Robert P.

Evans
Advisory Engineer, Idaho National Engineering
and Environmental Laboratory,
P.O. Box 1625, Idaho Falls, Idaho 83415
e-mail: [email protected]
Flow Rate Measurements Using
Jonathan D. Blotter Flow-Induced Pipe Vibration
Associate Professor, Mechanical Engineering
Department, This paper investigates the possibility of developing a nonintrusive, low-cost, flow-rate
Brigham Young University, measurement technique. The technique is based on signal noise from an accelerometer
Provo, Utah 84602 attached to the surface of the pipe. The signal noise is defined as the standard deviation
e-mail: [email protected] of the frequency-averaged time-series signal. Experimental results are presented that
indicate a nearly quadratic relationship over the test region between the signal noise and
Alan G. Stephens flow rate in the pipe. It is also shown that the signal noise–flow rate relationship is
Professor, College of Engineering, dependent on the pipe material and diameter. 关DOI: 10.1115/1.1667882兴
Idaho State University,
Pocatello, Idaho 83206

1 Introduction various researchers using analytical, numerical and experimental

methods 关2,5–10兴. This research builds on these efforts and the
Flow measurements are used in many applications for various
signal noise concepts and presents experimentally based relation-
purposes. Some of these include providing data for system con- ships between signal noise and the flow rate in a pipe. This paper
trol, process analysis, accounting of yield, and consumption. Re- first describes an empirical approach to the basic theory on which
cent developments in technology have provided improved sensor the accelerometer-based flow sensor is founded. The paper then
designs and measurement techniques. Coriolis, magnetic, and ul- describes the experimental test set up and procedure used to ac-
trasonic flow meters are a few examples of this improved technol- quire the data. The data are then presented and discussed. The data
ogy. Although many high-quality pipe-flow sensors and measure- indicate that there is a strong relationship between the accelerom-
ment techniques exist, there is a need for the development of a eter based signal noise and the flow rate in the pipe. Over the
low-cost, nonintrusive, flow sensor. The geothermal industry is in range of flow rates investigated, it is shown that the relationship
particular need of such a sensor for the measurement of brine between signal noise and flow rate is nearly quadratic and a func-
flows. Geothermal brine typically consists of hot, pressurized liq- tion of pipe material and pipe diameter.
uid, carrying dissolved solids. As the pressure of the liquid drops,
some of the dissolved solids precipitate out of solution coating
any surface in contact with the fluid. This coating renders an in- 2 Theoretical Overview
trusive flow meter inoperable after a short time. Similar non-
When describing turbulent flow, it is convenient to recognize
intrusive measurements are also needed in the nuclear industry.
that the local velocity at a point may be regarded as superposition
This paper explores the possibility of developing a low-cost, of an average value and instantaneous fluctuating value. The in-
nonintrusive, flow-rate sensor based on the signal noise from an stantaneous velocity, u, can then be written in terms of the time
accelerometer attached to the outside surface of the pipe. Signal average velocity, ū, also called the mean velocity, and a fluctua-
noise in this paper refers to the standard deviation of a time-series
tion velocity u ⬘ as shown by Eq. 共1兲.
signal acquired from the transducer. This signal noise is a result of
turbulence-induced pressure fluctuations in the fluid flow which u⫽ū⫹u ⬘ (1)
are imparted to the pipe wall. This type of signal noise analysis
The time average velocity ū is defined as shown by Eq. 共2兲
has shown promise in providing nonintrusive flow measurements.
where T 1 is a time large enough that ū is the same for any longer
The method was first considered at the Idaho National Engineer- time for steady flow. Therefore, the mean is independent of time
ing and Environmental Laboratory 共INEEL兲 as a diagnostic tool and the time-averages of the fluctuations are equal to zero.
for pump performance in nuclear applications 关1兴. During the
Loss-of-Fluid Test nuclear reactor safety testing program, mea-
surements were analyzed to determine their uncertainty and limi-
tations. It was noticed that in turbulent single-phase flow, the stan-
ū⬅
1
T1 冕
0
T1
udt (2)

