Professional Documents
Culture Documents
net/publication/326801696
Pressure Wave Velocity in Fluid-Filled Pipes with and without Deposits in the
Low-Frequency Range
CITATION READS
1 1,647
4 authors:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Kjetil Daae Lohne on 13 December 2018.
FINAL DRAFT
Citation:
Mosland, E. N., Lohne, K. D., Ystad, B. and Hallanger, A., (2018). “Pressure
Wave Velocity in Fluid-Filled Pipes with and without Deposits in the Low-
Frequency Range”, Journal of Hydraulic Engineering, ASCE, Vol. 144, Issue
10 (October 2018), Online Publication date: 20 July 2018,
https://doi.org/10.1061/(ASCE)HY.1943-7900.0001517
1 PRESSURE WAVE VELOCITY IN FLUID-FILLED PIPES
3 FREQUENCY RANGE
5 ABSTRACT
6 Solid pipeline deposits can be mapped by analyzing reflections of hydraulic pressure pulse
7 waves in the low frequency range. To determine the amount and position of the deposits an
8 inverse problem should be solved: The pressure pulse is measured in one or several positions,
9 and the pipeline that generated this pulse must be determined. To solve this problem
10 estimates of the pressure pulse velocity are essential. A finite element axisymmetric model for
11 fluid-filled pipelines has been implemented and applied to numerically study low frequency
12 estimates for the pressure wave velocity in pipes with deposits that appear frequently in
13 pipelines of the petroleum industry. Finite element simulations include deposit thicknesses
14 in the range up to, and covering, 90 % of the internal diameter of the pipe. The results show
15 how deposits with different stiffness affect the pressure wave velocity. Input to the model
16 are the geometry and parameters (thickness, Young’s modulus and Poisson’s ratio) of the
17 pipe and the deposits.
18 Keywords: Finite element modelling, Fluid-structure interaction, Pressure wave velocity.
1
Christian Michelsen Research AS, Fantoftvegen 38, Bergen, Norway. Email: eivind.mosland@cmr.no
2
Christian Michelsen Research AS, Fantoftvegen 38, Bergen, Norway (corresponding author). Email:
kjetil@cmr.no
3
KTN AS, Gravdalsveien 262, Bergen, Norway. Email: by@knt.no
4
Christian Michelsen Research AS, Fantoftvegen 38, Bergen, Norway. Email: anders.hallanger@cmr.no
147 Lafleur and Shields (1995) studied axisymmetric propagation in a liquid-filled elastic
148 tube waveguide and used previous work by Del Grosso (1971) as the basis for the theoretical
149 description of modal phase velocities. These works are based on exact longitudinal and shear
150 wave equations for tubes of arbitrary thickness, and dispersion relations are obtained for
151 axially propagating waves in fluid-filled pipes surrounded by vacuum. For the zero frequency
152 limit, Lafleur and Shields (1995) derived an analytical expression for the fundamental and
153 low frequency fluid mode. Later, Baik et al. (2010) extended the work to include the complex
2c2f Ê
157 c2 = 1/2 (1)
−F̂ − F̂ 2 − 4Ê Ĝ
159 Ê = 3β̂ − 4α̂ ŷ (2)
h i
160 F̂ = −4γ̂ β̂ − α̂ − ˆ
3β − 4α̂ (1 + γ̂) + β̂ β̂ − α̂ ŷ (3)
161 Ĝ = β̂ 2 γ̂ + β̂ β̂ − α̂ (1 + γ̂) ŷ (4)
162 α̂ = (cf /cL )2 , β̂ = (cf /cS )2 , γ̂ = β̂ρf /ρp , ŷ = [(R + tp )/R]2 − 1 (5)
163 cL and cS are the longitudinal and shear wave velocities for the elastic pipe, respectively.
