UNDERSTANDING AND USING ULTRASONIC GAS FLOW METERS

John Lansing
William Freund ,Dan Hackett,
Winsor Letton, Kevin Warner & Klaus Zanker

Daniel Industries
Houston, Texas, USA

meter’s health. The primary purpose of this paper is to

ABSTRACT
discuss basic gas ultrasonic meter operation, diagnostics,
This paper discusses fundamental issues relative to review the fundamentals of field maintenance, discuss
ultrasonic gas flow meters used for measurement of some test results and provide the reader with an examples
natural gas. A basic review of an ultrasonic meter’s of good and not-so-good piping designs.
operation is presented to understand the typical operation
of today’s Ultrasonic Flow Meter (UFM). The UFM’s ULTRASONIC METER BASICS
diagnostic data, in conjunction with gas composition,
Before looking at the main topic of integral diagnostics, it
pressure and temperature, will be reviewed to show how
is important to review the basics of ultrasonic transit time
this technology provides diagnostic benefits beyond that
flow measurement. In order to diagnose any device, a
of other primary measurement devices. The basic
relatively thorough understanding is generally required.
requirements for obtaining good meter performance,
If the technician doesn’t understand the basics of
when installed in the field, will be discussed with test
operation when performing maintenance, at best they can
results. Finally, recommendations for installation will be
only be considered a “parts changer.” In today’s world of
provided, including an example of a good piping design.
increasingly complex devices, and productivity demands
on everyone, companies can no longer afford this type of
INTRODUCTION
service.
During the past several years, the use of ultrasonic flow
meters for natural gas custody transfer applications has Fortunately for everyone, the basic operation of an
grown significantly. The publication of AGA Report No. ultrasonic meter is relatively simple. Consider the meter
9, Measurement of Gas by Multipath Ultrasonic Meters design shown in Figure 1. Even though there are several
[Ref 1] in June 1998, has further accelerated the designs of ultrasonic meters on the market today, the
installation of ultrasonic flow meters (UFMs). Today principle of operation remains the same.
virtually every transmission company is using this
technology, either for custody transfer, or for operational
applications. Transducer 2
Since the mid-1990’s the installed base of UFMs has X
grown by approximately 50% per year. There are many
reasons why ultrasonic metering is enjoying such healthy
sales. Some of the benefits of this technology include the Flow D
L
following:
• Accuracy: Can be calibrated to <0.3%, little or no
drift.
• Large Turndown: Typically 50:1, or more. Transducer 1
• Naturally Bi-directional: Measures volumes in both
directions with comparable performance. Figure 1 - Ultrasonic Meter
• Tolerant of Wet Gas: Important for production
Ultrasonic meters are velocity meters by nature. That is,
applications.
they measure the velocity of the gas within the meter
• Non-Intrusive: No pressure drop.
body. By knowing the velocity and the cross-sectional
• Low Maintenance: No moving parts means reduced area, uncorrected volume can be computed. Let us review
maintenance. the equations needed to compute flow.
• Fault Tolerance: Meters remain relatively accurate
even if one or more sensors should fail. The transit time (T12) of an ultrasonic signal traveling with
• Integral Diagnostics: Data for determining a meter’s the flow is measured from Transducer 1 to Transducer 2.
health is readily available. When this measurement is completed, the transit time
(T21) of an ultrasonic signal travelling against the flow is
It is clear that there are many benefits to using UFMs. measured (from Transducer 2 to Transducer 1). The
Although the first several benefits are important, the most transit time of the signal travelling with the flow will be
significant may turn out to be the ability to diagnose the
less than that of the signal travelling against the flow due These equations look relatively simple, and they are. The
to the velocity of the gas within the meter. primary difference between computing gas velocity and
speed of sound is the difference in transit times is used for
Let’s review the basic equations needed to compute
computing velocity, where as the sum of the transit times
volume. Assume L and X are the direct and lateral (along
is used for computing speed of sound.
the pipe axis and in the flowing gas), distances between
the two transducers C is the Speed of Sound of the gas, V Unfortunately, determining the correct flow rate within
the gas velocity, and T12 and T21 are transit times in each the meter is a bit more difficult than it appears. The
direction. The following two equations would then apply velocity shown in Equation (3) refers to the velocity of
for each path. each individual path. The velocity needed for computing
volume flow rate, also know as bulk mean velocity, is the
L
T12 = (1) average gas velocity across the meter’s area. In the
X
C +V ⋅ pipeline, gas velocity profiles are not always uniform, and
L often there is some swirl and asymmetrical flow profile
and within the meter. This makes computing the average
velocity a bit more challenging.
L
T21 = (2) Meter manufactures have differing methodologies for
X
C −V ⋅ computing this average velocity. Some derive the answer
L
by using proprietary algorithms. Others rely on a design
Solving for gas velocity yields the following: that does not require “hidden” computations. Regardless
of how the meter determines the bulk average velocity,
L2  T21 − T12 
V =   (3) the following equation is used to compute the uncorrected
2X  T21 ⋅ T12  flow rate.
Solving for the speed of sound (C) in the meter yields the
Q =V *A (5)
following equation:
This output (Q) is actually a flow rate based on volume-
L  T21 + T12 
C =   (4) per-hour, and is used to provide input to the flow
2  T21 ⋅ T12  computer. A is the cross-sectional area of the meter.
Thus, by measuring dimensions X & L and transit times In summary, some key points to keep in mind about the
T12 & T21, we can also compute the gas velocity and speed operation of an ultrasonic meter are:
of sound (SOS) along each path. The speed of sound for
• The measurement of transit time, both upstream and
each path will be discussed later and shown to be a very
downstream, is the primary function of the
useful parameter in verifying good overall meter
electronics.
performance.
• All path velocities are averaged to provide a “bulk
The average transit time, with no gas flowing, is a mean” velocity that is used to compute the meter’s
function of meter size and the speed of sound through the output (Q).
gas (pressure, temperature and gas composition). • Because the electronics can determine which transit
Consider a 12-inch meter for this example. Typical time is longer (T21 or T12), the meter can determine
transit times, in each direction, are on the order of one direction of flow.
millisecond (and equal) when there is no flow. The • Speed of sound is computed from the same
difference in transit time during periods of flow, however, measurements as gas velocity (X is not required).
is significantly less, and is on the order of several
nanoseconds (at low flow rates). Thus, accurate Transit time is the most significant aspect of the meter’s
measurement of the transit times is critical if an ultrasonic operation, and all other inputs to determine gas velocity
meter is to meet performance criteria established in AGA and speed of sound are essentially fixed geometric
Report No. 9. (programmed) constants.

