Sei sulla pagina 1di 24

282

Vol. 1, No. 4 HVAC&R Research October 1995


Introduction
Motivation and Background
With recent concern about improving air quality and energy efficiency in buildings, accurate monitoring
and control of the airflow rate in ducts has become increasingly important. Energy efficiency is improved by
delivering the appropriate amount of heated or cooled air to each room in a building, and air quality is
improved by carefully monitoring the amount of ventilation air taken into the building. Widespread flow rate
sensing could also lead to improved performance monitoring and automated fault detection and diagnostics.
Vortex-shedding flow meters provide a very promising alternative to pitot-static probes (Drees et al.
1992) and heated thermistor sensors for measuring air flow in HVAC ducts that could lead to greater appli-
cation of flow sensing. When a bluff body is placed in a flow, vortices form and shed alternately from each
side, as depicted in Figure 1. The frequency of the shedding is proportional to the mean flow velocity (and
bluff body diameter), and the relationship is linear over a wide range of Reynolds numbers.
A vortex-shedding flow meter consists of a bluff body and a means of sensing the vortex-shedding fre-
quency. Typically, the shedding frequency is determined through spectral analysis of pressure measurements
made at the surface of the bluff body. Vortex-shedding flow meters offer high accuracy (even at low flow
velocities) and linearity, and can be produced at a low cost. Their calibration is independent of flow temper-
ature, density, and humidity, making them ideal for use in HVAC applications. These meters have no moving
parts, so they are very reliable and can offer many years of service.
Evaluation of Vortex-Shedding Flow
Meters for Monitoring Air Flows in
HVAC Applications
Mark C. Wolochuk James E. Braun, Ph.D., P.E.
Member ASHRAE
Michael W. Plesniak, Ph.D
This paper describes the evaluation of vortex-shedding flow meters under highly disturbed flow condi-
tions that occur in HVAC ducts. Field data were acquired from a typical HVAC duct to quantify the dis-
turbed flow conditions, and to set the bounds for controlled wind tunnel experiments. Wind tunnel
experiments were then used to demonstrate that vortex shedding flow meters can be used for measuring air
flows in the highly disturbed conditions that occur in HVAC ducts, but that care must be taken in their
design and application. The results imply that vortex shedding flow meters should be calibrated in turbu-
lent flows, and that the turbulence length scale at the bluff body must be controlled to ensure accurate
meter calibration. Periodic unsteadiness in the mean flow caused the vortex shedding to lock-on to a sub-
harmonic of the disturbance frequency for disturbances at various multiples of the shedding frequency. No
lock-on was observed for a forcing frequency at half the shedding frequency, indicating that lock-on can be
avoided by sizing the bluff body small enough so that its shedding frequency is always much greater than
any unsteadiness frequency present in the flow. When a long bluff body was spanned across a non-uniform
velocity profile, single-frequency shedding cells occurred along the bluff body. The single-frequency cells
broke down with increasing velocity gradient and bluff body aspect ratio. Therefore, to measure an aver-
age flow rate, a number of short bluff bodies must be used.
Mark C. Wolochuk is a Research and Development Engineer for Huntair, Portland, Oregon. James E. Braun and Michael W. Plesniak
are professors at the School of Mechanical Engineering, Purdue University, West Lafayette, Indiana.
Table of Contents
VOLUME 1, NUMBER 4, OCTOBER 1995 283
Although vortex-shedding flow meters have been commercially available since 1968, they have not yet
been used in HVAC ducts, and their performance in such environments has not been documented. Typically,
they are used in laminar pipe flows, and a long section of straight pipe upstream of the flow meter is required
to ensure clean approach flow. ASME (1987) publishes a standard for using vortex flow meters to measure
fluid flow in pipes, but it does not address the performance of these devices when measuring air flows in
ducts. In HVAC ducts, a number of disturbances exist, such as free stream turbulence, periodic unsteadiness,
and non-uniform velocity profiles, which can potentially degrade flow meter performance. These distur-
bances are illustrated schematically in Figure 1.
All real flows contain some degree of free stream turbulence and unsteadiness, and in many locations in
HVAC ducts, disturbance levels can be quite high. Turbulence is generated in the wake of obstructions in the
flow, by jets emerging from small openings (e.g., partially closed dampers), by fans, and by secondary flows
caused by bends in the duct. Free-stream turbulence can affect both vortex shedding from a bluff body and
the signal-to-noise ratio (SNR) of the pressure measurements.
Periodic unsteadiness in duct flow may be produced by a fan where blade passage forces the flow peri-
odically; the periodic velocity component is superimposed on the mean flow velocity. Unsteadiness can also
be generated by vortex shedding from blunt objects (e.g., reinforcing rods) within the duct. Periodic
unsteadiness can interfere with vortex shedding from a bluff body and alter the shedding frequency.
Most flows in ducts are non-uniform to some degree, so the velocity may vary widely across a duct cross
section. This is especially true downstream of a bend or partial blockage in the duct caused by dampers. Non-
uniform flow can cause metering errors if the volumetric flow rate is to be calculated from a single-point
measurement. A vortex-shedding flow meter with a very short bluff body is essentially a point-velocity
meter. On the other hand, a meter with a very long bluff body (spanning the entire duct cross section, for
example) is definitely not a point-velocity meter. Ideally, this long bluff body would act to average the
non-uniform flow, generating a single-shedding frequency which corresponds to the average flow rate. At
the other extreme, the shedding frequency may vary continuously along the span of the bluff body, with the
local shedding frequency corresponding to the local flow velocity. This latter case again represents a point-
velocity meter, since a number of sensors would be required along the span of the bluff body, and the output
of each sensor would represent only the local (point) velocity at that location.
If a flow meter is to be designed for use in disturbed flow environments, it is important to quantify the
effects that these disturbances will have on meter performance. It is also important to define the range of con-
ditions for which a given calibration is valid. The next section describes previous studies involving the
effects of flow disturbances on vortex shedding from bluff bodies.
Literature Review
The phenomenon of vortex shedding from bluff bodies has been studied since the pioneering work of
Strouhal (1878) and Von Krmn (1912). The non-dimensional shedding frequency, the Strouhal number, is
Figure 1. Disturbances affecting vortex-shedding flowmeters
284 HVAC&R RESEARCH
defined as: St = f D/U, where f is the vortex-shedding frequency, D is the bluff body diameter (width), and
U is the mean free stream velocity (see Figure 1). A constant Strouhal number implies a linear relationship
between shedding frequency and mean velocity for a given bluff body geometry, making the vortex-shed-
ding phenomenon attractive for use in flow metering.
The idea of building a flow meter based on the assumption of a constant Strouhal number was first pro-
posed by Roshko (1953, 1955), who studied vortex shedding from a circular cylinder in a Reynolds number
range of 40 < Re < 10,000. For Re > 300, the Strouhal number remains at an almost constant value of 0.2,
independent of Reynolds number. The normal operating range (i.e., constant Strouhal number range) for a
flow meter based on vortex shedding from a circular cylinder is 300 < Re < 2 10
5
. However, the value of
the Strouhal number in this range, as well as the extent of the range itself, is dependent on bluff body geom-
etry, end conditions, and flow blockage within the test section.
Free stream turbulence has been reported to weaken vortex shedding, decrease the signal to noise ratio
(SNR) and alter the shedding frequency. Turbulence, which is random three-dimensional velocity fluctua-
tions in a flow stream, can be characterized by its intensity and length scale. Typically, turbulence intensity
above about 10% is considered high turbulence, while that lower than about 0.5% is considered low tur-
bulence. Bearman and Morel (1983) noted that the ratio of the turbulence length scale to bluff body char-
acteristic length (L
x
/D) is the most important parameter in characterizing the interaction of free stream
turbulence with bluff body vortex shedding. Longitudinal and transverse integral length scales, L
x
and L
y
,
represent the average size of the energy-containing eddies in the flow.
Malard et al. (1991) showed that for small-scale free stream turbulence (L
x
/D 0.25), an increase in the
turbulence intensity (Tu) from 3.3% to 10.3% increased the Strouhal number by 2% for a trapezoidal bluff
body with the large face oriented normal to the flow. Nakamura and Ohya (1984) studied vortex shedding
from rectangular cylinders for various turbulence integral length scales (from 0.21 to 14 bluff body diame-
ters). They found that small-scale turbulence tended to strengthen vortex shedding for shorter (i.e. in depth)
cylinders, while it weakened shedding for longer cylinders. Large-scale turbulence weakened shedding for
all cylinders, and decreased the spanwise coherence of vortex formation. This effect was attributed to the
three-dimensional nature of turbulence, and was most pronounced for L
x
/D of about 3.
Periodic unsteadiness in the mean flow can cause the vortex shedding to lock-on to a subharmonic of the
perturbation frequency. While the shedding is locked-on, the shedding frequency remains constant over a
range of changing velocities. Periodic unsteadiness is characterized by its frequency f
p
and amplitude A. In
this work, the amplitude is defined as the ratio of the RMS streamwise velocity fluctuation to the mean veloc-
ity: . Many experimental and numerical investigations of the lock-on phenomenon are summa-
rized in the review article by Griffin and Hall (1991). In general, lock-on is found over a range of frequencies
corresponding to forcing at the shedding frequency or its harmonics. The range of lock-on is amplitude
dependent and the range of frequencies over which shedding remains locked is not symmetric about the res-
onant frequency f
so
.
Recently, Hall and Griffin (1993) performed a direct numerical simulation of the incompressible Navier-
Stokes equations to investigate vortex shedding from a circular cylinder in a flow with a sinusoidal velocity
perturbation superimposed on the mean flow. They investigated the range of lock-on frequencies as a func-
tion of perturbation amplitude, and their results agree quite well with previous experimental results. The
numerical results confirmed the existence of a transition range between lock-on and no lock-on, in which the
dominant frequency in the flow is neither the natural shedding frequency nor half the perturbation frequency.
Non-uniform flow may affect the Strouhal number and have an adverse effect on signal quality, due to a
break-up of the coherent vortex structure along the span of a long bluff body. Non-uniform flow can be char-
acterized by a steepness parameter:
(1)
where dU/dz is the velocity gradient in the spanwise direction of the bluff body (z), and U
REF
is a refer-
ence velocity, usually measured at the midpoint of the bluff body span.
A u
RMS
U