dard deviation of the signal increased with flow rate for flow- As the molecules of the fluid approach the wall, they have
measuring instrumentation 关2兴. Based on these results, initial kinetic energy. This kinetic energy must be converted to another
laboratory tests were performed by the authors. These tests dem- form of energy as the molecules reach the pipe wall. According to
onstrated that it was possible to use signal noise from an acceler- the first law of thermodynamics, some of the kinetic energy is
ometer to sense flow noise in a pipe 关3,4兴. These preliminary converted to heat as the turbulent eddies dissipate, but most is
results were used to obtain funding from the Department of En- converted into potential energy in the form of pressure 关11兴.
ergy 共DOE兲 through an INEEL Laboratory Directed Research and Consider a turbulent flow through a horizontal pipe of circular
Development 共LDRD兲 award to construct an air-water flow loop cross section. The velocities of the fluid can be expressed in terms
and continue to investigate the relationship between signal noise of a time average and a fluctuation as shown by Eqs. 共3兲 and 共4兲.
and flow rate. This paper presents the results of this investigation. u⫽ū⫹u ⬘ (3)
Pipe vibration due to internal flow has been investigated by
v ⫽ v̄ ⫹ v ⬘ ⫽ v ⬘ (4)
Contributed by the Fluids Engineering Division for publication in the JOURNAL
OF FLUIDS ENGINEERING. Manuscript received by the Fluids Engineering Division
In Eqs. 共3兲 and 共4兲, u is the velocity in the direction of the
April 28, 2003; revised manuscript received November 7, 2003. Associated Editor: primary pipe axis and v is the velocity perpendicular to the pipe
J. Lee. axis. Since there is no net flow in the direction perpendicular to

280 Õ Vol. 126, MARCH 2004 Copyright © 2004 by ASME Transactions of the ASME

Downloaded 16 Sep 2008 to 129.241.141.144. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 the pipe axis, the time-averaged flow in that direction is zero and where:
so the instantaneous flow is just equal to the fluctuation.
Although the time averages of the fluctuations in any direction g
C⫽
are zero, the time average of the products, such as u ⬘ v ⬘ are not A␥
equal to zero. Prashun 关12兴 states that in general, it can be shown Equation 共14兲 indicates that the acceleration of the pipe is pro-
that the time average of the product of the velocity fluctuations is portional to the pressure fluctuations in the fluid. This experimen-
less than zero as shown by Eq. 共5兲. tal research is based on this premise.
The last relationship to be demonstrated is the relationship be-
u ⬘ v ⬘ ⬍0 (5)
tween the standard deviation of the pipe vibrations and the mean
Prashun 关12兴 also demonstrated that for a circular conduit of flow rate of the fluid in the pipe. Blake 关15兴 stated that the gen-
radius r, the shear stress, ␶, at the wall can be related to the eration of vibrations by fluid motion involves the reactions of
pressure, p, as shown by Eqs. 共6兲 and 共7兲. fluids and solids to stresses imposed by time-varying flow. For
dynamically similar flows, the ratio of the flow fluctuations to the
r dp average flow is constant. Bird et al. 关16兴 clarified this relationship
␶ ⫽⫺ (6)
2 dx by noting that the oscillatory term is the time average of the ab-
solute magnitude of the oscillation, given by 冑m̄ where m⫽u ⬘ 2 .
dp 2␶ They defined this relationship as ‘‘intensity of turbulence,’’ which
⫽p ⬘ ⫽⫺ (7)
dx r is a measure of the magnitude of the turbulent disturbance, and is
given by Eq. 共15兲.
From the Navier-Stokes equations, Prashun also demonstrated
that for turbulent flow, the turbulent shear stress can be expressed 冑m̄
as shown by Eq. 共8兲 ⫽intensity of turbulence (15)
ū
␶ ⫽⫺ ␳ u ⬘ v ⬘ (8) From the definition of turbulent flow given in Eq. 共1兲 the inten-
Combining Eqs. 共7兲 and 共8兲 shows that the pressure fluctuation sity of turbulence expression can be rearranged as shown by Eq.
is proportional to the fluctuation of the fluid velocity as shown by 共16兲
Eq. 共9兲 1
N

p ⬘ ⬀u ⬘ v ⬘ (9) 冑m̄ m̄ N 兺 关 u 共 t 兲 ⫺ū 兴

i⫽1
i
2

⫽ 2⫽ ⫽C (16)
It can also be shown that the pipe vibration is proportional to ū ū ū 2
the pressure fluctuations in the fluid. For this analysis, the fluid-
Multiplying both sides by the number of points N and ū 2 and
filled piping system can be depicted as a one-dimensional model
dividing by N⫺1 results in Eq. 共17兲 where the left hand side is the
of a beam. It is well known from structural mechanics that the rate
definition of the sample standard deviation.
of change of the moment along a beam is equal to the shear and
the rate of change of shear, dV/dx, along the length of the beam N
1 NC
is equal to the pressure fluctuations, p ⬘ (x), per unit length as
N⫺1 兺 关 u 共 t 兲 ⫺ū 兴 ⫽ N⫺1 ¯u ⫽Kū
i
2 2 2
(17)
shown by Eq. 共10兲 关13兴. i⫽1

