You are on page 1of 27

4 ^ Practical Developments in Gas Flow Metering Workshop

A Review of Gas Ultrasonic Flowmeter Performance, Paying Particular


Attention to Installation Effects, and Proposals for Future Research

Chris Duffell

TUV NEL

7'^ November 2001


4 Practical Developments in Gas Flow Metering Workshop

A REVIEW OF GAS ULTRASONIC FLOWMETER PERFORMANCE, PAYING


PARTICULAR ATTENTION TO INSTALLATION EFFECTS.
AND PROPOSALS FOR FUTURE RESEARCH

Chris J. Duffell MPhys CPhys MInstP


Flow Centre, Nafional Engineering Laboratory, East Kilbride
Dept. of Electronic and Electrical Engineering, University of Strathclyde, Glasgow

SUMMARY

Offshore platforms place a high importance on maintainability and long-term performance to


minimise metering costs. For this reason transit fime ultrasonic flowmeters are being used for
many crifical applications in North Sea installations. Potenfial uncertainties of less than 0.5%
and an independence of variable fiuid properties are claimed for most transit time ultrasonic
flowmeters, indicating suitability for allocafion metering duties in the Oil & Gas flow
measurement market. However, we are only just starting to see reports of long-term
peri'ormance, which are required to insfil further confidence in this relafively new technology.

One disadvantage with ultrasonic flowmeters is their relative inability to meter asymmetric
flows accurately. If the required, fully developed, fiow profile Is perturbed by unsuitable
upstream installations, then the fiow profile entering the metering section can become
asymmetric. As a result some of the ultrasonic flowmeter paths will over-read while others
under-read. When weights are applied to these paths and the values combined to infer a total
volumetric flowrate there will be an error.

Ultrasonic transit fime flowmeters, for the measurement of high-pressure gas. normally
employ multiple measurement paths to reduce the effects of swirl and distortion of the axial
velocity profile on metering uncertainty. The overall sensitivity of the meter to these
installation effects depends not only on the transducer configurafion, but also on the method
used to combine the individual path velocifies into one total flowrate measurement.

This paper reviews recent ultrasonic fiowm.eter development with respect to installation
efl'ects and details work currenfiy underway, ufilising mathematically formulated velocity
profiles to determine the optimum configurafion for ultrasonic transducers.

There Is also a discussion on future work planned to categorise the profiles experienced
downstream of a number of common installation condifions for use with neural network
packages. This will enable the meter diagnostics to determine not only the type of profile
entering the metering section, but also to apply ttie appropriate corrective weights to the path
velocities to compensate for the asymmetry.

The results obtained from this work could be used in the design of future ultrasonic gas
flowmeters, which ideally would be completely insensitive to variafions in the flow velocity
distribution.
4"" Practical Developments in Gas Flow Metering Workshop

1 INTRODUCTION

The last two decades have seen a dramafic increase in the number of conferences, seminars
and workshops in the field of Oil & Gas Flow Measurement. This high level of Technology
Transfer not only highlights the need to find accurate, self-contained flowmeters that are
unsuscepfible to a variety of working condifions. but also acknowledges the seriousness with
which the industry is taking this quest. The Nafional Engineering Laboratory alone has many
courses and conferences aimed at members of the Oil & Gas industry this year (2001)
ranging from one-day seminars at NEL to a four-day international conference in Malaysia [1].
The following statement, made by Mark Leigh of Conoco (UK) Ltd, in the proceedings of the
1999 Pracfical Developments in Gas Flow Metering Workshop, confirms this need:

"A key factor in the confinued successful production of hydrocarbons into the 21st century will
be the development of low cost, fit for purpose metering and allocation systems," [18].

Companies are sharing their pracfical field experiences, as well as their expertise, with those
who are traditionally their competitors in the bid for a greater understanding of modern,
complex measurement systems. There is a large amount of Technology Transfer in this
sector, due mainly to the high value of the fluids concerned and the regulatory involvement of
the Government in terms of taxafion. This involvement has meant that much of the transfer of
knowledge is lead by DTI. Their Oil & Gas Office not only produce the Brifish Standards for
correct manufacture, installafion and use of flowmeters, but also publish Independent data on
many aspects of fiowmetering.

Much of this work Is sub-contracted to companies such as the National Engineering


Laboratory in the Flow Programme. This is a three-year programme producing relevant and
independent literature in research and development as well as providing intercomparisons
between Europe's main test and calibration facilities. This is an advantageous arrangement
as individual companies may not wish to pay for large R&D programmes or readily publish the
results of their own, possibly cosfiy, findings.

As the pace of technological advances and financial pressures on marginal oil & gas fields
increase, all areas of fiow measurement are affected. Out-of-date methods can lead to
significant losses in revenue as well as unacceptable uncertainty of measurement. Operators
also wish to minimise producfion line downtime, which may occur for meter maintenance or
calibrafion.

Many papers have been published detailing research work, relafing pracfical experiences and
exposing shortfalls using various metering technologies. This often takes the form of either
improvement to an established method, or support/concerns regarding a new technology.
One of the new technologies attracting much interest from metering engineers is that of
ultrasonics. This is illustrated by the high percentage of papers presented at conferences on
this subject.

The results of a survey carried out by NEL in 1996 indicated a high level of Industrial interest in
ultrasonic flowmeters (USMs) and a general demand for more Information. From further
aspects of the review process It has been established that the need for information remains
unfulfilled yet clearly evident. This was verified by, for example, the degree of interest in
ultrasonic flowmeters shown at the Practical Developments in Gas Fiowmetering Workshop
held in Aberdeen In June 2000 [13,15,17, 22, 26].

Offshore platforms place a high importance on maintainability and long-term performance.


Damaged components may increase the uncertainty of measurement and replacement may
demand a re-calibration of the meter. With ultrasonic flowmeters the effect of mechanical
wear is eliminated as there are no moving parts, and electronic drift has been overcome since
the advent of digital technology. However, we are only just starting to see reports of long-
term performance.
4 Practical Developments in Gas Flow Metering Workshop

Transit time ultrasonic flowmeters are being used for many crifical applications in North Sea
installations. In practice their performance can deviate from the ideal due to installafion
conditions such as upstream pipework or changing viscosity in the velocity distribution.
Potential uncertainfies of less than 0.5% and an independence of variable fluid properties are
claimed for most transit time ulfi-asonic flowmeters, indicating suitability for allocafion metering
dufies in the Oil & Gas flow measurement market.

One disadvantage with ultrasonic flowmeters is their relative inability to meter asymmetric
flows accurately. If the required, fully developed, flow profile is perturbed by unsuitable
upstream installations, then Uie flow profile entering the metering section can become
asymmetric. As a result some of the ultrasonic flowmeter paths will over-read while others
under-read. When weights are applied to these paths and the values combined to infer a total
volumetric fiowrate fiiere will be an error.

Ultrasonic flowmeters normally employ mulfiple measurement paths to reduce the effects of
swiri and distortion of the axial velocity profile on metering uncertainty. The overall sensifivity
of the meter to these installation effects depends heavily on the transducer configurafion.

The aim of the research project underway Is to review recent ultrasonic flowmeter
development with respect to installation efi'ects and then to ufilise mathemafically formulated
velocity profiles along with computafional fluid dynamics software to determine the optimum
configuration for transducers. It is also planned to categorise the profiles experienced
downstream of a number of installation condifions using neural networks. This will enable the
meter diagnosfics to not only determine fiie ^ype of profile entering the metering section, but
also apply appropriate corrective weights to the path velocities to compensate for the
asymmetry.

