See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/267495047

Internal Turbulent Flow Induced Pipe Vibrations With and Without Baffle Plates

Conference Paper · January 2010

DOI: 10.1115/FEDSM-ICNMM2010-30997

CITATIONS READS
11 1,479

3 authors, including:

Daniel Maynes Jonathan D Blotter

Brigham Young University - Provo Main Campus Brigham Young University - Provo Main Campus
140 PUBLICATIONS 2,425 CITATIONS 100 PUBLICATIONS 531 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Graduate Thesis View project

All content following this page was uploaded by Daniel Maynes on 09 May 2016.

The user has requested enhancement of the downloaded file.

 Proceedings of the ASME 2010of3rd
Proceedings theJoint US-European
7th International Fluids Engineering
Symposium Summer Meeting
on Fluid-Structure and
Interactions
8th International Conference on Nanochannels, Microchannels, and Minichannels
FEDSM2010
FEDSM-ICNMM2010
August 1-5, 2010, Montreal, Quebec, Canada
August 1-5, 2010, Montreal, Canada

DRAFT FEDSM-ICNMM2010-30997
FEDSM-ICNMM2010-3

Internal Turbulent Flow Induced Pipe Vibrations with and without Baffle Plates

Andrew S. Thompson, Daniel Maynes, Jonathan D. Blotter

Department of Mechanical Engineering
Brigham Young University
Provo, UT, USA

ABSTRACT

The induced vibration in pipes due to turbulent flow through them is important in many industries and
applications. This paper presents the results of an experimental investigation that characterizes pipe wall vibration
caused by turbulent internal flow. Experiments were conducted using a water flow loop to characterize how the
pipe wall vibration depends on the average flow speed, the pipe diameter, and the pipe thickness for fully-
developed turbulent pipe flow. Experiments were also conducted to characterize the influence on the pipe
response of turbulence generation due to the flow passing through baffle plates with various hole sizes and
constant through area. All experiments were conducted using PVC pipe with diameters ranging from 51 mm to
102 mm and diameter to thickness ratios ranging from 8.9 to 16.9. Average flow speeds for the experiments
ranged from 3 to 11.5 m/s and the baffle plates employed exhibited hole sizes ranging from 1.6 to 25 mm.
Accelerometers mounted on the pipe walls were used to characterize the pipe vibrations. The results show that for
fully developed turbulent flow the rms of the pipe wall acceleration scales nominally as the square of the average
fluid speed and increases with decreasing pipe wall thickness. Based on the data, a non‐dimensional parameter
describing the pipe wall acceleration for the fully-developed turbulent flow scenario is proposed and its
dependence on relevant independent nondimensional parameters is presented. Lastly, when turbulence was
induced using baffle plates the localized turbulence intensity was greatly increased. For the largest holed baffle
plates, cavitation was observed to occur, significantly increasing the rms pipe wall acceleration. As baffle plate
hole size decreased, vibration levels were observed to approach levels that were measured when no baffle plate
was employed. For all baffle plate experiments the magnitude of the vibration was observed to decrease with
increasing downstream distance from the turbulence source, approaching the baseline no baffle plate case.

1. INTRODUCTION Therefore the characterization of turbulence induced

pipe vibrations is important in many industries and
The conveyance of fluids through pipes is an integral applications.
part of world-wide economic activity and has For example, related work with relevance to the
contributed to the progress that has occurred over the nuclear power industry, explored external turbulent flow
last century. These pipes are prone to cyclic loading, in past cylindrical rods. This work showed that rod
the form of pipe vibrations, induced by internal turbulent vibrations are due to turbulent pressure fluctuations in
flow. Vibration loading becomes a problem as piping the boundary layer. It was also shown that the vibration
infrastructure ages, contributing to fatigue induced levels are proportional to the average fluid velocity
failure. For instance, in the U.S. about 60% of the squared2. Several researchers have proposed correlations
pipelines have been in use for over 25 years and are between the vibration level, in terms of the characteristic
becoming prone to failure 1. This may lead to economic amplitudes of displacement or acceleration, and the fluid
loss as well as environmental and societal damage. dynamic and geometric parameters3-5. These dependent

1 Copyright © 2010 by ASME

Turbulence Induced Vibrations…..by Thompson et al.

