Baker R.C. Flow Measurement Handbook: Industrial Designs, Operating Principles, Performance, and Applications

Подождите немного. Документ загружается.

13.6 ACCURACY

327

and that the received signal

then sensed by its current pulse. The path is then

reversed.

13.6 ACCURACY

Van Dellen (1991) suggested that the ultrasonic meter could meet custody transfer

accuracies, and this appears to be happening (cf. Selvikvag 1997).

Reports indicate that good accuracy can be achieved by dry calibration

the

use

flowmeter dimensions

obtain calibration (de Boer and Lansing 1997).

Zanker and Freund (1996) even suggest that the difficulties

calibrating

large

(30-in.

75O-mm)-diameter meter may make dry calibration preferable. This will,

in part, reflect the limitations in the flow calibration facilities at large volume flow

rates.

13.6.1

REPORTED ACCURACY

LIQUIDS

Cascetta (1994) compared

clamp-on transit-time ultrasonic meter with

elec-

tromagnetic meter

situ

in a

water distribution network with 400 mm ID. The

ultrasonic meter generally overestimated the flow rate, and the difference between

the ultrasonic and electromagnetic flowmeter was mainly in the range

±1-5%.

Delsing (1991) tested the zero-flow performance

of a

sing-around ultrasonic

flowmeter with multiple paths. The meter was

in a

25-mm line, with transducers

set at 20° to the flow direction, and of 15 mm diameter. The sound path was 96 mm,

giving a period of about 64 /xs for a sound speed of 1,500 m/s. The single-period reso-

lution was about 15 ns, giving a velocity resolution of about 0.18 m/s. This could be

increased by using multiple-period averaging techniques. For instance, using 1,000

sing-around loops, a theoretical period measurement resolution of about 15 ps and

velocity resolution of about 0.18 mm/s could allow measurement down to 1.8 cm/s

with an uncertainty of

±0.5%.

The experimental data showed an overall zero-flow

error

less than ±0.6 mm/s, which would allow uncertainty

±1% for flows

low as 6 cm/s, and Delsing suggested that this would open up the meter's range to

100:1.

Manufacturers of commercial meters tested at

NEL

(Brown 1996) seemed to claim

for uncertainty, in the main, about ±0.5% of rate, but there appeared to be quite

lot of variation in actual performance between designs.

13.6.2

REPORTED ACCURACY

GASES

Holden and Peters (1991) found that, for undisturbed flow, the British Gas 300-mm

meter readings lay within 0.4 and -0.8%

the reference turbine meters. For

150-mm meter, the variation was between 1.4 and

-0.35%,

and

500-mm meter

appeared to be within about 0.3%.

6-in. (150-mm) ultrasonic meter calibration against sonic nozzles (with repeata-

bility of 0.04%, Erdal and Cabrol 1991) gave a day-to-day repeatability of

0.1%.

Grimley (1996) tested Daniel and Instromet meters and concluded that they were

capable of accuracies within 1% tolerance and with repeatability better than 0.25%.

Pressure change (e.g., 4.5 MPa) caused shifts of 0.4%.

also suggests that, for dry

calibration, there will be a need to understand the effect of parameter changes.

328 ULTRASONIC FLOWMETERS

13.6.3 MANUFACTURERS' ACCURACY CLAIMS

Typical performance claims for single-path meters for liquids with wetted transducers

are ±1% of full scale. Repeatability of ±0.2% full scale may be available. Ability

to cope with entrained solids or gas bubbles is increasing. Wetted transducers are

available for temperature ranges from —200 to +260° C and pressures up to 200 bar.

Pipe sizes range from less than 50 mm to 2 m. Fluid velocity in the pipe can range

from 0.03 to 10 m/s or higher.

For

gases,

the accuracy is likely to be of order ±0.5-5% of reading, and repeatabil-

ity will range from 0.25 to 0.5% of reading, with a range of velocity of 0.3-45 m/s.

Duct sizes can range up to 10 m. Temperature range may be as much as -50 to

+260°C.

