1

ANALYSIS OF DIAGNOSTIC DATA FROM AN 8-PATH ULTRASONIC METER

Dr Gregor J Brown, Director of Application Engineering, CameronWilliam R Freund, Principal Engineer, Cameron

1 INTRODUCTION

Multipath ultrasonic meters were first developed for gas custody transfer applications in the mid to late1980’s. The technology offered significant benefits over traditional orifice metering in terms of increasedrangeability and reductions in pressure loss, upstream straight length requirements, and routinemaintenance. It was also hoped that based on measurement on meter geometry and correction for non-fluid timing errors in the signals, determined during factory bench testing, it would be possible to useultrasonic meters without flow calibration, in the same way that orifice meters are still used today.

In practice, technology and design limitations, coupled with a drive by the industry towards lowermeasurement uncertainties, have resulted in a situation where not all of the potential benefits of ultrasonictechnology have yet been harnessed.

In particular the elimination of the need for long upstream straight lengths has generally been achieved byuse of flow conditioning devices, typically of the perforated plate design. This in turn negates a large partof the reduction in pressure loss, and also introduces a maintenance requirement, as either the plate canbecome blocked with debris, or a filter is required upstream to protect the plate. In the latter case thefilter then introduces additional pressure loss and maintenance requirements.

Other issues that have been reported are the failure of transducers, particularly those made with epoxyparts exposed to the gas, and concern over the effects of corrosion and/or deposition or fouling on theinterior of the meter body.

At the 2013 AGA Operations conference a paper [1] was presented showing test results obtained as partof the process of certifying an 8-path gas ultrasonic meter to the requirements of AGA9, ISO17089 andOIML R137. Of these standards, in terms of installation effects the requirements of OIML R137Accuracy Class 0.5 are the most stringent. On the basis of those tests the 8-path meter has been certifiedas meeting the requirements of OIML Accuracy Class 0.5 when installed only 5 diameters downstream ofbends, tees and reducers, including single bends and out-of-plane combinations, without use of a flowconditioner.

There is an argument that says that flow conditioning is useful for two purposes: firstly to reduce theinfluence of upstream conditions on measurement uncertainty; and secondly to provide a known‘baseline’ for velocity diagnostic analysis, so that changes can used to identify potential problems. The 8-path meter configuration without a flow conditioner has been shown to outperform 4-path (and other)meter configurations even when latter are used with a flow conditioner. This challenges the first point ofthe argument above, i.e. it demonstrates clearly that there are ways to reduce installation effects by virtueof employing a first-principles approach to meter design. In other words, a good meter design can, bothin principle and practice, achieve better results without flow conditioning than a poorer design that isreliant on a flow conditioner.

In terms of addressing the second point of the argument in favour of flow conditioning, we could assertthat with the improved meter design monitoring for apparent velocity profile changes becomes of much

2

lower importance. However, even if we accept the argument that diminishing the influence of upstreamhydraulics reduces the need for monitoring, it is still worthwhile considering if something is lost in termsof velocity profile diagnostic monitoring capability when an 8-path meter is employed and therequirement for the flow conditioning is removed.

This paper aims to explore that issue through detailed examination of the velocity diagnostic dataobtained during the testing that was described in the 2013 AGA paper. Relative to the 2013 paper there issome repetition and summarisation here in order that this paper can be read without requiring access tothe 2013 paper. However, for more details of the testing the reader should refer also to that paper [1].

2 LIMITATIONS AND ADVANCES IN MULTIPATH ULTRASONIC DESIGN

Multipath ultrasonic meters have been in continuous development since the 1960’s. In early publicationsand patents, it was noted how multipath meters that employ numerical integration methods couldsignificantly reduce the sensitivity to distortions in the axial velocity profile caused by upstream hydraulicdisturbances. Studies of the accuracy of the numerical integration methods have shown that chordalmeters with four chords spaced according to the rules of Gaussian integration could typically be expectedperform with errors of less than one or two tenths of a percent.

In the earliest implementations of chordal integration schemes, it was common to place only onemeasurement path at each of the prescribed chord locations. In the patents and papers of Westinghousepublished in the 1970’s [2, 3], the paths of their Leading Edge Flow Meters (LEFM) were shown asresiding a single plane, typically angled at 45° to the pipe axis, as illustrated in Figure 1 below.

Figure 1 Illustrations of the Westinghouse multipath meter patent

An individual path at an angle of 45° is sensitive not only to the axial flow velocity but is equallysensitive to any non-axial component of flow such as that generated by pipe bends. The result is that indisturbed flow conditions where swirl or non-axial flow exists, the inputs to the integration method are inerror, and this in turn results in poorer flow rate measurement accuracy than can be achieved in a non-swirling flow. In some special cases, such as a single-vortex flow that is centred between the two insidepaths of the Westinghouse arrangement the errors cancel, but in general they do not.

3

In the mid 1980’s British Gas (BG) began development of a chordal multipath ultrasonic flow meterintended for custody transfer measurement of natural gas. This design was based on a similararrangement of four horizontal chords to that used by Westinghouse, but with the paths criss-crossed suchthat the first and third paths were at +45° to the pipe axis and the second and fourth paths were at -45°, asillustrated in Figure 2. This design variation has been justified by technical arguments regardingsensitivity to cross-flow, but the fact that the 1976 patent of Westinghouse [3] was still in force in 1986when BG filed for their patent [4], suggests that patent considerations may also have come into play.

Figure 2 Illustrations of the British Gas multipath meter patent

One particular form of disturbance which it has been claimed the BG arrangement is insensitive to, is aform of cross-flow where the relative magnitude and direction of the cross-flow is equal at each of thechord locations in the cross-section [5]. With a Westinghouse arrangement of all chords at the same anglerelative to the pipe axis this would result in a systematic over or under reading, whereas it is shown thatfor the criss-crossing arrangement of paths in the BG design this cancels. However, this is a hypotheticalform of non-axial flow, which is unlikely to occur in practice in closed pipes, as in reality any disturbancethat creates a cross-flow in one part of the cross-section likely to create a counter-rotation in another part.

A more realistic form of cross-flow is that produced downstream of a single bend, where there is a strongcross-flow in the plane of the bend in the form of two counter rotating vortices. The BG design differsfrom the Westinghouse design in its response to this situation in that the BG design would in principle beinsensitive to the presence of these two counter rotating vortices if those vortices were symmetrical aboutthe line that is centred between paths B and C. However, in practice, owing to a combination of factorsincluding effects from components further upstream, small asymmetries in bend geometry and the factthat the flow wants to recover to a fully developed condition, it is virtually impossible to create twosymmetrical counter-rotating vortices. This is borne out in the results presented in the 2013 AGA papermentioned in the introduction. In the case of the single bend, with both the bend and the paths of themeter aligned horizontally, the resulting errors for the BG 4-path arrangement were significant, and muchlarger than for the Westinghouse 4-path arrangement [1].