Since, as was demonstrated above, the flow fluctuations are

d 2 M dV
⫽ ⫽p ⬘ 共 x 兲 (10) proportional to the pressure fluctuations and the pressure fluctua-
dx 2 dx tions are proportional to the pipe vibrations, it follows that the
When a beam is subjected to bending, one side of the beam is in standard deviation of the pipe vibrations is proportional to the
tension while the other side is in compression. Differentiating the average flow rate.
well known Flexure Equation, given by Eq. 共11兲, twice with re-
spect to x, then using the relationship for p ⬘ from Eq. 共11兲 gives
the results in Eq. 共12兲. 3 Experimental Test Facility
The air-water flow loop at Idaho State University 共ISU兲 was
d2y used as the experimental test facility. A schematic of the flow loop
M ⫽EI (11)
dx 2 is shown in Fig. 1. This flow loop will be described as two main
components. These are the water system and the test section.
d2M d4y
2 ⫽EI ⫽p ⬘ 共 x 兲 (12) 3.1 Water System. Although this research is focused on
dx dx 4
single phase 共all water兲 flow, the ISU flow loop was designed to
In order to relate the pressure fluctuations to the pipe accelera- have two phase 共air-water兲 flow capability. The water system sup-
tion d 2 y/dt 2 , consider the differential equation of motion for plies the liquid phase of the test fluid and consists of a primary
transverse vibration of a beam as given by Seto 关14兴 in Eq. 共13兲. pump, air- and manually-operated flow control valves, a test loop,
a bypass loop with a water clean-up system, two air-water sepa-
⳵2y EIg ⳵ 4 y g ⳵4y rator tanks with an air discharge system, plus the interconnecting
2 ⫽⫺ 4 ⫽⫺ EI 4 (13)
⳵t A␥ ⳵x A␥ ⳵x piping, loop pressure, and pump speed instrumentation. These
components are shown in Fig. 1.
where: Water is circulated through the flow loop by a Goulds 10 in.
A ⫽ cross sectional area of the beam diameter impeller, centrifugal pump driven by a Reliance 75 HP,
␥ ⫽ specific weight of the beam 2,500 rpm, Model Super RPM DC, direct current motor with a
g ⫽ acceleration of gravity variable-speed controller. Water from the pump can be routed
EI ⫽ flexural rigidity through the test section, the by-pass section, or both so that the
Since g, A, and ␥ are constants, Eq. 共13兲 can be rewritten as amount of water passing through the test section can be con-
shown by Eq. 共14兲. trolled. The flow path for the water is controlled by two air-
operated Valtek, size 3, Class 150 valves at the outlet of the pump.
⳵2y ⳵4y Each valve opening can be controlled to allow the desired flow
2 ⫽⫺CEI ⫽⫺Cp ⬘ 共 x 兲 (14)
⳵t ⳵x4 through the test section and the bypass.

Journal of Fluids Engineering MARCH 2004, Vol. 126 Õ 281

Downloaded 16 Sep 2008 to 129.241.141.144. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Fig. 1 Flow loop schematic