It is interesting to note in Equation (3) that gas velocity is INTEGRAL DIAGNOSTICS OF UFMs
independent of speed of sound, and to compute speed of One of the principal attributes of modern ultrasonic
sound (Equation (4)), gas velocity is not required. This is meters is their ability to monitor their own health, and to
true because the transit time measurements T12 and T21 are diagnose any problems that may occur. Multipath meters
measured within a few milliseconds of each other, and gas are unique in this regard, as they can compare certain
composition does not change change significantly during measurements between different paths, as well as
this time. Also, note the simplicity of Equations (3) and checking each path individually.
(4). Observe that only the dimensions X and L, and the
transit times T12 and T21, are required to yield both the gas Measures that can be used in this online “health
velocity and speed of sound along a path. checking” can be classed as internal or external
diagnostics. Internal diagnostics are those indicators
derived only from internal measurements of the meter. the face) will attenuate the transmitted (and received)
External diagnostics are those methods in which signals. The reader might assume that this buildup would
measurements from the meter are combined with cause the meter to fail (inability to receive a pulse).
parameters derived from independent sources to detect However, this is not generally the case. Even with
and identify fault conditions. Some of the common excessive buildup of more than 0.050 of an inch of an
internal meter diagnostics used are as follows. oily, greasy, and/or gritty substance, today’s UFMs will
continue to operate.
GAIN
One of the simplest indicators of a meter’s health is the The reader may wonder what impact on transit time
presence of strong signals on all paths. Today’s multipath accuracy could be attributed to transducer face
UFMs have automatic gain control on all receiver contamination. It is true the speed of sound will be
channels. Any increase in gain on any channel indicates a different through the contaminated area when compared
weaker signal, perhaps due to transducer deterioration, to the gas. Let us assume a build-up is 0.025 of an inch
fouling of the transducer ports, or liquids in the line. on each face, and the path length is 16 inches. Also
However, caution must be exercised to account for other assume the speed of sound through the contamination is
factors that affect signal strength, such as pressure and twice that of the typical gas application (2,600 fps vs.
flow velocity. 1,300 fps). With no buildup on the transducer, and at zero
flow, the average transit time would be 1.025641
Gain numbers vary from manufacturer to manufacturer. milliseconds. With buildup the average transit time
Thus, recommendations may also differ. However, would be 1.024038 milliseconds, or a difference of
regardless of design or methodology for reporting gain, it 0.16%. This would be reflected in the meter’s reported
is important to obtain readings on all paths under speed of sound (more on that later). However, it is the
somewhat similar conditions. The significant conditions difference in transit times that determines gas velocity
to duplicate are metering pressure and gas flow rate. (thus volume). This is the affect that needs to be
Gain readings are generally proportional to metering quantified.
pressure (and to a much lessor extent, temperature). That Maybe the easiest way to analyze this is assume the
is, when pressure increases, the amount of gain transit time measurements in both directions are reduced
(amplification) required is reduced. If an initial gain by 0.16% (from the previous example). Remembering in
reading were taken at 600 psig, when the meter was Equation (3) that gas velocity is proportional to a constant
placed into service, and subsequent readings taken at 900 (L2/2X) multiplied by the difference in transit times, all
psig, one would expect to see a change. This change in divided by the product of transit times. The decrease in
reading (assuming gain values are linear, not in dB) transit times will occur for both directions, and this effect
would decrease by the ratio of pressures (600/900). appears to be negated in the numerator. That is, the ∆t
Understanding that pressure affects gain readings helps will remain the same. However, the error in both T12and
guard against making the false assumption something is T21 will cause the both values to be effected. The result is
wrong. the volumetric error will be is the same as the percentage
Fortunately, most applications do not experience a of transit time (0.16%). Therefore the meter’s output will
significant variation in metering pressure. If pressure increase by 0.16%. However, this amount of buildup is
does vary, the observed gain value can be adjusted abnormal, and not typical of most meter installations.
relatively easily to allow for comparison with baseline Concluding the discussion on gain readings, UFMa all
values. This method of adjustment varies with have more than adequate amplification (gain) to
manufacturer, so no discussion will be incorporated here. overcome even the most severe reductions in signal
Gas velocity can also impact the gain level for each path. strength. The amount of buildup required to fail today’s
As the gas velocity increases, the increased turbulence of high-performance transducers and electronics generally
the gas causes an increase in signal attenuation. This exceeds pipeline operational conditions. Periodic
reduction in signal strength will be seen immediately by monitoring of this parameter, however, will help insure
increased gain readings. These increases are generally good performance throughout the life of the meter.
small compared to the amount of gain required. Typical Metering accuracy (differences in transit time velocity
increases might be on the order of 10-50%, depending computation) can be affected, but only when significant
upon meter size and design. Thus, it is always better to buildup of contamination occurs.
“baseline” gain readings when gas velocities are below 30 SIGNAL QUALITY
fps. Using velocities in excess may provide good results,
This expression is often referred to as performance (but
but it is safe to say that lower velocities provide more