164 They can be expressed as function of Kf (Kinsler et al. 1999) and Ep and νp (Cheeke 2002)
Kf
165 c2f = (6)
ρf
166
Ep (1 − νp )
167 c2L = (7)
ρp (1 + νp )(1 − 2νp )
168
Ep
169 c2S = (8)
2ρp (1 + νp )
170 Eq. (1) is denoted here as Model 1 and is for a fluid-filled pipe with arbitrary wall
171 thickness. The typographic error corrected by Baik et al. (2010) is in the expression for γ̂.
172 In Lafleur and Shields (1995), this quantity is set equal to the inversion of the correct ratio.
185 cf (P, t) = 1402.7 + 488t − 482t2 + 135t3 + (15.9 + 2.8t + 2.4t2 )(PG /100) (9)
186 and is a function of the gauge pressure, PG , in bar and the temperature, t = T /100, with
187 T in degrees Celsius. At 10 bar and approximately 19.1◦ C, Eq. (9) shows that the sound
188 velocity in distilled water is 1481 m/s.
190 Several authors have studied theoretically the propagation of elastic waves in liquid filled
191 tubes, most of them using various approximations where the tube is treated as a thin shell.
192 Tijsseling (2007) states that if the usual thin-wall assumption is made, in which tp /R is
193 negligible with respect to unity, the thick-wall FSI four-equation model in Tijsseling (2007)
194 is reduced to the four-equation model of Skalak (1956). Skalak’s mathematical model is
195 valid for axisymmetric thin-walled tubes and includes the effects of radial inertia of liquid
196 and pipe, and longitudinal stress waves in the pipe (Tijsseling et al. 2008). The low frequency
197 pressure wave velocity, given also in Tijsseling et al. (2008), can be shown to reduce to the
c2f
201 c2 = K D
(10)
1 + Ψ Epf tp
202 Ψ is a coefficient that accounts for different pipe supports. Eq. (10) with Ψ = 1 is denoted
203 here as Model 2 and was first presented by Korteweg in 1878. It has since seen widespread
204 use both in classical waterhammer analysis (Tijsseling 1996) and in acoustic leak detection
205 (Long et al. 2003). Ψ = 1 corresponds to a pipe where each pipe section is anchored with
206 expansion joints or to pipeline materials with a high elastic modulus such as steel (ESDU
207 2007). Eq. (10) reduces to Eq. (6) for compressible fluids in a pipe with entirely rigid walls,
208 i.e. when Ep >> Kf , (Tijsseling 1996; Tijsseling et al. 2008).
209 When (Kf D)/(Ep tp ) < 1, Eq. (1) can be shown to reduce to Eq. (10) by using νp = 0
210 and the approximation that the pipe wall thickness is thin in comparison to the inner radius
211 (i.e. by only keeping the leading orders of tp /R). The definition for thin-walled pipes differs
212 somewhat between sources. The definitions in Jones et al. (2008) and ESDU (2007) are
213 tp /D < 0.025 and tp /D < 0.1, respectively. Tijsseling (2007) states that the requirement for
214 a thin-walled pipe is that tp /D is small with respect to 0.5, and points out that exact solutions
215 from the thick-wall FSI four-equation model show that the corrections are important only
216 for pipes with tp /D > 0.25.
219 Only one analytical expression which may describe a fluid-filled pipe with free outer
220 boundaries and with a layer inside the pipe has been revealed. The Engineering Sciences
221 Data Unit publication, ESDU (2007), presents this analytical expression as the pressure wave
v
Kf /ρf
u
c=t (11)
u
223 2 +t2 )(1−ν )+2ν D t ]
(Dm −tp )(1−νp2 )[(Dm p p p m p
1+ Kf ( Ep Dm tp (Dm −tp )(1−νp )+Ed td [(D 2 2 )
m +tp )(1−νp )+2νp Dm tp ]
224 where Dm = D + tp is the mean diameter. Eq. (11) is here denoted as Model 3. In the
225 derivation it is assumed that Poisson’s ratio for the lining, νd , is equal to νp (ESDU 2007). It
226 is stated in ESDU (2007) that the assumption is valid for most pairs of isotropic materials as
227 the expression is relatively insensitive to variation in the Poisson’s ratio. Additional details
228 on the assumptions used when deriving the expression are not given.