D
U
REF
--------------
z d
dU
=
VOLUME 1, NUMBER 4, OCTOBER 1995 285
Griffin (1985) reviewed the subject of vortex shedding from bluff bodies in shear flow. A commonly
reported feature is the formation of cells along the span of the bluff body, in which the frequency is con-
stant for a particular cell. Griffin noted that cell length is typically limited to about 5 bluff body diameters,
and that the strongest cells form near the ends of a bluff body. In one experiment with a very long circular
cylinder placed in a linear shear flow (L/D = 48, = 0.015), no cells were observed, except those within 4
to 8 diameters from each end. Away from the ends, shedding frequency varied linearly with distance across
the bluff body. In this case, no averaging was achieved for flow metering.
Typically, vortex-shedding flow meters are constructed with a trapezoidal cross-section bluff body ori-
ented with the large face normal to the flow and with a diameter of about 1/4 to 1/3 that of the pipe or shroud
in which it is placed. Also, end walls are often used to terminate the bluff body. For the most part, these geo-
metrical considerations have been derived empirically (White et al. 1974).
Some work has been done to systematically optimize bluff body geometry. Igarashi (1985) studied six dif-
ferent bluff body geometries: circular, rectangular, triangular, semi-circular, trapezoidal, and circular with a
two-dimensional slit through the center. He found that the circular cylinder with a slit near the trailing edge
produced the strongest, most regular fluctuating pressure signals. He attributed the strong signal to commu-
nication between the two boundary layers through the slit.
Pankanin (1985) devised a method for optimizing bluff body geometry by defining a Quality Coeffi-
cient, based on the power spectral density of the vortex-shedding signal. Pankanin (1991) also performed
experiments aimed at optimizing the signal for a vortex-shedding flow meter based on a circular cylinder
with a slit. For a range of s/D (ratio of slit width to bluff body diameter) from 14 to 18%, he found that the
signal was most stable for s/D = 14%.
Popiel et al. (1992) modified the circular cylinder with a slit to produce a special-shaped cylinder, which
had a concave base in addition to the slit. They varied the slit width s and concave base radius r, in an attempt
to produce a bluff body geometry that minimized the variation in Strouhal number and maximized the signal-
to-noise ratio (SNR). They also placed a small (0.4D

0.4D) splitter plate at each end of the cylinder to
decouple the vortices from any end effects. The optimal geometry was for s/D = 0.09 and r/D = 2. With this
design, the Strouhal number varied by less than 1% for 300 < Re < 4.5

10
4
and the SNR was about 3.5
times greater than that for an equilateral triangle under the same flow conditions. They also showed that, to
maximize the SNR, the effect of the slit was more important than that of the concave base radius.
Davis et al. (1984) studied confinement effects of rectangular cylinders both numerically and experimen-
tally for blockage ratios (D/H) of 0, 1/6, and 1/4 (H is the height of the channel). They found that the Strouhal
number increased with increasing blockage ratio. Flow visualizations showed that moving recirculation
zones appeared between the wake and the confining walls.
Mukhopadhyay et al. (1992) numerically modeled flow past a square-section cylinder in a confined chan-
nel for Reynolds number ranging from 60 to 312. They found that the Strouhal number increased from 0.16
to 0.35 as the blockage ratio was increased from 0.125 to 0.375. They also showed that the channel walls had
a damping effect on the wake, so that at a certain distance downstream of the bluff body, no periodicity could
be detected in the wake.
Gerich and Eckelmann (1982) studied the effects of end plates and free ends on vortex shedding from a
circular cylinder with a laminar wake in the range 40 < Re < 250. When end plates were fixed to a long cyl-
inder, end cells were observed, with the shedding frequency in each end cell 10 to 15% lower than the fre-
quency at the center of the cylinder. The span of cylinder dominated by these end effects ranged from 6 to
15 bluff body diameters. When the distance between the two plates was less than 20 to 30 cylinder diameters
(i.e., the length of two adjacent end cells), a single shedding frequency was observed along the entire bluff
body; this frequency was less than the Strouhal frequency, but greater than the frequency of the end cells.
Finally, they showed that the presence of these two, closely-spaced end plates had a stabilizing effect on the
wake, increasing the critical Reynolds number for the onset of shedding from 40 to 60, and the Reynolds
number for laminar-turbulent transition in the wake from 150 to 250. This stabilizing effect is desirable for
flow metering, since it may help to decrease adverse effects of flow disturbances on the wake.
Objectives and Scope
Most of the previous research studies have been related to understanding the physics associated with
vortex shedding and have not adequately addressed issues related to flow meter design and application in
286 HVAC&R RESEARCH
disturbed environments. In fact, most studies have focused on circular cylinders, whereas sharp-edged
bluff bodies are used for flow metering. No references were found that report actual flow disturbances in
HVAC duct applications and evaluate their effects on flow meter design and performance. For the most
part, vortex-shedding flow meters have been designed using trial-and-error approaches.
The overall objective of this work is to evaluate the potential for using vortex-shedding flow meters in typ-
ical HVAC duct flow applications, to determine design and application guidelines, and to examine the accu-
racy of the devices under a wide range of flow conditions. Field data were acquired to define a realistic range
of conditions encountered in typical ducts. Detailed measurements of shedding frequency and signal quality
were then performed in a wind tunnel, which was used to simulate the disturbances present in the field in a
controlled manner. By systematically varying the type and level of flow disturbances, their independent and
combined effects on vortex shedding were documented. The flow velocity, turbulence intensity and length
scale, unsteadiness frequency and amplitude, and non-uniform velocity gradient were varied independently
over the range of interest. The wind tunnel results, combined with field data from HVAC ducts, were used
to formulate guidelines for the design and application of vortex-shedding flow meters for measuring air flow
in ducts.
Experimental Facility and Techniques
HVAC Duct Flow Measurements
Velocity measurements were performed in HVAC ducts to establish the nature of the disturbances
present. The results were used to define the range of parameters to be varied systematically in the wind tun-
nel. The hot-wire probe was calibrated in the wind tunnel by the method described in the upcoming section,
Measurement Techniques and Uncertainty.
Measurements were taken at two locations along the same duct, on a 3

