Baker R.C. Flow Measurement Handbook: Industrial Designs, Operating Principles, Performance, and Applications
Подождите немного. Документ загружается.
12.8 INSTALLATION CONSTRAINTS
-
FLOW PROFILE CAUSED
BY
UPSTREAM PIPEWORK
297
12.8 INSTALLATION CONSTRAINTS - FLOW PROFILE CAUSED
BY UPSTREAM PIPEWORK
12.8.1
INTRODUCTION
In this chapter, four examples
of
rectilinear weight function distributions have been
given (a fifth is given
in
Figure 12.A.l(b)). From these,
it
should be possible to deduce
the performance
of
different types
of
flowmeters
for
various installations.
Rather than use actual fittings, some experiments have used localized jets and ori-
fice plates to distort the flow profile in the meter. Shercliff (1955) used a localized wall
jet
to
demonstrate the validity
of
the weight function distribution
in
Figure 12.4(a).
Cox
and
Wyatt (1984) used this method
to
show
the
insensitivity
of a
contactless
design
to
severe profile changes.
It
is
necessary
to
remember that
in
real installations
the
designer seeks
a
meter
that
is
affected
as
little
as
possible
by
changes from
the
calibration profile. These
changes,
in
most cases, represent small perturbations
of
the datum profile. The elec-
tromagnetic flowmeter is insensitive (Baker 1973)
to a
number
of
profiles that have
reverse symmetry about diametral planes. Thus if such a profile is added to the datum
profile, there should be
no
change
of
signal.
12.8.2
THEORETICAL COMPARISON
OF
METER PERFORMANCE
DUE
TO UPSTREAM FLOW DISTORTION
Al-Khazraji
et al.
(1978) used
an
upstream orifice
to
distort
the
flow profile (Table
12.4).
The
diameter
of the
orifice was half that
of the
pipe,
and the
center
of the
eccentric orifice was
at
half the pipe radius.
Where experiments were done, they appeared
to
confirm these trends. These
computed values suggest that
a
disturbance upstream
of the
flowmeter will have
minimum effect
if
its plane
of
symmetry is perpendicular
to the
electrode plane.
Approximate calculations based
on
eccentric orifice profile and Shercliff's (1962)
weight function distribution [Figure 12.4(a)] suggest that errors
for a
uniform field
flowmeter would be ±1% depending
on
the
jet
orientation.
Halttunen (1990) used data
on
flow profiles and combined them with magnetic
field data obtained using
a
Hall probe
and a
computerized traversing
and
measure-
ment system. He also gave experimental data from installation tests. The tests appear
to have been confined
to a
single elbow test
for one
configuration
of
electromag-
netic meter.
The
agreement
is
quite good
and
suggests that this type
of
combined
approach
to
installation testing may
be
worth pursuing. The conclusions that
can
Table 12.4. Calculated signal change
(%)
from uniform profile
(Al-Khazraji
et
al. 1978)
Orientation of
Eccentric Orifice
Peak Velocity
At electrodes
45° displaced
90° displaced
Small Electrode
Rectangular Coil
[Figure 12.4(b)]
3.3
2.4
1.4
Diamond Coil
[Figure 12.4(c)]
1.5
0.9
0.0
Large Electrode
[Figure 12.4(d)l
0.03
-0.03
-0.02
298
ELECTROMAGNETIC FLOWMETERS
Table 12.5. Conclusions that can be drawn from the plots obtained by
Halttunen (1990) on signal change due to upstream disturbance
2D (%) 5D (%) 10D (%) 15D (%) 20D (%)
Single elbow
Double elbow
90°
-1.5
to 0.5
±2.5
±0.5
±1
±0.5
±1
small
±1
small
±1
Note: that at about 40D error becomes small.
be drawn from the plots are given in Table 12.5. The flowmeter's (electrodes and
magnetic field) orientation relative to upstream orientation of fittings may cause an
additional ±0.5% at 5D or less spacing. I was pleased to see a meter in these tests
named The Shercliff after someone whose contribution to the development of this
meter has been so important!
Luntta and Hattlunen (1989) confirmed the effect of a bend but noted that four
different meters (400 mm ID) gave very different results when downstream of an
eccentric orifice plate and of a reducer.