should not be confused with meter accuracy). All
consistent, repeatable results.
ultrasonic meter designs send multiple pulses across the
So, what else causes reductions in signal strength meter to another transducer before updating the output.
(increased gain) you ask? There are many sources other Ideally, all the pulses sent would be received and used.
than gas velocity and pressure. For instance, However, in the real world, sometimes the signal is
contamination of the transducers (buildup of material on distorted, too weak, or otherwise the received pulse does
not meet certain criteria established by the manufacturer. others can. The important thing to remember is the best
When this happens the electronics rejects the pulse rather time to deal with control valve noise is during the design.
than use something that might distort the results. Today’s technology has improved significantly in dealing
with extraneous noise. Reducing it in piping design is
The level of acceptance (or rejection) for each path is
always the best choice (more on this later).
generally considered as a measure of performance, and is
often referred to as signal quality. Meters provide a value Other sources can cause reduced signal to noise values.
describing how good signal detection is for each Typically they are poor grounding, bad electrical
ultrasonic path. connections between electronics and transducers,
extraneous EMI and RFI, cathodic protection interference,
As mentioned above, there are several reasons why pulses
transducer contamination and in some instances, the
can be rejected. Additional causes may include
meter’s electronic components. However, the major
extraneous ultrasonic noise in the same region the
reason for decreased signal to noise ratios remains
transducer operates, distorted waveforms caused by
pressure drop from flow control or pressure reducing
excessive gas velocity, and to some degree, contamination
valves.
on the face of the transducer.
Concluding this discussion on signal to noise, the most
Typically, the value of acceptance for each path, under
important thing to remember is high-pressure drop
normal operating conditions, will be 100%. As gas
(generally in excess of 200 psig) across a control valve
velocity increases to near the meter’s rating, this
can cause interference with the meter’s operation. If the
percentage will begin to decrease. Depending upon
noise is isolated to a transducer or pair of transducers, the
design, this percentage may decrease to below 50%.
cause is generally not control valve related. Here
Generally, this reduction in performance will have little
probable causes are poor component connections or a
impact on meter accuracy. However, if the percentage of
potential failing component. Control valve noise usually
accepted pulses is this low, it is safe to say the meter is
causes lower signal to noise levels on the transducers that
not operating at top performance, and investigation may
face the noise source (all would be affected).
be warranted (assuming the meter isn’t operating at
110+% of rated capacity). VELOCITY PROFILE
Concluding the discussion on performance, this parameter Monitoring the velocity profile is possibly one of the most
should be monitored periodically as poor performance on overlooked features of today’s ultrasonic meter. It can
a path may be an indication of possible impending failure. provide many clues as to the condition of the metering
Lower than expected performance can be caused by system, not just as a monitor of the meter. AGA Report
several factors. Besides excessive gas velocity, No. 9 requires a multipath meter to provide individual
contamination on the transducer face and excessive path velocities. As mentioned previously, the output used
extraneous ultrasonic noise can reduce signal quality. by the flow computer is an average of these individual
However, by monitoring gains, this condition can be readings.
easily identified before it becomes a problem. Once the UFM is placed in service, it is important to
SIGNAL-TO-NOISE RATIO collect a baseline (log file) of the meter. That is, record
the path velocities over some reasonable operating range,
This parameter is another variable that provides
if possible. Good meter station designs produce a
information valuable in verifying the meter’s health, or
relatively uniform velocity profile within the meter. The
alert of possible impending problems. Each transducer is
baseline log file may be helpful in the event the meter’s
capable of receiving noise information from extraneous
performance is questioned later.
sources (rather than its mated transducer). In the interval
between receiving pulses, meters monitor this noise to Many customers choose to use a “high performance flow
provide an indication of the “background” noise. This conditioner” with their meter. This conditioner is
noise can be in the same ultrasonic frequency spectrum as intended to isolate any upstream piping effects on gas
that transmitted from the transducer itself. profile. In reality, they don’t totally isolate the
disturbance, but do provide a reasonably repeatable
Noise levels can become excessive if a control valve is
profile. The important issue here is the velocity profile is
placed too close and the pressure differential is too high.
relatively repeatable. Once a baseline has been
In this scenario the meter may have difficulty in
established, should something happen to the flow
differentiating the signal from the noise. By monitoring
conditioner, it can be identified quickly by comparing
the level of noise, when no pulse is anticipated, the meter
path velocities with the baseline. Many things can happen
can provide information to the user, warning that meter
to impact the original velocity profile. Changes can be
performance (signal quality) may become reduced. In
caused by such things as:
extreme cases, noise from control valves can “swamp” the
signal to the point that the meter becomes inoperative. • partial blockage of the flow conditioner,
• damage to the flow conditioner,
All meters can handle some degree of noise created from • or upstream piping affects, such as a change in a
this condition. Some UFM designs can handle more than
valve position.
Of course, something could have also occurred with the Pressure & Temperature Effects