Single-vortex swirl is another basic ‘test case’ for the path layout in an ultrasonic meter. In theWestinghouse 4-path arrangement, if the single-vortex swirl is symmetrical about the centre of the pipe,then the effect on path 1 would exactly cancel with that on path 4 and likewise the effect on path 3 wouldcancel with that on path 3. This is because the magnitude of the swirl would be the same in the top andbottom of the pipe but the swirl direction would be opposite relative to the path angle. For the BG designthe effect of single-vortex bulk swirl cancelation relies upon a mathematical quirk of the design, wherebyif a solid-body rotation of the flow is assumed, the combined effect on the outside paths (paths A and D)cancels with the effect on the inside paths (paths B and C). The two inside paths in the BG design see theswirl from the same direction and the two outside paths see the swirl from the opposite direction.However, true cancellation does not result in the case of single-vortex swirl, even when that properlycentred and symmetrical as BG design relies on the magnitude of the swirl effect on the inside paths

4

versus the outside paths being in inverse proportion to the weighting factors. Owing to mathematics ofcircular geometry that assumption holds true if the swirl is a solid-body rotation of the fluid. However asZanker has pointed out, in practice that particular case is unrealistic as the swirl must have its ownboundary layer and go to zero velocity at the pipe wall [6].

In the 1990’s a gas ultrasonic meter with five chords was jointly developed by Statoil and Fluenta (then asubsidiary of Christian Michelsen Research). The original design had a criss-crossing arrangement ofpaths, with paths 1, 3 and 5 in the same plane and paths 2 and 4 in the opposite plane. However, a fewyears later the meter design was altered to a 4-chord, 6-path design in order to account for the adverseeffects of symmetrical double-vortex swirl. The new Fluenta/FMC design placed two crossed paths ineach of the chord locations in the top half of the pipe, and one path in each of the chord locations in thebottom half of the pipe. This configuration has the benefit of tackling both a single-vortex swirl and thecross-flow caused by symmetrical double-vortex swirl, but similar to some of the 4-path cases discussedabove it is truly insensitive only if the vortex pattern is symmetrical about the diametric plane that isparallel with the chord arrangement.

Throughout the 1990s and into the 2000s numerous laboratory tests were carried out on ultrasonic metersfor the natural gas industry. Particularly notable are the programmes of the Gas Research Institute in theUSA [7] and GERG in Europe [8]. These tests exposed the weakness of 4, 5 and 6-path configurations insome installation configurations and demonstrated that for these particular designs, using either direct orreflected paths, a flow conditioner is generally needed if the requirements of today’s standards are to bemet.

Despite the clear recognition in the natural gas industry of the importance of installation effects onultrasonic meters, it appears that developments in other industries either went unnoticed by the gas metermanufacturers, or if developments were noted by some, they chose not act to improve their meter designsowing to other considerations. The use of flow conditioners has therefore become a de facto standard inmany parts of the industry today despite the stated aim in the BG patent to have a solution that “causes noblockage to the flow and generates no pressure loss”. Moves towards including the ‘end treatments’ ofthe metering package in the calibration in addition to the meter run and flow conditioner represent afurther departure from the original promise of ultrasonic technology.

As mentioned in the start of this section, the advantage of the Gaussian integration method, if a sufficientnumber of chords are used, is that it is relatively insensitive to distortions of the axial velocity profile. Itwas also stated that the main problem that prevents the method from achieving its potential is theinfluence of non-axial flow or swirl on the individual paths that are used to provide the axial velocityestimate to the integration method. This problem was recognised early on by Westinghouse andORE/Accusonic who were deploying their ultrasonic meters for large-scale measurements in rivers,hydroelectric and nuclear plants.

A description of the solution can be found as far back as the 1977 publication by Lowell [9] where theauthor highlighted the influence of non-axial flow and stated that the resulting error “can be reduced bythe addition of one or more acoustic paths, at the same elevations as the original ones but installed at theopposite angle. Exact cancelation of errors can be accomplished on the crossed paths and an estimatedof the cross-flow component used to adjust the readings on the non-crossed paths.” The significance ofthis statement is that it encourages pairs of crossed paths at each elevation used in the integration method.It also highlights that for paths that are not crossed in the same elevation the cross-flow can only beestimated by making some assumptions.

The way that swirl or cross-flow interferes with the measurement of axial velocity and how a pair ofcrossed paths solve the problem can be described quite simply. Swirl or cross-flow introduces an

5

unwanted non-axial component of velocity to measurement path. This unwanted component of velocitycan be additive or subtractive. If the non-axial flow velocity is going in the same direction as theultrasound when it travels from the upstream transducer to the downstream transducer then the effect willbe to increase the measured velocity, as illustrated in Figure 3 below. If the non-axial velocity is oppositein direction to the downstream travel of the ultrasound then the effect will be to decrease the measuredvelocity.

Figure 3 The influence of non-axial flow on an ultrasonic measurement path

As a result, a crossed pair of paths located on the same chord allows the true axial velocity data to berecovered, as illustrated in Figure 4 below.

Figure 4 An illustration of how crossed paths cancel the effects of swirl

With this understanding of the fundamentals of how these meters work, it is relatively simple to examinedifferent non-axial flow scenarios or swirl patterns and evaluate whether or not the interfering non-axialflow components would cancel partly or fully. This exercise has been performed for a variety of directpath chordal meter designs and the results are shown in Table 1 below. From this table it can be observedthat meters with only single paths in each chordal plane, whether all in the same angled plane with respectto the pipe axis, or in a non-planar criss-crossing arrangement, only cope properly with one particularform of symmetrical swirl. With the addition of a second crossing path at each of the top two chordalplanes, the 6-path arrangement is able to cope with both forms of symmetrical swirl but still has problemswith asymmetric swirl patterns. However, it is only when a second crossing path is added to each of the

Actual velocity

Upstream transducer Downstream transducer

Axial component (wanted)

Transversecomponent(unwanted)

Measured velocity

1 up5 down

1 down5 up

Actual velocity

Axial component (wanted)

Transverse component (unwanted)

Measured velocity

Path 1 Path 5

Path 1 + Path 5

Path 1 + Path 5

2

Key:

=

6

chordal planes and every crossed pair works together to cancel the effects of swirl that the meter design isable to cope with swirl of any form.

Table 1 Ability of chordal path configurations to correct for different forms of swirl

The ability of the 8-path configuration to cope with a wide variety of disturbed installation conditions hasbeen evaluated in numerous analytical, computational and laboratory studies by the meter manufacturersand third parties. In circular pipes both ORE/Accusonic and Westinghouse deployed 8-path meters withpairs of crossed paths in each of four chordal planes from around 1980. These meters were designedinherently insensitive to the swirl and cross-flow that exists in applications where flow conditioning wasnot practical. Caldon, as successor to Westinghouse having acquired the LEFM technology fromWestinghouse in 1989, then went on to use the 8-path concept in high accuracy liquid meters first innuclear applications and later for liquid hydrocarbon custody transfer. As a result of this heritage there isa wealth of data validating this design in a wide range of hydraulic configurations, including almost 100meters for nuclear plants that have been calibrated in a grand total of more than 400 installationconfigurations.

In Caldon 8-path meters, a crossed pair of paths located in each of four chordal planes, those chordalplanes being located in accordance with the Gaussian integration methods described in the originalWestinghouse patents. The four chord selection made by Westinghouse was based on extensive researchand although further gains could be made by adding more chords, others have also concluded that a four-chords integration is sufficient to obtain an appropriately small error in integration of the axial velocityprofile.