variables include the average fluid speed, the rod proposed by several researchers and has broad
diameter, the fluid density, the mass per unit length of application throughout industry4,10,12. There is thus a
the rod, etc. clear need to understand how the vibration levels depend
Concerning fully developed turbulent pipe flow on both the flow dynamics and the pipe characteristics.
several research groups, using both experimental and In general these include: average fluid speed, fluid
numerical approaches, have shown that pipe vibration density, turbulence levels, pipe diameter, pipe wall
levels increase with increasing flow dynamic pressure 6- thickness, pipe length, pipe material, etc.
11
. Although results from each investigation have shown The focus of this paper is to present results of an
differences, each study has concluded that the pipe experimental investigation that uses wall mounted
vibration is a direct result of the inherent spatially and accelerometers to characterize pipe wall vibration due to
temporally varying pressure at the pipe wall. It is well internal, fully-developed turbulent pipe flow. From the
known that the flow field is made up of eddies of various accelerometer measurements, the influence of the
sizes. The turbulent kinetic energy of these eddies is average fluid speed, the pipe diameter, and the pipe wall
transferred from large eddies into smaller eddies. As thickness on the pipe vibration are all explored. For all
these eddies approach the pipe wall, most of their energy results presented here the pipe was hung supported with
is converted into pressure fluctuations which induce pipe flexible cables. Specifically, experiments were
vibrations 4, 6, 12. conducted in PVC test sections of internal diameters of
Large amplitude vibrations have also been observed 51 mm – 101.6 mm, and pipe wall diameter to thickness
in piping systems in French nuclear power plants, as ratios ranging from 8.9 – 16.9. The experiments were
flows passed through single hole orifices. Experiments conducted with average fluid speeds ranging from 0 –
to determine the cause of these unwanted vibrations 11.5 m/s. Further, data published previously6 are also
concluded that they were caused by supercavitation at examined to characterize the influence of pipe material
the orifice13. This cavitation induced vibration can be on vibration levels. This previous work studied the
identified by computing the power spectral density of relationship between pipe vibration and pipe flow in
wall mounted acceleration measurements. Specifically, aluminum, stainless steel and PVC pipes.
cavitation adds a broad increase to the spectrum with an Also presented are results characterizing the
amplitude that depends on the incipient cavitation14. influence of baffle plates of varying hole size on the
Because piping systems incorporate other components, vibration levels. Baffle plate hole size varied between
such as orifices or valves, it may be important to identify 1.6 to 25 mm. Wall mounted acceleration measurements
how they affect the vibrational response of the system. were taken at various streamwise distances downstream
Although excessive vibrations can lead to unwanted from the from the baffle plates in each PVC test section.
consequences, the monitoring of vibration levels can
also be implemented to provide non-intrusive flow
sensing. Most flow sensors in use today require
interrupting the flow, which, for some applications, is
not always feasible. A non-intrusive technique has been

Rubber Coupler Baffle Plate Insertion

Point
Support Support Test Section

Flow Conditioner
Flow Meter Instrumentation
Vent Flow Meter
Vent

Clear Plastic

Pump Rubber Coupler

Figure 1: Schematic illustration of the BYU water loop used for experiments.

2 Copyright © 2010 by ASME

Turbulence Induced Vibrations…..by Thompson et al.

2. METHODOLOGY described in further detail in section 2.2. After the test

section, the flow passes through another 6.096 m section
2.1 Water Flow Loop after which it is returned to the pump. On the return
Experiments were conducted in a water flow loop portion of the flow loop, a clear section of schedule 40
facility at Brigham Young University (BYU). A PVC is mounted in line to allow visual inspection of the
schematic illustration of the flow loop is shown in Fig. 1. flow.
Water was circulated via a Bell and Gossett 75 hp, 1800
RPM centrifugal pump driven by a 75 hp Marathon 2.2 Test Sections and Instrumentation
Electric 365T motor. Water was used as the working The test sections consist of 6.096 m interchangeable
fluid and the loop was filled by two open vertical vent sections of 50.8 mm, 76.2 mm, and 101.6 mm diameter
columns. These vents extend above the level of the rest schedule 40 and 80 PVC pipe (see Figure 2) hung
of the system and keep the flow loop slightly pressurized supported by ceiling mounts. The actual pipe diameters,
to keep air from leaking in. The vertical columns also D, and wall thicknesses, t, are shown in Table 1.
serve to vent air bubbles that develop during the filling
process and to maintain nearly atmospheric pressure at Table 1: Internal pipe diameters and wall thicknesses
the pump inlet. The pump inlet is fed by 203.2 mm for experiments with PVC pipes.
schedule 80 PVC pipe. The pump outlets to 101.6 mm Pipe Schedule D (m) t (m) D/t
schedule 80 pipe, which divides into bypass and main 50.8 mm Sch 80 0.049 0.0055 8.899
branches. Each branch is controlled by a hand-actuated 76.2 mm Sch 80 0.074 0.0076 9.672
valve, allowing the flow through each section to be 101.6 mm Sch 80 0.097 0.0086 11.355
controlled. The bypass line provided a way to control
50.8 mm Sch 40 0.053 0.0039 13.427
flow rate without changing pump speed.
The main line expands to 203.2 mm schedule 80 76.2 mm Sch 40 0.078 0.0055 14.215
PVC to accommodate a flow conditioner to minimize the 101.6 mm Sch 40 0.102 0.0060 16.944
influence of pump induced vibration. The flow
conditioner was constructed from 76.2 mm thick PCB 352B68 accelerometers with nominal
aluminum honeycomb and several layers of polyethylene sensitivities of 10.2 mV/g were used to measure the pipe
mesh to facilitate dissipation of coherent turbulent wall acceleration. Accelerometers were placed on
structures and swirl induced by the pump. After the flow opposite sides of the test section at six discrete axial
conditioner, the pipe contracts to 101.6 mm schedule 80 locations. A fluctuating pressure transducer (PCB
pipe. After the contraction, the outlet is connected to a 102A02S) with a nominal sensitivity of 7.3 mV/kPa was
flexible rubber coupler. The coupler reduces structural also placed on the side of the test section. The pressure
vibrations transmitted to the test section from the pump transducer was located at the same axial locations as the
and pipe components and is connected to the building accelerometers. On the return leg of the flow loop, two
wall by two wall mounts (one up-stream and one down- Omega FP6500 paddle wheel flow meters with a range
stream) to absorb low frequency pipe swaying. The flow of 0.1-12 m/s and an accuracy of ±1.5% were used to
is then allowed to develop over 6.096 m of pipe (L/D > measure the average velocity of the water through the
60). The water then passes into the test section, which is

Figure 2: Photograph of 101.6 mm and 50.8 mm test sections.