229 A similar model for the pressure wave velocity of a tunnel through rock which is lined
230 with concrete and steel is also given in ESDU (2007). The same Poisson’s ratio of steel lining,
231 concrete and rock is assumed. It can be shown that if the Young’s modulus of rock is equal
232 to zero, results from that expression agrees with results calculated with Eq. (11). Halliwell
233 (1963) also presents an expression for steel and concrete lined tunnels. If the radii and
234 thickness used in the expression in Halliwell (1963) are doubled, agreement to the similar
235 expression in ESDU (2007) is obtained. In the model presented in Halliwell (1963), it is
236 stated that the thickness of the lining represents the thickness of a thin steel lining.
238 An alternative method to estimate pressure wave velocities in fluid-filled pipes which are
239 nearly clogged with deposit is to set the properties of the pipe wall equal to those of the
240 deposit and use the analytical expression for a fluid-filled pipe with arbitrary wall thickness
241 to calculate c. This yields Model 4 as Eq. (1) using Ê, F̂ and Ĝ as defined in Eqs. (2), (3)
242 and (4), but using α̂, β̂, γ̂ as
243 α̂ = (cf /cLd )2 , β̂ = (cf /cSd )2 , γ̂ = β̂ρf /ρd , ŷ = (Ro /Ri )2 − 1 (12)
247 and where Ro and Ri represents the outer and inner radius of the deposit cylindrical shell,
248 respectively. When approaching a fully clogged pipe, i.e. the ratio of the inner diameter to
249 the pipe thickness approaches zero, the equations reduce to the pressure wave velocity in a
250 circular tunnel made of the deposit material (Tijsseling 2007),
−1
2 ρf 2ρf
251 c = + (1 + νd ) (15)
Kf Ed
252 FE MODEL
253 Commercial FE software can be used for numerical calculations of the pressure wave veloc-
254 ity in fluid-filled pipes with and without deposits. The two-dimensional (2-D) axisymmetric
255 FE model implemented here is based on using the Acoustics Module with acoustic-structure
256 interaction in the commercial software COMSOL MultiphysicsTM . The acoustic-structure
257 interaction couples fluid and structural domains. The model solves two equations coupling
258 the displacement in the elastic media to the pressure in the fluid. The elastic wave equation
259 in the time domain is (COMSOL 2015a; COMSOL 2015b; Aanes et al. 2017a)
∂2
260 ρ u = ∇ · σ̃(u) (16)
∂t2
261 where u is the mechanical displacement, σ̃(u) is the Cauchy stress tensor and all source
262 terms have been set equal to zero (Aanes et al. 2017a). For a time-harmonic wave, Eq. (16)
263 reduces to an inhomogeneous Helmholtz equation (COMSOL 2015a; COMSOL 2015b)
1 ∂2
268 p = ∇2 p (18)
c2f ∂t2
ω2
269 (∇2 + )p̂ = 0 (19)
c2f
270 respectively. The boundary conditions at the different interfaces are i) zero normal and shear
271 stress at the elastic-vacuum interface, ii) continuity of normal and shear displacement and
272 stress at the elastic-elastic interface, and iii) continuity of normal displacement and stress,
273 and zero shear stress at the elastic-fluid interface.
274 The FE model is general and simulations can be performed for arbitrary frequencies.
275 In this study the analysis is performed in the frequency domain. A central fluid column
276 is enclosed by either a pipe wall or a deposit layer and a pipe wall. The fluid is modelled
277 without losses and is defined by ρf and cf . The deposit and pipe materials are modelled as
278 isotropic linear elastic materials without losses, defined by their Young’s modulus, Poisson’s
279 ratio, and density. A rectangular mapped mesh is used throughout. One end of the fluid
280 column is exited by an axial acceleration, yielding guided waves. At the end of each domain,
281 a perfectly matched layer (PML) is used to attenuate the incoming sound waves. The PML
282 prevents the creation of reflected waves that could create standing waves and other artefacts
283 in the steady state solution. Different PMLs are defined for the fluid and elastic domains.