The water for the test section goes from the pump, through the 3.1.1 Control System. The pump and valves in the flow loop
control valve and then through an air-water mixer before entering are controlled from a control panel in an adjacent room. The con-
the test section. In this research, air was not mixed with the water trol panel has an on-off switch for the pump motor as well as a
but the flow still passes through the mixing chamber. After pass- variable speed control. A dial gage, which receives input from the
ing through the test section, the water enters the first of two air- pump speed controller, displays the speed in revolutions per
water separator tanks which allow any entrained air to separate minute. The air-operated valves are controlled by electric potenti-
from the water by gravity. The pressure and water level in the tank
ometers on the control panel. The panel is situated so that the loop
are monitored by a pressure gage and level indicator, respectively.
The water from the first air-water separator tank passes out of the can be viewed during operation.
bottom of the tank into a second, smaller air-water separator tank, 3.1.2 Test Measurement Transducers. The flow loop instru-
where any additional air can be removed from the water. The mentation consisted of: a兲 a PCB Piezotronics Model 352B68 pi-
water then goes to the pump to start the cycle again.
ezoelectric accelerometer to measure pipe vibration in the test
The bypass loop provides two functions: it provides an alternate
path for the water so that the amount of water flowing through the section, b兲 a Flow Technology FT Series full-bore turbine flow
test section can be controlled without changing the pump speed, meter with a 0.0762-m 共3 in兲 nominal diameter, operating range of
and it also provides a system to cool and clean the loop water. The 250 to 2,500 liters/min 共l/m兲, repeatability of ⫾0.5% of reading,
water can become heated as it passes through the pump. After and linearity of ⫾0.5% of reading to measure water flow through
several passes through the pump, the water can become relatively the test section, c兲 a 0 to 1.034 MPa 共0 to 150 psi兲 Heise pressure
warm. Some of the water flowing through the bypass loop can be gage to measure test section pressure, and d兲 a bimetallic ther-
diverted through a system to cool and clean the water. The cooling mometer temperature gage to measure the test section tempera-
is accomplished by passing the water through a heat exchanger ture.
which transfers some of the heat to secondary cooling water. The
water also passes through an ion exchanger and filter to clean the 3.1.3 Data Acquisition and Recording System. Data from the
water before re-entering the flow path between the first and sec- flow loop transducers were recorded simultaneously on a
ond air-water separator tanks. The amount of water passing HP35670 Spectrum Analyzer and a PC based data acquisition sys-
through the cooling and cleanup system is monitored by a tem as shown in Fig. 2. As shown, the accelerometer data were
variable-area, rotameter and controlled by a hand-operated valve recorded using the spectrum analyzer while all other data were
in the bypass line. The remainder of the water from the bypass recorded using the PC data acquisition system.
loop re-enters the flow path at the outlet of the first air-water
separator tank. The loop water piping is primarily 0.0762 m or 3.2 Test Section. The flow loop test section consists of a 6.1
0.102 m 共3 or 4 in兲 nominal pipe size and primarily of 304 stain- m 共240 in兲 interchangeable section of pipe. In this effort, five
less steel and polyvinyl chloride 共PVC兲. different test sections were used. The first three sections, were
The structural support for the test section was designed to mini- nominal 0.0762-m 共3 in兲 diameter schedule 40 pipe made of clear
mize any external vibration. The test section was connected to the PVC, 304/304 L stainless steel, and aluminum. These three sec-
flow loop through rubber isolators and the pump was mounted to tions were used to determine the effects of material properties on
isolate as much vibration as possible. Background noise tests in-
the signal noise–flow rate relationship. Two other test sections,
dicated a quiet system and the data required minimal filtering for
extraneous noise. one of 0.1016-m 共4 in兲 nominal diameter clear PVC and the other
For normal operation, the system is started from an uncircu- of 0.0381-m 共1.5 in兲 diameter stainless steel were used to inves-
lated, cold water condition and brought up to a flow rate that tigate the effects of pipe diameter. The various test sections were
completely fills the pipe. The system is then allowed to run until inserted into the flow loop one at a time by flanged connections.
the piping is free of air and a steady temperature and flow rate The accelerometer was mounted on the top of the pipe test section
have been reached. 2.34 m 共92 in兲 downstream of the inlet.

282 Õ Vol. 126, MARCH 2004 Transactions of the ASME

Downloaded 16 Sep 2008 to 129.241.141.144. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Fig. 3 Frequency domain plots of the accelerometer data with
various flow rates
Fig. 2 Data acquisition schematic