5 grid of points (see Figure 2).
The characteristic width W of the duct is defined using the height and width of the rectangular cross-section:
(2)
The first location (section #1) was 3W downstream of a bend and dampers, and the second location (sec-
tion #2) was 8W downstream of a diffuser within a straight duct. The probe was traversed manually through
holes in the side of the duct. The two locations were chosen to provide two extremes in disturbance levels.
The flow immediately downstream of the bend and dampers was expected to be highly disturbed, whereas
the flow downstream of a long straight section of duct was expected to be less disturbed. While the measure-
ments were taken, the dampers were locked completely open, parallel to the flow.
Wind Tunnel Tests
Experiments were performed in an open-circuit, low-speed wind tunnel. Flow was conditioned upstream
of the test section by a turbulence management section consisting of a bell-mouth inlet, followed by a hon-
eycomb flow straightener, four screens, and a 5:1, two-dimensional contraction section based on a 5th order
polynomial design. The test section is 30 cm wide by 30 cm high by 91 cm long. Flow through the test section
has a mean uniformity of 2.4% and a free stream turbulence intensity of less than 0.7% over a velocity range
of 2 to 30 m/s.
Four different bluff bodies were used in this study: three triangular-section bluff bodies (an equilateral tri-
angle and two right triangles) and a square-section bluff body. The bluff bodies spanned the entire width of
the test section (30.5 cm), with the tunnel walls terminating each bluff body The geometric characteristics
of the bluff bodies are presented in Table 1. The bluff body diameter is given by D, aspect ratio (length over
diameter) by L/D, and tunnel blockage ratio (bluff body diameter over test section height) by D/H. Each of
the faces with dimension D shown in Table 1 were oriented normal to the direction of flow. All four bluff
bodies were used in the turbulence experiments, but only bluff body #1 was used in the unsteadiness exper-
iments.
Turbulence was generated with square-mesh biplanar grids of circular-section bars, with a solidity of
approximately 50%. Turbulence intensity at the bluff body was varied by changing the distance of the grid
relative to the bluff body x, and integral length scale L
x
was varied by using three grids of different mesh sizes
W 0.5 height width + ( ) =
VOLUME 1, NUMBER 4, OCTOBER 1995 287
(grid bar diameters of 3.175, 6.35, and 9.525 mm and grid bars separations of 12.7, 25.4, and 38.1 mm). By
varying the combinations of the three grids and the four sharp-edged bluff bodies, the ratio of integral length
scale to bluff body diameter, L
x
/D, was varied from 0.25 to 11.7.
Periodic unsteadiness was produced by a set of 5 continuously rotating shutters downstream of the test
section. As the shutters rotated, the variable blockage caused fluctuations in the static pressure, which in
turn drove the velocity fluctuations, generating a near-sinusoidal velocity variation in the test section. The
shutters were driven by a 1.5 hp (1 kW) dc motor that allowed continuous adjustment of the rotational
speed to produce a longitudinal velocity perturbation between 10 and 180 Hz. An EMF feedback motor
controller was used to keep the frequency constant to 1%. Perturbation amplitude was varied by changing
the amount of blockage (i.e., the shutter width). Four sets of shutters were used, producing maximum
blockages of 51%, 61%, 71%, and 82%; RMS perturbation amplitude of the velocity A ranged from 1 to
16%. The amplitude of the fluctuation was dependent on perturbation frequency and increased linearly
with mean flow speed.
A fine-mesh screen oriented at an angle to the approach flow was employed to generate a non-uniform
velocity profile. The main advantage of this method is that it does not introduce significant levels of free
stream turbulence (turbulence is generated, but it is of very small scale, so it decays in a very short down-
stream distance x). Also, the velocity profile is easily adjusted by changing the shape and/or angle of the
screen. The screen mesh size and solidity also affect the velocity profile, with smaller mesh, higher solidity
screens generating steeper velocity gradients. In this study, a 150 square mesh, 37.4% open area screen was
stretched across the test section at an angle to the z-axis (direction along the span of the bluff). The mean
velocity gradient dU/dz was varied by changing the angle of the screen. The steepness parameter was fur-
ther varied by changing bluff body diameter D.
This arrangement produced reasonably linear velocity profiles without generating high levels of turbu-
lence over most of the span (Wolochuk 1994). Through various combinations of and D, the steepness
parameter when averaged across the span was varied from 0.0198 to 0.0834.
Since pressure sensors are commonly used to detect the shedding frequency, it is important to understand
the effects of approach velocity and flow disturbances on fluctuating pressure signals. Two measures of sig-
nal quality are the fluctuating pressure amplitude p' and the signal-to-noise ratio SNR. Fluctuating pressure
amplitude is simply the standard deviation of a pressure-time signal. The signal-to-noise ratio is computed
from the power spectrum of a signal, and is defined as:
(3)
where S
xx
( f
vsf
) is the amplitude of the spectrum at the vortex-shedding frequency (signal strength), and
is the mean amplitude at all other frequencies (noise). In cases where there was low-frequency
noise at higher amplitudes than at the shedding frequency, the signal was conditioned with a high-pass
filter prior to computing the SNR.
Table 1. Bluff Body Characteristics
Bluff # Geometry D (cm) L/D D/H
1 5.08 6.0 16.7%
2 2.54 12.0 8.33%
3 1.49 20.5 4.88%
4 0.325 93.8 1.07%
D
SNR 20log
10
S
xx
f
vsf
( )
S
xx
f ( )
--------------------- dB
S
xx
f ( )
288 HVAC&R RESEARCH
Measurement Techniques and Uncertainty
Hot-wire anemometry was used to characterize the mean flow, as well as to measure velocity profiles and
spectra in the wake for all four bluff bodies. A low-pass, anti-aliasing filter conditioned the anemometer out-
put signal, which was then acquired by an A/D board on a personal computer. The digitized signal was ana-
lyzed using data acquisition software. Typically, 2048 points were acquired for each realization at a sampling
frequency of at least twice the maximum frequency of interest to meet the Nyquist criterion. The vortex-
shedding frequency was inferred from the dominant peak in the power spectral density of the streamwise
fluctuating velocity signal. Bluff body #1 was instrumented with static pressure taps on its three sides, so
shedding frequency could also be determined by analyzing the fluctuating pressure on this bluff body. The
approach velocity was monitored with a pitot-static probe connected to a differential pressure transducer, the
output of which was also acquired by the computer.
For U greater than 10 m/s, the uncertainty in the Strouhal number determined from wind tunnel measure-
ments is less than 3% (see Appendix A for an uncertainty analysis). The uncertainty in the Strouhal number
results mainly from the uncertainty in the measured approach velocity. At low velocities, the uncertainty in
the pitot-static probe measurements is very large, leading to large errors in the Strouhal number. Most of the
results presented in this study are for U > 10 m/s.
Experimental Results and Discussion
Duct Flow Characterization
Results. In order to define the range of conditions to be generated in the wind tunnel, hot-wire velocity
measurements were made in actual HVAC duct flows. Turbulence intensity and integral length scale,
unsteadiness amplitude and frequency, and non-uniform flow steepness parameter were all measured at
two locations in a typical HVAC duct (see Figure 2). For calculations of the steepness parameter a value
of 1.27 cm (0.5 in.) was assumed for the bluff body diameter, since this is a typical bluff body size for flow
meters.
Mean velocity profiles at the two duct locations are illustrated in Figures 3a and 3b. Positions at which
measurements were made are indicated with stars (*), while the meshes represent interpolations of these
data. At the location downstream of the bend (section #1), the velocity is highly non-uniform, with a stan-
Figure 2. Duct flow measurement locations.
VOLUME 1, NUMBER 4, OCTOBER 1995 289
dard deviation of 30% from the mean over the cross-section. For this velocity profile, the shear flow