The results obtained could be used in the design of future ultrasonic flowmeters, which would
be effectively insensitive to variafions in flow velocity distribution.

2 INSTALLATION EFFECTS

The velocity distribufion across a cross secfion of pipe will follow a parabolic law if the flow is
laminar. The maximum velocity is at the centre of the pipe and is twice the mean velocity.
The velocity at fiie pipe wall is zero due to the no-slip condifion [3].

A general formula is:

V = V o [ 1 - ( r / R ) ^ l ; where Vo is the central velocity. (1)

In turbulent fiow a more uniform distribution of velocifies results. There are many forms of
velocity distribufion equation for turbulent flow, some of which take into account the pipe
roughness as well as the Reynolds Number.

A general formula is:

V = Vo[1-(r/R)]'""' (2)

The parameter n Is related to Reynolds Number. Experimental work by Schlichting in


'Boundary Layer Theory' (1979) detailed this relationship, which enabled Moore [20] to
produce the following table.

Table 1: Relationship between n and Reynolds Number

1 Re 4.0x10-' 2.3x10* 1 1.1 xlO-' 1.1 xlO'' 2.0 xlO*" 3.2x10"


n 6.0 6.6 7.0 8.8 10 10
4"^ Pracfical Developments in Gas Flow Metering Workshop

If a uniform velocity distribution enters a pipeline the velocity is inifially U everywhere. Once
inside the pipe we know that the velocity must be zero at the walls because of the no-slip
condition. A boundary layer will develop along the walls of the pipeline. The solid surface
exerts a retarding shear force on the flowing fluid. The effect of the solid boundary is felt
ftjrther Into the flow as we move down the pipe. Conservation of mass means that the
velocity at the centre of the pipe must increase with distance from the inlet. At a sufficient
distance fi-om the inlet the boundary layer will reach the centre of the pipe and the flow
becomes entirely viscous. The shape of the velocity profile is different once the Inviscid core
has disappeared. When the shape of the profile no longer changes with distance the profile is
fully developed. The distance required to produce fully developed flow depends on whether
the flow is laminar or turbulent. The lengths are shorter for turbulent flow as there is mixing of
fluid particles between adjacent layers.

For laminar flow the distance required for a fully developed flow profile is given by:

L = 0.06X o v D (3)
D ii

where pv D is \he Reynolds Number, Re.


M

If Re = 2000 (still in the laminar regime):

L = 0.06 X 2000 x D

L = 120D.

Turbulent flow causes more rapid growth of the boundary layer due to enhanced mixing
among fluid layers. It has been shown by experiment that the flow profile becomes fully
developed within approximately 25 to 40 D. However, even this is sometimes too short a
distance in practice where space is at a minimum.

Space is all too often at a premium on offshore platforms as well as onshore process lines
and test rigs. This means that there are sometimes insufficient straight lengths of pipework
upstream and downstream of the installed flowmeter to ensure a fully developed, symmetrical
flow profile exists at the meter. Most commonly we find single or double bends within 20D of
the meter as well as occasionally finding valves, expansions and contracfions. These
pipework configurations effect the way in which the fluid flows through the pipe and causes
the cross-sectional velocity proflle to become asymmetrical. Ultrasonic flowmeters infer a
total flowrate from one or more sample taken in a predetermined plane in the flow. If the
required, fully developed, flow profile is perturbed by unsuitable upstream installations, then
the fiow profile entering the metering secfion can become asymmefilc. As a result some of
the ultrasonic flowmeter paths will over-read while others under-read. When weights are
applied to these paths and the values combined to infer a total volumetric flowrate there will
be an error.

Presenfing at the North Sea Flow Measurement Workshop in 1999, Brown, Barton and Moore
[7] assessed the performance of different ultrasonic flowmeter designs, and they determined
the magnitude of various installation effects in pracfice. The distortions to the flow for
common pipeline components were found to be as follows:

1. A contracfion flattens the fiow profile Inifially then becomes more peaked than a fully
developed profile.
2. An expansion has the opposite effect, making the profile more peaked at first and then
flattening out.
3. Between 5 and 12D downstream of a single bend the profile Is quite complex. The flow Is
then pushed to the top of the pipe and around the sides. Two vortices spiral down the
pipe and 30D downstream the flow has not returned to a fully developed profile.
4* Practical Developments in Gas Flow Metering Workshop

V-
4. The double bend generates a single vortex that decays much more slowly than the double
vortices downstream of a single bend. The vorticity causes the region of high axial velocity
to move down the pipe in a corkscrew fashion.

Salami's analytical profiles, based on experimental LDV data, are usefiji for assessing
ultrasonic flowmeters since the velocity at any point can be calculated without the need for
interpolation. Another advantage is the profile function can be integrated over the area to
obtain the total flowrate exacUy.

The profiles used were evolved by Salami to resemble common profiles generally
encountered in pracfice [24]. The main requirements Salami had when developing profiles
was the velocity should change both along the radius and from one radius to another, as is
the case in practice. It was also required that the fijncfion be capable of being integrated over
the cross-section of the circular pipe.

Regardless of the extent of the asymmetry each profile must always have zero velocity at the
walls because of the no-sllp condifion of flow. This means there Is a laminar layer near the
wall that becomes turbulent as the centre of the pipe is approached.

The profiles are essenfially made up of two components.

U = Ub + U, (4)

The power law creates an Ideal, fully developed fiow profile: -

Ub = ( 1 - r ) " " . where n depends on Reynolds number. (4a)

The polynomial component produces the asymmefiy.

Uc = mr(1-r)'''f(e) (4b)

The second of these components. Uc, also has two parts: the polynomial weight function (m)
determines the strength of the asymmetry and the angular function f(0) determines the
behaviour of the asymmetry.

The funcfion f(9) can be any funcfion of 9. If f(9) = sin (9) then for half the circle the value of
Uc is added to the power law profile, Ut,, and subtracted for the other half. The severity of the
distortion is confi-olled by the magnitude of the factor m.

The shape of the distribution can be changed by varying n, m, k, and f(8). The values of
these parameters can be found in Table 4 in the appendices. This is not available with the
CFD data without interpolation, which can lead to errors. The actual flowrate is given by the
double integration:

A 27t1

Qact=:rl^-'-^'-^^
2n ^^ ' (5)
^^^ 0 0
Salami points out that because asymmetry is a random event the higher number of radial
traverses taking place the more accurate the result should be if obtained using an average of
these individual velocifies.

The values of the parameters have been chosen by Salami so that the maximum velocity
occurs between 1.0 and 1.2 of the central velocity. This was found to be the case in practice.
Contours containing dips were arranged so that the dip did not fall below 0.5 of the velocity at
the centre of the pipe. This was also to support experimental data.
4'" Pracfical Developments in Gas Flow Metering Workshop

Salami's profiles were used as some of them (originally 23 were produced) are quite similar to
velocity profiles after bends. A3 and A7 can be used to model flow downstream of a double
bend and A8 and A l l model flow downstream of a single bend. Salami also uses multi-
peaked profiles which Moore [21] refers to as "More complex". These were originally
designed to simulate fiow after heat exchangers and chimney stacks, as well as ensuring the
peaks were not always In regular posifions, i.e. diametrically opposite to each other.

On the following pages Figure 1 shows the 15 output profiles from the Mafiab programme
written by the author. These are shown In 3D format to give the reader a greater
understanding of the shape of the profiles. When the profiles are used for transducer
configurafion analysis we shall use 2D contour plots of the theoretical profiles to obtain path
velocity data.