would then be plausible. However, the changes in natural fre-

quency are typically very small 共on the order of tenths of a Hz or
4 Data Acquisition, Results, and Analysis less兲 for turbulent flow that is well below the critical flow rate
One focus of this research is to determine the relationship be- 共i.e., v/vcritical⬍0.1). Therefore, the sensitivity of such a technique
tween flow rate and the signal noise as measured by an acceler- for relatively slow-flow rates would be very poor. For flow rates
ometer attached to the pipe. The purpose is to determine if a that provide a v/vcritical ratio greater than 0.3 the natural frequency
relationship exists such that a nonintrusive flow-rate sensor or shifts become significantly larger. If estimates of the flow rate are
measurement technique could be developed. This section first pre- acceptable and relatively high-flow rates are expected it may then
sents the data acquisition process followed by the presentation and be plausible to develop a technique based on these concepts 关17兴.
discussion of the results. To illustrate the ability to measure the flow rate based on natu-
ral frequency shifts, frequency domain plots of the accelerometer
4.1 Data Acquisition. The data acquisition system previ- data for the various flow rates were generated. Portions of these
ously discussed and shown in Fig. 2 was used. The accelerometer are shown in Fig. 3. The data shown in Fig. 3 represents an aver-
data was initially acquired using various sample rates and time age of 10 data sets where 4,096 data points were acquired over 2
spans. Based on these results, a 2 s time span and a sample rate of s as previously discussed. By comparing the main peaks for each
2,048 samples/s were used to acquire the data presented. Anti- flow rate as shown in Fig. 3, the downward shift of the natural
aliasing filters were used with a cut off frequency of 800 Hz. The frequency due to the increase in flow rate is clearly apparent.
accelerometer time series data were transformed to the frequency However, for a flow rate of 1,311 l/m the peak occurs at 5.906 Hz
domain using a Fast Fourier Transform. Ten data sets were aver- and at a flow rate of 416 l/m the peak occurs at 5.937 Hz. There-
aged in the frequency domain and the data were transformed back fore, a factor of 3.15 decrease in flow rate results in a frequency
to the time domain. Initially, 50 data sets were used to compute factor increase of 1.094. This is a small increase in frequency for
the average. However, the data was such that the difference be- a large decrease in flow and would make for a system with a very
tween the signals based on 50 averages and the signals based on low sensitivity. An increase in the amplitude of the peak response
10 averages were indiscernible at all flow rates. Therefore, only as the flow rate increases is also shown to occur. Although this
10 data sets were used in the averaging. The standard deviation of data is shown to contain multiple peaks, which may be difficult to
the frequency domain average time series signal was then used to distinguish under various conditions, it does verify that the mea-
compare with the flow rate. It should also be noted that care was surements are responding according to theory and gives some feel
taken to vibrationally isolate the pump from the test section and that the accelerometer is indeed sensing changes in the flow. It is
the prominent pump frequencies were filtered from the accelerom- also noted that the first mode frequency peaks can be identified
eter data before the data were transformed back to the time do- because they follow the expected trends where the other small, yet
main. somewhat significant peaks, appear to have random shifts and do
Fourteen flow rates between 400 and 1,500 l/m were used for not follow the expected trends. To build on these results and in-
all data sets for the 0.0762-m 共3 in兲 test sections. The minimum vestigate techniques with higher sensitivity, this paper now con-
flow rate of 400 l/m was required to keep the pipe full of water. tinues with the focus of this research, which is to determine the
The 1,500 l/m limit was the highest flow rate obtainable with the possibility of using signal noise to measure the flow rate.
pump and system components. The flow steps were repeatable
within a ⫾2 l/m range. The flow rate was held relatively constant 4.2.2 Signal Noise Results and Analysis. The signal noise,
with a maximum fluctuation of ⫾1 l/m during testing. This results computed as the standard deviation of the frequency domain av-
in a less than 1% error in the flow rate input. The same acceler- eraged time-series signal, versus flow-rate plots for the three pipe
ometer position, sample rate, and flow rates were used in the materials are shown in Figs. 4 – 6. The diamond shaped points
testing of the five test sections. indicate the experimental data and the solid lines represent a
second-order least squared error fit to the data. The equation for
4.2 Results and Analysis. The results and analysis are pre- the curve fit and the R 2 value for the fit are also shown as part of
sented in two sections. First, experimental data will be compared each figure. In the curve fit equations shown, x, represents the
with a frequency analysis to show that the data follows expected flow rate in liters/minute. The PVC and stainless steel curves have
trends. The data that identifies the relationship between signal
R 2 values of 0.997 and 0.991, respectively, while the aluminum fit
noise and flow rate will then be presented and discussed.
drops to an R 2 value of 0.983. There is no apparent reasoning as
4.2.1 Frequency Domain Results and Analysis. It is well to why the aluminum pipe appears slightly less quadratic in the
known that the fundamental natural frequency of a pipe contain- measured flow range except for possible background noise de-
ing a flowing fluid decreases as the flow rate increases. A flow tected by the accelerometer. These curves clearly illustrate that
measurement technique based on this natural frequency shift there is a strong correlation between accelerometer based signal