13.6.4 SPECIAL CONSIDERATIONS FOR CLAMP-ON TRANSDUCERS

For clamp-on meters, manufacturers appear to be unwilling to claim better than

about 3% of rate at higher flow rates, or even 3% of full-scale deflection for low

flow rates. However, as in at least one case, the manufacturer may give figures

such as

• repeatability if nothing changed, ±1%;

• uncertainty (D < 25 mm), ±10%;

• uncertainty (25 < D < 50 mm), ±5%; and

• uncertainty (D > 50 mm), ±3%.

Sanderson and Torley (1985) described an intelligent clamp-on flowmeter and

claimed an uncertainty within ±2% of reading. They concluded that the wedge an-

gle of the transducers (estimated uncertainty of ±1°) could account for much of the

error. They gave plots of baseline drift equivalent to about 30 mm/s. They demon-

strated the value of Hemp's (1979) reciprocal drive system on baseline stability (e.g.,

where there is temperature change), which reduced drift to about one-eighth of that

for a conventional system. They found that an error of ±2.5% in transducer separa-

tion resulted in errors within 0.5%. The meter used additional transducers to locate

optimal axial separation of transducers and for self-calibration. The transducers al-

lowed pipe wall thickness to be measured and also internal dimensions of the pipe.

They also gave a very useful analysis of the likely errors in the system.

Cairney's (1991) experience of dedicated transducers, where there is a specific

pair of clamp-on transducers for each different size of pipe, with a scale card within

the portable electronics unit, was that, within the limitations, accuracy and repeata-

bility were good for pipes with OD greater than 35 mm. An example is the measure-

ment of lubricating-oil flows at near ambient temperature in a 60.4-mm OD line

with 5.5-mm carbon steel wall. Reynolds number was 50-350 (laminar). Readings

were within ±3% of a volumetric tank, where ±3% was typical above 35 mm OD,

but below this errors tended to increase. It seems that Cairney also implied that

removing and replacing the transducers led to larger errors, typically in the range

5-10%.

Cairney's (1991) experience of multipurpose transducers, which allowed the use

of one pair of transducers for a wide range of pipe sizes, was that the unit would

operate with almost any homogeneous, Newtonian liquid and on any pipe that

13.6 ACCURACY 329

Table 13.4. Pipe material and sizes used by Cairney (1991)

Materials

Carbon steel

Stainless steel

Aluminum

Unplasticized

polyvinyl chloride (UPVC)

Outside Diameter

(mm)

60.4-61.0

89.1

33.3-33.7

33.4

33.5

63.7

33.5 &33.7

114.3

Wall Thicknesses

(mm)

5.55-8.6

6.5

3.89-6.21

1.74

2.86

6.6

5.10

2.39

5.4

conducts sound. Relevant details are fed into the electronics unit: pipe geometry,

material, and fluid properties. The manufacturer claimed ±3% uncertainty without

calibration. Cairney tested the meters to check this claim. The tests were all in hori-

zontal pipes with the transducers mounted on the horizontal centerline. Pipe details

for his tests are given in Table 13.4.

For most of the tests flowmeter uncertainty was within ±5%. Particular notes

follow:

• Flow rate and pipe material had no noticeable effect, and coupling to the pipe

was not a problem.

• Performance was similar in direct and reflex modes.

• Rust on the inner surface is more likely to affect the cross-sectional area of the

pipe than the signal transmission.

• The ratio of pipe diameter D and wall thickness t should be greater than 15;

D/t

> 10 is acceptable, and D/t < 10 may cause problems due to ultrasound

transmission around the pipe wall. The low values may increase the uncertainty

to order ±10%.

• Removal and replacement of transducers caused, on some occasions, a severe

shift in reading traced to an incorrect measure of sound speed. This appeared to be

a fault with the electronics. Cairney recommended checking the computed speed

of sound. Provided care is taken, the removal and replacement of transducers

should contribute only about ±2% to the error.

• Air bubbles can cause loss of signal strength, which should, therefore, be moni-

tored.