12.8.3 EXPERIMENTAL COMPARISON OF METER PERFORMANCE DUE
TO UPSTREAM FLOW DISTORTION
Scott (1975b) was the first to publish data on electromagnetic flowmeter operation
downstream of pipe fittings. As Mr. D. Halmi commented in the discussion of this
paper, it is not possible to generalize from such data. This is mainly because of the
variation in meter design.
Much of the useful data is inaccessible to the public. For instance, de Jong (1978)
has published a few results from his tests, which omit full flowmeter design detail.
His most useful conclusions appear to be that, for a gate valve 5D upstream and at
least 50% open, errors were within 0.5% of specification.
Tsuchida et
al.
(1982) gave the requirements in Table 12.6 for minimum upstream
straight pipe to ensure uncertainty within ±0.5% of full scale.
r
Table 12.6. Tsuchida et al.'s (1982) values for minimum
upstream straight pipe to ensure uncertainty within ±0.5% of
fll l
full scale
Pipe Fitting Minimum Upstream Straight Pipe
Reducer
Expander
1
Ball valve
J
Single elbow
Two elbows in perpendicular
planes
Gate valve
Butterfly valve
3D (direct connection)
5D
10D
15D
12.8 INSTALLATION CONSTRAINTS - FLOW PROFILE CAUSED BY UPSTREAM PIPEWORK 299
Table 12.7. Deacon's (1983) values of maximum error over all
tests for upstream components
Spacing Max. Error (%)
Reducer 2.5D 0.8
Gate valve (at least 50% open) 5D 1.2
One bend
Two bends in perpendicular planes
J
Deacon (1983), after tests with a gate valve upstream (at least 50% open), one
swept bend (radius/D = 1.5), two swept bends in perpendicular planes, and a reducer,
was able to conclude that the worst error with 5D spacing was less than 2% from
calibration (Table 12.7). For gate valve 50% open and for one bend, the orientation
of the flowmeter electrodes relative to the pipe fittings did not affect the magnitude
of the error.
These data indicate considerable variation for a particular installation. One bend
with 5D spacing caused about 0.3% error for a large electrode flowmeter and up to
2.0%
error for a small electrode flowmeter. The small electrode flowmeters exhibited
errors that could be, in some cases, a factor of 2 different. For one small electrode
flowmeter, opening the gate valve from 50 to 75% increased errors by a factor of more
than 4 when the orientation caused a jet past the electrodes.
12.8.4 CONCLUSIONS ON INSTALLATION REQUIREMENTS
Tentative conclusions on the likely influence of installation profiles follow.
a. The error due to a reducer installed next to the flowmeter is likely to be less
than 1%.
b.
An error of up to 2% should be assumed for other fittings separated from the
meter with 5D of straight pipe.
c. An error of 1% may occur for other fittings separated from the meter with 10D
of straight pipe.
d. The orientation of the flowmeter at 5D spacing or greater does not simply corre-
late with the size of the error.
e. At least 3D spacing downstream should be allowed.
Manufacturers may be less conservative, but spacing such as 3D to 5D upstream,
with 3D downstream, should be treated with caution. Despite data such as de Jong's
(1978),
valves should not be installed close upstream of flowmeters and, even if fully
open, are likely to have an effect of ±0.5% even if at least 15D upstream. In all cases,
it should be recognized that, because of the design differences between magnetic
flowmeters, it is not possible to give absolute guidelines that are equally valid for
all designs.
A crude rule of thumb to retain errors within ±0.5% would be to use the instal-
lation data for a p = 0.2 orifice plate for "zero additional uncertainty/' although
the very different operation of the meters makes this very conservative for, say, two
bends in perpendicular planes.
300 ELECTROMAGNETIC FLOWMETERS
12.9 INSTALLATION CONSTRAINTS
-
FLUID EFFECTS
12.9.1
SLURRIES
The electromagnetic flowmeter has found
a
special niche (almost
a
universal solu-
tion) for slurry flow measurement and also for some complex non-Newtonian fluids.
A slurry will affect the flowmeter because
of
flow distortion, conductivity variation,
and magnetic permeability. The work quoted
in
Section 12.9.3 enables
us to
make
an informed guess
as to
the
errors,
if
we
have some idea
of
the
size
of
the con-
ductivity variation
due
to
the
presence
of
slurry.
In
addition,
the
distribution
in
Figure 12.4(b) suggests that
a
flattened profile, due
to
a
non-Newtonian fluid,
say,
could cause negative
errors.
The likely performance on an untried fluid should, there-
fore,
be undertaken with caution.
In a vertical line, which is a common installation position in slurry service, it will
also be necessary to allow
for
any slip that may occur between the conducting fluid
and
the
nonconducting solid. For instance,
a
settling slurry
in a
vertical pipe with
zero net flow will record the movement upward of water displaced by the settlement
downward of the solid matter (as noted by a colleague when working for a flowmeter
manufacturer).
A magnetic slurry (Baker
and
Tarabad 1978) will result
in a
change
of
size
and
distribution
of
magnetic field.
If
the flow signal
is
compensated
for
change
in
field
current size, then
the
increase
in
fluid permeability will cause
an
uncompensated
increase
in
flux density and flowmeter signal.
Where
the
flow signal compensation
is
achieved
by
a
search coil monitoring
magnetic flux density, then careful design should ensure that magnetic fluids have
a negligible effect
on
flow signal.
A magnetic slurry may also
be
deposited
in
the
flowmeter magnetic field.
The
cure
for
this may be
to
run the flowmeter with
a
lower field strength.
Eren (1995) provided
an
example
of
the use
of
frequency analysis
of
the output
signal generated by DC-type electromagnetic flowmeters. The frequency and ampli-
tude components are dependent
on
slurry characteristics
of
the flow. Eren implied
that these signals can be related to the density and other characteristics of the slurries.
12.9.2
CHANGE
OF
FLUID
Although conductivity
of
the fluid should not affect the flowmeter signal, there has
been some suggestion that electrochemical effects
of
the fluid may do so. There has
been some suggestion that this may affect
the
performance
of
the modern square
wave field excitation flowmeters. However, as noted earlier, Ginesi and Annarummo
(1994) commented
on
the
effect
of
severe variations
in
conductivity
in
AC-type
magnetic flowmeters
but
suggested that
DC
types should
be
unaffected above
a
minimum threshold. Few data
are
available,
but
Baker
et
al. (1985) compared
the
performance
of an
AC-excited
and
a
DC-field flowmeter
for
a
change
in
fluid from
pH
=
7
to
pH
=
4
and found
no
significant effects.
12.9.3
NONUNIFORM CONDUCTIVITY
Nonuniformity
of
conductivity
in
the
flowmeter
can
cause changes
in
flowmeter
signal due
to the
changes
in the
size
of
the shorting currents which result. Thus
a
higher conductivity layer near the flowmeter wall will cause higher shorting currents
and
a
reduced flow signal. For certain conditions
of
turbulent flow, field shape,
and
12.11 ACCURACY UNDER NORMAL OPERATION 301
conductivity profile, a signal change of up to 3% may occur (Baker 1970a), and it
will be higher for laminar profiles. This is probably most significant in multiphase
flow applications with a second nonconducting phase (cf. Ginesi and Annarummo
1994) because the continuous conducting component will generate the same signal
as for a uniform fluid provided the fluid is homogeneous.
12.10 MULTIPHASE FLOW
Baker and Deacon (1983) have reported experiments on the behavior of an elec-
tromagnetic flowmeter in a vertically upward air-water flow and have shown that
volumetric flow of the mixture was measured with errors of less than
-1%
up
to void fractions of 8%. For void fractions greater than this, the negative error
increased.
Bernier and Brennen (1983) have analyzed the flowmeter's performance with
uniform field and obtained in addition to the following equation for a dispersed
second phase without slip:
AI/EE =
BDV
m
= -^- -r^— (12.5)
TTD (1
—
a)
where q
L
is the liquid (or conducting phase) volumetric flow rate, and a is the void
fraction, the same equation for annular flow. They also show that, for small values
of a, the expression is the same for "cylindrical voids
;;
parallel to the flowmeter axis.
Their data for bubbly and churn flows fall within ±2% of the predictions up to void
fractions of 18%. Murakami et al. (1990) developed an electromagnetic flowmeter
with two pairs of electrodes for studying gas-liquid two-phase flow.
The use of the electromagnetic flowmeter for slurry flows has been mentioned
earlier in the context of its early development. Peters and Schook (1981) have devel-
oped the idea of using the signal fluctuation as a measure of slurry concentration.
It is also possible that special design of the magnetic field could be used to obtain
concentration information.
Most recently, Krafft et al. (1996) have looked at the passing of gas bubbles
through the flowmeter in terms of the electric dipoles that result and have made
a start on relating the theoretical signal pattern to experiment (cf. Baker 1977).
12.11 ACCURACY UNDER NORMAL OPERATION
Manufacturers' claims for uncertainty (often referred to as accuracy) tend to lie some-
where near the following:
Uncertainty Flow range as % of FSD
±0.3%
rate 100-50%
±0.5%
rate 50-15%
±1.0%
rate 15-5%
±1.5%
rate 5-2.5%
In other cases, the value may be given as a percentage of rate or a velocity (e.g.,
±0.015 m/s), whichever is greater.
302 ELECTROMAGNETIC FLOWMETERS
Repeatability is likely to be of order ±0.2% rate. The analogue output may be less
precise than this (e.g., ±0.1% FSD).
One manufacturer, at least, offers meters in the range 2.5-6 mm with uncertainty
of order 0.8% rate at the upper end of the flow range.
There has been some discussion in the technical press relating to the accuracy of
this instrument over large turndown ranges. There may, therefore, be some research
needed to consider the fundamental limits to accuracy related to
a. flow steadiness and turbulence, and the sensitivity of the meter to profile varia-
tion over the Reynolds number range;
b.
the field stability and the usually neglected size of currents flowing in the
liquid;
c. the effect of charge distribution, current size, and conductivity variation; and
d. the dimensional stability of the flowmeter
;
s components, and compensating
systems with variation in ambient conditions.
12.12 APPLICATIONS, ADVANTAGES, AND DISADVANTAGES
12.12.1 APPLICATIONS
The applications of this meter for liquid flow measurement are extensive. It
is
suitable
for essentially any conducting liquid, and to my knowledge, it has been unsuccessful
on hardly any to which it has been applied. One industrial expert once claimed to me
that his only major problem had been icing sugar! Failure is more likely to be due to
flow problems or incompatibility. If it is applied to a two- or multicomponent flow,
then the continuous component must be conducting, and the signal
will,
essentially,
be due to the velocity of this component. If it is applied to a liquid metal, the physics
becomes more complex.
A
description of this application is beyond the scope of this
book, but a review has been given by Baker (1977; cf. Tarabad and Baker 1982).
Applications include viscous fluids, corrosive chemicals, erosive slurries, and
start-up and shut-down operations, but the flow tube should run full (some manu-
facturers offer versions capable of part-filled tubes), and the electrodes should not
be open circuited by air bubbles (Ginesi and Annarummo 1994). The tube should,
if possible, be vertical with upward flow. If horizontal, the electrodes should be in a
horizontal diameter. If the meter is at a low point in the line, then the danger of slur-
ries or other fluids coating the electrodes should be monitored. Coatings can have
a conductivity that differs from the fluid and will then behave like a partially con-
ducting wall, as well as changing the internal diameter of the meter. Likelihood of
deposition is reduced if the velocity is kept above about 2-3 m/s (6-10 ft/s) through
the meter. Electrodes of conical shape reduce deposits, and electrode-cleaning sys-
tems can be used. Non-Newtonian fluids may alter the response. Wear of the liner
can also result from abrasive slurries from close elbows, etc., and can be reduced by
liner protection. The cleaning fluid must also be compatible. Additives can cause
nonuniform conductivity.
The AC technology may be preferable for applications with large amounts of
entrained air, with slurries that are nonhomogeneous or have nonuniform particle
size,
high solids content or solids with a tendency to clump, and with pulsating
flows. This includes about 15% of the industry including pulp and paper flows. The
12.12 APPLICATIONS, ADVANTAGES, AND DISADVANTAGES 303
DC output can be noisy in these applications, but it is increasingly competitive as
an option. Damping of 1-3 s may help.
Effects of radiofrequency interference (RFI) should be virtually eliminated in
new meters. However, signal cables must be screened and earthed according to the
manufacturer's instructions.
Rose and Vass (1995) discussed adaptations of electromagnetic flow technology
to difficult processes:
Chemical acids, bases, polymers, emulsions, latex
solutions
Pharmaceutical spray coatings, flavoring, hygienic
Mining and minerals slurries of iron ore, taconite, magnetite,
pyrite, copper, alumina
Food and beverage beer, soda, toothpaste, milk, ice cream,
sugar, juices
Water and waste water, wastewater, raw sewage, primary
sludge, digester flows
Pulp and paper white and black liquor, brown stock,
bleaching chemicals, additives
Nuclear fuel processing on both radioactive and nonradioactive
plant liquids (Finlayson 1992)
To this list may be added slag, cement, slurries (abrasive), reagent charging, and
special applications like ultra low rates, custody transfer, liquids with steam tracing,
blast furnace flows, batching, and erosive liquids.
At high (120 measurements per second) rates, AC metering allows measurement
of pulsating flows from pumps, etc.
One manufacturer offers a range of
sizes
from 2 to 25 mm for milk, and others will
offer their own range and specification for hygienic and sanitary applications. The
meters are suitable for high speed batch processing of these and other applications
and can achieve 0.2% repeatability.
12.12.2 ADVANTAGES
a. The theory shows that the response is linear, and the only reason why the meter
may not register satisfactorily down to zero flow is zero drift. It is one of the few
meters capable of this performance and has been unfairly criticized as a result,
for the zero drift that is observable. Modern designs often have a low flow cutoff
that avoids this problem.
b.
Clear bore is of most value in fluids that contain solids or in fluids that can be
damaged by passing through constricted flow passages.
c. There are no moving parts.
d. Sensitivity to upstream fittings is comparable to other meters and only seriously
bettered by positive displacement and Coriolis meters or by ultrasonic meters
with more than two beams.
12.12.3 DISADVANTAGES
The major disadvantage is its restriction to conducting liquids only, and although
laboratory designs have operated on nonconducting liquids (e.g., transformer oil),
304 ELECTROMAGNETIC FLOWMETERS
to my knowledge only one or two commercial designs have attempted to operate in
this regime. However, those interested in this application should consult Al-Rabeh
et al. (1978) and Al-Rabeh (1981), although it is unlikely that such a meter could
operate with normal gases.
Its sensitivity to upstream distortion has sometimes been suggested as
a
weakness.
In my view, it is one of its strengths. Very few meters operate with less detrimental
effect from upstream disturbance than this one. The other feature that was also
suggested as a disadvantage from time to time was zero drift, the fact that early
designs were prone to large errors at very low flows. Again, there are virtually no
flowmeters capable of operating remotely linearly over the range of this one or down
to the flow rates of which this is capable. Indeed, at least one commercial meter is
claimed to be capable of
1,000:1
turndown.
12.13 CHAPTER CONCLUSIONS
More high-level mathematics has probably been applied to this design of flowmeter
than any of the others. The theoretical analysis of the electromagnetic flowmeter is
well understood and has been developed by
Shercliff,
Bevir, Hemp, and others with
some contributions from the present author. This analysis allows the response of
the flowmeter to various flow profiles to be deduced but also, through the weight
function analysis, allows a flowmeter to be designed with a predetermined level of
insensitivity to profile change.
The weight function analysis
is,
itself,
very important as it offers an approach that
Hemp has exploited for ultrasonic, thermal, and Coriolis meters.
A
similar approach
may be possible in other types of meter.
This meter must be a prime contender in any flow measurement application
with a conducting liquid. I have always taken the view that its successful commer-
cial development for nonconducting liquids, although possible, was highly unlikely
because the voltages needed to drive the field coils would be incompatible with the
intrinsic safety requirements of many nonconducting (oil-based) liquids. However,
Al-Rabeh et al. (1978) have given the theory for nonconducting fluids, and Al-Rabeh
(1981) tested such a meter. Further work may overcome some of the expected limi-
tations for this meter.
It is unlikely that the limitation of the meter to liquids will be overcome, and we
have to look elsewhere for a technique for gas flow measurement with this meter's
advantages.
What other developments are likely? The discredited idea of dry calibration could
still be developed to provide condition monitoring of the field. For some time I
have wondered if the inclusion of a vortex-shedding bar in the meter, to provide
another approach to condition monitoring, would be worthwhile, even though it
would remove the clear bore. It would provide two independent velocity measures,
the electromagnetic steady signal, with a fluctuation superimposed related to the
vortex shedding frequency. It might also be possible to include additional electrodes
in the shedder to provide a guide to upstream flow distortion.
One very elegant device, which could well offer applications other than that for
which it was developed, should be mentioned.
12.A.1 INTRODUCTION 305
Gray and Sanderson (1970) developed a differential flowmeter using two adjacent
rectangular flow channels so that the same magnetic field traversed both and the
electrodes were so connected as to give the difference of the two signals due to
the two flows and hence the differential flow. They claimed to be able to measure
differences of 1-10 cm
3
/min in a flow of 500 cm
3
/min to ±10% and possibly even
to measure down to 0.5 cm
3
/min.
Another line of development, exemplified by Horner et al. (1996), is the use of
rotating fields or other multiple coil arrangements. In their case, two pairs of coils
and 16 electrodes appear to have been used to improve performance on nonaxisym-
metric flows.
With the developments in design, materials technology, manufacture, and signal
processing, this meter has become of increasingly competitive price, usually lower
maintenance than those with moving parts, and steadily improving performance.
This has included
• improved accuracy,
• extended turndown,
• reduced size and weight,
• reduced power, and
• microprocessor-based intelligence.
This has led to the possibility of battery operation with time periods up to 10
years or more and turndown claims of
1000:1.
The potential of the meter has also
encouraged the use of new materials, especially ceramics, to increase stability and,
by careful design, to reduce noise from radiofrequency and electrochemical sources.
Independent verification of the remarkable performance claims for these devices
would be very valuable.
APPENDIX
12.A
Brief Review of Theory
The theory set out in this appendix has been extended to other flowmeters, and the
relevant papers are given elsewhere in this book.
12.A.1 INTRODUCTION
We make use of a simplified form of Maxwell's equations
V x B = /xj (12.A.1)
V
•
B = 0 (12.A.2)
V x E = -B (12.A.3)
and Ohm's law with velocity included
B) (12.A.4)
306 ELECTROMAGNETIC FLOWMETERS
where B is magnetic flux density vector,
/x
is permeability, j is current density vector,
E is electric field vector, a is conductivity, and V is velocity vector.
We may ignore B because the effect of this term may be eliminated by careful
mechanical and electrical design. This allows the introduction of the scalar electric
potential U:
j = a(-V[/ + VxB) (12.A.5)
Taking the divergence of this equation and making use of the fact that the magnetic
field produced by external coils will be virtually unaltered by the very small currents
in the fluid so that
V x B = 0 (12.A.6)
we obtain the flowmeter equation
V
2
L/ = B- V x V (12.A.7)
and boundary conditions for a nonconducting wall of
? = (Vx Bh
(12.A.8)
on
when n is the coordinate perpendicular to the wall. Since the velocity at the wall is
normally parallel to the wall, this will become, for cylindrical coordinates,
This condition is assumed to exist everywhere on the wall of a conventional
flowmeter with small electrodes. For a nonslip condition on velocity, the right-hand
side is zero. However, if the electrodes are large, then the simplest assumption is
that the fluid at the electrode surface is at the same potential as the electrode. The
electrode is assumed to have the same potential at every point on its surface, and
the total current flow into the electrode is assumed to be zero. If a contact resistance
exists at the electrode surface, this will have to be allowed for in relating the value
of the voltage on the electrode and in the fluid.
One important observation from Equations (12.A.2) and (12.A.6) is that B in the
fluid may be expressed as the gradient of a scalar potential O; hence
V
2
<D = 0 (12.A.9)
and we observe that the magnetic field distribution obeys Laplace's equation.
The boundary conditions on the magnetic field will usually be those existing
on the pole pieces and at the surface of the excitation coils or the slots into which
they fit. Thus, O will have a constant positive value on the surface of one pole and
an equal but negative value on the other. The size of the surface potential on the
remainder of the yoke may be more complicated unless the coil winding slots are
thin and the value of O can be assumed to be zero on the yoke with discontinuous
changes at the poles.
So far we have assumed steady conditions. Often, however, the excitation is si-
nusoidal, and the voltages induced by transformer effect may not be negligible. We
can demonstrate that the analysis so far is valid by noting that if we associate a sinu-
soidal excitation of known phase with the applied magnetic field, all the equations