steepness parameter ranges from 0.0016 to 0.1074. The velocity is highest in the center, and lowest near
the side walls of the duct. The velocity profile is nearly two-dimensional, but it is not possible to determine
its precise shape, due to the coarseness of the measurement grid (3 5). This profile shape may result from
secondary flows caused by the bend, although the open dampers and the frame may also have an effect. It
would be difficult to measure the average flow rate in a duct with such a highly non-uniform velocity pro-
file. In the long straight section of duct (section #2), the profile is much more uniform, with a standard
deviation of 4.7% of the mean velocity over the cross-section. For this profile, the mean value of is
0.0126.
Turbulence intensity profiles for the duct measurements are illustrated in Figures 4a and 4b. Again, the
profile in section #1 is highly non-uniform. Turbulence intensity ranges from 7% to 47%, with the higher lev-
els near the sidewalls of the duct. This highly non-uniform distribution of Tu may be due to secondary flows
caused by the bend. In the long straight section, Tu ranges from 6.7% to 10.2%, with a spatially-averaged
mean value of 7.6%. Thus, the turbulence level is significant, even after a long straight section of duct.
A time trace of the velocity in the center of section #2 is presented in Figure 5, along with its power density
spectrum and autocorrelation function. Also shown in the figure are values of the time-averaged mean veloc-
ity, the turbulence intensity, and the turbulence integral length scale. Some periodicity is evident in the time
trace and autocorrelation. The total disturbance level at this location (9.1%) results from the combination of
turbulence and unsteadiness, even though it is presented in Figure 5 as Tu. For complete flow character-
ization, it would be necessary to deduce the periodic and the random velocity components from the
total signal. In this way, the turbulence intensity, Tu, and unsteadiness amplitude A could be calculated inde-
pendently for flows containing both turbulence and periodic unsteadiness.
From the spectrum in Figure 5, it is obvious that most of the energy in the flow is contained at low fre-
quencies, below 50 Hz. A turbulence integral length scale of 22 cm (8.7 in.) was calculated from the auto-
correlation zero-crossing and mean convection velocity. The length scale is close to the width of the duct
(25 cm), and is fairly typical of all the data recorded. These results indicate that turbulence length scale is
on the order of the duct width. Similar plots are shown in Figure 6 for data taken at the center of section #1.
At this location, the disturbance level is 10.7%. Again, the length scale, 23 cm (9.2 in.), is of the order of
the duct width, and most of the energy in the spectrum is at very low frequencies. In order to increase the
resolution at low frequencies, the sampling rate was decreased from 1000 to 150 Hz. Thus, the frequency
resolution was improved from 0.49 Hz to 0.073 Hz. Results for the hot-wire probe positioned off-center
in the duct showed that most of the energy in the spectrum occurred at frequencies below 20 Hz, and there
was a distinct peak at 2.7 Hz.
Discussion. The results presented above can be used to form a reasonable representation of duct flow,
although more field data are necessary to depict the wide range of flow conditions that exist in ducts. Based
on these results, duct flow is expected to contain high-intensity, large-scale turbulence. Turbulence length
scale is on the order of the duct size, which is very large, relative to typical flow meter dimensions. Also,
the flow contains low-frequency, periodic unsteadiness. Finally, the velocity and turbulence intensity pro-
files can be quite non-uniform, and are strongly dependent on location in the duct.
From these limited field data, it is clear that the three flow disturbances discussed in this work (turbulence,
unsteadiness, and non-uniform flow) are present in duct flow. Additional field data should include all of the
above measurements (Tu, L
x
, U and Tu profiles, and spectra) for a wide range of duct locations and operating
conditions. Also important is the evolution of each disturbance as a function of upstream or downstream dis-
tance from the disturbance generator (e.g., bend, dampers, fan). This information is necessary in order to
define application guidelines for locating the flow meters relative to disturbance generators.
Wind tunnel experiments were performed to systematically model the types of disturbances found to exist
in ducts: turbulence of various length scales and intensities, periodic unsteadiness, and non-uniform velocity
profiles.
Low-Turbulence Wind Tunnel Test Results
In order to establish a basis for comparison, approach velocity and shedding frequency were measured
for each bluff body over a velocity range of 2 to 25 m/s in the absence of grid-generated turbulence or peri-
odic unsteadiness. Since the working fluid was air, and bluff body diameter ranged from 0.325 to 5.08 cm,
u ( ) u' ( )
290 HVAC&R RESEARCH
Figure 3a. Velocity profile 3W downstream of a bend (section #1)
Measurement points are indicated with stars (*).
Figure 3b. Velocity profile downstream of 8W of straight duct (section #2)
Measurement points are indicated with stars (*).
VOLUME 1, NUMBER 4, OCTOBER 1995 291
Figure 4a. Turbulence intensity profile 3W downstream of a bend (section #1)
Measurement points are indicated with stars (*).
Figure 4b. Turbulence intensity downstream of 8W of straight duct (section #2)
Measurement points are indicated with stars (*).
292 HVAC&R RESEARCH
the Reynolds number range was 480 < Re < 83,000. The background turbulence level was below 0.7%, and
the mean flow uniformity was within 2.4%. For a given bluff body, the Strouhal number was fairly con-
stant over the entire Reynolds number range tested in the absence of controlled disturbances. Wolochuk
(1994) reported the following values of the Strouhal number for undisturbed flows: 0.127, 0.124, 0.131,
and 0.111 for bluff bodies #1 through #4, respectively.
Free Stream Turbulence Effects
Results. The effect of elevated turbulence intensity on the Strouhal number is shown in Figure 7 for L
x
/
D = 0.5. After an initial drop of 15% in the Strouhal number at Tu of 2.5%, further increasing the turbu-
lence intensity to 10% results in only a slight (2.4%) increase in the Strouhal number. The trend of a
slightly increasing Strouhal number with increasing turbulence intensity is consistent with the results of
Malard et al. (1991). Since an intensity of 2.5% was the lowest that could be generated with grids in the
wind tunnel (due to the short test section), it is not known whether St drops suddenly at some threshold
value of Tu (lower than 2.5%), or if the decrease is continuous at very low intensities. Even though the
background turbulence level, 0.7%, was lower than 2.5%, the length scale of the turbulence was not the
same with and without the turbulence generating grid in place. It is possible that the decrease in Strouhal
number could occur at lower turbulence intensities than 0.7% for L
x
/D = 0.5. The behavior of St with vary-
ing Tu should be investigated in the low-turbulence range (< 2.5%) in future work.
Figure 5. Time trace, spectrum, and autocorrelation for point
in center of section #2
VOLUME 1, NUMBER 4, OCTOBER 1995 293
Figure 8 is a plot of percent decrease in the Strouhal number (as compared to a low-turbulence base case)
versus length scale of the incoming turbulence (L
x
/D). Data from bluff bodies #1 through #4 were used to
generate this plot, so that a wide range of L
x
/D could be covered. The base case for each bluff body is rep-
resented by the point (0,0), and Tu = 10% for all turbulence cases. For L
x
/D < 3, the Strouhal number
decreases with increasing length scale. Increasing L
x
/D beyond 3 causes the Strouhal number to increase
toward its undisturbed value. Nakamura and Ohya (1984) found that the relationship between mean base
pressure and turbulence length scale exhibited the same trend, with a minimum base pressure occurring for
L
x
/D 3.
In order to evaluate the effect of turbulence on sensor requirements, time traces of pressure were recorded
from a transducer connected to a static pressure tap on the back of bluff body #1 and were processed to obtain
the power spectral density of the pressure signals. The values of the SNR and fluctuating pressure amplitude
for the three cases tested are tabulated in Table 2.
Table 2. Pressure Tap Signal Parameters (Tap #14, U = 5.1 m/s)
L
x
/D Tu p (Pa) p (psi) SNR (dB)