• Nominal pipe dimensions should not be used. A 1-mm error in a pipe of 50-mm

diameter results in about 8% error in cross-sectional area.

• Incorrect axial spacing of the mounting tracks is not significant, but incorrect

angular spacing can cause errors up to ±10%.

Cairney concluded that errors should be containable within about ±5%.

Commercial devices allow a range of materials for the pipe including steel, stain-

less steel, cast iron, vinyl chloride, fiber reinforced plastic (FRP), and asbestos, in

sizes from 6 mm or less to 9,000 mm. The presence of a lining of tar, epoxy, mortar,

or Teflon appears to be surmountable. They may be available in both reflex (one or

more internal reflections off the pipe wall) or direct mode.

330 ULTRASONIC FLOWMETERS

Sensitivity as small as 0.3 mm/s and zero drift of 3-6 mm/s has been suggested

with uncertainty claims of 1.5% reading down to

m/s. The meter should be capable

of measuring flows in continuous liquids with low attenuation of the beam and no

air bubbles, etc.

NEL (1997) reckoned that, with care, clamp-on meters could achieve spool piece

precision, but diametral paths were a limitation. An error in the measurement of

diameter of, say,

would result in a cross-sectional error of

2%.

However, the point

was also made that temperature change could lead to changes in the refraction angle.

13.7 INSTALLATION EFFECTS

In addition to the following effects, it should be remembered that with these pre-

cision instruments, in which the accuracy of mounting transducers and general

dimensional stability is important, temperature variation within the measuring re-

gion of the flow tube, stresses in pipe fixtures, unsuitable mounting, and even Joe

Bloggs who regularly taps the pipe with his spanner as he walks by or climbs on it

as a step, may have detrimental effects on precision.

Zanker and Freund (1994) found that less than 1% condensate had little effect

(+1%) and that the meter survived in site trials, despite adverse conditions. NEL

(1997) reckoned that the maximum air-in-water for satisfactory operation was 0.5%

by volume, but that oil-in-water was a lesser problem.

Heritage (1989) reported an important series of tests on the performance of

transit-time ultrasonic flowmeters using water as the test fluid. The first part of her

work indicated the surprisingly large failure rate of new flowmeters. Out of 11 dif-

ferent flowmeters from seven manufacturers, 3 failed to complete the test program,

only

performed within the manufacturer's specification, and differences of

much

as 1.7% of rate were observed between the analogue and pulse outputs for certain

meters. Based on analogue readings, 2 meters were within 2%, 3 were within the

range 2-5%, 2 were within the range 5-10%, and 2 were over 10%.

13.7.1 EFFECTS OF DISTORTED PROFILE BY UPSTREAM FITTINGS

Some early reported installation effects (Al-Khazraji et al. 1978) obtained by com-

bining measured flow data for an eccentric orifice discharge at 5.5D with theoretical

calculations suggested that a single-path meter could have errors of 5-16%; a twin

path meter, ±1.6%; and a four-path meter could have errors of less than 1%.

Tables 13.5-13.7 summarize the installation tests of Heritage (1989) on water. In

the original paper, the results of two single-path meters are reported separately, but

they are combined in Table 13.5. The effect of the gate valve is taken regardless of

orientation. This is partly because the differences did not appear to be substantial,

and the safe assumption is the range given. The same is true for the two bends,

which were in perpendicular planes. However, in this case, the variation between

the meters is so great that it appears to outweigh any effect of orientation.

It appears from Tables 13.5 and 13.6 that a single-path meter will probably re-

quire at least the spacing for 0.75 beta ratio orifice. For dual-path flowmeters, the

performance (Table 13.7) appears to be close to that for an orifice with p = 0.6 with

±0.5%

additional uncertainty.