The first profile is called Velocity Profile Number 0. This is a fully developed profile, which is
the ideal condifion at the meter.

The next seven profiles were grouped together by Moore as they each have one velocity
peak. Profiles 8 - 1 1 have two peaks and the final three profiles are termed "More complex".
The numbers used In this review are the same as the A-numbers used by Moore in her work.
These differ from the original P-numbers used by Salami.
4*^ Practical Developments in Gas Flow Metering Workshop

Figure 1: The Tifteen theoretical profiles used for analysis

-.wv^-^.tn.. rfHc a

i
*!
1
Al.
; - - j - i r L j ^ , ' j - - .i-h^i-i ^
U-i
VXv

m
.ii W-'-

- J t i f l ^ >. * > Hn->v Scb. H n > 1

*' '. . . - - ' ^ " *

v'iB's'M W ^ M .J*vv='-'l - i r

"-.
-.
ft! . 119..

mk-

' J I . # T S ' W . HWW t -Jmn bi/h .1,

At.;

0. -. i^:---'
-' hi
4' Practical Developments in Gas Flow Metering Workshop

VMWv Drrm <t ( I M M '.Mivliaii^b V n r w *

91.'

UJ-.

i l t ,r

UWvTf Bvi*. M.rJw i l UBAfli) k n t W M m M '

ai.
ut-.

flp-'--

V*f^.Tf B w ^ l U f f W * ' ./ fipVT^ ^ i " * f t ^ O

fi-
4 * Practical Developments in Gas Flow Metering Workshop

TRANSIT TIME ULTRASONIC FLOWMETERS

3.1 Introduction

There is a diverse range of ultrasonic flow instrumentation available on the current market
ft-om relatively simple fiow switches to highly accurate multipath flowmeters. Bragg [6] makes
file important point that not all ultrasonic flowmeters are based on the same measurement
technique. It is important to choose the most suitable ultrasonic meter for any given
applicafion.

Today's ulfi-asonic flowmeters are based on three different basic technologies:

Transit fime . . ^ :
Doppler Effect
Cross correlation

Of these technologies, it is only the transit time method that will be looked at in great detail
with respect to flow measurement. The Doppler technology is principally used for medical
appllcafions and flow field analysis. Doppler Effect is, In general, not sufilcienfiy accurate for
industrial applications. Cross correlafion has not yet achieved wide commercial usage and Is
still impaired by fundamental problems concerning accuracy and range of measurement, [12].
Transit time ultrasonic flowmeters are being used for many crifical applications in North Sea
installations. In practice, their performance can vary due to Installafion condifions such as
upstream pipework or changing viscosity in the velocity distribufion. The four basic methods
by which transit fime velocity measurement is performed are direct time differenfial; phase
differenfial; phase control and frequency differential. In modern ultrasonic flowmeters the
direct time differenfial method is most common and may be referred to by the terms leading-
edge, pulse, time-of-flight, or sing-around.

3.2 Principle of Operation

Ultrasonic transit fime flowmeters are based on measurement of the propagafion fime of
acousfic waves in a flowing medium. A description of how we obtain the flowrate is found in
most papers or books on the subject. The following method is as described in BS 7965 [10].
The velocity of sound in the medium, plus the component of fluid velocity along the path, give
the velocity along a path. Considering the general ray geometry shown below, the upstream
and downstream transit times are given respectively by:

L
Kt- and l^^ (6)
(c-vcosd) (c + vcos^)

Figure 2: General path geometry for acousfic velocity measurement


4"^ Practical Developments in Gas Flow Metering Workshop

This method is used to eliminate the velocity of sound from the analysis. An advantage of
removing the velocity of sound from the equafions is that it eliminates the temperature
dependence.

In the direct time ditferential method, short acousfic pulses are propagated upstream and
downstream and the fime duration between excitafion and recepfion Is determined from:

. 2Lv cos^
\c^ -y^cos"^)
^ ^ (7)

if we assume that v^ c^, this leads to the following expression for the average path velocity
v:

2LAt
V =
^osd{t,,+tJ ^^^

The volumetric flowrate is determined by weighfing and summing path velocity measurements
and applying a calibration factor. In the case of a single path diametrical meter, the weight is
equal to unity and the calibration factor is replaced by the cross secfional area, A, and the
Reynolds number dependent velocity profile correcfion factor, k^

qv^k.Av^ (9)

Flow meters based on the transit time principle are the most widely accepted form of
ultrasonic flow measurement instrumentation but even within this generic group there is wide
diversity. Non-invasive clamp-on meters are advantageous as they do not come into contact
with the process fluid and so are ideal for metering of harmful or corrosive materials, but they
have been shown to give rise to many factors that have a detrimental effect on performance.
For example, It has been shown by Brown [8], that since clamp-on meters always use a
diameter path configuration they are particularly sensifive to changes In viscosity.
Performance is also heavily dependent on the pipe geometry and how accurately this is
measured.

Wetted transducers can provide manufacturers with more freedom and control during the
design stages. Brown has shown dual mid-radius chord path configurations have a lesser
effect from changes In viscosity than the dual diameter path configuration. This ability to
position the measurement paths at locations other than the diameter can Improve
performance of the meter. Gas custody transfer meters currently use this to enhance
performance employing up to 5 paths and achieving less than 0.5% uncertainty.

Ultrasonic flowmeters Infer a total flowrate from one or more sample taken In a predetermined
plane in the flow. A meter factor is then used, assuming the flow to be symmetrical, to
calculate the total flowrate.

3.3 Advantages of USMs

Ultrasonic flowmeters have many potenfial advantages over other fiow metering technologies
such as the more conventional orifice plate, turbine and positive displacement meters. The
following points should assist those in the indusfi^ that require a new metering system.

10
4 * Practical Developments In Gas Flow Metering Workshop

1. An ultrasonic meter does not Intrude into tiie path of the fluid and so does not contribute
to head loss. Nor does the meter obstruct flow, lowering flowrate, and hence suffer from
erosion.

2. The clamp-on type does not require the pipe wall to be altered so pipeline Integrity is
maintained. The meter is also isolated fi'om any hazardous materials that may be in the
pipe.

3. The claimed turndown ratio of an ultrasonic meter can be as much as 400:1. which
removes the need for multiplicity for measuring large flow ranges.

4. The inherent fast response time In ultrasonic meters means they can accurately measure
transient or pulsating flow.

5. A 'dry' calibration can be performed to find the calibrafion factor calculated on the basis of
measurements made which do not require fiie presence of any flow.

6. The lack of moving parts means ultrasonic meters can be used for applications where
minimum maintenance access is available. This obviously affects the total life cost
considerably.

7. The cost of directing a reverse flow through a meter using additional pipes and valves is
avoided by the bi-direcfional measurement available with ultrasonic meters.

8. The component parts of an ultrasonic metering system are not scaled-up in proportion to
pipeline size as in many technologies. This makes ultrasonic systems increasingly
competitive at larger diameters.

9. An ultrasonic meter also provides speed of sound measurements that can be used for
inferred mass flow and even direct calorific value measurements. This can also be used
for self-checking meter health.

10. Ulfi-asonic meters can be used for a wide range of fiows including corrosive, non-
conducting, extreme temperature, gaseous, wet gas and mulfiphase.

11. Remote operation can reduce the need for manned metering stations, hence reducing
costs.

3.4 Disadvantages of USMs

It can be argued that despite most potential users being aware of the benefits of ulfi"asonlc
technology acceptance is not widespread. The industry's reluctance to employ the
technology can be related to some of the factors detailed below.

