very significant, as is the sudden pressure rise for the differential pressure measurement device. This observation will also apply to other flowmeters that would act as a target for the slug. Recalling equation (7), the data has been collected together in Figure 13 according to the relation [Ao12[ /1 F = constant x p A U 2 [ ~ ] [ ~ ] (9) Because only two slug lengths and just the one pipe diameter have been used in the impact tests, equation (9) cannot be applied universally. However, it can be seen from Figure 13 that the present data

Peak pressure = 52.6 bar ( 20 bar/division)

Slug length = 2.16m Tank pressure = 4.8 bar ,40/4 = 0.5

I

0 *If this reference is consulted, a subsequent letter to the editor of the Int. J. Multiphase F l o w from Bendiksen et al. ~ should also be consulted.

I

I0

I

I

I

20

I

30

I

I

40

Time (ms)

Figure 10 Pressure recorded at orifice plate

I

50

F l o w Meas. Instrum.

103

Vol 2 A p r i l 1991

30C

300,

(AO/A )

Orifice

o o

(0.250)

z~ ( 0 . 3 7 5 ) 25C

[]

x (0,500)

25C

o [] 0 []

200

200

[]

2 v o

o

o

~

150

150 [3

E

E z~

I00

A A

[]

Zi

I00

x

Li

x

Z~ A

x,,

X xX x

x

x x xx

50 /k

x

x

x I

I

0

x

50

x

I

4.

2

a

I

6

8

I

I

I0

Ton k pressure ( bo r )

0

I

3 4 Tonk pressure (bar)

I

I

I

5

6

b 7

Figure 11 M a x i m u m i m p a c t pressure on orifices: (a) 2. 16 m slug; (b) 3.15 m slug 60

60 Orifice ( / I O / A ) o

(0.250)

/x ( 0 . 3 7 5 ) 50

x

50

(0.500)

40

40

O0

[]

O []

o

[]

0[3

[]

n

o

0

30

0

O

[]

[3

3o A Z~

/k

Z~

Z~

Z~ /k

20 /x

A

A

A

,X

Z~ X

X

I0

o 25

X

X xX

Z~

2O

X>jK

X

X

X

X

X

X

X

X

X

X

X

I0-

i 30

~

~

I

I

i

35

40

45

50

55

a 60

Slug velocity (ms- I)

Figure 12 M a x i m u m (b) 3.15 m slug

0 2O

i

i

i

25

30

35

Protection of the flowmeter by a separator As demonstrated in Figure 11, the pressures that can be generated at the orifice plate are substantial. Figure 14 demonstrates convincingly how a water separator

b 45

Slug velocity ( m s - I )

i m p a c t pressure o f slug on orifice as a ratio o f initial tank pressure:

correlates well with this expression and further experiments would confirm this or yield the correct relationship. This, however, was beyond the scope and resources of the present work.

~ 40

(a) 2.16 m slug;

can protect a flowmeter. Regardless of whether the separator is already flooded or not, the protection it provides is very significant, reducing the impact pressures to about one-tenth of those experienced by the unprotected orifice plate. Taken together with the fact that the separator also increases the accuracy of the flowmeter by reducing the uncertainty of the water mass fraction, the case for installing a water separator upstream of the flowmeter is irrefutable. The positioning of the separator, whilst not critical, does need

104

I. Owen et al - The impact of water slugs on wet steam flowmeters

50 A

401

/(m)

~o/,4

2.16

0.250

•

2.16

0.375

O

2.16

0.500

0

3.15

0.250

x

3.15

0.375

[]

3.15

0.500

60

[]

Without separator []

5O

A 0 A

4O ~C

[]

A 0

30-

A Slug with flooded separator

zxO A

xx

[ ] Slug

E

0

x

O Flooded separator

Q. "6

0

/k

E -- 2 0 -

x Slug with emptyseparator

3C

20[]

z~

O

x

With separator

I0-

10

f 0

x

Q

0

I I

I 2 pNU2(A0/A)2(//D)

I 3 x 106kg ms- I

2

I 4

Ix 3

5

Figure 13 Correlated impact force some thought. It has to be sufficiently far upstream as to comply with the flowmeter installation requirements, and care has to be taken that between the separator and the flowmeter there is no opportunity for more water to collect to form another slug. Thus, as is recognized as good practice, the pipework should have a slight fall towards the separator, and no components that might cause water to collect (e.g. a valve) should be installed between the separator and the flowmeter. Conclusions This work has been concerned with possible damage that can be inflicted onto steam flowmeters as the result of the acceleration of an isolated liquid slug along an empty pipeline by an expanding high pressure gas. The velocity of the slug has been successfully predicted and compared with experimental data. It has been observed, however, that because of liquid being shed from the rear of the slug, if it does not collect additional water from ahead of itself, it will disintegrate after a certain time and its potential for causing damage will diminish accordingly. The pressure generated by the slug impacting on an orifice plate has been measured and has been

shown to be sufficiently large to have significant consequences for flowmeters in steam distribution systems or in any piping systems containing condensing or wet gas. Whilst this feature is well known from practice, steam flowmeters are often not preceded by a separator. The present study has quantified the problem and shown how good practice can eliminate any problems.

A x

,~

A

©

O

I

I

4 Tank pressure (bar)

5

Z~ O

Figure 14 Maximum impact pressure on orifice plate (Ao/A = 0 , 5 ; slug = 2 . 1 6 m )

Acknowledgements The authors wish to acknowledge the financial support of the Science and Engineering Research Council and Spirax Sarco Ltd. The contribution made through helpful discussions held with Dr J. W. Cleaver is also gratefully acknowledged. References 1 Sharp, B. B. 'Water Hammer: Problems and Solutions', Arnold, London (1981 ) 2 Dukler, A. E. and Hubbard, M. G. A modei for gas-liquid slug flow in horizontal and near horizontal tubes. Ind. Eng. Chem. Fund. 14(4) (1975) 337-347 3 Taitel, Y. and Dukler, A. E. A model for predicting flow regime transition in horizontal and near horizontal gasliquid flow. Am. Inst. Chem. Eng. J. 22(1)(1976)47-55 4 Dukler, A. E., Maron, D. M. and Brauner, N. A physical model for predicting the minimum stable slug length. Chem. Eng. Sci. 80(8) (1985) 1379-1385 5 Sakaguchi, T., Ozawa, M., Hamaguchi, H., Nishiwaki, F. and Fujii, E. Analysis of the impact force by a transient slug flowing out of a horizontal pipe. Nucl. Eng. Design 99 (1987) 63-71 6 Taitel, Y. Effect of gas expansion on slug length in long pipelines. Int. J. Multiphase Flow 13(5) (1987) 629-637 7 Barnea, D. and Brauner, N. Holdup of the liquid slug in two-phase intermittent flow. Int. J. Multiphase Flow 11(1) (1985) 43-49 8 Bendiksen, K., Hayer, N. and Malnes, D. Effect of gas expansion on slug length. Int. J. Multiphase Flow 14(5) (1988) 653-654 9 BS 1042 (1964) Methods of measurement of fluid flow in closed conduits. British Standards Institution