No Grid 0.7% 3.9 4.0 10
4
48
0.25 10% 2.4 3.8 10
4
46
0.75 10% 1.4 3.4 10
4
39
Figure 6. Time trace, spectrum, and autocorrelation for point
in center of section #1
294 HVAC&R RESEARCH
For the base case, the time trace was highly periodic, and its spectrum contains a large peak at the shedding
frequency, giving a SNR of 48 dB. For L
x
/D = 0.25 and Tu = 10%, there is a slight decrease in the fluctuating
pressure amplitude (from 3.9 to 2.4 Pa), but its periodic nature was still apparent. The decrease in amplitude
is accompanied by a slight decrease in the SNR (from 48 to 46 dB). When L
x
/D was increased to 0.75, the
fluctuating pressure amplitude decreased further, to 1.4 Pa, and the SNR decreased to 39 dB. This case rep-
resents a 64% decrease in fluctuating pressure amplitude, and a 9 dB decrease in the SNR, compared to the
low-turbulence base case. Thus, the presence of turbulence in the free stream has a significant effect on the
quality of a fluctuating pressure signal. The trend of a decreasing SNR with increasing L
x
/D is consistent with
the results of Nakamura and Ohya (1984).
Figure 7. Strouhal number versus Reynolds number for
various turbulence intensities (L
x
/D = 0.5)
Figure 8. Decrease in Strouhal number vs. turbulence length scale (Tu = 10%)
VOLUME 1, NUMBER 4, OCTOBER 1995 295
The results summarized in Table 2 are consistent with those shown in Figure 8. As L
x
/D approaches 3,
the turbulence exerts a greater influence on the near-wake, resulting in a decrease of the Strouhal number,
and a decrease of the spectral power at the shedding frequency.
Implications for Flow Meter Design and Application. The turbulence results have important implica-
tions for the design and application of vortex-shedding flow meters. The presence of turbulence does not
inhibit vortex shedding, nor does it destroy the linear relationship between shedding frequency and
approach velocity. However, turbulence can affect meter calibration and signal quality. Therefore, vortex-
shedding flow meters should be calibrated in flows with large-scale free stream turbulence, with Tu above
a threshold value (between 0 and 2.5%). This may seem contrary to conventional calibration practices,
since instruments are normally calibrated in ideal environments (i.e., laminar flow). However, since the
presence of turbulence shifts the calibration curve, metering errors would result when the sensor is placed in
turbulent duct flow. The results from the Duct Flow Characterization section indicate that turbulence inten-
sity levels in ducts are quite high, so it is not expected that Tu would be lower than the threshold value. For
turbulence intensities between 2.5% and 10%, meter calibration is fairly independent of Tu, so it is not nec-
essary to suppress the turbulence. However, it is necessary either to know or control the turbulence integral
length scale at the bluff body.
Several methods can be used to control the turbulence length scale. A shroud around the bluff body,
which is already a common feature of vortex-shedding flow meters, has the effect of shielding the near-
wake from large-scale turbulence. This is similar to the effect of honeycomb flow straighteners which con-
strain the lateral components of the fluctuating velocity. If turbulence length scales in the duct flow L
x
are
always greater than the shroud size H, then the bluff body will always be exposed to the smaller length scale
H. For L
x
> H, calibration is independent of L
x
, and no further design modifications are necessary. However,
if length scales are smaller than the shroud (L
x
< H), then the calibration is dependent on L
x
, and metering
errors are likely.
Another method of controlling length scale at the bluff body is to mount small screens, grids, or flow
straighteners upstream of the bluff body. However, the mesh can become clogged with dust and debris, so
it may require periodic cleaning. Also, screens and grids can cause considerable pressure drop, increasing
required fan power.
These experiments have shown that the presence of turbulence decreases the SNR. Thus, flow meter
geometry, sensing, and electronics must all be designed to optimize meter performance in turbulent flows.
A flow meter designed and tested for laminar flows may not work well in flows where turbulence is present.
It should also be noted that the fluctuating pressure amplitude decreases with decreasing flow velocity and
bluff body diameter. Thus, the bluff body should be sized so that it is large enough to generate a strong sig-
nal for the entire range of flow speeds to be measured under turbulent flow conditions.
Periodic Unsteadiness
Results. A constant Strouhal number implies that the relationship between shedding frequency and mean
velocity is linear. However, the presence of periodic unsteadiness can cause the shedding frequency to lock-
on to the forcing frequency, disturbing the linear relationship as illustrated in Figure 9. For these results, the
free stream was oscillated at a frequency f
p
of 40 Hz, and an RMS amplitude A of 5% of the mean velocity.
Vortex shedding locked on to 20 Hz (half the perturbation frequency) for a mean velocity range of 7.4 to 9.2
m/s. The locked frequency, 20 Hz, corresponds to a velocity of 8.0 m/s for unforced shedding. Therefore, the
meter output indicates 8.0 m/s, while the actual flow speed is between 7.4 and 9.2 m/s. As indicated in the
figure, this could result in metering errors up to 14% for the case shown.
In addition to lock-on, there is a transition range in which the forced shedding frequency is neither the nat-
ural shedding frequency nor half the perturbation frequency. This is most apparent for an approach velocity
between 9.2 and 10.0 m/s. In this range, the shedding frequency rises gradually from the locked-frequency
(20 Hz) to the unforced value f
so
. The presence of a transition range between lock-on and no lock-on is con-
sistent with the findings of Hall and Griffin (1993) and Barbi et al. (1986). Metering errors also occur in this
range, although they are not as great as in the lock-on range.
The free stream was oscillated at a number of forcing frequencies in order to determine which harmonics
and subharmonics of the Strouhal frequency would cause lock-on. The results are tabulated in Table 3. No
lock-on was observed for forcing at half the Strouhal frequency (f
p
0.5f
so
), which is also consistent with the
296 HVAC&R RESEARCH
findings of Malard et al. (1991). This result is significant in that it implies that lock-on will not occur for
(for A up to at least 16%), since lock-on in this frequency range would most likely occur for forcing
at the first subharmonic of f
so
.
It was found that as disturbance amplitude increases, the range of lock-on increases. Thus, greater meter-
ing errors are likely as the amplitude of the periodic disturbance increases. The extent of the lock-on range
is not symmetric about the forcing frequency; consistent with the findings of Blevins (1985) and Hall and
Griffin (1993), greater shifts in shedding frequency were obtained on the low-frequency side.
Implications for Flow Meter Design and Application. For flow meter design, an important result is
that no lock-on occurs for subharmonic forcing, i.e., for f
p
f
so
(for A up to at least 16%). Therefore, lock-on
is easily avoided by sizing the bluff body small enough so that the range of its shedding frequencies is
always greater than any disturbance frequencies in the flow. For most duct locations, disturbances are typi-
cally of low frequencies (see Duct Flow Characteristics), so this criterion is easily met. The potential for
lock-on to occur may be greater near a fan, however, where disturbances may have higher frequencies and
amplitudes. Thus, it is important to develop application guidelines so that the flow meter is not installed in
locations where lock-on is a potential problem.
The influence of flow meter geometry on lock-on was not investigated in this work. As indicated in the
literature, lock-on is dependent on bluff body geometry, with the extent of lock-on roughly twice as great for
a circular cylinder as for sharp-edged bluff bodies. The special-shaped cylinder of Popiel et al. (1992) may
be even less susceptible to lock-on than the triangular wedge used in this study. Also, the presence of a shroud
around the bluff body may have an effect.
Table 3. Lock-On Results Summary
f
p
/f
so
A Lock-on?
0.5 16% no
1 3% yes
2 1-16% yes
3 4.5% no
4 2.5% yes
8 3% no
Figure 9. Shedding frequency vs. approach velocity in unsteady flow for
forcing at twice the Strouhal frequency ( f
p
= 40 Hz, A = 5%)
f
p
f
so