3 Copyright © 2010 by ASME

Turbulence Induced Vibrations…..by Thompson et al.

pipe. The total pressure drop across the length of the test the sensors were sampled for 10 second intervals at a
section was also measured. As noted previously, the sample rate of 5000 Hz.
present results are also compared to data obtained and Experiments were conducted in the following
reported on previously by Pittard et al.11 where three manner. For each test section, the pump was set at a
different pipe materials were employed. The pipe desired speed and the flow was allowed to become
characteristics for the results of Pittard et al. are shown steady. Then 10 seconds of time series data were
in Table 2. acquired at 5000 Hz. The pump speed was then adjusted
and the process repeated to acquire 24-29 discrete flow
Table 2: Pipe material, diameters, and wall thicknesses for speeds. Subsequently, the baffle plates were inserted and
data of Pittard, et al 11. experiments were repeated in the same fashion.
Pipe Schedule Material D (m) t (m) 2.5 Pump Effects
101.6 mm Sch 40 PVC 0.102 0.00602 Accelerometers were mounted directly on the pump
76.2 mm Sch 40 PVC 0.0779 0.00548 to characterize the spectral content of the pump
76.2 mm Sch 40 Aluminum 0.0779 0.00548 vibration. In general the dominant vibration frequencies
Stainless were relatively low, with the exception of a spike at the
76.2 mm Sch 40 0.0779 0.00548 impeller rotation frequency. This spike did not appear in
Steel
Stainless the frequency spectrum of the accelerometer
38.1 mm Sch 40 0.041 0.00368 measurements on the pipe test section.
Steel
Although not attributed to the pump, low frequency
2.3 Baffle Plates drift in the accelerometer measurements was observed at
In order to produce various levels of turbulence in low and no flow. Therefore, these data were filtered with
the test sections, holed baffle plates were inserted a high pass filter of 2 Hz. Thus, except at flow rates
between the flanges that connected the end of the lower than presented in this paper the influence of the
developing region and the test sections, with the holes pump on the acquired data is minimal.
parallel to the pipe axis. Five baffle plates were
machined from 6.35 mm thick aluminum plate with 25.4 3. RESULTS
mm, 12.7 mm, 6.35 mm, 3.18 mm, and 1.59 mm holes
drilled into them. The center pitch of the holes (distance 3.1 Wall pressure Fluctuations
between the center of one hole and the center of the next The rms of the pressure fluctuations, P ′ , as a
hole) was 32 mm, 16 mm, 8 mm, 4 mm, and 2 mm function of the average fluid speed for the six test
respectively. The through area of the holes in each baffle sections listed in Table 1 was measured and is shown in
plate was constant and equal to 3548 mm2. This results Fig 3. At low speeds some scatter exists in the data due
in seven holes for the 25.4 mm baffle plate and 1793 to resolution limits of the sensors. At higher speeds
holes for the 1.59 mm baffle plate. however, (Vf > 2 m/s) the trend in the data is similar for
all test sections. Namely, the P’ vs. Vf trend exhibits a
2.4 Data Acquisition power law relation, P’ ~ Vfm. A least squares fit to each
A PC based data acquisition system consisting of a data set shown in Fig. 3 over the range Vf > 2.5 m/s
multi channel National Instruments data acquisition reveals that m varies from 1.91 to 2.07 with an average
module was used to collect acceleration, flow rate, and value of 2.02. There appears to be no systematic
fluctuating pressure time series data. For the variation in m. This result shows that P’ scales
accelerometer and pressure fluctuation time series data directly with the average fluid dynamic pressure
the rms values of the time series were computed. These (ρVf2).
values are referred to here as A ′ and P ′ respectively, When plotted as a function of Vf, the P’ data for
and represent typical magnitudes in the pipe wall the 10.16 cm diameter schedule 40 test section shows the
acceleration and internal surface pressure fluctuations. largest magnitude at a given Vf. The magnitude of the P’
The accelerometer data were also integrated to yield pipe data 101.6 mm test sections appears to be larger than the
velocity (integrated once). Subsequently the rms value of data for the 76.2 mm test sections, which are larger than
the pipe velocity, V ′ , was also computed. However, the the data for the 50.8 mm test sections. As expected,
results presented in this paper will focus on A ′ . All of

4 Copyright © 2010 by ASME

Turbulence Induced Vibrations…..by Thompson et al.

however, there seems to be no systematic variation in P’

with diameter to thickness ratio (D/t). Figure 5 shows A’ vs. Vf for each of the test sections
considered. Again, these data exhibit a power law
relationship of the form ′~ , where m varies from
1.94 to 2.19, with an average value of 2.06.
Modest variation between schedule 40 and 80 data
1000 sets exist for each pipe diameter, with the general trend
P' (Pa)

being an increase in with decreasing schedule size.