meter to cause a significant profile change. Generally The speed of sound in gas can be easily computed in the
speaking, this is unlikely as all components are securely field. There are several programs used for this purpose
mounted. However, the velocity of a given path could be (like SonicWare) and they are generally based upon the
affected by other problems. When considering that only equation of state provided in AGA Report No. 8,
X and L dimensions, and transit times, impact path Compressibility and Supercompressibility for Natural
velocity, it is relatively easy to eliminate these. If a Gas and Other Hydrocarbon Gases [Ref 2]. When
problem develops within the meter that impacts only one computing speed of sound, there is always some
or more paths, other performance indicators, such as gain, uncertainty associated with this operation. It is important
path performance, and speed of sound will also be to realize that the speed of sound is more sensitive to
indicating problems. temperature and gas composition than pressure. For
Concluding this discussion on path velocities, most good example, a one degree F error in temperature at 750 psig,
installations produce somewhat symmetrical velocities with typical pipeline gas, can create an error of 0.13%, or
within the meter. Comparing each path’s velocity with about 1.7 fps. An error of five psig at 750 psig and 60
the average, and sometimes to other paths, depending degrees F only contributes 0.01% error. Thus, it is very
upon the UFM design, can give the user confidence the important to obtain accurate temperature information.
profile has not significantly changed. Today’s UFM can Knowing the temperature contributes error in computing
handle some relatively high levels of asymmetry within SOS is important. However, if the temperature is in error
the meter. It should not be assumed that the meter’s by the amount in the previous example, a more significant
accuracy is significantly impacted just because the question might be “what error is this causing in the
velocity profile has changed. It is usually an indication, volumetric measurement?” A quick calculation shows a
however, that something within the meter set, other than one degree F error will cause the corrected volumetric
the meter itself, is probably causing the effect. Careful calculation to be incorrect by 0.28%. Having a history of
review of other diagnostic parameters can determine if the calculated SOS vs. measured may actually be a good
meter is at fault, or not. “health check” on the stations temperature measurement!
SPEED OF SOUND Gas Composition Effects
Probably the most discussed and used diagnostic tool is Sensitivity to gas composition is a bit more difficult to
the meter’s speed of sound (SOS). The reader may recall quantify as there is an infinite number of sample analyses
that speed of sound is basically the sum of the transit to draw from. Let’s assume a typical Amarillo gas
times divided by their product, all then multiplied by the composition with about 90% methane. If the
path length (Equation (4)). As was discussed earlier, the chromatograph were in error on methane by 0.5%, and the
primary measurement an ultrasonic meter performs to remaining components were normalized to account for
determine velocity is transit time. If the transit time this error, the resulting effect on speed of sound would be
measurement is incorrect, the meter’s output will be 0.03%. Thus, minor errors in gas composition, for
incorrect, and so will the speed of sound. Thus, it is relatively lean samples, may not contribute significantly
important to periodically verify that the meter’s reported to the uncertainty.
speed of sound is within some reasonable agreement to an
independently computed value. However, lets look at another example of a Gulf Coast
gas with approximately 95% methane. Suppose the
Modern UFMs use high frequency clocks to accurately methane reading is low by 0.5%, and this time the
perform transit time measurements. In a typical 12-inch propane reading was high by that amount, the error in
meter, the average transit time may be on the order of one computed speed of sound would be 0.67% (8.7 fps!).
millisecond. When there is no flow within the meter, the Certainly one could argue this may not be a “typical”
difference between T12 and T21 will be zero. As flow rate error. There are many scenarios that can be discussed and
increases, the difference will be detected, and a resulting each one would have a different effect on the result. The
flow rate computed. To obtain a perspective on this uncertainty that gas composition contributes to the speed
differential time, values start out in the 10’s of of sound calculation remains the most elusive to quantify,
nanoseconds and typically increase to maybe 100 and, depending upon gas composition, may prove to be
microseconds at the highest velocities. the most significant.
Obviously accurate meter performance requires A typical question is “what difference can be expected
consistent, repeatable transit time measurements. between that determined by the meter, and one computed
Comparing the SOS to computed values is one method of by independent means?” It has been shown [Ref. 3] that
verifying this timing. This procedure would be the expected uncertainties (two standard deviations) in
considered an external diagnostic technique. Let’s speed of sound, for a typical pipeline gas operating below
examine the affects (or uncertainties) on computing speed 1,480 psig, are:
of sound in the field.
• UFM measurement: ± 0.17%
• Calculated (AGA 8): ± 0.12%
Since the UFM’s output is independent of the calculation TYPICAL SPEED OF SOUND FIELD RESULTS
process, a root-mean-square (RMS) method can be used This section provides actual data from two different
to determine the system uncertainty. Thus, when using meters. Figures 2 and 3 show trended vs. time. Data is
lean natural gas below 1,480 psig, it is expected that 95% shown for an eight-inch meter in Figure 2. It compares
of readings agree within 0.21% (or about 2.7 fps). the average speed of sound over the four paths with the
Therefore, it may be somewhat unrealistic to assume the AGA 8 calculated value.
meter will agree within 1 fps under typical operating
conditions.
Concluding this discussion on speed of sound, this 407.5