VOLUME 1, NUMBER 4, OCTOBER 1995 297


Combined Turbulence and Unsteadiness
In order to examine the combined effects of turbulence and unsteadiness, the free stream was oscillated
at a frequency near twice the Strouhal frequency ( f
p
2f
so
), and a turbulence grid was placed upstream of
the bluff body. The unsteadiness conditions ( f
p
, A) were the same as for the case presented in Figure 9, while
the free stream turbulence characteristics were varied. It was reported by Wolochuk et al. (1995) that the
effects of the unsteadiness and turbulence are additive: turbulence shifts the Strouhal frequency, and
unsteadiness causes lock-on about the shifted Strouhal frequency. When the results are normalized to remove
the turbulence effect, the lock-on range is the same for all length scale cases.
Non-Uniform Flow
Results. To study the effects of non-uniform approach flow, bluff body #2 was placed in the test section
with the screen installed at an angle of = 25, generating a shear flow with a steepness parameter of
0.0391. A plot of shedding frequency versus spanwise position is presented in Figure 10. Two shedding cells
are present, a low-frequency cell at 44.9 Hz for 1.0 z/D 6.0, and a high-frequency cell at 58.5 Hz for 6.5
z/D 11.0. In the high-frequency cell, the Strouhal number is centered about 0.133, and in the low-fre-
quency cell it is centered about 0.122.
The data represented in Figure 10 were analyzed to determine whether an average flow velocity could be
calculated from the constant shedding frequency in each cell. The mean velocity over each spanwise region
was determined from the velocity profile measured using the hot-wire anemometer. The mean velocity was
also estimated using the Strouhal number of 0.136 from the uniform-flow case and the constant shedding
frequency f of each cell:
(4)
The actual and estimated mean velocities are compared in Table 4.
For cell #2, there is good agreement (2.4% difference) between the actual and calculated mean velocity
over the length of the cell. In this cell, the local Strouhal number (0.133) is only 2.2% lower than the uniform-
flow Strouhal number (0.136). The higher discrepancy of the calculated velocity for cell #1 (10.2%) results
Figure 10. Shedding frequency vs. spanwise location for bluff body #2
screen at = 25 ( = 0.0391)
U
calc
f D
St
------ =
298 HVAC&R RESEARCH
from the lower local Strouhal number (0.122), which is 10.3% lower than the uniform-flow Strouhal number.
Thus, accurate calculation of the average flow rate is obtained only when the local Strouhal number is close
to the calibration Strouhal number.
The presence of more than one shedding cell along the span of a bluff body has the effect of decreasing
the signal-to-noise ratio (SNR). Spectra of hot-wire velocity signals were recorded for the conditions plotted
in Figure 10, with the hot-wire located just behind and above the bluff body trailing edge. The values of the
SNR computed from the spectra are compared to a uniform-flow case in Table 5. It is apparent from the table
that when the number of cells spanning the bluff body increases from one to two, the SNR decreases. Thus,
a single shedding cell across the entire span of the bluff body leads to the highest SNR, and is therefore the
optimum case for a flow meter.
To study the effect of a higher velocity gradient, the screen angle was increased to = 52, generating a
of 0.0834 (roughly twice that of the first case). A plot of local shedding frequency is presented in Figure
11. In this case, three regions are apparent along the bluff body span. At the low-velocity end of the bluff
body (1.0 z/D 4.5), there is a single-frequency shedding cell. In this region, the local Strouhal number is
centered about 0.091. This low value for the Strouhal number may be due to the high turbulence level in
this region. In the center of the bluff body (5.0 z/D 6.5), there is a region where the shedding frequency
increases somewhat linearly with z/D. This is reflected by the almost constant local Strouhal number in this
Table 4. Comparison of Actual and Calculated Mean Velocity
Cell # Cell Length (z/D) f (Hz) Discrepancy
1 5.0 44.9 9.34 8.39 10.2%
2 4.5 58.5 11.20 10.93 2.4%
Table 5. Dependence of SNR on
Number of Cells Spanning Bluff Body