Journal of Fluids Engineering MARCH 2004, Vol. 126 Õ 283

Downloaded 16 Sep 2008 to 129.241.141.144. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Fig. 7 Comparison of the curve fit of the standard deviation of
the vibrations in the test section with mass flow for various test
Fig. 4 Signal noise and curve fit for 0.0762-m „3-in… PVC pipe section materials

noise and flow rate. Although these curve fits are good it is noted two diameters are shown in Fig. 8. Figure 8 indicates that the
that if the flow rate went to zero this curve would most likely still dependence on pipe diameter is not constant over the flow domain
be deterministic but not quadratic. These curves also illustrate that investigated. It also shows that the larger diameter pipe has less
a system based on signal noise potentially has a much higher signal noise than the smaller pipe as might be expected.
sensitivity than a frequency shift based method. The second case consisted of comparing the 0.0762-m 共3 in兲
Figure 7 is a superposition plot of the signal noise versus flow stainless steel pipe to a 0.0381-m 共1.5 in兲 diameter stainless steel
rate for all three materials. This plot clearly indicates that the pipe pipe. The signal noise data for this case is shown in Fig. 9. Be-
material properties significantly change the shape of the curve and cause of limitations in the flow system, the maximum flow rate for
would need to be accounted for in the development of a measure- this case was approximately 1,200 l/m. The difference between
ment technique. These curves indicate that as the density and stiff- the curves in this case is more significant as is the change in pipe
ness of the material are increased, the slope of the signal noise— diameter. Furthermore, as in the previous case, the larger pipe
flow rate curve decreases. A focus for future work would be to diameter has lower signal noise values for the same flow rate.
determine relationships between density, stiffness, and signal To further investigate the significance of the data a nondimen-
noise. sional plot with all the data was developed as shown in Fig. 10.
Pipe diameter was another system parameter that was consid- The vertical axis is a nondimensional form of the standard devia-
ered to likely shift or change the shape of the signal noise–flow tion of the accelerometer data and the horizontal axis is the Rey-
rate curve. To investigate this, two cases were investigated. nolds number. This plot clearly indicates a strong relationship be-
The first case consisted of comparing the results of a 0.0762-m
and a 0.1016-m 共3 and 4 in兲 diameter pipe. The raw data for these

Fig. 8 Signal noise versus flow rate for 0.1016 m and 0.0762 m
Fig. 5 Signal noise and curve fit for 0.0762-m „3-in… stainless diameter PVC pipe
steel pipe

Fig. 6 Signal noise and curve fit for 0.0762-m „3 in… aluminum Fig. 9 Signal noise versus flow rate for 0.0762 m and 0.0381 m
pipe diameter stainless steel pipe