1. Unless potenfial users can idenfify the various benefits offered by ultrasonic
technology, the motivafion to change from the established methods will be negligible. An
engineer may understand that USMs produce very little pressure drop. Being unaware of the
diagnosific capabilifies offered by USMs. the technology may be disregarded on the basis that
pressure drop can be tolerated in the applicafion under considerafion.

2. There is a low level of confidence in manufacturers' claims due to insufficient reliable,


up-to-date and Independently derived USM performance information. The current lack of
supporting International Standards is also of concern.

3. With regards to long-term duty a lack of confidence may exist as there are few
available reports of long-term performance. Although in support of the USM the effect of
mechanical wear is eliminated due to the absence of moving parts and electronic drift has
been overcome since the advent of digital technology.

11
4"' Practical Developments in Gas Flow Metering Workshop

4. A potential user may have a lack of understanding about the technology and which
manifestation would best suit, and any a mistake could be costly. There are several USM
Instruments based on different principles of operation, each In several configurations and
available fi^om a number of manufacturers at varied costs. In this situation the cost of
investigating each of the available opfions may outweigh the expected benefits.

5. Potential users may disregard ultrasonic technology on the basis of bad experience
they have had in the past with a different sort of USM. They may be unaware that the
limitations of a particular USM are not the llmitafions of ultrasonic technology or that the
technology may have progressed sufficienfiy to be reconsidered, (e.g. transit time clamp-on
meters).

4 INTERNATIONAL STANDARDS

Spearman and Sattary [25] identified that In 1997 no ISO standards relafing to the installafion,
operafion and calibration of ultrasonic meters were published. The situation has not improved
much In the last few years. There are three best practice/guidance notes but no standard
available yet. There is the AGA9 document Measurement of gas fiow by mulfi-path ultrasonic
meters, a BSI document number BS7965:2000 Selecfion, Installafion. Operation and
Calibration of diagonal path transit fime ultrasonic flowmeters for Industrial gas applications
and the ISO best pracfice guide number BS ISO TR 12765.

The BSI document BS7965: 2000 [10] applies only to industrial gas applicafions where the
entire stream of gas flows through the t>ody of the meter and the transducers are wetted.
Sub-clause 4.2 highlights flow velocity profile as an extrinsic factor that effects fiowmeter
performance Secfion 4.3.3.2 states that crossed paths or double traverses in a single plane
are essentially Insensitive to non-axial velocity components. Other configurafions (for
example, the triple traverse) may be sensitive to non-axial components but may be used in
combination to eliminate the effects of swirl and cross-flow.

The document recommends at least 10D upsfi-eam straight lengths and 3D downstream if the
meter has four or more paths. Less than this requires at least 20D upstream (or 10D with a
fiow condifioner) and 50 downstream. It is also noted that the meter manufacturer should be
consulted to determine if there is a preferred meter orientation for the given pipeline
configuration that may cause fiow profile distortions.

BS ISOrrR 12765:1998 [11] is a Type 2 Technical Report rather than a standard. This type
of document is issued when the subject is still under technical development or where for any
other reason there is the future but not Immediate possibility of an agreement on an
International Standard. BS ISO/TR 12765 Is seen as a "prospective standard for a provisional
application" in the field of USMs. It was felt there Is an urgent need for guidance on how
standards in this field should be used to meet an Identified need. A review of this document is
currently due. Section 7.2.1 highlights the existence of swiri and the shape of the axial fiow
profile as factors that can infiuence the uncertainty of the meter reading. Possible methods of
reducing these effects include using multi-patii meters with integration techniques suited for
the actual conditions, and carrying out flow calibrafions under condifions similar to actual
conditions. It is recognised that in a multi-path arrangement, the number of chords, chord
positioning and the integration technique used can reduce the measurement uncertainly
considerably, as well as reducing the effect of changes in the flow profile.

The AGA Report No.9 [2] was issued as a recommended practice and not as a standard. The
document warns potential users that various combinations of upstream fittings, valves and
lengths of straight pipe can produce velocity profile disturbances at the meter inlet. These
may lead to an error in the flowrate measurement. The amount of meter error will depend on
the magnitude of the distortion of the velocity proflle and, more significanfiy for this project,
the meter's ability to compensate for this distortion. It is also highlighted that for bi-directional
applications both ends of the meter should be considered upstream.

12
4' Practical Developments in Gas Flow Metering Workshop

5 PREVIOUS WORK UNDERTAKEN IN THIS FIELD

Delsing [12] presented a review of development trends In ulti-asonic flowmeters at the Global
Conference on Flow Metering and confi-ol at FCRI in India in 2000. The tiiree main areas of
development found were installation effects, measurement elecfi-onics and transducers. A
conclusion of the review was that ultrasonic meters are as good or better than any other
meter technology available. Ultrasonic meters are the current and future fiowmeter
technology of choice. Delsing nofices recent conferences show a multitude of work in the
direction of installation effects on ultrasonic fiowmeters.

At the NEL's Non-intrusive Flow Measurement Workshop in November 2000 Moore confirmed
[21] that a mid-radius chord is able to measure fully developed fiow more accurately than a
diametrical chord. Moore tried three configurations of mid-radius chord. The first conclusion
reached was that the orientafion sensitivity is reduced when the number of paths is increased.
Moore concluded by suggesfing a more sophisticated approach to ultrasonic flow metering
might be to employ adaptive weighfing of tiie individual path velocities. The Individual weights
would depend on additional information obtained from the distribution of these path velocities.

As menfioned in Section 2, Brown, Barton and Moore [7] assessed the performance of
different meter designs in 1999. They determined the magnihjde of various installation effects
In practice using CFD to model fiows in 3D and they found significant installation effects 20
diameters downstream of common components. The high sensitivity of single diameter paths
to velocity profile was demonstrated. It was also shown that introducing a second diametrical
path reduced this sensitivity by a factor of 2 or more in most cases. The dual mid-radius . .. -.
meter performed slightly better than the dual diametrical meter, but the 4-path Gaussian ':*
outperforms the dual mid-radius configuration reducing fiie error by a factor of 2 or 3. with a - . '.%
few exceptions.

The distortions to the flow for common pipeline components were found to be as follows:

1. A contraction flattens the flow profile initially then becomes more peaked than a fully
developed profile.
2. An expansion has the opposite effect, making the profile more peaked at first and then
flattening out.
3. Between 5 and 12D downstream of a single bend the profile is quite complex. The flow Is
then pushed to the top of the pipe and around the sides. Two vortices spiral down the
pipe and 30D downstream the flow has not returned to a fully developed profile.
4. The double bend generates a single vortex that decays much more slowly than the
double vortices downstream of a single bend. The vorticity causes the region of high
axial velocity to move down the pipe in a corkscrew fashion.

Barton and Brown carried out similar work under the DTI's Flow Programme in the 1999
report Velocity Distribution Effects on Ulfi-asonic Flowmeters. Part 2 [4]. CFD was again used
to model the effects of certain Installations on ultrasonic fiowmeters. The report concluded
that CFD and numerical modelling can be used to accurately model performance of
flowmeters in asymmetric fiow, even stating that results obtained demonsfi^ate these
techniques' promise for future development of improved measurement schemes and
enhanced self-diagnosfics.