A recent paper by Zanker and Mooney [10] re-examined the choice of the number of chords from theperspective of velocity profile integration in fully developed and asymmetric flows. The analysis isbroadly in line with work carried out by Westinghouse and others, and the conclusion the authors appearto reach is that increasing the number of chords beyond four is of questionable valve when it comes toobtaining a representative average of the axial velocity profile. However, although the Zanker andMooney paper discusses fully developed, distorted asymmetric and symmetric axial flow profiles and

4 paths, 4 chords,planar

4 paths, 4 chords,non-planar

5 paths, 5 chords,non-planar

6 paths, 4 chords,two crossed chords

8 paths, 4 chords,four crossed chords

1 up 1 down

2 down

3 down

4 down

2 up

3 up

4 up

1 up5 down

1 down5 up

2 down6 up

3 down

4 down

2 up6 down

3 up

4 up

7

factors such as the effect of steep velocity gradients, transducer cavity effects, it neglects to examine theeffects of swirl or transverse flow and gives these only a passing mention. The paper opens with adiscussion involving a 32-path meter design and states later that that eliminating the need for a flowconditioner would be an advantage. In the absence of a discussion of non-axial flow there is a risk thatthe reader could assume that the authors have concluded that increasing the number of paths brings littlebenefit. Adding paths arbitrarily does not necessarily bring a benefit but doing it in a particular way toaddress a problem using a first-principles approach is different. As the purpose of the additional paths inthe 8-path design is to cancel the unwanted effects of non-axial flow and allow the numerical integrationmethod to properly do its work of evaluating the mean velocity, the Zanker and Mooney paper is in factsupportive of the 4-chord integration method employed in the 8-path meter.

3 THE 8-PATH ULTRASONIC GAS METER

The 8-path ultrasonic gas flow meter used for the tests we describe here was a Caldon LEFM 380Ci. TheCaldon brand covers a family of ultrasonic meters manufactured by Cameron with heritage directly fromthe Westinghouse multipath Leading Edge Flow Meters first developed in the late 1960’s.

The arrangement of paths adopted for the Caldon LEFM 380Ci ultrasonic gas custody transfer meter issimilar to that used in Caldon 8-path liquid meters, with the exception that a steeper path angle is used toallow for the effects of high Mach numbers. As illustrated in Figure 5 below, the meter employs 16transducers to form eight measurement paths which are grouped in crossed pairs of paths at each of thechordal locations associated with the 4-chord Gaussian integration method.

Figure 5 An illustration of the path layout in the 8-path Caldon LEFM 380Ci

8

When introducing the LEFM 380Ci product for gas custody transfer, three steps were taken in an effort toadvance the technology in some of the areas where it had previously been lacking in gas meters.

First, the adoption of the 8-path configuration previously described was seen as a necessary step to enablethe meter to perform with high accuracy without the need for a flow conditioner. Eliminating the flowconditioner would not only reduce energy losses, but would also allow metering stations to be reduced insize, and remove the requirement for maintenance of the flow conditioner and the frequently reportedproblem of partial blockage.

Secondly, the meter body and transducers were designed such that each transducer capsule is placed in ametal alloy housing that is integrated into the meter body and fully isolates the transducer from theprocess fluid and pressure. This not only results in a very robust transducer by eliminating failure modesassociated with aggressive chemical components or rapid depressurisation, it also means that if necessarytransducer can be easily removed and replaced without requiring depressurisation of the line. Each metalalloy transducer housing is fully pressure retaining and all work required to replace the transducer is doneon the low pressure side. There is no breach of the pressure boundary and therefore no special extractortools are required; transducer replacement can be performed quickly and safely.

A third enhancement is provided in the form of a proprietary coating that is applied to the inside of themeter to inhibit corrosion and reduce contamination build up inside the meter body. The coating isapplied to the bore of the meter and to the wetted surfaces of the transducer housings. The obvious aimhere is to minimise changes to the interior of the meter would otherwise result in changes to its calibrationover time.

4 PERFORMANCE TESTING REQUIRED BY THE STANDARDS

In order to be accepted for use in custody transfer applications, it is necessary that ultrasonic gas meterscomply with the requirements of the relevant standards. In this case the relevant standards underconsideration are AGA9 (2007) [11], ISO 17089-1 (2010) [12] and OIML R137 - 1&2 (2012) [13].

The above standards describe the performance expectations that have been set for gas ultrasonic metersfor custody transfer applications. In terms of installation effects, AGA9 requires that the “manufacturershall ... recommend at least one upstream and downstream piping configuration without a flowconditioner or one configuration with a flow conditioner, as directed by the designer/operator, that willnot create an additional flow rate measurement error of the meter of more than 0.3% due to theinstallation configuration. This error limit should apply for any gas flow rate between qmin and qmax. Thisrecommendation shall be supported by test data.”

ISO 17089-1 prescribes a series of disturbance tests that are intended to cover a range representative ofthe type of conditions that may be encountered in practice. These include a single bend, out-of-planebends, contractions, expansions and steps. The manufacturer is allowed to specify the length between themeter and the disturbance at which the meter will be tested, and then the meter should be tested at thatdistance and at a second distance that is ten pipe diameters further away. The requirement in ISO 17089-1 is that above qt, all calculated mean additional errors shall be within 0.3 %. For ISO 17089-1, the testshave to be performed at one flowrate below qt and two flowrates above qt. In addition to the installationtests, ISO 17089-1 requires that tests should be performed to evaluate repeatability, reproducibility, theeffect of transducer change out and simulated transducer failure. The general performance requirementsin ISO 17089-1 are very similar to those required by AGA9.

9

A new edition of OIML R137 was issued in 2012. Although the 2012 edition has been partly harmonisedwith ISO 17089, some differences remain, not only in terms of the tests required, but also in theevaluation criteria by which the flow meter is deemed to pass or fail. Unlike the other standards, OIMLR137 allows classification of the meter performance to different levels, the most demanding beingAccuracy Class 0.5. In terms of the installation effect testing, the test configurations have a large degreeof overlap with those in ISO 17089-1, but for OIML the requirement is that “the shift of the error due tothese disturbances shall not exceed one third of the maximum permissible error”; which means in thecase of Accuracy Class 0.5 the shift of the error should be within +/- 0.167 %, which is approximatelyhalf that allowed by AGA and ISO.

In addition to the general requirements of these standards, and the flow tests mentioned above, thestandards also require a series of ‘environmental’ tests be performed to ensure the that metrologicalcharacteristics of the meter are immune to factors such as radio frequency interference, damp heat,vibration and surges on electrical supply lines.

5 PERFORMANCE TEST RESULTS

A comprehensive test programme jointly prepared by Cameron and NMi, the weights and measuresauthority of the Netherlands was performed to cover all the requirements of the AGA, ISO and OIMLstandards, with minimum duplication.

The majority of the flow testing was performed at the CEESI high pressure natural gas calibration facilityin Iowa, USA. All tests were witnessed by NMi as a notified body (issuing authority) for the typeapproval of gas meters according to the requirements of OIML and the European MeasurementInstruments Directive (MID).

The results of the flow tests were described in detail at the 2013 AGA Operations Conference and willonly be selectively summarised here.

The tests were performed with three different upstream pipe arrangements between the prescribeddisturbance and the meter: 5D of straight pipe with no flow conditioning, 15D of straight pipe with noflow conditioning, and an arrament where the disturbance was followed by 5D then a CPA 50E perforatedplate flow conditioner then a further 10D before the meter, as illustrated in Figure 6 below.