284 Õ Vol. 126, MARCH 2004 Transactions of the ASME

Downloaded 16 Sep 2008 to 129.241.141.144. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 References
关1兴 Evans, R. P., and Goodrich, L. D., 1992, ‘‘Determination of Void Fraction in a
Fluid Through Analysis of the Flow-Induced Noise on a Drag Disk Signal,’’
38th ISA International Instrumentation Symposium, Instrument Society of
America, Paper #92,0178.
关2兴 Lassahn, G. D., 1983, ‘‘LOFT Experimental Measurements Uncertainty
Analysis-Methodology and Summary,’’ I, NUREG/CR-0169, EGG-2037.
关3兴 Evans, R. P., Blotter, J. D., and Stephens, A. G., 1998, Determination of Flow
Parameters in Two-Phase Flow Through Analysis of Flow-Induced Noise on
an Accelerometer Signal; Geothermal Mass Flow Measurement Feasibility Re-
port, Idaho National Engineering and Environmental Laboratory internal re-
port.
关4兴 Evans, R., 1999, ‘‘Two-phase Mass Flow Measurement Using Noise Analy-
sis,’’ 45th International Instrumentation Symposium, Albuquerque, NM, ISA-
International Society for Measurement and Controls.
关5兴 Weaver, D., Ziada, S., Au-Yang, M., Chen, S., Paidoussis, M., and Pettigrew,
Fig. 10 Nondimensional plot of acceleration data versus Rey- M., 2000, ‘‘Flow-Induced Vibrations in Power and Process Plant Components-
nolds number for all data sets Progress and Prospects,’’ ASME J. Pressure Vessel Technol., 123, pp. 339–
348.
关6兴 Saito, N., Miyano, H., and Furukawa, S., 1990, ‘‘Study on Vibration Response
tween flow noise and flow rate and provides promise for the of Pipes Induced by Internal Flow,’’ Pressure Vessels and Piping Conference,
ASME Pressure Vessels Piping Div Publ PVP, pp. 233–238.
development of a flow sensor based on the flow noise approach 关7兴 Durant, C., Robert, G., Filippi, P., and Mattei, P., 2000, ‘‘Vibroacoustic Re-
presented in this paper. sponse of a Thin Cylindrical Shell Excited by a Turbulent Internal Flow:
Comparison Between Numerical Prediction,’’ J. Sound Vib., 229, pp. 1115–
5 Summary 1155.
关8兴 Durant, C., and Robert, G., 1998, ‘‘Vibroacoustic Response of a Pipe Excited
In summary, it can be concluded that noise from an accelerom- by a Turbulent Internal Flow,’’ Flow, Turbul. Combust., 61, pp. 55– 69.
eter mounted to the surface of a pipe has a strong 共deterministic兲 关9兴 Brevart, B., and Fuller, C., 1993, ‘‘Effect of an Internal Flow on the Distribu-
relationship with the flow rate. Based on measurement sensitivity, tion of Vibration Energy in an Infinite Fluid Filled Thin Cylindrical Elastic
it can also be concluded that a flow rate measurement technique, Shell,’’ J. Sound Vib., 167, pp. 149–163.
关10兴 Kim, Y. K., and Kim, Y. H., 1996, ‘‘Three Accelerometer Method for the
based on the signal noise method presented, has significantly more
Measurement of Flow Rate in Pipe,’’ J. Acoust. Soc. Am., 100, pp. 717–726.
potential, than a frequency shift based technique. It was also 关11兴 Pittard, M. T., and Blotter, J. B., 2003, ‘‘Numerical Modeling of LES Based
shown that pipe material and the diameter have effects on the Turbulent-Flow Induced Vibration,’’ ASME International Mechanical Engi-
relationship. Based on this research, it can also be concluded that neering Congress & Exposition, Washington, D.C.
a signal-noise based, flow-measurement technique has potential 关12兴 Prashun, A. L., 1980, Fundamentals of Fluid Mechanics, Prentice Hall, Engle-
for development but that continued research into calibration tech- wood Cliffs, NJ, pp. 202–222, Chap. 7.
关13兴 Thomson, W. T., and Dahleh, M. D., 1998, Theory of Vibration with Applica-
niques and methods to account for parameters that affect the rela- tions 5th ed., Prentice-Hall, Inc., Upper Saddle River, NJ, pp. 258 –286, Chap.
tionship such as pipe material and diameter needs to be per- 9.
formed. 关14兴 Seto, W. W., 1964, Theory and Problems of Mechanical Vibrations, Schaum
Publishing Co., NY, pp. 128, Chap. 5.
关15兴 Blake, W. K., Mechanics of Flow-Induced Sound and Vibration, Academic
Acknowledgment Press Inc., Harcort Brace Jovanokich Publishers, Orlando, FL, pp. 1– 43, 1,
Chap. 1.
The authors would like to thank the DOE and the INEEL for
关16兴 Bird, R. B., Stewart, W. E., and Lightfoot, E. N., 1960, Transport Phenomena,
the support of this research through the LDRD program. They John Wiley and Sons Inc., NY, pp. 153–179, Chap. 5.
would also like to thank Dr. Dan Maynes and Matt Pittard for 关17兴 Blevins, R. D., 1977, Flow-induced Vibration, Van Nostrand Reinhold Co.,
their suggestions to the paper and involvement with the research. NY, pp. 287–311, Chap. 10.

Journal of Fluids Engineering MARCH 2004, Vol. 126 Õ 285

Downloaded 16 Sep 2008 to 129.241.141.144. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm