| Author
:
Dr.RA Furness -Consultant
JD+F Associates
Summary
Water supply management is becoming increasingly
important in India as population grows and sources
diminish. Accurate balancing and measurement
in large supply lines are particularly difficult
as a small error represents a large volume of
water, and reports of tests in such applications
are rarely found in the open literature. Any
reduction in leakage and unaccounted-for water
(UWF) figures depends critically on the correct
amounts of water going into supply being known.
The Author presents some simple modelling work
aimed at trying to predict flow effects in pipelines.
Two series of experiments have been performed.
The first of these are in India using the 2400mm
transmission lines that are part of the BMC
network in Greater Mumbai. Here the hi-tech
multi-path ultrasonic flowmeters installed on
some of the pipelines do not balance, even though
the pipelines are above ground and any leakages
can be seen. The second and more comprehensive
series were undertaken in Brazil using a pumping
station run by COPASA as part of water supply
network for the City of Belo Horizonte. Studies
in both 1500mm and 1800mm water pipelines were
used. The Author attempts explain the imbalance
reported from the modelling studies and tries
to verify the predictions using high accuracy
pressure and differential pressure measurements
around the pipe circumferences at various locations
within the system in India and different radial
velocity measurements using state-of-the-art
portable ultrasonic meters in Brazil. Tentative
conclusions are drawn that may have major implications
in water balancing strategies.
Introduction
Studies of flow in large diameter lines are
rarely made. In many major cities however, bulk
water supplies are made in pipelines of 150mm
or greater. Balancing networks in these cases
is extremely difficult and reports of the largest
imbalances have been made for such applications.
Flow measurement standards and manufacturers
literature claims that relatively short distances
are required for disturbances from fittings
such as valves, bends and tee-junctions to have
settled sufficiently for good flow measurements
to be made. However is this true for the largest
diameters? It does not seem logical that small
swirling masses of fluid decay over the same
lengths of pipe as flows of 3.5m3/s, typical
of the flow rate in the experimental work reported
here. In this regard we question the validity
of installation length statements in both manufacturers
literature and ISO standards for large pipes.
Simple linear theory (ref 1) will show that
it is possible that swirl does not decay as
previously thought. If the pipe inside surface
is very rough, the retarding effect at the wall
is significantly reduced and swirl decays at
a far slower rate. This hypothesis has been
tested in some of the bulk transmission mains
in the Greater Mumbai water supply system. In
the test section used, 4-beam ultrasonic flowmeters
are installed. The pipes run above ground and
any leaks can be easily detected. In one area,
a pair of these meters some 4 kms apart do not
balance to within 10% or more of each other
and this area was chosen to qualitatively validate
the theoretical work. It is important that any
fluid dynamic effects existing within the pipe
are understood so that transmission lines can
be correctly balanced. Any bias effects that
may exist make correct assessment of water production
(and reported UFW figures) very difficult. Theoretical
studies The behaviour of flowmeters in non-ideal
flows has received very considerable experimental
verification, but so far attempts to predict
the effects have largely been unsuccessful.
The theoretical work (Reference 1) has looked
at two areas. Firstly a simple linearised theory
has been used to look at the tangential and
axial motion of fluid flow. The work is based
on turbulence mixing length principles. Secondly,
a non-linear theory has been used to predict
flow fields as a function of Reynolds number,
pipe roughness and the amount of swirl present
as characterised by the swirl number (S). The
basic equations for the mathematical model were:
Where Q is flowrate, P is pressure,
ρ is density ε is the eddy viscosity and L is the length of
pipe. By transposing these into polar co-ordinates
and neglecting the second order terms in order
to simplify the computation, these equations
can be reduced to the form:
The eddy viscosity ε can be recast in cylindrical polar co-ordinates
as:-
A stream function can be used
to remove equation (1) from the set to be iteratively
solved. These stream functions can be defined
as:
Where U is the velocity, Z is the axial position and
R is the radial position. The flow field is
this study is therefore represented by three
scalar quantities (ψ, Uθ and P) each being a function of radial and axial positions
(r and Z). The boundary conditions for the solution
were set as follows:
In order to solve the set of equations
6 to 11 using these boundary conditions, it
is assumed the velocity is given from the log
law relationship and Nikuradse’s mixing
length equation gives the eddy viscosity ε. Two solution methods can be used. In the first
a velocity V can be defined on theoretical grounds
to obtain the value of ε from shear stress relations. In a second method,
experimental values of ε can be used and from them velocity can be calculated.