Brown's paper at the Non-intrusive Flow Measurement Workshop 1999 [9] points out that It is
important to realise the arrangement of the paths and the scheme employed to combine the
measurement is generally more important than the number of paths employed. Generally
ultrasonic meters have self-checking and fault diagnostics which indicate problems such as
entrained gas attenuating the ultrasonic signals.

Zanker [27] showed that a Daniel's 4-path meter does an excellent job of Integrating the
velocity profile to give the correct flow rate over a wide range of both Re and roughness. If
the velocity increases in the centre of the pipe It must decrease near the wall to maintain
confinuity. With the 4-path meter when the velocity on chord B increases that on chord A

13
4 Practical Developments in Gas Flow Metering Workshop

decreases. The decrease on chord A is greater than the increase on chord B but this is
compensated for by the weighfing factors to give virtually the same fiow. The 4-path meter
showed a maximum error of 0.06% for fully developed profiles.

Zanker also made use of the ratio VBCA/AD = (V9 + Vc) / (VA + VQ) to examine the velocity
profile that exists in the pipe, and used the rafios VBA/C and VA/VQ to check for symmetry. In
reasonably well-developed axis-symmetric fiow conditions we would expect VBA/C = 1, VAA/D
= 1 and VBC/VAD to vary ft-om 1.10 to 1.18.

The paper concludes by saying that four velocity measurements of the profile are a very
useful diagnosfic tool provided by the meter. The integration technique still gives the correct
fiowrate while the path veloclfies indicate changes in velocity profile. The better accuracy and
superior diagnosfic ability of the 4-path meter enabled Zanker to claim that it is suitable for
fiscal measurement, while the single diametrical path cleariy is not.

Moore found that although some profiles had a discontinuity at 9 = 0, 2iz, they were sfill useful
in obtaining informafion about meter performance. One aspect of future work is to replace
this with a continuous funcfion. Moore concluded that increasing the number of traverses
decreased the sensitivity of the configuration to orientation within the profile. The 5PTST and
the DTRI were both particulariy insensitive to orientation and asymmetry. The same
configuration but using more transducers could reveal much more diagnostic information
about the flow at the meter.

In the article published by Moore, Brown and Sfimpson in the MST journal [19] it is also
pointed out that the weighfing of individual path velocities could be chosen to control errors,
especially for extremely non-uniform or asymmetric flow profiles. Although numerical profiles
are advantageous for many reasons CFD and/or experiment must support the results.
Discrepancies were encountered between experimental data and CFD results downstream of
a doutile bend. The model of the ultrasonic meter used assumed negligible thickness for the
ultrasonic path instead of a volume, which is the real case. The model of the ultrasonic path
could be improved to reduce this anomaly.

Barton and Moore [5] used CFD to model Installation effects on ultrasonic meters and then
looked at ways to use the data ft'om multi-path meters. The individual path velocifies were
found to differ downsti^eam of the pipe configurations and this Information was used to classify
the type of flow. A simple neural network of the single layer perceptron type was used In the
ciasslficafion using ratios of velocities as described by Zanker above. Barton and Moore were
able to successfully classify parallel single bend, perpendicular single bend, double out-of-
plane bend, expansion, contraction and fully developed profiles through their multi-pati^
meter.

Although classification was successftjl there was no discussion of a method of compensafion


for the error with respect to each pipe configuration. The long-term aim of the work was to
develop an ultrasonic flow metering system, which can automatically identify and compensate
for installation effects.

The aim of this current work is to confinue the work of Moore and enable the diagnostics to
not only identify the regime but also to apply corrective weights to necessary path velocities to
compensate for the disturbance. Moore used two connected perceptrons which is much the
same as using only one but the outputs available are increased from two to four. Moore used
the ratio of path velocities in a plane as inputs to the perceptron network and successfully
classified four profiles: double bend, single bend in parallel plane, single bend in
perpendicular plane and fully developed flow. This method was only useful however in
asymmetric fiow; axisymmelric fiow, such as expansions and contracfions, would not stand
out in this way. Moore added an extra, diametrical, chord at 45 to measure the flatness of
the velocity profile. A contraction in pipe diameter flattens the profile, from the Ideal fully
developed case; and an expansion has the opposite effect

14
4**' Practical Developments in Gas Flow Metering Workshop

6 CURRENT METHODS IN USE

The current work began by planning to reproduce tiie model of an ulfi-asonlc flowmeter used
by Moore combined with the asymmefi-ic velocity profiles suggested by Salami. Brown had
also created a similar model, which was useful to consult in the generation of the new
program. The language used was Mafiab 5.3. which is a programming package that treats all
variables as matrices and hence performs many operafions at once without much need of
Torioops[163.

The model of the velocity profiles had to have V(r,9)=0 at r=1 due to the no-slip condition. A
grid of points had to be declared so the values of r and theta could be used to determine
values of the velocity at each point in the circle. The ultrasonic paths would then be
superimposed on the velocity distribution and the velocity along each path could be
calculated. The paths were in the samie positions as in currently available ultrasonic
flowmeters, and were weighted accordingly, therefore the path velocities could be combined
to produce a total fiowrate figure. The advantage of the Salami profiles is that they can be
integrated over their area to give a total value equivalent to flowrate. From these two
numbers an error could be calculated for each configuration of ultrasonic paths.

Another factor to be considered is tiie orientation. In pracfice, the alignment of the paths with
the asymmetric profile could be at any of 180 orientations. This sensitivity would also be
included in the model by rotating the ultrasonic path arrangement through 180*. The result
would be an output of error values at each orientation, which would not only show the
sensitivity of the path arrangement to asymmetry but would also highlight any sensitivity to
orientafion.

Since the work will be building on the work performed by Moore the first step was to
reproduce the results of Moore's model. Moore used the fourteen Salami proftles which best
simulated various characteristics of asymmetric fiow. For comparison with ideal behaviour
the author has also included a fully developed profile with no asymmetry. The equations for
these profiles can be found in Table 4 in the appendices.

Arrays of r and 9 values were produced as input arguments for the velocity profile equafion.
This gave a much better distribution of data points than a Cartesian x-y grid.

Each profile has a value for Its total flowrate. Although this could be computed accurately,
since the profiles are integrable analytically, they all have different values of total flowrate
and so make comparisons between profiles unproductive. To eliminate this problem the
profiles were all normalised with respect to total flowrate. This was performed by dividing the
total flowrate value by itself, which had the result that all the profiles now had a total flow
equivalent of unity.

The profile plots displayed in Section 2 are in 3D format to enable the user to obtain a feel for
their nature. 2D contour plots of the velocity values will be used in the actual analyses.

A problem associated with the integration is that the values of V(r,0) are only available at the
predetermined values of r and 6 within the grid. This means the accuracy of the .model will
only ever be as good as the definition of the grid. i.e. the co-ordinate spacing. However, if
this were increased to make the model more accurate the grid definition would be increased
over the entire profile, which would increase computation time needlessly.

Another problem to overcome was finding a suitable method of obtaining V(r,e) values. This
was originally achieved by defining a path by where the transducers would be on the outside
of the pipe. This was done using angular displacement around a 360 circumference. For
example, a diametrical path may have its first transducer at 91=0" and the second at 92=180.
A mid-radius path would have transducers at 9i=30 and 92=150". The straight-line path
between these points could then be divided into a set number of intervals, which is controlled
by the user depending on the required accuracy.

15
4^^ Pracfical Developments in Gas Flow Metering Workshop

The above method of defining a path allowed the program to be upgraded to trapezoidal
Integration, which is more accurate than rectangular rule. A problem still existed however in
as much as the user had to input the desired accuracy at the start of the run and all
calculations were then performed to this accuracy. This would increase computational time
needlessly as not all paths would need the same number of Iterations to reach the same level
of accuracy. Paths crossing steep gradients would need more iterations than paths
traversing relatively flat portions of the profile.