Figure 6 An illustration of the 5D, 15D and 5D-CPA-10D upstream pipe configurations

5D

15D

CPA

10

As explained previously the 8-path meter comprises two planar sets of 4 paths with the paths set at thesame chordal heights as in a 4-path design. By making a selection of only some of these paths it istherefore possible to use the 8-path meter to replicate other path arrangements such as a single-plane 4-path arrangement (Westinghouse) or a 4-path criss-crossing arrangements (BG). Figure 7 shows the patharrangements that were evaluated; Plane A and Plane B being of the Westinghouse type, BG1 and BG2being of the British Gas type. In all these evaluations, the abscissa (path chordal heights/locations) andweighting factors, were the same as prescribed by the 1976 Westinghouse patent [3] and later adopted byBG [4] and others.

Figure 7 4 and 8-path configurations selected for evaluation

Arguably the most important of the tests prescribed by ISO17089-1 and OIML R137 are thosedownstream of single and double bends as they are broadly represented of a range of typical pipingconfigurations. The results of the installation effect tests downstream of bends were summarised in the2013 AGA paper in terms of the shift in the flow weighted mean error (FWME) relative to the straightpipe baseline calibration of the same meter configuration. That method of summarising the results is thesame as was used for the data from the GRI and GERG projects referred to in the introduction andenables comparison of different installation/meter combinations on the basis of a single number.

The FWME summary of the data obtained with the Caldon meter in both 4-path and 8-path format isreproduced in Table 2 below. For each meter type and upstream meter run arrangement (i.e. 5D, 15D,CPA), the outer extremes of error shift have been highlighted. This table clearly shows that the flowweighted mean error shifts are lowest for the 8-path meter at 0.08% or less and are typically around onethird of the 4-path planar arrangement. The 4-path non-planar arrangement produces the largest flowweighted mean error shifts, typically around 4 or 5 times greater than the 8-path meter, but larger still inthe 5D configuration. In terms of the flow weighted mean errors, the benefit for the 4-path meters whenmoving from 5D to 15D and then including the CPA flow conditioning plate is fairly clear, but theimprovement for the 8-path meter is not very significant, showing the extremes of +/- 0.06 at 5D reducingto a range of -0.04 to +0.06 % in the 5D-CPA-10D case.

8-path

Plane A Plane B BG 1 BG 2

11

Table 2 Bend summary data in terms of flow weighted mean error shift for 4 and 8-path meters

Perhaps the most important finding when looking at the data in Table 2 is that even when a flowconditioner is used the 4-path meters show FWME shifts that are larger than the results obtained with the8-path meter at 5D with no flow conditioner, as illustrated graphically in Figure 8 below.

Figure 8 FWME performance comparison for the 8-path meter at 5D with no flow conditionerversus the 4-path meters with 15D inclusive of flow conditioning

Disturbance Upstream Path orientation A B 1 2

Horizontal 0.06% -0.08% 0.21% 1.02% -0.90%

Vertical 0.03% 0.00% 0.07% -0.86% 0.93%

Horizontal 0.02% -0.10% 0.15% 1.17% -1.12%

Vertical -0.06% -0.26% 0.14% 0.45% -0.57%

Horizontal -0.08% -0.04% -0.13% 0.30% -0.46%

Vertical -0.05% -0.02% -0.08% -0.61% 0.51%

Horizontal -0.05% -0.24% 0.13% 0.09% -0.20%

Vertical -0.08% -0.06% -0.11% -0.12% -0.05%

Horizontal -0.02% -0.06% 0.02% -0.12% 0.07%

Vertical -0.04% -0.01% -0.07% -0.14% 0.06%

Horizontal 0.03% -0.05% 0.11% -0.11% 0.17%

Vertical 0.06% -0.08% 0.20% 0.12% 0.00%

Planar 4-path

(Westinghouse)

Non-planar 4-path

(British Gas)

5D - CPA - 10D

Single Bend

Double Bends

8-path

meter

Flow Weighted Mean Error Shift

5D

15D

Single Bend

Double Bends

Single Bend

Double Bends

12

6 COMPARISION WITH PUBLIC DOMAIN PERFORMANCE TEST RESULTS

Given the fact that ultrasonic meters are commonly used today with flow conditioners, and that this isoften put forward as ‘best practice’, the results shown in Figure 8 may challenge some preconceptionsabout using meters with or without flow conditioners. It is mainly practical experience that has broughtabout the common usage of flow conditioners, and that experience is valid, but it is valid only for themeter designs on which that experience is based.

The fact of the matter is that while flow conditioners do reduce non-axial flow velocities, they do notcompletely eliminate them. What the data shown in Figure 8 shows is that as the 8-path meter is designedto do a first-principles cancellation of non-axial flow, it fares better than a meter design that is adverselyaffected by non-axial flow, even when the latter is used with a flow conditioner.

Rather than relying solely on the 4 and 8-path data obtained with the Caldon meter, this can be validatedby comparing the 8-path results with the data from the GRI and GERG tests that were carried out undersimilar conditions.

Both the GRI and GERG projects conducted tests on multipath ultrasonic meters from the same threemanufacturers and included single bend and double-bend out-of-plane configurations in their tests. Themeters were a 4-path chordal design, a 6-path chordal design and a meter with reflected paths which wasa 5-path version of the meter for the GRI tests and a 4-path version for the GERG tests. The GRI testswere conducted on 12-inch meters at SwRI whereas the GERG tests were conducted on 8-inch meters atthe Advantica (now DNV GL) facility in the UK. The results were summarised in terms of the flowweighted mean error (FWME) shift relative to the calibration baseline, in the same way as was done toproduce the data in Table 2.

The shortest length of upstream pipe without flow conditioning was 10D in the GRI tests and 12D in theGERG tests. Figure 9 below compares the FWME results from the GRI and GERG projects with the 8-path data, all without flow conditioning. It can be observed that for 10 and 12 D without a flowconditioner the GRI and GERG results are typically in the range of +/- 0.5 to 1 % whereas for the 8-pathmeter the results are less than +/- 0.06 % for 5D and no flow conditioner.

Figure 9 Comparison of 8-path meter at 5D vs GRI and GERG results at 10 and 12 D

13

Both the GRI and GERG projects also included results where they tested the meters first in straight pipewith a CPA flow conditioner at a distance of 10D from the meter, and then downstream of the disturbancewith the 10D position of the conditioner relative to the meter unaltered. Figure 9 below compares theFWME results from the GRI and GERG projects with the 8-path data. It can be observed that althoughthe magnitude of error the GRI and GERG results is reduced with the CPA plate, they are typically in therange of +/- 0.3 to 0.6 %, still much larger than for the 8-path meter with 5D and no conditioner at +/-0.06 %.

Figure 10 Comparison of 8-path meter at 5D vs GRI and GERG results with CPA conditioner

7 FLOW CONDITIONING CONSIDERATIONS

The data presented in sections 5 and 6 shows that the improvements in performance achieved by the 8-path design outweigh the improvements obtained when a 4, 5 or 6-path meter is coupled with a flowconditioner. That in itself should be sufficient to challenge any notion that all ultrasonic meters must beused with flow conditioners. However, the following additional considerations add further strength to theassertion that improving the meter performance with respect to upstream effects has advantages relative toemploying flow conditioning.

Flow conditioners create pressure loss. While this is not always a larger concern, in some cases, forexample when summed over many measurement points, it can have a significant operational costimplication.