If it is assumed the perturbation velocity is
small with respect to the axial velocity (this
may or may not be valid), then some linearised
equations are obtained:
Such a linearised theory can only
deal with perturbations inn the flow where the
axial and tangential components are independent.
If any interaction exists between these (as
would be the case for swirling flow), then a
second theoretical method must be used for the
modelling. Here we can take equations 6 through
11, non-dimensionalise them and substitute the
following expressions for the three scalar quantities.
Similarity theory can also be
used to determine the magnitude and form of
the axial profile that is inherently associated
with a particular level of swirl. The same assumptions
can be made for the expressions given above,
namely:
From these equations, a relationship
between the swirl angles in two directions as
function of decay rate along the pipe is obtained.
These expressions for two directions, the tangential
velocity and the stream function, are solved
iteratively. In the first pass, the stream function
is assumed zero and a preliminary value for
the tangential component is found. This is then
re-substituted into the stream function equation
and a preliminary value for ψ
is given. This calculated value is finally substituted
into the tangential velocity equation before
the whole cycle is repeated many times until
the solution of the N iterations is made to
converge within preset values. The full set
of equations is presented in reference (1).
The theories, summarized for the
purposes of this paper, now give three expressions,
which can be compared: linear and non-linear
expressions and a similarity expression. Some
of the computed results using the similarity
theory are shown in figure 1. Here decay rates
are plotted as a function of Reynolds number
for both smooth and rough pipes for varying
amounts of swirl number (S).
The results from this are very
interesting. It shows that swirl and Reynolds
number are related, but not in the linear manner
that might be expected. It also shows that the
swirl will decrease much more slowly in pipes
with high internal roughness. This prediction
is important because one interpretation could
be that swirl persists for very long distances
in larger diameter lines that have large amounts
of deposits or calcification on the inside surfaces.
Such pipes are traditionally found in municipal
transmission systems. Furthermore in systems
where high levels of swirl form (at pumping
stations for example), the swirl may actually
increase during normal flow along the pipe.
The positions and spacing of pumps can affect
swirl intensity and therefore flowmeter readings.
This has just been verified in very large diameter
Libyan installations (report in preparation).
The theory can also predict the axial and tangential
velocities as a function of the Reynolds number.
Examples are given below.
Recent modelling work using the
‘Phenix’ CFD code at the Mexican
Technical Water Institute (IMTA) at Cuernavaca
has shown that using very high values of pipe
roughness promotes the formation of cross-flow,
and enhances swirl that forms in bends and other
pipes fittings (ref. 2). Little modelling work
has been published using the size of roughness
that may be present in some old large water
pipes, and the IMTA work suggests support for
the ideas put forward in this paper. The theory
can also be applied to the behaviour of electromagnetic
and ultrasonic type meters. Computations have
been made of the expected errors downstream
of two bends in offset planes. There have been
some reported cases in the literature of meters
being less affected near to a disturbance than
further away from it. This might at first sight
sound strange, but the model predicts such behaviour.