The integration method was improved to adaptive trapezoidal rule, which repeats each
integration with an Increased number of intervals unfil the desired tolerance has been met.
This tolerance was set to 1.0E-4, which equates to 0.01% of total. This was chosen as a
suitable value for the accuracy as transit-time ultrasonic fiowmeters In Ideal installations have
uncertainties in the order of 0.1% and the model needs to be at least a factor of ten better
than in practice to be worthwhile.

The user was given the opfion to choose from six widely used transducer configurafions:
double triangle (DTRI), 5-pointed star {5PTST), Tailored 4-path (TA4). Gaussian 5-path (G5),
Tchebychev 5-path (TCH5) and Gaussian 8-path (G8). The individual path velocities would
then be calculated as described above. These path velocities would then be divided by the
chord length, which results in the Integrated measured velocity. For the first two
configurations the mean of the measured velocities Is used to obtain a total flowrate figure.
For the latter four the measured velocifies are weighted and summed and the total divided by
TT to obtain the final value of flowrate.

The original angular positions of the transducers are shifted around the circle by two degrees
and the whole process is repeated unfil results are obtained for 90 orientations. Example
graphs of error against orientation can be found In the appendices.

The model had to be checked in order to show it was producing worthwhile results. One
method used was to compare results within those published by Moore. Using profile A2 and
configurafions TA4, G5 and TCH5 and comparing the results from the model with those
produced by Moore most points are within 0.04% of Moore's values. The largest deviation
from Moore's results is 0.07%. All points lie well within 0.1%. The TA4 results show the best
correlation. The discrepancies could be due to minor differences in tolerances in the
integration routines or different start/finish points when rotating the transducer model through
180 degrees. A more suitable check had to be used.

To ensure the model was working correcfiy it was tested against profiles where path
velocifies could be checked analytically. The test profiles used for validation were a cone
where V(r,9)=(1-r) and a hemisphere where V(r,e)=(1-r^).

The table below shows the path velocifies across a cone and a hemisphere at certain
abscissa against those obtained with the model.

The maximum deviation from true is 0.0158%. This is above the 0.01% tolerance for the
model we set ourselves above. This is found in the measurement of the Gaussian 5-path
meter on paths 3 and 5.

It Is thus concluded that although the model is performing sufficiently well to be used for
continuing the study of installation effects on ulfi-asonic fiowmeter performance the
integration routine needs to be improved.

16
4"" Practical Developments in Gas Flow Metering Workshop

Table 2: Verification of trapezoidal routine

Conf. Path Abscissa Chord V(r,9) cone V(r,9) hemisphere


theory model theory model
DTRI 1-6 0.500000 1.732050 0.309916 0.309910 0.500000 0.499992

5PTST 1-5 0.309017 1.902113 0.407490 0.407485 0.603006 0.602996

TA4 1&3 0.406735 1.827094 0.359790 0.359782 0.556378 0.556368


2&4 0.889275 0.914746 0.072946 0.072955 0.139459 0.139479

G5 1 0.000000 2.000000 0.500000 0.500000 0.666667 0.666656


2&4 0.538469 1.685290 0.288369 0.288335 0.473368 0.473355
3&5 0.906179 0.845788 0.061929 0.061939 0.119226 0.119247

TCH5 1 0.000000 2.000000 0.500000 0.500000 0.666667 0.666656


2&4 0.500000 1.732050 0.309916 0.309910 0.500000 0.499992
3&5 0.866025 1.000000 0.088021 0.088019 0.166667 0.166664

G8 1&5 0.183435 1.966064 0.459257 0.459253 0.644234 0.644224


2&6 0.525532 1.701548 0.295658 0.295686 0.482544 0.482579
3&7 0.796666 1.208840 0.132436 0.132454 0.243548 0.243578
4&8 0.960290 0.558003 0.026367 0.026363 0.051895 0.051890

7 IMPROVEMENT OF THE MODELS

Although the adapfive trapezoidal integration was an improvement on the original model, and
was used by Brown and Moore in their separate works, Mafiab has built-in functions called
"quad" and "quadS" fiiat use adaptive Simpson's rule and adaptive Newton Cotes 8 panel
rule respectively. These algorithms are more accurate than trapezoidal rule and more
efficient. They also use fewer lines of code as long as the correct syntax can be implemented
[23].

The correct syntax for calling "quad", (or "quadS"), is:

Vpath = quad ('velocity', min, max, tol, h'ace, a, b);

where Vpath is the path velocity required by the integration,


quad is the call to in-bullt fijnction "quad",
'velocity' is the function containing the velocity profile equations for V(r,6),
min Is the lower limit of the integration variable,
max is the upper limit of the integrafion variable,
tol is the tolerance or accuracy required,
trace plots the integrand evaluafions, and
a and b are additional input variables Uiat may be required within 'velocity*.

The problem that arose when trying to implement this code was that a suitable co-ordinate
system had to be defined so that min and max would only be functions of a worthwhile, single
variable.

If we kept to our polar co-ordinate system then it sounds sensible to have r being the
integration variable, which works well if we have a diametrical path as r varies from +1 to - 1
(keeping usual sign convention). This falls down, however, if we have a path that does not
pass through the centre. As the integration is performed from r^ax to rn^n the values would

17
4*^ Practical Developments in Gas Flow Metering Workshop

pass through zero, but this is not tiie case of r itself for a non-diametrical path. The geometry
could be converted back to an x-y co-ordinate system which would work if for example x was
0.5 and y varied from +V(1-x^) down to -V (1-x^). Again, however, if the path is at an angle x
varies as well as y when traversing along the path.

It was decided that because the x-y system would work for all vertical paths a similar
Cartesian method should be used. If we redefined our co-ordinate system in terms of u, v
and (J) we could define a path by its offset from the centre, u, and its angle to the vertical, ^. v
would then be a measure of our position along this path.

The co-ordinate transformafion is such that:

x=u*cos((t)) - v*sin(tt)), and


y=u*sin((l)) + v*cos(it).

As with the x-y system fiie maximum and minimum values of v are calculated using Vmax=
V(l-u^) and v,TOi= -^ (1-u^). It can also be seen that the chord length is simply 2*Vmax-

This enabled the use of "quad" and "quadS" but both returned error messages similar to:
"Warning: Recursion level limit reached in quad. Singularity likely." This is due to the velocity
profile functions having the form:

U = (1-r) ^'" + m r (1-r) ^^ f(9), where n depends on Reynolds number and Is usually between
5 and 12 (see Section 2). Even with no asymmetry, and the second term being zero, the
Integral contains a singularity because of r being raised to the (1/n).

The current work is planned to continue by defining more suitable profiles without
disconfinulties and singularifies but this did not help us with the above problem. The
algorithms did produce answers so we had to see how good the answers were and If any
other, seemingly more robust, Integration algorithms existed.

In the paper "Adaptive Quadrature - Revisited" Gander and Gautschl draw attention to
deficiencies in quad and quad8 [14]. They suggest two new methods for adaptive quadrature
in Matlab. The first, 'adaptsim.m', is an implementation of adaptive recursive Simpson's rule
and the second, 'adaptiob.m' is based on a four-point Gauss-Lobatto formula and two
successive Kronrod extensions. The programs, which were published In BIT, have been
incorporated in the new version of Matiab 6.