The principles of chordal integration used either explicitly or implicitly in all multipath ultrasonic metersfavour a relatively smooth velocity profile. The job that the multipath design is doing (once non-axialvelocity effects are accounted for) is akin to attempting to curve fit a function with only a limited numberof points on the curve. If the velocity profile has lumps and bumps, then it will be difficult to account forthese. In that respect the way that flow conditioners divide the flow into a number of discrete jets iscontrary to the desired velocity profile characteristics according to the principles of the design. This is thereason it is always advisable to have some distance between the conditioner and the meter to allow theprofile to recover to a smoother form. It also means that when a flow conditioner is to be use it is

14

advisable for the meter and conditioner to be calibrated together and maintained that way as reflected, forexample, in the advice given in ISO 17089-1:

“Installing a flow conditioner at any position in the meter run upstream of the USM will cause a changeof the meter’s indicated flowrate. This change depends on many factors (e.g. flow conditioner type, metertype, position relative to the USM, flow perturbation upstream of the flow conditioner, etc.)” . . . “Toavoid this additional uncertainty, the best option is that the USM is calibrated with the actual flowconditioner and meter tube as one package (USMP).”

The practical implications are that the meter and conditioner must now be calibrated (and recalibrated) asone package with associated logistical challenges and costs. It also means that operationally, any partialblockage of the conditioner will have an immediate, sustained and serious effect on the accuracy of themeasurement. While it is of course possible to protect a conditioner with a filter or even anotherconditioner upstream, the alternative approach of improving meter performance and eliminating theconditioner should be more attractive than placing further burdens on system design.

As mentioned in the introduction, it is often argued that flow conditioning is required to provide abaseline for flow profile diagnostics during calibration and in service. For a meter that is sensitive to non-axial flow, such as the 4-path Westinghouse and BG type designs that makes some sense, but it is worthre-evaluating in light of the benefits of the 8-path design.

First and foremost, the question to ask is this: With a meter design that uses a first-principles approach toreduce the effects of swirl and cross-flow, is monitoring of the flow profile still as important as it is for 4-path meters? Recent presentations CEESI workshops have shown that different ‘end treatments’ can havesignificant effects on some meter designs [14], and these might be detected by means of velocity profilemonitoring [15]. These effects are clearly similar to those that appear in the GRI and GERG testing, inthat the flow conditioner is not eliminating all of the non-axial flow and profile distortion. This supportsthe conclusion that reducing the performance deficiency also reduces the need for monitoring. Secondly,aside from upstream effects that the conditioner does not completely eliminate, what is it that velocityprofile monitoring is being used for? It would appear from many of the presentations and papers on thistopic that flow profile monitoring is primarily being used to detect flow conditioner blockage. It istherefore easy to conclude that if the conditioner can be eliminated with no detrimental effect onperformance with varying upstream conditions, the primary reasons for monitoring velocity profile areeliminated at the same time.

Flow conditioners can of course be used with 8-path meters, and although the flow weighted mean erroranalysis of Table 2 does shows limited additional benefit when the conditioner is used, a slightimprovement could be seen in terms of the reduction of error shifts at different flowrates on a point bypoint basis. So for the user that insists on a flow conditioner, the impact on the 8-path meter performanceitself is only marginal but the performance benefit of the 8-path meter over the other meter designsconsidered in sections 5 and 6 is still significant.

8 USM DIAGNOSTIC CONSIDERATIONS

If the data is reviewed and the arguments made earlier in this paper are accepted, it seems there is indeedmuch less need for velocity profile monitoring when an 8-path meter is used than is the case of someother meter designs. However, as we are advocating elimination of the flow conditioner, it is still worthexamining if anything is lost in terms of velocity profile diagnostic monitoring capability when an 8-pathmeter is employed and requirement for the flow conditioning is removed.

15

Velocity profile diagnostics are of course only one aspect of a suite of ultrasonic meter parameters thatcan be evaluated as part of a condition monitoring or condition based maintenance system.

Similar to other multipath ultrasonic meters, an 8-path meter can provide a variety of ‘path leveldiagnostics’, some associated with signal detection (such as gain, SNR and performance performance)and others that relate to the process and flow conditions such as velocity of sound and ‘turbulence’. At apath level the majority of these diagnostics are relatively insensitive to flow conditioning. The relativestandard deviation of the path velocity measurement, often called ‘turbulence’, will display sensitivity toflow conditions and in the absence of a flow conditioner has its greatest use in monitoring conditionsrelative to an installation baseline rather than against laboratory conditions.

Some parameters such as gain and velocity of sound per path are sensitive to process conditions and arebest used in a comparison with the other paths in the meter. In that respect the 8-path meter has theadvantage that it has a larger population of paths which can be inter-compared: four long ‘inside’ pathsand four shorter ‘outside’ paths.

At the ‘meter level’ most summary diagnostics are concerned with charactering the velocity profile and/orthe presence of non-axial flow. The exception is the average sound velocity which finds its greatest usein a comparison with a ‘theoretical’ sound velocity value determined from composition, temperature andpressure by means of an appropriate equation of state.

The meter level diagnostics used to characterise velocity profile and/or non-axial flow often have similarnames but can be calculated differently and will respond in different ways to the same flow conditions.Diagnostic terminology and how different path configurations react will now be discussed.

8.1 Profile Factor/Flatness

The term ‘profile factor’ (PF) is commonly used in the gas industry describe the ratio of the inside pathvelocities over the outside path velocities and is therefore a measure of how flat or ‘peaked’ the flowprofile is. For Caldon LEFM flow meters the term ‘flatness ratio’ has been used for this purpose formany years, the difference being that the convention was to take the outside path over the inside paths, i.e.1/PF. In this respect, having both the terms ‘flatness’ and ‘ratio’ in the term, and having the value start atless than 1 in fully developed flow at low Reynolds numbers and increase towards 1 as the profile flattenswith increasing Reynolds number would seem to be preferable. However, as ‘profile factor’ is wellestablished in the gas industry we will use that terminology with the slight modification to call it ‘profileflatness’ in order to avoid potential confusion with a velocity profile related correction factor.

The definition of profile flatness (PF) for 4-path single-plane (Westinghouse), 4-path criss-crossed (BG)and 8-path meters is shown below in Figure 11 where the numeral represents the velocity measured onthat path.

16

Figure 11 Definitions of profile factor/flatness

Figure 12 shows how profile flatness would be expected to vary versus Reynolds number for four and 8-path meters typical of the Westinghouse and BG design using the Gauss-Jacobi path spacing.

Figure 12 Profile flatness versus Reynolds number

In a fully developed flow or even a distorted profile free of non-axial velocity components, we wouldexpect the two 4-path designs and the 8-path design to produce the same (or in practice, very similar)values of profile flatness. However, when various forms of non-axial flow are present, this can have anadverse effect on the profile flatness values registered by the 4-path meters. For example, in the case of asingle-vortex swirl, this would cause either the inside or the outside paths of the BG design to read highor low and the other pair to do the opposite. Thus swirl can fool the 4-path meter and result in a change inPF when in fact the axial velocity profile may be unchanged. The single-plane Westinghouse 4-patharrangement is not fooled in the same way, but for that design two counter rotating vortices, one in the topof the pipe and the other in the bottom, would produce a similar effect whereby the indicated value offlatness would change.

17

Zanker summed up this issue nicely in a paper at the NEL America’s Workshop in 2009 when he said:“In general four paths are not sufficient to resolve any arbitrary 3-dimensional flow field containingasymmetry, swirl, peaked or flat profile and cross flow.”

In a way that mirrors the discussions about performance earlier in this paper, the accuracy of the‘diagnosis’ of the velocity profile is also adversely affected by the fact non-axial flow interferes with thesingle paths at each chord location. For 4-path meter designs the solution to this problem is to install aflow conditioning plate to try to reduce the number of degrees of freedom by one by eliminating non-axialflow. The 8-path meter addresses this same concern by cancelling the non-axial flow by combining pairsof paths on each chord, thus removing the interfering effect of non-axial flow from the determination ofthe profile flatness.

Practical examples showing this are presented in section 9 of this paper.

8.2 Asymmetry/Symmetry Ratio

The term asymmetry or symmetry is commonly used to describe the ratio of the velocities in one half ofthe pipe over the other half. The definition of asymmetry ratio (AR) for 4-path single-plane(Westinghouse), 4-path criss-crossed (BG) and 8-path meters is shown below in Figure 13 where thenumeral represents the velocity measured on that path.

Figure 13 Definitions of asymmetry ratio

As suggested by the name the intention of the asymmetry ratio is to register changes in the symmetry ofthe distribution/profile of axial velocity. In a manner similar to the discussion in section 8.1, in the caseof 4-path meters, this parameter can be fooled when non-axial flow is present. This time if we consider asingle-vortex swirl and the single-plane Westinghouse arrangement, then it is clear that if the direction ofswirl were such that paths 1 and 2 were to read high, then paths 3 and 4 would read low at the same time.This would then affect the asymmetry ratio, making it impossible to separate effects due to profileasymmetry from those due to swirl. Similarly, for the BG design if the swirl was in the form of twocounter-rotating vortices, one in the top of the pipe and one in the bottom, the asymmetry ratio wouldregister a spurious change. In the case of the BG design the clockwise rotation causing an over-readingon paths A and B would be accompanied by an under-reading on paths C and D due to the accompanyinganti-clockwise vortex.

Yet again the swirl cancelling nature of the 8-path configuration allows a change in asymmetry to beregistered correctly without having to resort to flow conditioning to reduce the effects of non-axial flow.

18

8.3 Cross-flow and plane balance

Terms that attempt to quantify cross-flow and swirl are arguably of less value in terms of diagnosingmeter performance issues than profile flatness and asymmetry. The reason for making such an assertionis that in the case of 4-path meters it is clear that non-axial flow adversely influences any attempt tocharacterise the axial flow profile and therefore a clear separation of axial profile and non-axial floweffects is not possible. For 8-path meters, the aim is to cancel swirl effects by design and thereafter ameasure of swirl is of relatively importance though it can be used to make second-order correctionsrelated to cavity and boundary layer effects.

In the case of the BG design, a cross-flow term is defined by dividing the sum of the paths that reside inone angled plane by the sum of the paths in the other. As suggested above, this cross-flow term ispotentially susceptible to being fooled by asymmetry in the flow profile. For the Westinghouse single-plane arrangement, clearly a cross-flow calculation of this type is not possible. For the 8-path meter across-flow or plane balance term can be defined by taking the ratio of all the paths in one angle plane (1 –4) to the paths the second plane (1 - 8). For those familiar with the ‘4 + 4’ concept of two 4-path metersin a single body, the 8-path plane balance diagnostic gives an equivalent measure of the differencebetween two 4-path results. Relative to the BG design the 8-path plane balance changes only with cross-flow and is unaffected by changes in axial profile symmetry.

Figure 14 Definitions of cross-flow or plane balance

8.4 Transverse velocity per chordal plane

As discussed in the preceding subjections, for 4-path meters it is not possible to properly separate non-axial flow effects/swirl from changes in axial flow profile. The simple reason for this is that when weonly have single paths in each plane it does not permit separation of the axial and non-axial velocitycontributions to the measurement of velocity in that plane. The beauty of the pairs of cross paths in the 8-path meter is that they allow exactly that as the transverse velocity can be calculated from the differencein single-path velocities at each height multiplied by a simple geometric factor. As illustrated in Figure15, this gives the 8-path meter a capability that is not available in either of the 4-path arrangements.

As an aside, it should be noted that while use of single-bounce paths where each leg of the path traversethe flow in the same chordal plane will result in some in-built cancellation of the effects of axial velocityit does not have the same diagnostic capability as a pair of crossed paths. Simply put it cannot supply ameasure of transverse flow as the single bounce path measurement does not provide information on whathappens in each leg of the path.

19

Figure 15 Transverse flow velocity calculation

9 PROFILE DIAGNOSTICS DOWNSTREAM OF BENDS

In this section of the paper we present diagnostic data for a selection of installation configurations toillustrate the issues that were discussed in section 8 above.

9.1 Baseline Straight Pipe

Figure 16 shows a photograph of a baseline straight pipe set up.

Figure 16 Straight pipe test set up

Figure 17 shows the velocity profile diagnostics for the two 4-path meter configurations and the 8-pathmeter. Figure 17 (a) and 17 (b) show the profiles for 4-path Westinghouse and BG arrangementsrespectively. Figure 17 (c) shows the profiles for the 8-path meter, along with the derived non-axial flowrepresented in the right hand figure showing the results at the corresponding path locations in the meter.

20

(a) (b)

(c)

Figure 17 Flow diagnostics from the straight pipe test: (a) Westinghouse Plane A; (b) BG 1; (c) 8-path

Also shown in Figure 17 are the corresponding values of profile flatness (PF) and asymmetry ratio (AR)measured by each meter configuration. It can be observed that in this case, that of a long straight pipe, aswould be expected, there is good agreement in the diagnostic indicators between the three different metertypes.

9.2 5D downstream of double bends out-of-plane

Figure 18 below shows the installation 5D downstream of the double bends out-of-plane, with no flowconditioning and the paths in the meter orientated horizontally. Figure 19 shows the flow diagnostics forthis case in the same format as Figure 17. It can be observed that the single-plane Westinghouseconfiguration interprets its velocity measurements as a strong asymmetry with AR = 1.347 whereas theBG design interprets its measurements as a strongly inverted profile with PF = 0.794. When the 8-pathmeter results are examined it can be observed that in actual fact the axial velocity profile is relatively flatand quite symmetrical with PF = 1.035 and AR = 1.009. The cause of the inaccurate profilerepresentation by the 4-path meters is revealed in Figure 19 (c) as a strong, clockwise single-vortex swirl.It can even be observed that the swirl itself is asymmetric, which contributes to the measurement errors inthe 4-path meters. This asymmetry in the non-axial flow also results in some error in the more accurate ofthe two profile indicators of each 4-path meter (PF for the Westinghouse and AR for the BG design),which can be seen to be several percent different from the more accurate 8-path result.

21

Figure 18 Double bends out-of-plane at 5D upstream with no flow conditioner

(a) (b)

(c)

Figure 19 Diagnostics from the out-of-plane bends test (a) Westinghouse Plane A; (b) BG 1; (c) 8-path

22

9.3 5D downstream of a single bend

Figure 20 below shows the installation 5D downstream of the single bend, with no flow conditioning andthe paths in the meter orientated horizontally. Figure 21 shows the flow diagnostics for this case in thesame format as Figures 17 and 19. It can be observed that this time the single-plane Westinghouseconfiguration interprets the measurements as a strongly inverted profile with PF = 0.836 whereas the BGdesign interprets the measurements as a strong asymmetry with AR = 1.248. When the 8-path meterresults are examined it can be observed that in actual fact the axial velocity profile is again relatively flatand quite symmetrical with PF = 1.024 and AR = 1.012. The cause of the inaccurate profilerepresentation by the 4-path meters is again revealed in the plot of transverse velocities: Figure 21 (c). Inthis case the single bend has produced a strong, counter-rotating double vortex swirl. It can be observedthat the double vortex pattern is asymmetric, which again means that even the more accurate of the twoprofile indicators from each of the 4-path meters is in error by a few percent relative to the more accurate8-path result. This result illustrates that even with the nominally symmetrical geometry of the upstreamsingle bend, the resulting swirl pattern is likely to exhibit asymmetries.

Figure 20 Single bend at 5D upstream with no flow conditioner

23

(a) (b)

(c)

Figure 21 Diagnostics from the single bend test test (a) Westinghouse Plane A; (b) BG 1; (c) 8-path

9.4 Test results with the 5D-CPA-10D arrangement upstream

Figure 22 below shows the installation of the double bends out-of-plane with the 5D-CPA-10Darrangement upstream of the meter. Figure 23 shows the flow diagnostics. In this case only the 8-pathresult is shown as it can be inferred from this graph that each of the 4-path results, irrespective of whichtype, are very similar. It can be observed that the introduction of the flow conditioning plate has reducedthe swirl to a negligible level and produced a symmetrical profile similar to that seen downstream of along straight pipe. The profile factors in this case show close similarity being 1.158, 1.164 and 1.153 forthe 8-path, 4-path Westinghouse and 4-path BG meters respectively and the corresponding asymmetryratios are 0.998, 1.003 and 0.994.

24

Figure 22 Double bends out-of-plane upstream in the 5D-CPA-10D set up

Figure 23 Flow diagnostics for the double bends out-of-plane with the 5D- CPA-10D arrangement

25

10 DIAGNOSTIC AND PERFORMANCE ANALYSIS

The proposition in terms of use of velocity profile diagnostic data is that if the parameters stay with setlimits then it is a good indication that the meter is performing properly; or perhaps more correctly, that ifthe parameters go outside the set limits there is a potential problem.

This proposition can now be examined using the performance and diagnostic data acquired during theOIML and ISO certification testing of the 8-path meter and the 4-path Westinghouse and BG subsets.This exercise is particularly relevant as it allows us to examine and compare the usefulness of 4 and 8-path diagnostics under an identical set of installation conditions.

The comparison in this paper is performed in terms of profile factor/flatness (PF) and asymmetry ratio(AR), as these two parameters can be calculated for all three configurations.

Diagnosis and monitoring based on velocity profile indicators can be used in various ways. It can first beused to validate the transfer of the meter’s calibration from the laboratory to the field installation.Thereafter profile changes can be monitored alongside other diagnostics in an effort to detect the onsetproblems, such as might be indicated by a sudden, unexpected change in profile. The diagnostic limits wewill discuss below are appropriate for calibration transfer, whereas once in service, monitoring to tighterlimits may be considered, alongside monitoring of the other parameters mentioned above.

The installation effect data used for this comparison is the same data that was summarised in section 5 ofthis paper, i.e. the single-bend and double bend out-of-plane data. This includes the configurations of 5Dand 15D without flow conditioning and the arrangement of 15D total length with the CPA plate includedat 10D from the meters. Data for both the horizontal and vertical orientations of the paths is alsoincluded. In addition to the single and double bend data we have now added data obtained for the OIMLR137 (2012) severe perturbation coupled with the three meter tube configurations described above. TheOIML R137 (2012) severe perturbation comprises two out-of-plane bends with a half-moon blockageinstalled between them and is a more severe disturbance than what would normally be encountered in acustody transfer metering system.

The default values of PF were set using the baseline calibration data to 1.169 and 1.136 for theWestinghouse and BG 4-path arrangements respectively, and the default value of AR was set to 1 forboth. The limits around these default values were set as follows by giving consideration to variouspublications on this topic such as those by Zanker & Floyd [16] and Lansing et al [17]:

PF: +/- 5% AR: +/- 3%

26

Figure 24 below shows the asymmetry ratio plotted versus profile flatness for the 4-path Westinghousearrangement (plane A). Installations without flow conditioning are shown with coloured symbols,whereas installations inclusive of the CPA plate are shown as open symbols in black. The 5 % PF and 3% AR limits around the baseline conditions are shown as a red ‘diagnostic box’, which in this case isrectangular owing to the extremely wide span of asymmetry registered on the graph by the Westinghouse4-path arrangement. Next the data points, which represent each individual test run from the CEESI datafiles, the flow weighted mean error shift caused by the upstream installation change is shown.

A number of useful observations can be made by examining Figure 24. Firstly, all of the data fromconditions without a flow conditioner lie outside of the diagnostics box, the only exception being the longstraight pipe condition. It can also be observed that in general, those results that lie furthest from thediagnostics box correspond with the largest errors, up to the maximum of 3.2 % corresponding to theOIML severe disturbance with an extreme asymmetry ratio of greater than 2.5. While the largest errorslie at the extremes, it can be seen that the relationships are not proportional, making it difficult in fieldapplications to interpret what a result that lies outside of the diagnostic box would mean in terms of error.This is most obvious when looking at the double bend results with paths horizontal in Figure 24 at both5D and 15D. In that case the 5D diagnostic indicators lie further from the diagnostics box than the 15Dresults but with a corresponding FWME of - 0.1% that is smaller than the -0.24 % that is associated withthe 15D location.

Figure 24 Diagnostic data plot for the Westinghouse 4-path arrangement (plane A)

27

Figure 25 shows the asymmetry ratio plotted versus profile flatness for the 4-path British Gasarrangement (BG 1), with data in the same format as Figure 24. Similar to the other 4-path meter, themajority of the data from conditions without a flow conditioner lie outside of the diagnostics box, anadditional exception this time being the single bend at 15D with paths horizontal, which for the BGdesign falls inside the box. Like for the Westinghouse design, those results that lie furthest from thediagnostics box correspond with the largest errors, in this case the maximum of 2.7 % corresponding tothe OIML severe disturbance, with extreme profile factors registered at less than 0.6. Again it can beseen that although the largest errors lie at the extremes, the relationships are not proportional, making itdifficult in field applications to interpret what a result that lies outside of the diagnostic box would meanin terms of error. Again focusing on the double bend results with paths horizontal it can be observed thatin terms of the diagnostic indicators both lie about the same distance outside the diagnostics box, but thatthe FWME is 1.2 % for the 5D location and only 0.09 % for the 15D location.

Figure 25 Diagnostic data plot for the British Gas 4-path arrangement (BG 1)

28

Figure 26 shows a zoom the asymmetry ratio plotted versus profile flatness for the 4-path British Gasarrangement (BG 1) highlighting some of the FWME values inside or close to limits of the diagnosticsbox. What this figure illustrates is that it is possible to be inside the box, or outside but close to the limitsof the diagnostics box, and have FWME error values of the order of 0.3 %.

Figure 26 Zoom in on the data in and around the diagnostics box for the British Gas 4-path arrangement

29

Figure 27 below shows the asymmetry ratio plotted versus profile flatness for the 8-path meter with thedata in the same format as in Figures 24 to 26. In this case the baseline asymmetry ratio is again set to 1but now the baseline profile factor is set to 1.096 to reflect the fact that the 8-path meter is intended to beused without a flow conditioner, and hence is expected to see the flatter profiles that are typicallyproduced by downstream of bends, tees and headers etc.

The diagnostic box shown in red on Figure 27 is plotted with limits of +/- 10 % for profile factor and +/-6 % for asymmetry ratio. The use of wider limits for the 8-path meter compared to the 4-path meter canbe justified on several grounds:

The 8-path meter is intended to be used without a flow conditioner and hence is expected to see awider variety of conditions

The 8-path meter does a better job of accurately quantifying profile flatness and asymmetry With the influence of non-axial flow greatly diminished, the 8-path meter performance is

relatively insensitive to the range of profile flatness and asymmetry changes that are observed

Comparing Figure 27 with Figures 24 and 25 we can make a number of informative observations. Firstwe can see that with only one exception, that one being the OIML R137 severe disturbance at 5D, all ofthe results both with and without flow conditioning lie inside the 8-path diagnostic box. Second, all of theFWME values are less than 0.08 % for the conditions inside the diagnostics box. Given that for the 4-path meters FWME values of 0.2 % to 0.3% are typical of performance within the 4-path diagnostic box,this comparison favours the 8-path meter, as staying inside the 8-path diagnostic box is associated withtighter performance limits. This is also confirmed by observing the one result that lies outside the box.The OIML R137 severe disturbance at 5D produces a FWME shift of 0.21 %. Although this takes the 8-path meter outside of the limits associated with its diagnostic box, the 0.21 % FWME result comparesvery favourably with the 2.7 % and 3.2 % errors associated with this condition for the 4-path meters, andis on par with the 4-path FWME results that fall inside their corresponding diagnostic box.

Figure 27 Diagnostic data plot for the 8-path meter

30

11 DISCUSSION AND CONCLUSIONS

The 8-path meter design discussed in this paper addresses weaknesses of previous multipath meterdesigns by employing a first-principles method of non-axial flow cancellation. Results have beenobtained showing that the 8-path meter meets the ISO 17089-1, AGA 9 and OIML R137 Class 0.5performance requirements downstream of bends at 5D with no flow conditioner.

Comparing like-for-like installation conditions, the installation effects for the 8-path meter are typicallybetween 3 and 5 times lower than that for 4-path meters. Futhermore, at 5D with no flow conditioner, themaximum errors and flow weighted mean error shifts for the 8-path meter are less than those for the 4-path meters with the 5D – CPA – 10D upstream package, confirming that custody transfer accuracy canbe achieved by the 8-path meter without having to resort to the use of a flow conditioner.

Diagnostic principles have been discussed and data analysed. It has been shown that 4-path meterscannot accurately quantify both flatness and asymmetry changes downstream of disturbances in theabsence of a flow conditioner, owing to the interfering effects of non-axial flow. The 8-path meter on theother hand can accurately quantify flatness and asymmetry without the need for a flow conditioner. The8-path meter can also quantify and display information regarding non-axial flow in a way that is notpossible for the 4-path chordal meter designs.

A combined analysis of diagnostic and performance data shows that in order to stay within the 5 % limitfor profile factor and 3 % for asymmetry ratio normally set for 4-path meters, flow conditioning is anecessity, and then the FWME shifts range up to between 0.2 and 0.3 % in magnitude. For the 8-pathmeter it has been shown that even with more generous limits in terms of asymmetry and flatness,operation within the 8-path diagnostic box confines the magnitude of FWME shifts to less than 0.08 %.

Clearly, in principle, it would be preferable to use ultrasonic meters without flow conditioning; for anumber of reasons including pressure loss, blockage and other maintenance concerns, and the logistics ofhaving the conditioner installed for calibration. This is of course only acceptable on condition that it doesnot expose the user to additional measurement uncertainty or risk. While the data in this paper confirmsthe need for flow conditioning with the 4, 5 and 6-path meter designs considered, the combinedperformance and diagnostic analysis shows that the 8-path meter can overcome these limitations. Thisallows us to conclude that the 8-path configuration can be used to achieve reduced measurementuncertainty and that this reduced uncertainty can backed up by meaningful velocity profile diagnostics, allwithout having to resort to use of flow conditioning.

31

REFERENCES

[1] Brown, G J, Freund, W R, and McLachlan, A (2013) “Testing of an 8-path ultrasonic meter tointernational standards with and without flow conditioning” AGA Operations Conference,21 – 24 May 2013

[2] Malone, J T and Whirlow, D K (1971) Fluid Flow Measurement System, US Patent no.3,564,912, Assignee: Westinghouse Electric Corporation, Filed Oct 1968, Issued, Feb 1971

[3] Wyler, J S (1976) Fluid Flow Measurement System for Pipes, US Patent no. 3,940,985,Assignee: Westinghouse Electric Corporation, Filed April 1975, Issued, March 1976

[4] O’Hair, J and Nolan, M E (1987) Ultrasonic Flowmeter, US Patent no. 4,646,575, Assignee:British Gas Corporation, Filed July 1986, Issued, March 1987

[5] Zanker, K J and Mooney, T (2013) “Celebrating quarter of a century of gas ultrasonic custodytransfer metering” Presented by M Schlebach at the 2013 European Ultrasonic User’sWorkshop, Lisbon, Portugal, April 2013

[6] Zanker, K J (2000) “Installation effects on single and multipath ultrasonic meters” Flomeko,Salvador, BRAZIL, June 04-08, 2000

[7] Grimley, T A (2000) “Ultrasonic Meter Installation Configuration Testing,” AGA OperationsConference, 7 – 9 May 2000, Denver, Colorado

[8] Delenne, B et al (2004) “Evaluation of flow conditioners – ultrasonic meter combinations”,North Sea Flow Measurement Workshop, St. Andrews, Scotland, October 2004

[9] Lowell, FC (1977) “The design of open channel acoustic flowmeters for specified accuracy:sources of error and calibration test results” Flow measurement in open channels and closedconduits, NBS Special Publication 484, Vol. 1

[10] Zanker K J and Mooney, T (2013) “Limits on achieving improved performance from gasultrasonic meters and possible solutions” North Sea Flow Measurement Workshop, St.Andrews, Scotland, October 2013

[11] AGA9 (2007) Measurement of Gas by Multipath Ultrasonic Meters

[12] ISO 17089-1 (2010) Measurement of fluid flow in closed conduits - Ultrasonic meters for gas -Part 1: Meters for custody transfer and allocation measurement

[13] OIML R137 - 1&2 (2012) Gas meters - Part 1: Metrological and technical requirements - Part2: Metrological controls and performance tests

[14] Hanks, E and Miller, R (2013) “Installation Testing NAFFMC Research”, CEESI UltrasonicMeter Users Workshop, Denver, 2013

[15] Hackett, D (2012) “Specifying upstream meter tube lengths for gas ultrasonic meters”, CEESIUltrasonic Meter Users Workshop, Colorado Springs, 2012

[16] Zanker, KJ and Floyd, A (2010) “Trending diagnostics from SMART ultrasonic meters”, NELAmerica’s Workshop, Houston, 2010

[17] Kneisley, G, Lansing, J and Dietz, T (2019) “Ultrasonic meter condition based monitoring – afully automated solution” North Sea Flow Measurement Workshop, Norway, October 2009