DD==10.16
101.6 cm,
mm,Schedule
Schedule80
80 The wall thickness for the schedule 80 test sections
DD==7.62
76.2 cm,
mm,Schedule
Schedule80
80 is about 40% greater than for the schedule 40 test
DD==10.16
101.6 cm,
mm,Schedule
Schedule40
40 sections while the diameters differ from 3-6%,
DD==7.62
76.2 cm,
mm,Schedule
Schedule40
40
DD==5.08
50.8 cm,
mm,Schedule
Schedule40
40
respectively. Figure 6 presents as a function of pipe
DD==5.08
50.8 cm,
mm,Schedule
Schedule80
80 diameter to thickness ratio, D/t, for each test section
100 considered, at a constant fluid velocity of 6.7 m/s.
2 20
Although there are only two points for each pipe
Vf (m/s)
diameter, the trend clearly shows that increases with
Figure 3: P' as a function of the average fluid speed, Vf
for flow through the six test sections. increasing D/t.

3.2 Accelerometer Measurements

Figure 4 shows A’ as a function of Vf along the
length of the 101.6 mm schedule 40 test section, where 0.4
x/D represents the ratio of the distance from the test
section entrance to inner pipe diameter. Data are shown
A' (m/s2)

at x/D = 3, 6, 9, 15, 21, 30, and 57. Like the pressure DD==10.16
101.6cm,
mm, Schedule
Schedule 80 80
fluctuation measurements, these data exhibit a power DD==7.62
76.2cm,
mm, Schedule
Schedule 80 80
law behavior of the form ~ . Above a flow speed of DD==10.16
101.6cmmm, Schedule
Schedule 40 40
DD==7.62
76.2cm,
mm, Schedule
Schedule 40 40
3.5 m/s, n varies from 1.91 to 2.39 for the various x/D
DD==5.08
50.8cm,
mm, Schedule
Schedule 40 40
locations, with an average value of 2.14. The data
DD==5.08
50.8cm,
mm, Schedule
Schedule 80 80
displayed in this figure exhibit very little variation in A’ 0.04
3 15
with x/D. Similar behavior is observed for the other test Vf (m/s)
sections explored. Because of this, A’ for each test Figure 5: A' as a function of Vf for flow through the six
section is averaged over all x/D locations. test sections considered.

0.4

x/D = 3
x/D = 6
A' (m/s2)

x/D = 9
x/D = 15
x/D = 21
x/D = 30
x/D = 57

0.04
33 6 9
Vf (m/s)
Figure 4: A' as a function of Vf at seven x/D locations Figure 6: A' as a function of D/t at Vf ≈ 6.7 m/s for each
along the length of the 101.6 mm schedule 40 test the three diameter pipes considered.
section.

5 Copyright © 2010 by ASME

Turbulence Induced Vibrations…..by Thompson et al.

3.3 Dimensionless A’ the dimensionless variables, except ρ*, have a very weak
It has been shown above that A’ scales influence on A*. In the present experiments it is
nominally as Vf2 and D/t. In general, the rms of the pipe impossible to hold all but one of the pipe dimensionless
wall acceleration can be written as a function of all the independent parameters constant. However, it is still
variables that exert influence: useful to explore how A* depends on each parameter.
Table 4 lists the values of m corresponding to
, , , , , (3) A*~ Rem power law fits to the present experimental data
for each of the six test sections considered. There
where ρeq is the equivalent density, which accounts for appears to be no systematic pattern in the variation of the
the combined mass of the pipe and the fluid. 15 values of m among the test sections and the average
value of m suggests a very weak dependence of A* on
Re. The dependence of A* on Re is slightly different than
2 that observed in the numerical simulations of Shurtz 15.
2 2 2 (4)
This is likely due to the fact that Shurtz considered a
2
2 2 hydraulically smooth pipe and the results of the present
Recasting this set of dimensional variables into experiments show that the pipes employed exhibit
dimensionless form following the standard approach behavior more characteristic of rough pipes where the
yields the following set of dimensionless variables. influence of Re is less pronounced.
The present data also show that the
dimensionless pipe wall vibration, A*, is a weak function
(5) of t* following the general pattern shown by the
numerical results of Shurtz 15.
The results of the numerical investigation of
Shurtz suggests that to a first order the pipe wall
(6) vibrations should scale as A’~ Vf2/t*ρ* for an unsupported
pipe 15.
(7) Table 3: The values of m corresponding to A*~ Z*m power
law fit determined by a numerical simulation of flow
induced pipe vibrations presented by Shurtz15.
(8)
m
Re -0.18
t* 0.04
The dimensionless pipe acceleration can then be
ρ* -1.00
expressed as a function of the dimensionless variables
listed in Eqs. 6 to 8:
Table 4: The values of m corresponding to A*~ Rem power
law fit for each of the six test sections considered.
, , (9)