If the weight functions for magnetic and ultrasonic
meters are subjected to the profile variations
predicted by the model, again some interesting
pictures emerge. The predictions are shown below
for 3 types of meters commonly used in water
pipeline balancing. The first shows the theoretical
behaviour of magnetic meters, where the errors
oscillate in a damped sinusoidal manner. There
is a Reynolds number dependence, (errors reducing
as flow increases) but the largest predicted
errors would occur within the first 10 diameters
at all Reynolds numbers. Such behaviour has
been well reported in the open literature (references
3 and 4). The other prediction is that the errors
would depend on whether the electrodes are in
the horizontal or vertical planes. This is not
surprising since the flow is biased away from
the axi-symmetric condition and lines of symmetry
do not occur on the centre line in swirling
or distorted flows.
Single- and multi-beam ultrasonic
meters behave in a similar way but with some
significant differences. Single beam meters
tend to under-record, presumably because the
single beam is centre-line located whereas the
highest velocities are located much closer to
the wall. The behaviour is still a sinusoidal
wave because the line of symmetry will rotate
as the swirling flow moves in an axial direction.
Multi-beam meters behave in the damped sinusoidal
manner but here theoretical errors increase
as Reynolds numbers go up, the reverse of magnetic
meter behaviour. We can now write the equation
for such damped sinusoidal behaviour. This is
of the form:
The size of pipe, swirl intensity, roughness and period
of rotation all influence the decay rate of
the perturbations. If we now measure the amplitude
of the perturbation (x) in terms of the distance
along the pipe (τ), then from the theory we can estimate the position
where the perturbation decays to zero. The simple
theory of exponential decay however does not
include the effects of roughness. The earlier
theoretical work indicates that as roughness
increases, the decay will increase with distance.
The implication of this for large diameter rough
surface pipelines is that the swirl decay distance
will significantly increase. In extreme cases
it may never decay at all and could actually
increase with high mass flowrates found
in pipes such as the Great man Made River system
(Libya), the BMC and COPASA sites studied here.
Experimental verifications
1. India: There are virtually no reported results
for the decay of swirl in large transmission
lines and this series of tests was made in mid
2003. The test section was a section of the
Mumbai Corporation main 2400mm line located
within the Bhandup complex in the north of Mumbai
City. Here we find a combination of five fittings
all within 10 diameters of each other. Such
combinations were thought to generate sufficient
bulk swirl, which could be easily studied. At
Bhandup, the 2400mm line is one of the three
main feeders from the water treatment plant.
It takes water south into Mumbai City centre.
The five fittings were one bend of 35°,
followed by 4 fittings each of 45°,
that are close coupled to enable the pipeline
to dip below the roadway as show in the sketch
below. Large pumps are found several hundred-pipe
diameters upstream of this bend combination.
From later work reported in the next section,
this might actually be the real source of the
perturbations. The close proximity of the multiple
bends shown in the figure below may amplify
the swirl intensity.
Test points were located at 2D from the final fittings
outlet, then 5D, 25D and 55D. Each set of test
points consisted of eight pressure tappings
located at 45° around the circumference.
The top pressure tapping at each location was
taken as the reference point and static pressure
plus differential from each point to the reference
were measured at each circumference positions
at all four axial locations. If fluid is moving
axially down a pipe (i.e. tangential and radial
components are zero), then the pressure measured
at the wall would be only static pressure. However
if swirl were present, then additional dynamic
pressures would be present. If motion were toward
the tapping, the total pressure would rise.
If it were away from the tapping point, pressure
would fall. The value of the differentials is
therefore a measure of the qualitative motion
within the pipe.
Swirl is present in all water
treatment plant piping systems to some degree.
Experience tells us that as line size goes up,
turbulence increases due to the increased momentum
involved and so that swirling eddies should
persist in greater strength and for longer distances.
This is borne out by the theoretical work in
this paper. A particularly good example of a
swirl producing combination is the location
of two bends in different planes next to each
other, so that changes in direction and height
may be made in the minimum of space. The swirl
usually forms and decays away rapidly on exit
from the fitting, an example being shown in
figure 6 above. This is the computed velocity
distribution at two positions, bend outlet D
= 0, and D = 4.93D. Notice the change in the
value of the non-uniform circumferential velocity
components at the exit of the second bend and
how quickly these have decayed within about
5 pipe diameters. The velocity vectors at the
wall are also not representative of the velocities
in the centre of the flow. The effect of these
mal-distributions on flow meter performance
can be very serious. Under such conditions,
turbine-type mechanical meters, some differential
pressure and most ultrasonic meters show large
errors if installed at such positions. The distortion
can be reduced by the use of flow straighteners.
Swirl may be generated with either right hand
or left hand motion, so errors can be obtained
with differing sign (i.e. either positive or
negative), depending on the orientation of the
fittings causing the problem. The above figure
was taken in a 150mm pipe, but theory suggests
that in 3000mm pipes, the intensity and rate
of decay of the swirl will be quite different.
Any combination of bends generates both profile
distortion and bulk swirl. Bends in the same
plane generate less severe swirl than bends
in offset planes. In all cases, distortion of
the profile occurs in the first fitting as the
fluid moves to the pipe inside (described above).
However in order to navigate the second fitting
with minimum of energy loss, cross-flow occurs
in the main body of the fluid, close to where
the bends join. Large radial forces form and
severe swirl now appears at the outlet of the
second bend - figure 6 left hand side above.
In large sizes (>1000mm) such effects are
exaggerated and the flow may remain perturbed
for long distances as the radial momentum slowly
dissipates. The theory also suggests roughness
increases such effects. The differential pressure
measurements made show swirl intensifying as
one moves away from the source. The top of the
pipe is taken as reference point 1, the right
side is point 3, the bottom of the pipe point
5 and the left hand side is point 7. There are
also intermediate points, making up 8 measurements
at each section. The data shows the following
trends:
The average fluctuations in differentials
at each section were 6mm at the 2D section,
8mm at 5D, 4mm at 25D and 5mm at 55D. The static
pressures were also measured within the pipe
and the readings corrected for position. Notice
the differentials oscillate around the pipe
and also axially away from the disturbance.
This mirrors the theoretical behaviour analyses
in the previous section. From these readings
we can draw a pictorial representation of the
results. This is shown below at the 5D and 55D
positions in figure 7.
Close to the bend combinations,
the rapid changes of differential with position
and the large fluctuations measured at each
point indicate a rather complex motion within
the flow. As some air was also present, and
we could easily hear this air passing down the
pipeline. The presence of air will not be helpful
to the measurements from the 4-beam transit
time meter located within this test section.
By 55D, the fluctuations and variations were
lessened and we can envisage the formation of
a double vortex by this point. It is surprising
however that the fluctuations and variations
persist for such a long distance.
Fitting the results to the theory
shows that the swirl would have decayed away
by 500D, some 20 times the manufacturer’s
recommended distance for this meter. Pairs of
meters located 5 kms from each other did not
balance within 10%. We can visualize that one
meter may be biased low by say 6% whilst the
other may be reading high by 4%. We would contend
therefore that the swirl was persisting for
very long distances in this pipeline. The presence
of swirl and some air is responsible for the
poor imbalance betweens pairs of meters within
this section. Meters located in other pipelines
within the network show similar behaviour. In
all some 60 metering systems have been installed
and balances between adjacent pairs is well
outside the specifications. Imbalances in the
totalised flow of between 7 and 13% are found
throughout the network.
We may also speculate that swirl
generated by the pumps is already present at
the inlet to the bend combinations. It is not
appreciated that in pipes of this size just
how far swirl travels. Some recent studies made
by IPT (Ref. 6) in large diameter water pipes
in central Brazil have revealed distorted profiles
several hundred-pipe diameters downstream of
a pumping station in a line of 1500mm diameter.
This is similar to some findings in a pumping
station in southern England, where the flow
signals from ultrasonic clamp-on meters show
instabilities. These were being used to analyse
the flow at various points within the piping
system. Fluid dynamic studies in large pipes
(>1500mm) are seldom published.
2. Brazil: The second site studied
in much more detail was close to Belo Horizonte
City in Minas Gerais Province, Brazil. This
study (Rio das Velhas pump station) took place
in June 2005 and built on the earlier work reported
above. The work looked at flows in both 1500
and 1800mm lines, the first site being a problem
for measurement using clamp-on wide beam flowmeters.
This had been installed in accordance with standards
and suppliers’ recommendations but the
readings were unstable at some flowrates and
un-usable at others. Survey work in 2003 by
IPT showed profile distortion in two planes
at more than 200D away from the pump station
outlet and experiments were devised in conjunction
with COPASA (the operating authority) to verify
the IPT work and explain the behaviour of the
metering systems used.
The following three pictures (Fig.
8) show the bottom, centre and top sections
of the pump station outlet, with the larger
1500mm line in the foreground of the second
two pictures. In the left hand picture there
are two parallel 1200mm lines. These were not
part of this study, but had been examined by
IPT in 2003.
In the centre of the right hand
picture, the tapping points for the velocity
traverse are clearly visible. The IPT measurements
here showed distortion in both horizontal &
vertical traverses. Similar measurements at
the same place in the smaller lines showed no
distortion, but this is to be expected, as it
was just downstream of the Venturi meters located
in the building shown in the middle photo above.
The Venturi meters act as capacitors, so that
pulsations and swirl effects (if present) would
be greatly reduced.
Two series of measurements (at
90 degrees to each other) were made at each
selected site. The axial positions were estimated
from calculations made with the model. Indications
were that disturbances would occur every 30D,
so positions were chosen on this basis, with
the reference position (position 0) being at
5D before the incline began, but with a focus
at 20D (position where the troublesome meter
had been installed). Measurements were also
made at 50D, 80D, 110D, 180D and 220D (IPT site).
Figure 10 below shows the data
taken upstream on the control section (position
0 shown above) for the first minute of one of
the tests. These results show the 4 pumps and
the layout of this pump station do cause abnormal
profiles. The data in figure 11 alongside is
most revealing. Plotted here are the average
velocities recorded on the same axis at different
positions along the pipe. At the inlet (position
0), this was around 1.8m/s. By position 1 (20D)
it had fallen more than 3% and by position 2
(50D) it had increased 6% upwards compared to
the averages recorded near the meter. This change
is in line with expectations. Data on axis B
at the same axial locations showed velocities
different to those measured on the A axis.
The preliminary overall results
are shown below in figure 12. The average velocity
changes measured in the two perpendicular axes
are plotted against distance along the pipe.
The presence of swirl on both axes is clear,
even after more than 200 diameters. Typical
average velocities change between +/-4% for
this installation. It is believed each installation
will have different characteristics, so this
data trend will only relate to this installation.
However the swirling flow model does predict
that as pipe size increases such an effect will
persist for greater distances.
Flow measurement standards &
manufacturers literature claim that relatively
short distances are needed to maintain performance
within stated specifications for disturbances
from fittings such as valves, bends and tee-junctions.
However is this true for the largest diameters?
It does not seem logical that small swirling
masses of fluid in say 25mm pipes decays over
the same lengths of pipe as flows of 3.5m3/s,
typical of the flow rate in the experimental
work reported here. It is therefore right to
seriously question the validity of installation
length statements in both manufacturers literature
and ISO standards for the large installations
(800-1000mm and above).
In the two figures that follow,
the instantaneously measured velocities over
a 10-minute period at the first 4 locations
are plotted. This data is most revealing and
quite remarkable. First it shows the average
velocity on the two perpendicular axes is not
the same at each location and that each set
of velocity data also changes with position
along the pipe. It is data such as this that
allowed the construction of figure 12 above.
Such sinusoidal behaviour is very similar to
the theoretical predictions in figures 3 and
4.