Number of Cells
Spanning Bluff
SNR (dB) at
Center of Each Cell
0 1 35
0.0391 2 31 (cell #1)
21 (cell #2)
U m/s ( ) U
calc
m/s ( )
Figure 11. Shedding frequency vs. spanwise location for bluff body #2
screen at = 50 ( = 0.0834)
VOLUME 1, NUMBER 4, OCTOBER 1995 299
region. Finally, at the high-velocity end of the bluff body (7.0 z/D 11.0), no dominant frequencies were
detected in the spectrum. The data points in this region represent the spectral peaks located by the software;
however, peaks were hardly discernible from the noise in the spectrum.
Spectra of hot-wire signals recorded behind the center (z/D = 6.0) of bluff body #2 are presented in Figure
12 for the three different velocity gradients: uniform flow ( 0), = 0.0391, and = 0.0834. The trend is
for the SNR to decrease with increasing (see Table 6). For the highest case, the peak at the shedding fre-
quency is very low and broad, barely discernible from the noise in the spectrum.
The decrease in the spectral power at the shedding frequency for high values of may be due to the pres-
ence of streamwise and transverse vorticity in the near-wake. Transverse vorticity is generated by the span-
wise velocity gradient of the approach flow dU/dz. Streamwise vorticity is present at the junction between
two shedding cells (Maull and Young, 1973, Williamson, 1989), and is caused by the turning of the (span-
Figure 12. Spectra of center-span hot-wire signals for bluff body #2
for varying values of (U
ref
= 10.1 m/s)
300 HVAC&R RESEARCH
wise) vortex filaments toward the streamwise direction. As increases, the oblique shedding angle increases,
thus increasing the streamwise vorticity. Maull and Young (1973) noted that a high amount of streamwise
vorticity in the wake has been shown to completely suppress vortex shedding.
Bluff body #3 was also tested in the shear flow, to examine the effect of aspect ratio on spanwise shedding
frequency. The aspect ratio of this bluff body is roughly twice that of bluff body #2. The local shedding fre-
quency is plotted versus z/D for bluff body #3 with = 52 in Figure 13. The shedding frequency increases
continuously along the span of the bluff body, and no single-frequency cells are present. The local Strouhal
number is centered about a value of 0.125 over the entire span of the bluff body. This bluff body has the same
shape as bluff body #2 (triangular), and is about the same for this case (0.0488) as for the results shown in
Figure 10 (0.0391). Thus, it appears that a bluff body with a smaller aspect ratio (12.0 versus 20.5) is more
likely to generate single-frequency cells than a more slender bluff body. This result is consistent with the
observation by Gerich and Eckelmann (1982) that two, closely-spaced end plates had a stabilizing effect on
the wake. Closely-spaced end plates may have the effect of limiting the streamwise and transverse vorticity,
in the same manner that honeycomb flow straighteners suppress lateral components of velocity.
Implications for Flow Meter Design and Application. The results of this section indicate that averag-
ing of a flow non-uniformity is only possible to a limited extent using a single bluff body that spans the flow.
A single shedding cell breaks down with increasing velocity gradient, as well as with increasing bluff body
aspect ratio, L/D. Also, the local Strouhal number may vary along the span of a very long bluff body. When
more than one shedding cell is present, a number of sensors are required, and the SNR is decreased.
Another problem with using a long bluff body is that the spanwise dimension of the shroud is large, so the
turbulent velocity fluctuations are not constrained in this direction. Therefore, the shroud is less effective in
shielding the near-wake from large-scale turbulence.
Table 6. Comparison of SNR for Different Values of
Screen Angle () SNR (dB)
No screen 0 35
25 0.0391 21
50 0.0834 11
Figure 13. Shedding frequency versus spanwise location for bluff body #3
screen at = 52 ( = 0.0488)
VOLUME 1, NUMBER 4, OCTOBER 1995 301
The flow meter should be designed and applied so that only one constant-frequency shedding cell is
present along the span of the bluff body. This would minimize the 3-dimensional effects and ensure the most
stable signal. The steeper the velocity gradient, the shorter the bluff body must be to maintain a single shed-
ding cell. For example, Williamson (1989) has quoted a maximum aspect ratio of about 28 for a single shed-
ding cell to occur in uniform flow. As the velocity gradient increases, this maximum allowable L/D
decreases. The optimum bluff body length would probably be close to the maximum cell length, measured
in this study to be about 5D. Since the bluff body diameter is likely to be very small relative to the dimensions
of the duct, an array of bluff bodies will be required to measure the average flow rate. A single long bluff
body separated into short segments with end plates could also be used.
Further Discussion and Conclusions
The effects of typical flow disturbances on the performance of vortex-shedding flow metering in HVAC
applications were investigated. Field data were acquired from a typical HVAC duct, and controlled wind tun-
nel experiments were performed to document the effects of turbulence, periodic unsteadiness, and non-uni-
form flow on vortex shedding and signal strength from typical sensors. The overall conclusion of this study
is that vortex-shedding flow meters can be used in HVAC ducts, but care must be taken in their design and
application to ensure accuracy under the wide range of flow conditions that exists in this application.
From hot-wire velocity measurements, it was found that duct flow was highly non-uniform (30% devi-
ation from the mean, steepness parameter up to 0.1074) and turbulent (up to 47% turbulence intensities) just
downstream (3W) of a bend. Accurate measurements of average flow rate would be difficult to obtain in a
flow with such a highly non-uniform velocity profile. Downstream of 8W of straight duct, the velocity was
much more uniform (4.7% deviation from the mean, steepness parameter of 0.0126), but the turbulence
intensity was still quite high (ranging from 6.7 to 10.2%). The integral length scale of the turbulence, typi-
cally of the order of the duct width, was very large relative to the dimensions of a vortex-shedding flow
meter. Low-frequency (2.7 Hz) unsteadiness was also present in the flow, with most of the energy in the spec-
trum occurring at frequencies below 20 Hz, These frequencies are very low compared to the typical shedding
frequencies for vortex-shedding flow meters.
Wind tunnel tests showed that turbulence at intensities up to 10% did not inhibit vortex shedding, nor did
it destroy the linear relationship between shedding frequency and mean velocity. Furthermore, it was found
that the Strouhal number (and therefore meter calibration) is fairly insensitive to changes in turbulence inten-
sity in the range 2.5% Tu 10% (2.4% increase in St over the range). Therefore, it is not necessary to sup-
press the turbulence for Tu < 10%. However, there was a threshold turbulence level below 2.5% at which a
dramatic drop in St occurred. Since the turbulence intensities in ducts are expected to be above the threshold
value, vortex-shedding flow meters should be calibrated in turbulent flows similar to what would be expe-
rienced in the field.
It was found that the Strouhal number exhibited a strong dependence on the integral length scale of the tur-
bulence (26% drop in St for L
x
/D 3). Therefore, turbulence length scale must be controlled at the bluff body
location so that the bluff body is always exposed to the same length scale, unless it is known a priori. This
can be achieved with a screen, grid, or honeycomb flow-straightener upstream of the bluff body. If turbu-
lence length scale is always much greater than the flow meter dimensions, then using a shroud around the
bluff body may filter the larger scales. With a shroud in place, the length scale at the bluff body is always
of the order of the shroud size, and meter calibration is constant.
Turbulence was found to decrease the amplitude and signal-to-noise ratio (SNR) of a fluctuating pressure
signal, compared to a low-turbulence base case (64% decrease in amplitude, 9 dB decrease in SNR). There-
fore, flow meter and sensor design should be optimized for turbulent flows.
Periodic unsteadiness was found to cause lock-on when forcing at the various harmonics of the shedding
frequency ( f
p
f
so
, 2 f
so
, 4 f
so
). For forcing at twice the Strouhal frequency ( f
p
2f
so
), the range of lock-on
increased with increasing perturbation amplitude. Thus, higher metering errors are likely as the amplitude of
a periodic disturbance increases. For an amplitude of 16%, the shedding frequency was forced to decrease
by up to 37.5% and to increase by up to 15% from its undisturbed value.
No lock-on occurred for subharmonic forcing ( f
p
0.5f
so
) for amplitudes up to 16%, implying that lock-
on does not occur for forcing frequencies in the range f
p
f
so
. Since unsteadiness in ducts is of very low fre-
quency, lock-on is easily avoided by using a small enough bluff body so that the range of shedding fre-
302 HVAC&R RESEARCH
quencies is always much greater than the unsteadiness frequencies in the flow. However, there is a lower
limit to bluff body size, since the amplitude of the fluctuating pressure signal decreases with bluff body
diameter.
When turbulence and unsteadiness occurred simultaneously, their individual effects were additive. The
turbulence caused a decrease in the Strouhal frequency, and lock-on then occurred around this shifted fre-
quency. Thus, the net effect of a given set of turbulence and unsteadiness conditions can be predicted by com-
bining Strouhal data from independent turbulence and unsteadiness experiments. This result is significant,
since duct flow contains both types of disturbances to varying degrees.
In an effort to examine whether vortex-shedding flow meters could be used to measure average flow rate
in ducts, a series of tests were run in which a long bluff body was spanned across a non-uniform approach
flow. It was shown that some degree of averaging is possible. In limited regions along the span, a single shed-
ding frequency occurs, corresponding to the average approach velocity in that region. However, the single
shedding cell breaks down with increasing velocity gradient, as well as increasing bluff body aspect ratio L/
D. Also, variation of the local Strouhal number along the span caused errors (up to 10.2%) in the calculated
average flow rate. Finally, the signal-to-noise ratio decreased with an increase in the number of cells span-
ning the bluff body, and with increasing velocity gradient. Thus, the optimum case for flow metering is a sin-
gle shedding cell spanning the entire bluff body.
Bluff body aspect ratio L/D should be kept as small as possible to ensure a single, strong shedding cell over
the entire bluff body span, even in non-uniform flows. For the velocity gradients generated in this study (
up to 0.0834), this would be accomplished with an aspect ratio of 5. It may also help to isolate the central por-
tion of the bluff body from end effects to ensure regular, parallel shedding. This can be accomplished through
the use of splitter plates at the ends, a step increase in D near the ends, or angled end plates.
To measure the average flow rate, an array of short bluff bodies is needed. One possible setup is illustrated
in Figure 14. A very long, shrouded bluff body is used, but separated with partitioning walls to give short sec-
tions. Each bluff body segment contains one sensor, since it is assumed that there is only one shedding cell
per section. A number of sensors are required with this scheme, but it may be possible to connect each pres-
sure tap to a single plenum, so that only a single transducer is necessary to measure the average pressure. The
number of partitions and sensors required would be dictated by the extent of the velocity non-uniformity in
the flow.
In closing, vortex-shedding flow meters are a promising alternative for measuring air flow in HVAC
ducts. When properly designed and applied, they can offer high accuracy even at low flow velocities and
independent of flow temperature, density, and humidity. In addition, it should be possible to produce reliable
vortex-flow meters at a low cost, since they are extremely simple and have no moving parts.
Nomenclature
Figure 14. Possible vortex-shedding meter design for measuring
average flow rate
VOLUME 1, NUMBER 4, OCTOBER 1995 303
A RMS perturbation amplitude
(A =
~
u
RMS
/U )
D bluff body diameter (width, see Figure 1 or Table 1)
D/H blockage ratio
f Frequency of vortex-shedding
f
p
perturbation frequency
f
so
natural (Strouhal) shedding frequency
H channel height or shroud width
L bluff body spanwise length
L
x
streamwise turbulence integral length scale
L/D bluff body aspect ratio
P static pressure
P differential pressure, P = P
0
P
RMS root-mean-square:
Re Reynolds number, Re = UD/
SNR signal-to-noise ratio
St Strouhal number, St = f D/U
Tu turbulence intensity, Tu = u
RMS
/U
U mean free stream velocity
U
REF
reference mean velocity
u total streamwise velocity

u random velocity fluctuation in the streamwise direction
u
RMS
root-mean-square random streamwise velocity fluctuation
~
u periodic velocity fluctuation in the streamwise direction
u
~
RMS
root-mean-square periodic streamwise velocity fluctuation
W rectangular-section duct characteristic width, W = 0.5(height + width)
(x,y,z) Cartesian coordinates
z/D ratio of shedding cell length to bluff body diameter
shear flow steepness parameter,
angle between screen and z-axis
density of a fluid
( )
RMS