284 PMLs do not remove all reflections, but their effect seem to be negligible in the current
285 study.
286 Although the pressure amplitude vary across the pipe cross-section, the pressure wave
287 velocity is uniform. The sound pressure distribution along the central axis of the pipe is
288 extracted from the simulation. From the pressure as a function of position, an automatic
314 The fluid is excited using an axial acceleration of 0.1 m/s2 . A frequency of 100 Hz is used
315 in the FE model to obtain the presented results, allowing for reduced pipe length and efficient
335 The material parameters used to model steel pipes are ρp = 8000 kg/m3 , Ep = 220 GPa,
336 and ν = 0.3, and the enclosed fluid is water with material parameters ρf = 1000 kg/m3 , and
337 cf = 1481 m/s. The material parameters of water and steel differ, depending on e.g. type
338 of steel, processing methods, salinity, dissolved air, temperature and pressure. The chosen
339 material parameters for steel are similar to parameters listed in Leinov et al. (2015), where
340 ρp = 7932 kg/m3 , Ep = 216.9 GPa, and ν = 0.286543.
341 Deposit parameters are chosen to span possible deposits that may be encountered in
342 oil and gas pipelines. These are often divided into inorganic and organic scale, where the
372 Geometry
373 Steel pipe dimensions are typically defined by a nominal pipe size and an outer diameter
374 in inches. The pipe schedule is related to the pressure rating and it defines the wall thickness
375 for a given outer diameter. Results in the present study are given for pipe dimensions 2”, 4”
376 and 8” and the pipe schedules 40, 80 and 160. The corresponding wall thicknesses are given
377 in Table 2. The inner diameter, D, can also be calculated from these data. The tp /D ratio
378 is given in Table 3. Note that all tp /D ratios given in Table 3 represents thick-walled pipes
379 according to Jones et al. (2008).
380 RESULTS
382 In Fig. 1 the results of the FE model are compared to different analytical expressions
383 for a water-filled steel pipe (without deposits) for different schedules and outer diameters.
384 The analytical models are Model 1 - the fundamental fluid mode for a pipe with arbitrary
385 pipe wall thickness given in Eq. (1), Model 2 - the basic analytical model of Eq. (10) for
386 thin-walled pipes using Ψ = 1, and Model 3 - the analytical expression for a lined circular
387 pipe, Eq. (11), using td = 0. The results in Fig. 1 are plotted as a function of the tp /D
388 ratio.
390 Simulations for water-filled steel pipes with deposits are shown in Figs. 2, 3 and 4.
391 Fig. 2 shows calculations of the pressure wave velocity using the FE model and Model 3.
392 Calculations are shown for 2”, 4” and 8” pipes and with schedules 40, 80 and 160. In the
393 calculations, the pipe is filled with water and the deposit thickness is varied in the range
394 from 0 to 90 % relative to the inner radius of the pipe. Calcium Carbonate with material
409 The results in Fig. 1 show that c decreases with decreasing tp /D and that the deviations
410 between the models are relatively small. The estimates of the FE model and Model 1 are
411 close, with deviations smaller than 0.17 m/s. Model 2 estimates of c are 0.68 % to 0.84 %
412 (or 9 m/s to 12 m/s) larger than FE model estimates. Overestimation is expected for the
413 tp /D values in Fig. 1 since the model is based on a thin-wall approximation with validity
414 for tp /D < 0.025 (Jones et al. 2008). The additional mass of the pipe wall neglected in
415 the thin-wall approximation reduce the pressure wave velocity (Tijsseling 2007). Model 3
416 overestimates in comparison to the FE model when the tp /D ratio is small, but the estimates
417 approach those of the FE model and Model 1 with increasing tp /D ratio. Details on the
418 underlying assumptions of Model 3 are not given in ESDU (2007), preventing further analysis.
419 The results from the FE model with a relatively stiff deposit (Ed = 80 GPa) in Fig. 2 show
420 an increase in c with increasing deposit thickness for most of the studied pipe dimension.