In a parallel numerical investigation using a Test Section m

Large Eddy simulation approach of this same 101.6 mm, Sch 40 0.012
phenomena, Shurtz observed that the A* normalization is
101.6 mm, Sch 80 0.137
the appropriate dimensionless of A’ 15. By holding all but
one of the dimensionless variables listed in Eqs. 6 to 8 76.2 mm, Sch 40 -0.09
constant, Shurtz 15 was able to determine the first order 76.2 mm, Sch 80 0.015
effects of each of the dimensionless variables on A*.
These results are listed in Table 3 as power law fits of 50.8 mm, Sch 40 -0.11
the data (A* ~ Z*m) that were obtained using the 50.8 mm, Sch 80 -0.07
numerical model, where Z*is one of the dimensionless
variables listed in Eqs. 6 to 8. The table shows that all of Average -0.018

6 Copyright © 2010 by ASME

Turbulence Induced Vibrations…..by Thompson et al.

Shown in Fig. 7 is A’ as a function of Vf2/t*ρ* for

each of the six unsupported test sections considered. 0.1

This figure also includes power law fits of the data with
a zero intercept that pass through the schedule 40 and 80
test section data for each pipe diameter (101.6 mm, 76.2
0.01 DD==7.62
76.2cmmm
PVCPVC
mm, and 50.8 mm). This functional relationship causes
the data from the 101.6 mm and 50.8 mm test sections to DD==10.16
101.6cmmm
PVCPVC

A'
collapse to nominally a single curve. The 76.2 mm DD==7.62
76.2cmmm
schedule 40 and 80 data also collapse onto each other, 0.001 Aluminum
Aluminum
however, the magnitude of this collapsed data is smaller DD==7.62
76.2cmmm
Stainless
Stainless Steel
Steel
than for the 101.6 mm and 50.8 mm data. Much of the DD==3.81
38.1cmmm
behavior in A’ appears to be captured by the parameter Stainless Steel
Stainless Steel
Vf2/t*ρ*. It should be noted however, that holding one 0.0001

dimensionless variable constant while changing the 1 10

Vf2/ρ*t* 100 1000

others was not possible for these experimental Figure 8: A’ vs. Vf2/ρ*t* for the data presented by Pittard
measurements, thereby introducing confounding et al. 11.
influences.
3.4 Baffle Plate Influence
As discussed in section 2.3 baffle plates were
inserted at the test section entrance as turbulence
0.80 inducers. Five plates were used, each with a different
diameter and number of holes drilled into them. The
0.40 D = 101.6
10.16 mm,40Sch 40
cm, Sch diameters of the holes were 25.4 mm, 12.7 mm, 6.35
D = 76.2
7.62 mm,40Sch 40
cm, Sch mm, 3.18 mm, and 1.59 mm, with the hole diameter
D = 50.8
5.08 mm,40Sch 40
cm, Sch being how each baffle plate is distinguished in this
D = 101.6
10.16 mm,80Sch 80
cm, Sch
paper. The plates were fabricated so that the through
A'

D = 76.2
7.62 mm,80Sch 80
cm, Sch
D = 50.8
5.08 mm,80Sch 80
cm, Sch
area of the holes was 3548 mm2, resulting in various
D10.16
= 101.6
Linear mm fit
(10.16)
cm fit
numbers of holes in each plate (e.g. seven holes for the
D7.62
= 76.2
Linear mmcm)
(7.62
cm fit fit 25.4 mm plate, 28 holes for the 12.7 mm baffle plate,
D5.08
= 50.8
Linear mmcm)
(5.08
cm fit fit 112 holes for the 6.35 mm baffle plate, 448 holes for the
0.04 3.18 mm baffle plate, and 1793 holes for the 1.59 mm
30 300 900
Vf2/ρ*t* plate). The 101.6 mm schedule 40 test section was the
Figure 7: A’ vs. Vf2/ρ*t* for each of the six test sections only test section used for the baffle plate experiments.
considered. Power law fit lines with zero intercept pass Figure 9 illustrates the influence of each baffle
through the schedule 40 and 80 pipe section data for plate on A’ at seven flow velocities 0.305 m downstream
each diameter test section. from the baffle plate. The data are plotted versus the
ratio of the baffle plate thickness to hole diameter.
The data of Fig. 9 show that the 25.4 mm and
12.7 mm baffle plates (tbaffle/Dhole = 0.25 and 0.5
Figure 8 shows A’ as a function of Vf2/t*ρ* for respectively) result in the largest increases in the
the data presented by Pittard et al. 11. The pipe properties magnitude of A’; although for all baffle plates the pipe
for this work are listed in Table 2. This functional acceleration increases. The largest increases prevail
relationship causes A’ for each pipe material and when cavitation occurs. Cavitation existed for the largest
diameter (except for the 38.1 mm stainless steel test baffle plates and was accompanied by audible noise.
section) to collapse onto one another. What is interesting Evidence of cavitation in these data is shown in Fig. 8
to note is that although these data were collected from a between a flow speed of 3.07 m/s and 3.72 m/s with the
different facility and the pipe moduli and densities vary 25.4 mm baffle plate. At this Vf the magnitude of A’
greatly, the functional relationship, A’~ Vf2/t*ρ*, works suddenly jumps. Although not as apparent, cavitation
quite well to collapse most of these data as well. appears to occur between 3.07 m/s and 3.97 m/s in the
12.7 mm baffle plate. What is also evident is that as
tbaffle/Dhole increases, the magnitude of A’ decreases;