“integral diagnostic” feature may be the most powerful 407

USM(min)
tool for the technician. Using the meter’s individual path 406.5
USM(max)
AGA8
speed of sound output, and comparing it to not only the 406

computed values, but also comparing within the meter 405.5

SOS, m/sec
405
itself, is a very important maintenance tool. Caution 404.5

should be taken when collecting the data to help minimize 404

any uncertainty due to gas composition, pressure and 403.5

temperature. Additionally, it is extremely important to 403

obtain data only during periods of flow as temperature 402.5

0 5 10 15 20 25 30

stratification can cause significant comparison errors. By Time

developing a history of meter SOS, and comparing with Figure 2 - Eight-inch Meter Measured vs. Calculated SOS
computed values, it can also be used as a “health check”
At each measurement point, ten successive values of the
for the temperature measurement used to determine
ultrasonic meter’s SOS were logged. The two curves that
corrected volumes.
show the minimum and maximum values in Figure 2
demonstrates repeatability in SOS measurements of better
IMPORTANCE OF SOS VERIFICATION
than 0.03%. The difference in the meter’s speed of sound
As was discussed earlier, SOS verification helps insure vs. computed values are also, for most points, less than
the meter is operating correctly. However, what other 0.3%.
changes in a meter can affect the reading? From the
previous discussion on gain, buildup on the face of a Figures 3 shows the AGA 8 calculated speed of sound
transducer will affect the speed of sound. Thus, if a pair trended against the individual SOS readings from the four
of transducers has a different value, when compared to the paths. Note that in each case the agreement on all chords
average (or to other paths, depending upon meter’s is roughly as expected (better than 0.3%). In the area
design), this might be an indication of contamination. where speed of sound deviations exceeded 0.3%, (Figure
3) low flow temperature stratification was likely the
One thing to remember is that the percent change in speed cause. In the event of significant contamination on one or
of sound, given the same buildup, will be greater for a more pairs of transducers, this graph would have shown
smaller meter than a larger one. As path length increases the impact.
from say 10 inches to 30 inches (or more), a buildup of
0.025 inches will affect the transit time less. By utilizing
gain information with SOS data for a given path, it can be 396.00