These two routines were able to integrate the velocity profiles in their current format without
returning error messages. They were also much quicker and more accurate than the
adapfive trapezoidal rule that had been used until now. The trapezoidal program took 9.7
minutes to perform the Integrations using four paths In 90 orientations and has a tolerance
set to 1.0E-4. The adapfiob roufine took only 4.2 minutes. This Is an Improvement In itself
but it can be improved further by adjusting the tolerances. The default tolerance in Adaplob
is set to 1 .OE-16. If we reduce this to 1 .OE-6 we are still a factor of two more accurate than
the trapezoidal program but the computational time for the exercise now only takes 12.8
seconds. Adaptsim takes approximately three times as long as adaptlob with this kind of
function.

However, the simple cases like the cone and hemisphere have analytical solutions that
enabled us to compare the results obtained with each of the five numerical methods
available: trapezoidal rule, adaptive Simpson's rule (quad), Newton-Cotes 8-panel rule
(quads), adaptive recursive Simpson's rule (adaptsim) and Gauss Lobatto formula
(adapfiob).

It can be seen from Figure 5 in the appendices that despite producing warning messages the
quad and quadS routines are more accurate than their alternatives when calculating Integrals
is this form, t was decided to proceed using the 'quadS' routine. Table 3 below can be
compared with Table 2 In Section 6. This fime the model used is the quadS roufine. The

18
4 * Practical Developments In Gas Flow Metering Workshop

largest error encountered using this integrafion method is 0.0088% which is below are
modelling target of 0.01 %. The largest error Is again found in the Gaussian 5-patii meter, but
this time it is on paths 2 and 4.

Table 3: Verification of quads routine

Conf. Path Abscissa Chord V(r,9) cone V(r,9) hemisphere


theory model theory model
DTRI 1-6 0.500000 1.732050 0.309916 0.309914 0.500000 0.500000

5PTST 1-5 0.309017 1.902113 0.407490 0.407490 0.603006 0.603006

TA4 1&3 0.406735 1.827094 0.359790 0.359787 0.556378 0.556378


2&4 0.889275 0.914746 0.072946 0.072944 0.139459 0.139460

G5 1 0.000000 2.000000 0.500000 0.500000 0.666667 0.666667


2&4 0.538469 1.685290 0.288369 0.288343 0.473368 0.473368
3&5 0.906179 0.845788 0.061929 0.061927 0.119226 0.119227

TCH5 1 0.000000 2.000000 0.500000 0.500000 0,666667 0.666667


2&4 0.500000 1.73205 0.309916 0.309914 0.500000 0.500000
3&5 0.866025 1.000000 0.088021 0.088021 0.166667 0.166667

G8 1&5 0.183435 1.966064 0.459257 0.459258 0.644234 0.644234


2&6 0.525532 1.701548 0.295658 0.295656 0.482S44 0.482544
3&7 0.796666 1.208840 0.132436 0.132436 0.243548 0.243549
4&8 0.960290 0.558008 0.026367 0.026366 0.051895 0.051895

8 FUTURE RESEARCH

The research project detailed In this paper utilises mathematically formulated velocity profiles
along with computational fluid dynamics software to determine the opfimum configuration for
transducers. It is also planned to categorise the profiles experienced downstream of a
number of installation condifions for use with neural networks. This will enable the meter
diagnostics to not only determine the type of profile entering the metering section, but also
apply appropriate weights to the path velocities to compensate for the asymmetry.

The theoretical profiles suggested by Salami and used by Moore will be re-examined to
detennine their suitability for this study. It is believed that it will be possible to approximate a
small range of disturbed profiles caused by common installation conditions such as a bend,
double bend, expansion, contraction and valve by these analytical profiles.

Software has been written that can model the Salami profiles as well as superimpose an
ultrasonic path at any offset and orientation. The user can also determine the number of
measuring paths used. The software will be developed further so that the ofi'set and
orientation are varied automatically and the configurations with the least error can be
selected for further examination. It is hoped that a particular configuration will prove to be
least sensifive to a number of asymmetric profiles, and this configuration could be used for
extending the work to include varying the weights applied to each path. The differences in
path velocities experienced when using asymmetric profiles could provide valuable
Information about the profile to the diagnosfics and possibly be used to idenfify the con-ective
weights required to produce a useful measuremenL

19
4"" Practical Developments in Gas Flow Metering Workshop

The following points are suggested as a way fonward for the next stage in continuing this
work:

1. It has been shown,ftromthe work by Moore, that the DTRI, 5PTST, G8, G5, TCS and TA4
configurations were least sensitive to asymmetry In most cases. It is proposed that these
six configurations be used as a starting point for the work planned by the author.

2. The theoretical profiles suggested by Salami and used by Moore will be re-examined to
determine their suitability for this study. New profiles could be designed that better model
the real flows encountered in practice. It is believed at this earty stage that it will be
possible to generate a range of profiles that model disturbances caused by common
installation conditions such as a bend, double bend, expansion, contracfion and valve.

3. A new performance measure will be designed to enable the use of simple optimisation
techniques, which will provide figures of merit for configurations against a variety of
asymmetries and orientations.

4. Once ranked in order of performance the three flowmeter configurations with the best
overall scores will be studied ftjrther. The first phase will be to determine correcfive
weights for the paths in a large number of asymmetric fiows and then to produce tables of
path velocities that can be used to determine which set of weights needs to be used to
calculate the correct flowrate. Neural networks will be used to classify the different flow
profiles.

5. The next stage of the project will be to use an optimisafion package that varies the
posifion of the paths as well as the weighting factors. This will provide us with a
configuration that Is relatively insensitive to variations In flow profile and orientafion.

6. The closing stage will be to use the neural network package alongside the optimisafion
package. This will provide a design for an ultrasonic transit-time flowmeter with opfimlsed
path positions, that give least sensitivity to asymmetric profiles, and a capability for
Identifying the nature of the asymmetry, using rafios of path velocities. Finally, utilising an
adaptive corrective weighting method, the flowmeter will produce the correct total flowrate
regardless of profile entering the meter.

9 ACKNOWLEDGEMENTS

Chris Duffell is an Associate of the Postgraduate Training Partnership (PTP) between Uie
Nafional Engineering Laboratory (NEL) and the University of Strathclyde. The PTP scheme is
a joint initiative of the UK's Department of Trade and Industry (DTI) and the Engineering and
Physical Sciences Research Council (EPSRC). The PTP scheme is supported by a grant
from the DTI. Chris Dufl'ell gratefully acknowledges grant support from both the EPSRC and
DTI.

Chris joined NEL In 1997 as a project engineer following successful completion of MPhys
(Hons) in Astrophysics from University of Wales, College of Cardiff. His training and work at
NEL has involved many areas of project management and flow measurement project work.
Chris was a key member of the team responsible for the development of the NEL high-
pressure wet gas test facility, which was commissioned in 1999. Chris has worked with many
of the test facilifies in the gas and oil laboratories. In November 2000, The Institute of Physics
awarded Chris Chartered Physicist status. In September of 2001 he left full fime employment
at NEL to become a PTP Associate to further his career in flow measurement by attaining a
PhD in elecfi'onic engineering research.

20
4 * Practical Developments in Gas Flow Metering Workshop

10 REFERENCES

1 "In the Pipeline". Forthcoming events list in NEL's quarteriy newsletter. Flow Tidings,
Issue 30. Nafional Engineering Laboratory, East Kilbride. Spring 2 0 0 1 .