7 Copyright © 2010 by ASME

Turbulence Induced Vibrations…..by Thompson et al.

apparently due to an upward shift in the velocity at case. This becomes even more apparent for the 1.59 mm
which cavitation occurs and a reduction in the size of baffle plate (Fig. 12). Here the magnitude of A’ at x/D =
turbulent eddies formed. 3 is only about two times greater than for the no baffle
plate case. The value of m also changes very little with
x/D, with an average value of 2.04. These values are also
included in Table 5. There appears to be little systematic
10 variation in the value of m with x/D for the 1.59 mm
VfV=f =5.83
5.83
m/sm/s
baffle plate scenario. However, for the 6.35 mm baffle
A' (m/s2)

VfV=f =5.03
5.03
m/sm/s
VfV=f =4.37
4.37
m/sm/s plate scenario, the value of m decreases with increasing
1 VfV=f =3.97
3.97
m/sm/s x/D. The values of m appear to exhibit the same behavior
VfV=f =3.72
3.72
m/sm/s for the 25.4 mm baffle plate except at the end of the test
VfV=f =3.07
3.07
m/sm/s
0.1 VfV=f =2.43
2.43
m/sm/s
section, where the values of m begin to increase again.
Comparing the values of m for the three baffle plate
scenarios presented in Table 5 the general trend is that m
0.01 increases with increasing baffle plate hole diameter.
0.1 1 10
tbaffle/Dhole

Figure 9: A' as a function of tbaffle/Dhole for various flow

velocities in the 101.6 mm schedule 40 test section 0.305
m from each of the five baffle plates. 10 x/D= =
x/D 33
x/D= =
x/D 66

(m/s ) 2)
Figure 10 presents A’ as a function of Vf at x/D= =
99
A'A'(m/s
x/D
2
1 x/D= =
1515
various x/D locations along the test section for the 25.4 x/D
x/D= =
x/D 2121
mm baffle plate. Also shown are data for the no baffle x/D= =
x/D 3030
x/D= =
5757
plate scenario. As previously stated, cavitation is 0.1
x/D
Average,
Average, NoNo Baffle
Baffle PlatePlate
occurring at the baffle plate holes, for this plate, and its
effect on A’ propagates down the entire length of the test
0.01
section. Cavitation appears to be initiated at a fluid speed 1 Vf (m/s) 10
of about 3 m/s causing A’ to rise rapidly with increasing
Figure 10: A' vs. Vf at seven x/D locations along the test
Vf. At a flow speed of nominally 4 m/s, the rate of
section length with the 25.4 mm baffle plate. A' for the
increase in A’ levels off and becomes similar at all x/D. test section with no baffle plate has been included for
The magnitude of A’ decreases with increasing x/D and reference.
decreases towards the vibration levels of the no baffle
plate scenario at large x/D. A power law curve fit (A’ ~
Vfm) to the data above a flow speed of 4 m/s results in
values of the power, m, ranging from 4.12 to 3.36 for the 10
various x/D positions and is included in Table 5. At x/D
= 3 and a flow speed of nominally 5.5 m/s, A’ is
x/D
x/D =
=3 3
observed to be about 300 times greater than for the no 1
A' (m/s2)

x/D
x/D =
=6 6
baffle plate case and at x/D = 57 (end of the test section), x/D =9
A' (m/s2)

x/D =9
A’ is nominally 20 times greater. x/D
x/D = 15
= 15
x/D
x/D = 21
= 21
Figures 11 and 12 show A’ as a function of Vf x/D = 30
0.1 x/D = 30
for the 6.35 mm baffle plate (tbaffle/Dhole=1.0) and 1.59 x/D
x/D = 57
= 57
mm baffle plate (tbaffle/Dhole=4.0), respectively. In Fig. 11, Average,
Average, No No Baffle
Baffle Plate
Plate
cavitation appears to be initiating only at the highest
flow speed. Further, the magnitude of A’ is significantly 0.01
3 10
lower than what was shown for the 25.4 mm baffle plate Vf (m/s)
data in Fig. 9. At a flow speed of nominally 5.5 m/s, A’ Figure 11: A' vs. Vf at seven x/D locations along the test
is only about 20 times greater than for the no baffle plate section length with the 6.35 mm baffle plate. A' for the
case at x/D = 3. A power law fit of the data with the test section with no baffle plate has been included for
exponents included in Table 5, show that as x/D reference.
increases the value of m approaches the no baffle plate

8 Copyright © 2010 by ASME

Turbulence Induced Vibrations…..by Thompson et al.