quickly determined if the change in SOS is due to 394.00

contamination, or other causes. 392.00

Another benefit in monitoring path SOS is to verify 390.00

SOS, m/sec

proper identification of reception pulses. In the section on 388.00 vosA m/sec

vosB m/sec
signal to noise, extraneous noise was noted to potentially vosC m/sec
386.00

interfere with normal meter operation. That is, if vosD m/sec

V(aga8) m/sec
ultrasonic noise within the meter (caused by outside 384.00

sources) becomes too great, meter performance will be 382.00

0 5 10 15 20 25 30 35

impacted. Time

Figure 3 - Ten-inch Meter SOS with Four Chords

As the noise level increases, there is the possibility that
the circuit detecting the correct pulse will have difficulty. Concluding this discussion on external calculations, the
Good meter designs protect against this and reject results demonstrate multi-path ultrasonic meters show
received pulses that have increased uncertainty regarding good correlation between the computed speed of sound
their validity. If this scenario occurs, it is unlikely all and the meter’s reported speed of sound. Even though
paths will be affected simultaneously, and by the same there are differences between computed and reported
amount. Monitoring variations in SOS from path to path values, these remain relatively constant though out the
will identify this problem and help insure the meter’s test period. This also suggests that when performing an
health is satisfactory. on-line comparison of speed of sound, an alarm limit of
about ± 0.3% between the meter and computed values, as model is the amount of swirl generated by the elbow.
recommended earlier, is reasonable. However, as shown According to research work performed at Southwest
in Figure 3, for a short interval the error exceeded 0.3% Research Institute (SwRI) by Terry Grimley, it would
(during periods of low (or no) flow and temperature take on the order of 100D for the profile to return to a
stratification). Since this situation can occur in the field, fully symmetrical, fully developed, non-swirling velocity
safeguards should be implemented to insure gas velocity profile [Ref 4]. More complex upstream piping, such as
is above some minimum value, and for a specified time, two elbows out of plane, create even more non-symmetry
before alarming occurs. Thus, the use of independent and swirl than this model shows. Today’s UFM must
estimates of gas speed of sound, derived from an analysis handle profile distortion and swirl in order to be accurate
of the gas composition, can be an effective method of and cost-effective. However, just as with orifice and
understanding how well an ultrasonic meter is turbine meters, installation guidelines should be followed.
performing.
In 1998 AGA released the Transmission Measurement
Committee Report No. 9 entitled Measurement of Gas by
BASICS OF UFM INSTALLATIONS Multipath Ultrasonic Meters. This document discusses
When installing ultrasonic flow meters, many factors many aspects and requirements for installation and use of
should be taken into consideration to insure accurate and ultrasonic meters. Section 7.2.2 specifically discuss the
trouble-free performance. Before discussing these issues, UFMs required performance relative to a flow calibration.
let’s review the basics of a good installation. It states the manufacturer must “Recommend upstream
BASIC PIPING ISSUES and downstream piping configuration in minimum length
– one without and flow conditioner and one with a flow
Ultrasonic meters require adhering to basic installation
conditioner - that will not create an additional flow rate
guidelines just as with any other technology. Primary
measurement error of more than ±0.3% due to the
metering elements, such as orifice and turbine, have
installation configuration.” In other words, assuming the
adopted recommendations for installation long ago.
meter were calibrated with ideal flow profile conditions,
These are provided through a variety of standards (API,
the manufacturer must then be able to recommend an
AGA, etc.) to insure accurate performance (within some
installation which will not cause the meter’s accuracy to
uncertainty guidelines) when installed. The reason for
these guidelines is the meter’s accuracy can be affected deviate more than ±0.3% from the calibration once
by profile distortions caused by upstream piping. One of installed in the field.
the benefits of today’s UFM is that they can handle a During the past several years a substantial amount of
variety of upstream piping designs with less impact on testing has been conducted by SwRI in San Antonio,
accuracy then other primary devices. Texas to determine installation affects on UFMs.
Installation effects have been studied in much more detail Funding for these tests has come from the Gas Research
than ever before. This is due in part to the available Institute (GRI). Much of the testing was directed at
technology needed for evaluation. Reducing uncertainty determining how much error is introduced in today’s
for pipeline companies has also become a higher priority UFMs when a variety of upstream installation conditions
today due to the increasing cost of natural gas. Let’s look are present. This was presented in a report entitled
at a typical velocity profile downstream of a single elbow. Ultrasonic Meter Installation Configuration Testing at the
2000 AGA Operations Conference in Denver, Colorado.
Following is an excerpt from this report that shows the
impact of upstream effects on an ultrasonic meter.
Installation Effect: One Elbow Elbows Out Elbows In
Meter Orientation: 0°° 90°° 0°° 90°° 0°° 90°°
No Conditioner, 10D 0.07 0.02 0.53 0.04 0.02 0.24
No Conditioner, 20D 0.13 0.11 0.05 0.10 0.11 0.12
19-Tube Bundle 0.35 0.37 0.13 0.37 0.15 0.22
Flow Conditioner #1 0.10 0.03 0.02 0.02 0.07 0.14
Flow Conditioner #2 0.15 0.13 0.23 0.30 0.03 0.04
Flow Conditioner #3 0.04 0.00 0.17 0.23 0.36 0.41