2 A.G.A. Report No. 9 Measurement of Gas by Mulfipath Ultrasonic Meters. 1998.

3 BAKER, R. C. Flow Measurement Handbook. Cambridge University Press. 2000,


UK.

4 BARTON. N. A, and BROWN, G. J. Velocity Distribufion Effects on Ultrasonic


Flowmeters, Part 2 - Determination by Computational and Experimental Methods.
Report No 358/99. National Engineering Laboratory, East Kilbride, September 1999.

5 BARTON. N. A., and MOORE, P. I. Modelling and evaluation of installation effects on


ultrasonic flowmeters. Proc. of Non-lntnjsive Flow measurement Workshop, NEL,
East Kilbride, 1999.

6 BRAGG. Proc. of Non-Intrusive Flow measurement Workshop, NEL, East Kilbride,


1999.

7 BROWN, G. J., BARTON. N. A., and MOORE. P. I. Installation effects on ulfi-asonic


flowmeters. Proc. NSFMW 1999, Nonway,

8 BROWN, G . J . Developments in Ultrasonic Flow Measurement. Flow Tidings, Issue


13. National Engineering Laboratory. East Kilbride, February 1997.

9 BROWN, G. J. Proc. of Non-infi-usive Flow Measurement Workshop. NEL, East


Kilbride. 1999.

10 BS 7965:2000. The Selecfion, Installation, Operafion and Calibration of Diagonal


Path Transit Time Ultrasonic Flowmeters for Industrial Gas Applications.

tt BS ISO/TR 12765:1398. Measurement of fiuid fiow In closed conduits - Methods


using Transit Time Ultrasonic Flowmeters.

12 DELSING, J. A review of developm.ent trends in ultrasonic flowmeters. Proc. Flow


Metering and Control, FCRI. India, Sept 2000.

13 FOULKES, G. Justification for selecfion of Ultrasonic Flowmeters. Proc. of


PDGFMW 2000.

14 GANDER, W., and GAUTSCHI, W. Adaptive Quadrature - Revisited.1998.. BIT Vol.


40, No. 1. March 2000, pp. 8 4 - 1 0 1 .

15 GRIFFIN, D. Offshore Gas Metering: The Regulator's View. Proc. of PDGFMW 2000.

16 HANSELMAN, D., and LITTLEFIELD, B. MATLAB Version 5 User's Guide. The


Math Works Inc. Prentice Hall, NJ 07458.

17 JAMIESON, A. Implementafion of New Techniques for Gas Metering. Proc. of


PDGFMW 2000.

18 LEIGH, M. Future of gas fiow metering ofl^shore. Proc. of PDGFMW 1999.

19 MOORE, P.I.. BROWN, G.J., and STIMPSON, B. P. Ultrasonic transit-time


flowmeters modelled with theoretical velocity profiles: methodology. Meas. Sci.
Technol. Vol 1 1 . lOP Publishing, 2000.

21

- ' 7 : '
4"^ Practical Developments in Gas Flow Metering Workshop

20 MOORE, P. I. Modelling of installation effects on transit fime ultrasonic flow meters in


circular pipes. PhD Thesis. University of Strathclyde. 2001.

21 MOORE. P. I. Modelling of installation effects on transit fime ultrasonic flow meters in


circular pipes. Proc. of Developments In Non-intrusive fiow measurement, NEL. East
Kilbride, Nov 2000.

22 PETERS. R. J. W. Practical Application and Problem Solving. Proc. of PDGFMW


2000.

23 PRATAP. R. Getting started with MATLAB 5. New York. Oxford. Oxford University
Press 1999.

24 SALAMI, L. A. Application of a computer to asymmetric fiow measurement in circular


pipes. Trans. Inst. Meas. Control Vol 6, No 4 pp197-206, July-Sept 1984.

25 SPEARMAN, E. P., and SATTARY, J. A. A Study on the effect of pipe installafions


on the accuracy of flowmeters. Report No 468/97 pp 27-34. National Engineering
Laboratory, East Kilbride. March 1997.

26 STOBIE, G. J. Wet Gas Metering In the Real Worid: Part 2. Proc. of PDGFMW 2000.

27 ZANKER, K. The effects of Reynolds number, wall roughness and profile asymmetry
on single- and multi-path ultrasonic meters. Proc. of NSFMW. Norway, 1999.

22
ith
..cfical Developments in Gas Flow Metering Work >p

11 APPENDICES

Table 4. Analytical description of fourteen flow profiles. Parameters and functions as described by equation (4).

Profile Radial Function Angular Function U

Group Code Code Exponent Exponent Exponent Exponent f(0 2T

(Moore) (Salami) n k m a \f{0)dO


0

Al P6 9 4 _ OS - (^\n9 -2n
A2 P17 7 9 _ M - ttelnty -271
n
A3 P8 9 4 ().04 - (f/-l)(l-cos(9)' 4;r' 2;"
A A4 PI 9 0.5 3.3170 0.5 e'^^sin^? (1-e-'") (1-0-)
(8^1) - ^ 1.26
One peak A5 P12 7 9 0.2 e^^sinO (i-''" _ , (I-B"")
2 (o'.l) 'W
A6 P13 7 9 0.2 e^^sinO [ l - e ' " ' ^ (l-*'*l
2 (fl'.l) ^ 'W
A7 P9 9 4 2 (f{2-^-0f 15

A8 P7 9 4 1 - ay-cos'-O) n^
^1

B A9 P10 9 4 2 - (\2n-0)^\W-0 2' ,


3 "^ 2
Two peaks A10 P5 9 9 0.6813 0.1 e^"s\n'e

All P16 7 9 00.1, 0.2 e'^^sin^^

A12 P2 9 0.5 -6.7501 0.5 e'^sini? (1,8-'") ^ (l-e-)


(a'.l) " " 125
C A13 P22 7 9 1 - {2TK-0)^(^\nZO 4r
2,' 9
More than A14 P20 4 9 0.3 e-^%\n^^O 50(1-<|-'") _ (I-B"'")
2 fl(a'.100t " 060054
two peaks

23
4'^ Practical Developments in Gas Flow Metering Workshop

Figure 3: Error vs orientation for profile A2

1.01

1.005

1
o..-^5lig:;5;555|m|;||;;;;555|5|5^^^
'^xi^^^'^y^xx><^o^
0.995
nAAAA' 'AAA
DTRI
0.99 ^^>?* 4 ,
T3 5PTST
C
ATA4
0.985 -
u XG5
5 XTCH5
G8
0.98

0,975
0.97

0.965

0.96
20 40 60 80 100 120 140 160 180
Orientation

24
th
4 ctlcal Developments in Gas Flow Metering Wor jp

Figure 4: Comparison against Moore's model, using profile A2.

1.001

1.0008
% . <
1.0006
-^.*V!^

1.0004 H

1.0002
o
o
"Gaussian 5
S 1 Tailored 4
i Tchebychev 5
o
> 0.9998

0.9996

0.9994 -

0.9992

0.999
20 40 60 80 100 120 140 160 180
Orientation

25
4 Practical Developments in Gas Flow Metering Workshop

Figure 5: Comparing Numerical Methods using a Cone

1.0000020000

1.0000015000

.| 1.0000010000
Quad
Quads
o -Adaptsim
a
o XAdaptlob
E 1.0000005000
>

1.0000000000 m-TTTTnT-''*rTnT'"TrrnnxrrnTrr^^
X

0.9999995000 I r

10 20 30 40 50 60 70 80 90 100
Abscissa

26

You might also like