421 The slope of the curve is correlated with tp /D. For these deposit parameters, the lowest
460 CONCLUSIONS
461 The low frequency pressure wave velocity in fluid-filled pipes is investigated in this paper,
462 with particular focus on deposits affecting this velocity. A finite element model has been
463 implemented for the study, and it has been validated against analytical models for pipes with-
464 out deposit and for pipes that are nearly clogged with deposits. For pipes without deposits,
465 calculations with the FE model are in good agreement with calculations performed with
466 an analytical expression valid for pipes with arbitrary thickness (no thin wall assumption).
467 Small deviations (< 1%) are observed when calculating with other analytical expressions.
468 These simplified models can yield satisfactory results in applications where the uncertainty of
469 the estimate is dominated by the uncertainties of the input parameters. For fluid-filled pipes
470 with deposits where the deposit thickness approaches a fully clogged pipe, the FE model
471 calculates a pressure wave velocity that tends towards the pressure wave velocity in a tunnel
472 made of the deposit material. Assuming monotonous change in pressure wave velocity with
473 increasing deposit thickness, the maximum change in the pressure wave velocity caused by
474 a deposit is thus given by the analytical equations for a pipe without deposits and a tunnel
475 made of deposit.
486 ACKNOWLEDGEMENTS
487 The authors acknowledge the support provided by the Regional Research Fund of Western
488 Norway in the grant number 257068. Special thanks go to Stian Stavland for valuable
489 discussions during in the project execution.
493 REFERENCES
494 Aanes, M., Kippersund, R. A., Lohne, K. D., Frøysa, K.-E., and Lunde, P. (2017a). “Time-
495 of-flight dependency on transducer separation distance in a reflective-path guided-wave
Material ρd Ed νd
(kg/m3 ) (GPa) ()
Paraffin wax (Rommetveit 2011; Wedvich et al. 1992) 800-1000 1-4 0.2-0.4
Gas hydrate (Helgerud 2001) 800-1700 6-11 0.3-0.4
Calcium Carbonate (Helgerud 2001; Greenberg and Castagna 1992) 2710 80 0.32
Stearin 950 3.1 0.34
∗ The Poisson’s ratio of Paraffin wax and density of Stearin are estimated values.
1440
1420
1400
1380
Schedule
1360 40 80 160
1 5 FE model
1340 2 6 Model 1
3 7 Model 2
1320 4 8 Model 3
FIG. 1. Calculated pressure wave velocities of water-filled steel pipes without deposits
for different pipe schedules and pipe diameters as listed in Tables 2 and 3. Comparison
of calculations with FE model and analytical models, i.e. Models 1, 2 and 3 using Eqs.
(1), (10) with Ψ = 1 and (11) with td = 0, respectively.
1440
1420
1400
1380
FE model Model 3
1360 2" 4" 8" 2" 4" 8"
Schedule
40
1340
80
1320 160
0 20 40 60 80
Deposit thickness relative to inner radius [%]
FIG. 2. Calculated pressure wave velocities for water-filled steel pipes with deposits for
different pipe schedules and pipe diameters as listed in Tables 2 and 3. Calculations
made for absence of deposit and increasing deposit thickness to 90 % relative to the
inner radius of the steel pipe. Deposit material parameters are Ed = 80 GPa, ρd = 2710
kg/m3 , and νd = 0.32.
110
100
90
80
Poisson's ratio
70 0.2 0.3 0.4
Schedule
40 1 4 7
60 80 2 5 8
160 3 6 9
50
0 20 40 60 80
Deposit Young's modulus [GPa]
1300
1200
1100
1000
900 FE model
Model 3
800 Model 4
0 20 40 60 80
Deposit thickness relative to inner radius [%]
FIG. 4. Pressure wave velocity in a water-filled 4” schedule 80 steel pipe with a stearin
deposit layer of varying thickness. Deposit material parameters are Ed = 3.1 GPa,
ρd = 950 kg/m3 , and νd = 0.34. FE calculations compared to Model 3 and to Model 4.