As the flow speed increases to 6.84 m/s, as

illustrated by the results of Figure 14, it is evident that
1 the decay in the magnitude of A’ is pushed further
downstream. In the cases where cavitation is occurring,
the magnitude of A’ doesn’t begin to level off until
x/D
x/D =3
=3 above x/D = 30. The magnitude of A’ with the 3.18 mm
A' (m/s2)

x/D
x/D =6
=6 and 1.59 mm baffle plate has decayed to the no baffle
x/D
x/D =9
=9
0.1 x/D = 15
plate levels by x/D = 15, and by x/D = 30 for the 6.35
x/D = 15
x/D
x/D = 21
= 21
mm baffle plate.
x/D
x/D = 30
= 30
x/D
x/D = 57
= 57 100
Average,
Average, No No Baffle
Baffle Plate
Plate
0.01
3 Vf (m/s) 10 10
25.4cmmm
2.54 Baffle
baffle platePlate

A' (m/s2)
12.7cmmm
1.27 Baffle
baffle platePlate
Figure 12: A' vs. Vf at seven x/D locations along the test 6.35 cm
0.635 mmbaffle
Baffle
platePlate
section length with the 1.59 mm baffle plate. 3.18 cm
0.318 mmbaffle
Baffle
platePlate
1
1.59 cm
0.159 mmbaffle
Baffle
platePlate
NoBaffle
No Baffle Plate
Plate
Figure 13 illustrates how the magnitude of A’
decays with x/D for each baffle plate case at a constant
0.1
flow speed of 3.61 m/s. As expected, the magnitude of 1 10 100
A’ for the no baffle plate case is nominally flat along the x/D
test section length. Although the flow velocity is Figure 14: The decay of A' with x/D for each baffle
relatively low, cavitation is occurring with the 25.4 mm plate case at a flow speed of 6.84 m/s.
and 12.7 mm baffle plates. For these cases, the decay in
the magnitude of A’ appears to be steeper than for the These figures indicate that A’ decays with
three other scenarios. The test section may not be long increasing distance from the baffle plate. At x/D = 30,
enough for A’ to return to the baseline levels this streamwise position is nearly far enough away from
characteristic of the no baffle plate case. The magnitude the baffle plates for the flow to be considered fully
of A’ for the 3.18 mm and 1.59 mm baffle plates has developed again. The implication is that sufficiently far
decayed to the no baffle plate values by nominally x/D = away from a turbulence source that is not inducing
9. For the 6.35 mm baffle plate A’ decays to the baseline cavitation, the magnitude of A’ approaches a condition
value at x/D = 15. representative of the baseline fully-developed turbulent
pipe flow.
1
Table 5: The value of m from a power law fit of the 25.4
mm, 6.35 mm, and 1.59 mm baffle plate data with x/D.

Baffle Plate Hole Size

25.4cmmm
2.54 Baffle
baffle platePlate
x/D 25.4mm 6.35 mm 1.59 mm No Plate
)
( 2)

12.7cmmm
1.27 Baffle
baffle platePlate
A' (m/s

0.1 6.35cm
0.635 mmbaffle
Baffle
platePlate 3 4.12 2.93 2.02
3.18cm
0.318 mmbaffle
Baffle
platePlate
1.59cm
0.159 mmbaffle
Baffle
platePlate 6 3.78 2.75 1.92
NoBaffle
No Baffle Plate
Plate 9 3.54 2.45 2.29
15 3.36 2.28 2.19 2.03
0.01 21 3.70 2.03 1.89
1 10 100
30 3.47 1.95 2.00
x/D
57 3.85 2.09 1.98
Figure 13: The decay of A' with x/D for each baffle
plate case at a flow speed of 3.61 m/s. Average 3.69 2.35 2.04

9 Copyright © 2010 by ASME

Turbulence Induced Vibrations…..by Thompson et al.