Meets AGA 9 Doesn’t meet AGA 9

Table No. 1 – GRI Installation Test Results

Figure No. 4 – Single-Elbow Flow Profile
The preceding table presents metering accuracy results
From this mathematical velocity profile model it is from a 4-path meter with a variety of upstream effects
apparent the velocity profile at 10D from the elbow is far (single elbow, two elbows in plane and two elbows out of
from being fully symmetrical. What isn’t apparent in this plane). Tests were conducted with no upstream flow
conditioner and four brands of flow conditioners, all manufacturers have different methods for handling
located at their manufactures recommended position. One control valve noise. Thus, a discussion on how this is
thing to note is the 19-tube bundle did not perform very done will not be presented.
well. Also of importance is the meter met the AGA 9
installation requirement test producing less than ±0.3%
shift with no flow conditioner when located a minimum
of 20D from the upstream effect.
In conclusion for basic piping issues, upstream piping
does have an effect on the meter’s performance. Many
customers choose to use a flow conditioner in order to
reduce potential upstream effects. The use of 19-tube
bundles is not recommended by most manufacturers today
as the results are not consistent and are generally not as
good as with other flow conditioners. Flow conditioners
are not always required. As can be seen in line two of the
table, this meter passed the installation affects test with no
flow conditioner when located 20D from the effect, and
passed all but one test when located at 10D.
OTHER PIPING ISSUES Graph No. 1 – Typical UFM Waveform
One aspect to keep in mind when designing an ultrasonic
meter station is the use of control valves (regulators).
Ultrasonic meters rely on being able to communicate
between transducers at frequencies in excess of 100 kHz.
Control valves can generate ultrasonic noise in this
region. How much depends upon several factors
including the type of valve, flow rate and differential
across the valve. Manufacturers have different methods
for dealing with control valve noise. Whenever an
ultrasonic meter is used in conjunction with a control
valve, the manufacturer should be contacted prior to
design. Following is a diagram of meter set with a flow
conditioner and control valve.
11D 42D 2D 23D 10D 6D 11D

12-18” 18-12”
Expansion Reducer
Flow
Conditioner USM Graph No. 2- UFM Waveform With Valve Noise
Double Mokveld /
‘T’
Valve
18” 27D
Jet Valve
A better design would be to locate the meter further
Filter
upstream (see Figure 6 following). By installing two tees
Figure 5 – Poor UFM Piping Diagram between the meter and control valve, much of the
In this design pressure reduction occurs at about 27D ultrasonic noise is reflected back downstream, helping
from the meter. There are two elbows between the meter isolate the meter from the noise source. Also, in this
and valve. At low flow rates this design would probably design, the meter has been located more than 70D from
work fine. However, as the flow rate increases, so does the valve. Ultrasonic noise, just like audible sound,
the amount of energy generated by the pressure reduction. becomes attenuated the further you get from the source.
The amount of noise generated is roughly proportional to 11D 23D 10D 2D 42D 6D 11D

the square-root of the pressure times the differential. 18-12”

12-18”
Thus, as flow rate (or differential pressure) increases, so Expansion Reducer

will the amount of noise generated. At some higher flow Flow

USM Double ‘T’ Mokveld /
Conditioner
rate the meter will be unable to identify the signal, and Valve 18” 71D
Jet Valve