13.7 INSTALLATION EFFECTS

331

Table 13.5. Installation tests on water for single-path meter

(after Heritage 1989)

Spacing Between Downstream Flange of

Disturbance and Upstream Flange of Meter

(Error Range in Percentages)

Disturbance 5D 10D 15D

Gate valve

50%

closed

by movement

Swept bend

(r/d=

1.5)

parallel

Swept bend

[r/d=

1.5)

perpendicular

Two swept

bends

Reducer

Not obtained -1 to -2.5 +1.5 to -3

Not obtained +4.5 to -6.5 -2 to -3.5

Not obtained -3.5 to -4.5 +3.5 to -3

Not obtained +7.5 to -11 -0.5 to -7

Not obtained +2 to -4.5 +2 to -1

Halttunen (1990) used data on flow profiles and applied them to the ultrasonic

flowmeter analyses to obtain the effect of flow distortion on the performance. He

also gave experimental data from installation tests presumably on water. Each of his

methods appeared to give a slightly different datum, a salutary reminder that all such

measurements and predictions are subject to some uncertainty. The conclusions that

can be drawn from the plots are given in Table 13.8.

Table 13.6. Installation tests on water for single-path

clamp-on meter (after Heritage 1989)

Spacing Between Downstream Flange of

Disturbance and Upstream Flange of Meter

(Error Range in Percentages)

Disturbance 5D 10D 15D

Gate valve Not obtained +3 to -2.5 +1.5 to -0.5

50%

closed

by movement

Swept bend Not obtained +3.5 to -3 -2.5 to -3.5

[r/d=

1.5)

parallel

Swept bend Not obtained -2 to -2.5 +2.5 to -3.5

[r/d=

1.5)

perpendicular

Two swept Not obtained -5 to -10 +0.5 to -6

bends

Reducer Not obtained -2.5 to -4.5 0 to -0.5

332

ULTRASONIC FLOWMETERS

Table 13.7. Installation tests on water for dual-path meter

(after Heritage 1989)

Spacing Between Downstream Flange

Disturbance

and

Upstream Flange

Meter

(Error Range

Percentages)

Disturbance

5D 10D 15D

Gate valve

50%

closed

by movement

Swept bend

(r/d= 1.5)

parallel

Swept bend

(r/c?= 1.5)

perpendicular

Two swept

bends

Reducer

0 to+1.5

Oto +1

+0.75

-1.5

to -3

+1 to +0.5

+0.75

+0.25

-0.75

Not obtained

+0.5 to -1.5 Not obtained

+1 to +0.5 Not obtained

The orientation for single-path ultrasonic flowmeters may cause an additional

±1%

at 10D or less spacing, and for dual-path ultrasonic flowmeters it may cause an

additional ±0.5% at 5D or less spacing.

The following claims of Vaterlaus (1995) appear, possibly, optimistic but probably

reflect the (surprising) variation found by Heritage (1989). For a single-path meter,

errors should be within 1% if at least:

10D is allowed between a reducer and the meter;

20D is allowed for an elbow or a T;

25 D is allowed for two elbows in one plane;

40D is allowed for two elbows in perpendicular planes; and

50D is allowed for a partially open valve or for a pump.

Table 13.8. Halttunen's (1990) installation data

10D 20D 40D 80D

Ultrasonic flowmeter

—

single path

Single elbow

Double elbow

90°

-5%

-7%

-2.5%

-5%

Ultrasonic flowmeter

—

dual path

Single elbow

Double elbow

90°

-1%

±1.5%

-0.5%

±1%

-2%

-3%

small

±1%

—

-1%

—

+0.5%

-1%

+0.5%

13.7 INSTALLATION EFFECTS 333

Hakansson and Delsing (1992) tested the effect of upstream disturbance on a 20-

mm-ID ultrasonic flowmeter for gas. Against a reference flow from a 100D upstream

straight pipe, the results indicated that the laminar/turbulent flow change at Re in

the range 2,500-4,000 caused a calibration shift of about 11% and that for the Re

range 4,000-11,000 calibration dropped by about 1%. They reckoned that, within

the turbulent region,

• for a single bend in the same plane as the ultrasound beam, about 40D upstream

straight pipe (from the transducer path) is required to contain the errors to within

±1%

of the reference. Note that longer lengths may not significantly reduce this

value at certain Re values.