( )
2
=
u U u' u + + = ( )
304 HVAC&R RESEARCH
References
ASME. 1987. Measurement of Fluid in Pipes Using Vortex Flow Meters. ANSI/ASME MFC-6M. New York: American
Society of Mechanical Engineers.
Barbi, C., D.P. Favier, C.A. Maresca, and D.P. Telionis. 1986. Vortex Shedding and Lock-on of a Circular Cylinder in
Oscillatory Flow. Journal of Fluid Mechanics 170:527-544.
Bearman, P.W., and T. Morel. 1983. Effect of Freestream Turbulence on the Flow Around Bluff Bodies. Progress in
Aeronautical Science 20:97-123.
Blevins, R.D. 1985. The Effect of Sound on Vortex Shedding from Cylinders. Journal of Fluid Mechanics 161:217-237.
Davis, R.W., E.F. Moore, and L.P. Purtell. 1984. A Numerical-Experimental Study of Confined Flow Around Rectangu-
lar Cylinders. Physics of Fluids 27(1):46-59.
Drees, K.H., J.D. Wenger, and G. Janu. 1992. Ventilation Air Flow Measurement for ASHRAE Standard 62-1989.
ASHRAE Journal 34(10):40-45.
Gerich, D., and H. Eckelmann. 1982. Influence of End Plates and Free Ends on the Shedding Frequency of Circular Cyl-
inders. Journal of Fluid Mechanics 122:09-121.
Griffin, O.M. 1985. Vortex Shedding from Bluff Bodies in a Shear Flow: A Review. Transactions of the ASMEJour-
nal of Fluids Engineering 107:298-306.
Griffin, O.M., and M.S. Hall. 1991. ReviewVortex-Shedding Lock-on and Flow Control in bluff body wakes. Journal
of Fluids Engineering, 113:526-537.
Hall, M.S., and O.M. Griffin. 1993. Vortex Shedding and Lock-on in a Perturbed Flow. Journal of Fluids Engineering
115:283-291.
Igarashi, T. 1985. Fluid Flow Around a Bluff Body Used for a Krmn Vortex Flowmeter. In Proceedings, Fluid Con-
trol and Measurement Conference. Tokyo. 1003-1008.
Kline, S.J., and F.A. McClintock. 1953. Describing Uncertainty in Single-Sample Experiments. Mechanical Engineer-
ing 73:3-8.
Malard, L., W. Wisnoe, A. Strzelecki, P. Gajan, and P. Hebrard. 1991. Air Visualizations and Flow Measurements
Applied to the Study of a Vortex Flowmeter: Influence of Grid Turbulence and Acoustical Effects. In Proceedings,
Flucome '91, ASME 1991: 689-704.
Maull, D.J., and R.A. Young. 1973. Vortex Shedding from Bluff Bodies in a Shear Flow. Journal of Fluid Mechanics
60:401-409.
Mukhopadhyay, A., G. Biswas, and T. Sundararajan. 1992. Numerical Investigation of Confined Wakes Behind a
Square Cylinder in a Channel. International Journal for Numerical Methods in Fluids 14:1473-1484.
Nakamura, Y., and Y. Ohya. 1984. The Effects of Turbulence on the Mean Flow Past Two-Dimensional Rectangular
Cylinders. Journal of Fluid Mechanics 149:255-273.
Pankanin, G.L. 1985. A New Approach to Bluff Body Design in Vortex Flow Meters. Fluid Control and Measurement
(Conference). Tokyo.
Pankanin, G.L. 1991. Investigation of Vortex Signal Stability as a Function of Vortex Meter Configuration. Flucome
'91, ASME 1991: 455-459.
Popiel, C.O., D.I. Robinson, and J.T. Turner. 1992. Vortex Shedding from Specially Shaped Cylinders. 11th Australian
Fluids Mechanics Conference, University of Tasmania, Hobart. 503-506.
Roshko, A. 1953. On the Development of Turbulent Wakes from Vortex Streets. NACA Technical Note 2913.
Roshko, A. 1954. On the Drag and Shedding Frequency of Two-Dimensional Bluff Bodies. NACA Technical Note 3169.
Roshko A. 1955. On the Wake and Drag of Bluff Bodies. Journal of Aeronautical Sciences 22:124-132.
Strouhal, V. 1878. ber eine besondere art der tonerregung. Ann. Phys. und Chemie, Neue Folge, Bd. 5, Heft 10:216-
251.
Von Karman, Th., and H. Rubach. 1912. ber den mechanismus des flussigkeits und luftwiderstandes. Phys. Zeitschr.,
Bd. 13, Heft 2:49-59.
White, D.F., A.E. Rodely, and C.L. McMurtrie. 1974. The Vortex-Shedding Flowmeter. Flow: Its Measurement and
Control in Science and Industry 1:967-974.
Williamson, C.H.K. 1989. Oblique and Parallel Modes of Vortex Shedding in the Wake of a Circular Cylinder at Low
Reynolds Numbers. Journal of Fluid Mechanics 206:579-627.
Wolochuk, M.C. 1994. Evaluation of Vortex-Shedding Flow Meters for HVAC Applications. M.S. Thesis, Purdue Uni-
versity, West Lafayette, IN.
Wolochuk, M.C., M.W. Plesniak, and J.E. Braun. 1996. The effects of turbulence and unsteadiness on vortex shedding
from sharp-edged bluff bodies. To appear in J. Fluids Engr.
Appendix A: Uncertainty Analysis
Using the methods of Kline and McClintock (1953), the relative uncertainty in estimating the Strouhal
number depends on the relative uncertainties in the measurements of velocity, bluff body diameter, and shed-
ding frequency according to:
(A.1) u
St
u
U
2
u
d
2
u
f
2
+ + ( )
1 2
=
VOLUME 1, NUMBER 4, OCTOBER 1995 305
Velocities were computed from pitot-static probe measurements using Bernoulli's equation. For the
equipment used in this study, Wolochuk (1994) showed that the relative uncertainty in velocity can be esti-
mated as:
(A.2)
At high flow speeds, the second term in Equation (A.2) dominates, and the uncertainty in the measured
velocity asymptotically approaches the uncertainty in the air density. As U approaches zero, the first term
dominates, resulting in very high uncertainty at low velocities. Table A-1 gives values of the constants asso-
ciated with the conditions of this study used to estimate the relative uncertainties in the velocity and the
Strouhal number. The relative uncertainty in velocity is less than 2% for U > 10 m/s, but increases dramat-
ically for lower velocities. According to Equation (A.1), the uncertainty in velocity dominates the uncer-
tainty in the Strouhal number at low velocities, since u
D
and u
f
are both less than 2%. The relative
uncertainty in Strouhal number is less than 3% for U > 10 m/s and is acceptable, because most of the studies
were conducted in the range where U > 10 m/s.
Table A-1. Data used to calculate uncertainties
Symbol Definition Value
P
max
Diaphragm maximum pressure 860 Pa
Air density (22C) 1.18 kg/m
3
u

Uncertainty in air density 0.02


c Transducer calibration constant 0.054 Pa
N Number of samples per realization 2048
U Velocity measured by pitot probe 230 m/s
u
U
1
2
-- -
2

--- 0.0025 P
max
c + ( )
U
4
--------------------------------------------------------------- u

2
+
1 2
=
ACKNOWLEDGEMENTS
Financial support for this research was provided
by Johnson Controls, Inc. The authors appreciate
this support and the guidance provided by Don
Showers.

V
Error in voltage = 0.25% E
max
0.037 V
u
D
Uncertainty in bluff diameter 0.02
u
f
Uncertainty in vortex shedding frequency < 0.005
u
U
Uncertainty in U = f (U, P
max
, U

, c)
u
St
Uncertainty in St = f (u
U
, u
D
, u

)
Table A-1. Data used to calculate uncertainties
Symbol Definition Value

Potrebbero piacerti anche