Filter
measurement will cease.
Figure 6 – Good UFM Piping Diagram
Following are two sets of waveforms. The first is a
typical signal received by a pair of transducers when there In conclusion, control valves can create enough noise that
is no extraneous ultrasonic noise. The second is an will over-power the UFMs signal. Control valves should
example of a meter experiencing noise from a control be located away from the meter. Install the meter
valve. In order to continue operation, the meter must be upstream of the control valve whenever possible as more
able to handle this type of noise (Graph No. 2). Different noise propagates downstream than upstream. Also, with
the higher pressure upstream, the UFM will obtain a One significant benefit in performing online comparisons
stronger signal from the transducers, making it easier to between the meter’s speed of sound and a computed value
detect the signal when in the presence of noise. Tees is to provide a “health check” for the entire system. If a
between the meter and the control valve are more variation outside acceptable limits develops, the probable
effective than elbows at reducing noise. (about twice the cause will be temperature, pressure, or gas composition
noise attenuation). Probably the most significant thing to measurement error rather than the UFM. In this regard,
remember is consult with the manufacturer during the the UFM is actually providing a “health check” on the
design phase. Testing of UFMs with control valve noise measurement system!
is ongoing with all manufacturers, and better methods of
Installation of an ultrasonic meter is important if proper
handling noise are constantly being developed.
operation is to be obtained. The two primary issues
relating to a good installation are upstream effects and the
CONCLUSIONS
potential impact of control valve noise. Upstream effects
During the past several years ultrasonic meters have are much better understood today then two years ago.
become one of the fastest growing new technologies in Testing conducted by Southwest Research Institute, under
the natural gas arena. The popularity of these devices has the guidance of the measurement community, and funded
increased because they provide significant value to the by the Gas Research Institute, have provided much the
customer by reducing the cost of doing business. One of information needed to help understand installation effects.
the most significant benefits is the reduction in
maintenance over other technologies. Control valve applications are much better understood
today than a few years ago. All manufacturers have
There are several factors which can be attributed to this methods to deal with this issue, and it varies depending
increased usage. First, as there are no moving parts to upon design. The manufacturer should be consulted prior
wear out, reliability is increased. Since UFMs create no to design to help insure accurate and long-term proper
differential pressure, any sudden over-range will not operation.
damage the meter. If the meter encounters excessive
liquids, it may cease operation momentarily, but no Today’s UFM is a robust and very reliable device with
physical damage will occur, and the meter will return to many fault-tolerant capabilities. It is capable of handling
normal operation once the liquid has cleared. Most a variety of pipeline conditions including contaminants in
importantly, ultrasonic meters provide a significant the natural gas stream. In the event of transducer failure,
amount of diagnostic information within their electronics. the meter will continue to operate, and some UFM
designs maintain excellent accuracy during this situation.
Most of an ultrasonic meter’s diagnostic data is used to When encountering contamination such as oil, valve
directly interpret its “health.” Some additional grease, and other pipeline contaminants, today’s UFM
diagnostics can be performed by using external devices will continue working and, at the same time, provide
and information (for example, computing speed of enough diagnostic data to alert the operator of possible
sound). This diagnostic data is available on a real-time impending problems.
basis and can be monitored and trended in many of
today’s remote terminal units (RTUs). UFMs support As ultrasonic metering technology advances, so will the
remote access and monitoring in the event the RTU can’t diagnostic features. In the near future UFM diagnostic
provide this feature. data will become even more useful (and user friendly) as
more intelligence is placed within the meter. They will
There are four commonly used diagnostic features being not only provide diagnostic data, but will identify what
monitored today. These include speed of sound by path the problem is. When this happens, ultrasonic meters
(and the meter’s average value), path gain levels, path may be considered “maintenance free.”
performance values (percentage of accepted pulses), and
signal to noise ratio. By utilizing this information, the References:
user can help insure the proper meter operation.
1. AGA Report No. 9, Measurement of Gas by
Probably the most commonly used tools are path speed of Multipath Ultrasonic Meters, June 1998
sound and gains. Speed of sound is significant since it
2. AGA Report No. 8, Compressibility and
helps validate transit time measurement, and gains help
Supercompressibility for Natural Gas and Other
verify clean transducer surfaces. When computing speed
Hydrocarbon Gases, July 1994
of sound in the field, care should be taken to collect data
only during periods of flow in the pipeline as temperature 3. Letton, W., Pettigrew, D.J., Renwick, B., and
gradients will distort comparison results. Additionally, as Watson, J., An Ultrasonic Gas Flow Measurement
shown in one of the graphical examples, low-flow limits System with Integral Self Checking, North Sea Flow
should be implemented to insure pipeline temperature is Measurement Workshop, 1998.
uniform and stable before comparing meter speed of
sound with computed values from gas composition, 4. T. A. Grimley, Ultrasonic Meter Installation
pressure and temperature. Configuration Testing, AGA Operations Conference,
2000.