• for two bends in perpendicular planes upstream, 80D upstream straight pipe does

not contain the errors to within +1% of the reference.

Multipath (apparently two paths, and not necessarily off-axis) flowmeters were

tested on water to obtain profile effects due to a bend. A 10D upstream length was

considered to cause a calibration shift of 1% (Johannessen 1993).

Holden and Peters (1990, 1991) concluded from tests on a four-path ultrasonic

flowmeter operating on high pressure gas, in both fully developed flows and also

with upstream disturbances, that 10D upstream and 3D downstream is sufficient for

the data to fall within ±1%. The meter can also deduce swirl and turbulence in a

disturbed flow and may operate satisfactorily with less than 10D upstream at the

lower end of the flow range.

The effect of

bends,

step changes in diameter, and pressure reduction were inves-

tigated by a joint industry project on multipath meters, van Bloemendaal and van

der Kam (1994) concluded that, in well-developed conditions, uncertainty of 0.6%

should be achievable without calibration but with careful determination of meter

dimensions and zero setting. They reckoned that an additional ±0.5% should be

allowed for 10D minimum spacing from a bend and that small changes in pipe di-

ameter had negligible effect but that swirl and noise from pressure reductions could

have severe effects resulting in 2-2.5% additional uncertainty.

Lygre et al. (1992) described a five-path gas ultrasonic flowmeter. The paths were

positioned so that three are on one diagonal plane, whereas the other two

lie

between

the three but on a reverse diagonal plane. It was designed to operate with 10D

upstream and 3D downstream. They claim that, for a bend upstream, the installation

length may be reduced to 5D. Their results suggested an uncertainty within ±0.5%

over a 4:1 turndown.

From these sources of data, the following overall conclusions are suggested.

a. A single-path meter will have a calibration shift of up to 33% for changes from

laminar to turbulent and 3% for changes in the turbulent range.

For a single-path meter to remain within 2% of calibration, the following spac-

ings should be allowed:

15D for a reducer

20D for a bend or a T

40D for two elbows in perpendicular planes.

334 ULTRASONIC FLOWMETERS

For distances less than these, the following values are conservative:

Fitting

Reducer

Bend

Two bends

Error

±5%

±4%

±7%

Spacing

10D

15D

two-path meter has a calibration shift of 0.7% (NEL 1997) or less for changing

Reynolds number and should have the following spacings to remain within 1%

of calibration:

10D for a reducer

10D for gate valve (50% closed) or bend

20D for two swept bends

four-path meter should retain its performance to within

±1%

in liquid or gas

with 10D upstream and 3D downstream.

The values appear to be reasonably confirmed as being on the cautious side by

Grimley (1997). However, Grimley's results have some puzzling changes in baseline,

which may result from the meter calculation method, and appear to make perfor-

mance better than 0.5% uncertain. His work suggested, also, that thermowells at 5D

or less may cause errors of order 0.6%.

NEL's (1997) rule of thumb that single-, two-, and multipath meters need 20D,

10D,

and 5D upstream between the meter and a fitting perhaps needs a bit more de-

tail.

The summary of values, which take the more pessimistic end of the published

data, suggests that to achieve less than 1% on a single-path meter would require sim-

ilar installation lengths or more than those for an orifice plate for the maximum beta

ratio and "zero additional uncertainty/' The two-path meter, to remain within 1%,

would require those for an orifice with p = 0.6 with ±0.5% additional uncertainty,

whereas the four-path meter is probably much better than these figures suggest. The

assumption made is that there should be similar effects for gas and liquid.

13.7.2 UNSTEADY AND PULSATING FLOWS

Mottram's (1992) key comment "if you can't measure it, damp it!" is balanced by his

advice that ultrasonic transit-time meters are probably not affected. However, there

are some dangers of which the user should be aware. Hakansson and Delsing (1994)

found that there were at least three error-creating effects due to pulsating flows in

ultrasonic flowmeters:

• Aliasing where the sampling frequency picks samples that do not give a true

average - This error should be avoidable provided it is recognized.