4. CONCLUSIONS observed that as the baffle plate hole size decreased, A’

would approach magnitudes shown with the no baffle
This paper presented the results of an plate baseline. A’ was also observed to decay to baseline
experimental investigation to characterize pipe levels as the distance from a non-cavitating baffle plate
vibrations induced by turbulent pipe flow. Experiments increased.
were conducted using a water flow loop to address two
general phenomena related to pipe vibration: 1) How the NOMENCLATURE
pipe vibration depends on the average flow speed, pipe
diameter, and pipe thickness. 2) How turbulence : Dimensionless pipe acceleration.
generation caused by holed baffle plates influence the A’ : Rms of the time series acceleration signal.
pipe response. D : Pipe diameter.
When comparing a power law fit of the average Dhole : Baffle plate hole diameter.
of P’ along the pipe length to Vf, it was found that P’ P’ : Rms of the time series pressure signal.
scaled nearly as Vf2, with the power from the Re : Reynold’s number.
experimental data varying less than 5% from an Vf : Fluid velocity.
expected value of 2.0. It was determined that A’ for each Z* : A generalized dimensionless variable.
of the test sections also scaled nearly quadratically with m : Power law curve fit exponent.
Vf, with an average power of 2.06 over all the test n : Power law curve fit exponent.
sections. Put differently, P’ and A’ are proportional to t* : Dimensionless pipe thickness.
the dynamic pressure in the pipe. t : Pipe thickness.
When comparing the dimensionless pipe wall tbaffle : Baffle plate thickness.
acceleration (A*) to the dimensionless parameters Re, t*, x : Axial pipe distance.
and ρ*, it was found that A* was weakly dependent on all ρ* : Dimensionless pipe/fluid density.
of them except ρ*. This was in good agreement to a ρ : Fluid density.
parallel numerical study performed by Shurtz 24. This ρeq : Equivalent density of the fluid and pipe material.
strong dependence on ρ* resulted in the scaling ρp : Pipe material density.
relationship, A’~ Vf2/t*ρ*. It was found that this
expression collapses A’ data for PVC, steel and ACKNOWLEDGEMENTS
aluminum pipes. However, use of this expression for The authors acknowledge CCI Control Components
pipe diameters and wall thicknesses deviating Inc. and Genscape Inc. for their generous support of this
significantly than those used in the present experiments work.
is not recommended. This scaling relationship is a first
order estimate of the expected level of pipe vibration in a REFERENCES
long pipe. Further studies are currently ongoing that will 1.Awawdeh, A., Bukkapatnam, S.T.S., Kumara, S.R.T.,
examine the effects of diameter, wall thickness, and fluid et al., “Wireless sensing of Flow-Induced
density. While the pipe modulus does not appear in this Vibrations for Pipeline Integrity Monitoring,”
scaling expression, recent numerical work has also IEEE SAM 06, Boston, July 2006.
shown that with regard to pipe wall acceleration in long
2. Paidoussis, M.P., “Fluidelastic Vibratio of Cylinder
pipes that the pipe modulus exercises only modest
Arrays in Axial and Cross Flow: State of the Art,”
influence and its influence is easily masked by varying
Journal of Sound and Vibration, Vol. 76, Number
pipe density, which is included in the scaling
3, 1981, pp 329-360.
expression15.
3. Burgreen, D., Byrnes, J.J., and Benforado, D.M., 1958,
It was found that placing baffle plates into the
“Vibration of Rods Induced by Water in Parallel
flow would induce turbulence downstream of the baffle
Flow,” Transactions of the ASME, 80, pp. 991—
plate. For large baffle plate hole sizes cavitation existed
1003.
at high fluid speeds. Cavitation would cause the
4. Reavis, J.R., 1969, “Vibration correlation for
magnitude of A’ to increase by up to 300 times. As the
maximum fuel-element displacement in parallel
baffle plate hole size decreased, it was observed that the
turbulent flow,” Nuclear Science and Engineering,
fluid speed at which cavitation would initiate would
38, pp. 63-69.
increase. Cavitation was not prevalent at all with baffle
plate hole sizes smaller than 6.35 mm. Further, it was

10 Copyright © 2010 by ASME

Turbulence Induced Vibrations…..by Thompson et al.

5. Basile, D., Faure, J., and Ohlmer, E., 1968,

“Experimental study on the vibrations of various
fuel rod models in parallel flow,” Engineering and
Design, 7, pp17-534.
6. Saito, N., Miyano, H., and Furukawa, S., 1990, “Study
on vibration response of pipes induced by internal
flow,” Pressure Vessels and Piping Conference,
ASME Pressure Vessels Piping Division
Publication PVP, pp. 233-238.
7. Durant, C., Robert, G., Filippi, F., and Mattei, P.,
2000, “Vibroacoustic response of a thin cylindrical
shell excited by a turbulent internal flow:
comparison between numerical prediction,”
Journal of Sound and Vibration, 229, pp. 1115-
1155.
8. Durant, C., and Robert, G., 1998, “Vibroacoustic
response of a pipe excited by a turbulent internal
flow,” Turbulence and Combustion, 61, pp. 55-69.
9. Brevart, B., and Fuller, C., 1993, “Effect of an internal
flow on the distribution of vibration energy in an
infinite fluid filled thin cylindrical elastic shell,”
Journal of Sound and Vibration, 167, pp. 149-163.
10. Kim, Y.K. and Kim, Y.H., 1996, “Three
accelerometer method for the measurement of
flow rate in a pipe,” Journal of Acoustical Society
of America, 100, pp. 717-726.
11. Pittard, M.T., Evans, R.P, Maynes, D., and Blotter,
J.D., 2004, “Experimental and numerical
investigation of turbulent flow induced vibration
in fully developed flow,” Review of Scientific
Instruments, 75, pp. 2393-2401.
12. Evans, R.P., Blotter, J.D., Stephens, A.G., “Flow
Rate Measurements Using Flow-Induced Pipe
Vibration,” Transactions of the ASME, Vol. 126,
March 2004, pp 280-285.
13. Caillaud, S., Gibert, R., Moussou, P., Cohen, J.,
Millet, F., “Effects on Pipe Vibrations of
Cavitation in an Orifice and in Globe-Style
Valves,” Proceedings of Pressure Vessels and
Piping Conference, July, 2006.
14. Moussou, P., “An Attempt to Scale the Vibrations of
Water Pipes,” Transactions of the ASME, Vol.
128, November 2006, pp 670-676.
15. Shurtz, T.P., Analysis of Induced Vibrations in Fully-
Developed Turbulent Pipe Flow Using a Coupled
LES and FEA Approach, Brigham Young
University, Provo, UT, USA, Master’s Thesis,
2009.

11 Copyright © 2010 by ASME

View publication stats