• Flattening of the flow profile due to the pulsating flows - This error is fundamen-

tal to pulsating flows and the incorrect averaging of the diametral path ultrasonic

flowmeter and, therefore, is difficult to avoid.

• Extreme flows during the fluctuation exceeding the upper and lower cutoffs of

the flowmeter and giving a spurious mean - This error should be avoidable with

care.

13.8 GENERAL PUBLISHED EXPERIENCE IN TRANSIT-TIME METERS 335

13.7.3 MULTIPHASE FLOWS

I have observed, in some field data from an ultrasonic flowmeter, a behavior that

could result from entrapment of air. A meter in a low head flow, where air was

entrained with the water, periodically failed. The possibility that in such a flow the

transducer cavities could cause small local vortices that would entrap the air and

block the ultrasonic beam offered a possible explanation.

Lenn and Oddie (1990) have looked at how the scattering of ultrasound from

secondary components in the flow, both liquid and solid, could be detected, and

they considered that it was possible to measure component size concentrations and

velocity using a variety of signal-processing techniques, particularly at low concen-

trations.

A commercial transit-time device has been claimed to be capable of operating

in flows of drilling mud, although my experience is that it is usually recommended

that commercial devices should not be applied to other than single-phase flows.

Brown (1997) appears to have suggested that ultrasonic meters may be able to

cope with oil-gas and oil-water flows, but Johannessen (1993) put a limit of 10%

oil-in-water and reckoned that they were unlikely to cope with more than 0.5% by

volume of air in liquid, in agreement with NEL's (1997) summary.

13.8 GENERAL PUBLISHED EXPERIENCE

IN TRANSIT-TIME METERS

13.8.1 EXPERIENCE WITH LIQUID METERS

Boer and Volmer (1997) described a five-path liquid meter claimed to be for use in

maintenance-free custody transfer applications. It had two paths to measure swirl

and inlet and outlet cones to condition the profile.

Reynolds (1994) discussed water industry specifications, and data sheets have

been produced for the industry on ultrasonic flowmeters.

Transit-time meters have been used in the electricity supply industry (Cairney

1991).

Transducers were grouted into the wall of an octagonal section culvert. It

was also possible to use radioactive-isotope dilution methods for in situ calibration

with an uncertainty of ±1% or better with a 95% confidence level. The ultrasonic

flowmeter was reckoned to have an overall error of

±1.5%.

difference of 5% from

the manufacturer's calibration was put down to error in the set-up procedure.

In the Ataturk Hydro Power Plant in Turkey, two systems were used to compare

flows for loss due to bursting (Vaterlaus 1995). The pipes were up to and over 6 m

in diameter. Taylor and Cassidy (1994) discussed the use of ultrasonic meters at BC

Hydro in Canada.

Baumoel (1994) described the main features of an ultrasonic meter including ve-

locity up to 30 m/s (±100 ft/s), an active zero flow, sensitivity of

0.3

mm/s (0.001 ft/s),

and uncertainty of up to ±0.5% of flow rate. Figure 13.8 shows diagrams of an ul-

trasonic clamp-on meter, and Figure 13.8(c) shows a reflect mode installation. Sig-

nal strength fluctuation indicates changes in the quality of the flow, like aeration.

Applications given are fuel mass metering, hydraulic fluid flow metering, leak detec-

tion, engine lubricant, ground support of all aircraft fluid systems, rocket fuel and

336

ULTRASONIC FLOWMETERS

Cables

(2)

994

Series

Row

Computer

•

4.10

0.750"

Tubing

(typ)

2.06"

*(typ.2Xdcrs)-

Mounting

Frame

(a)

Transducer

Clamping

Screw (typ.)

Local

Display

Digital

and/or

Analog

Outputs

Transducer

Figure 13.8. Clamp-on ultrasonic flowmeter (Baumoel 1994; reproduced with

permission of ISA and Controlotron Corporation): (a) Diagram of the clamp-on

meter; (b) Direct mode installation; (c) Reflect mode installation; (d) Photo of

clamp-on flowmeters.