In figure 13a (beam velocity A)
the average velocity trends are quite clear.
From the reference position at 5D before the
start of the incline, the average velocity first
falls, then rises to the maximum before falling
again. On the other axis (beam B at 900) the
trend is quite different, first rising from
the reference, then falling and rising again.
Taken together the data clearly shows the presence
of swirl, with the highest being recorded at
20D (the location of the existing meter). Flow
here is quite unstable, even at the highest
velocity chosen for these tests. The theory
suggests that such trends would alter as flowrate
reduces. The results certainly call into question
the recommendations of all suppliers on ultrasonic
meters. Nearly all do not qualify the recommended
distance as pipe size changes. This must be
so since the theoretical model shows that stabilizing
length should increase with flowrate and with
pipe size.
Data taken in a pair of 1800mm
lines at the same site shows different trends.
The two lines are shown below left, with a typical
installed below right. The permanently installed
meter is on the horizontal centre line (far
right) and the comparative measurements were
made in two planes downstream as shown.
As with the data reported above,
the measured velocities on the two axes shown
above yielded different velocities, one being
higher and one being lower than that recorded
by the second ultrasonic meter mounted on the
centreline. These results are shown below in
figure 15.
Simple calculations revealed the
average velocity measured on axis B was around
3.5% lower that that measured on axis A. The
average of all the data recorded in figure 15
was 1.5% higher than that recorded from the
centreline meter at the same time. Similar trends
were recorded in line 1 at lower flowrates (figure
16). In this line, the difference between the
two recorded average velocities was much lower
(less than 0.7%), but the average of all the
data was still 1% higher than that from the
permanently installed meter. In this line the
layout of the feed pumps was quite different
to that for line 2. All the pumps were more
than 60D away from the start of the straight
sections shown in figure 14a. As with the velocity
data taken from the 1500mm lines, the turbulence
from the pumps appears to be travelling long
distances. The other conclusion that may be
drawn is the layout of the pump station and
the angle the water is fed into the discharge
line will directly affect swirl and profile
effects at the meter. This may account for the
trends reported here.
Laboratory and field performance
of installed flowmeters
When purchasing any instrument,
the supplier usually gives a written specification
for that device. Some manufacturers give detailed
performances backed by independent testing whilst
a minority of specifications barely enable the
prospective user to determine what they are
purchasing. Flowmeters are a little different
to other instruments. They are tested in a flow
laboratory under reference conditions. This
means that standard flowrates are used in long
straight pipes under steady flow conditions.
Few manufacturers have comprehensive data on
installation effects and often are reluctant
to part with this data. An installation effect
is defined as the variation from laboratory
calibration to that obtained under field conditions.
Examples of flowmeter installation
effects are:
- Differences in pipe characteristics (roughness,
ovality, etc.)
- Proximity of fittings (valves/bends) which
are not present in reference testing
- Differences in temperature (fluid and ambient)
- Effect of local RFI which are not present
in reference testing
- Signal acquisition and processing errors
of the local system
The laboratory data are relevant for the meter
in the laboratory set-up only. Once the meter
is installed in the customer’s pipe, other
changes may be introduced and may include:
- Bore of the mating pipe is usually different
to that of the meter
- The flowrates in the transmission lines
may be different to the lab data
- There may be sediment, or calcification
effects within the network
- And many other reasons
All these variables may introduce
additional bias into the meter readings from
the day a meter is installed. This can only
be estimated or quantified from an in-situ verification
and this should form part of regular network
management activities. Many users, when purchasing
flowmeters, expect the manufacturer’s
specification to apply immediately and be stable
with time. This is a popular misconception and
is a key point in this paper. For example, small
mechanical meters are always tested at the manufacturer’s
premises in accordance with local Weights and
Measures regulations or international metrology
standards. This is the guarantee to the user
that when it leaves the factory it is within
predetermined calibration limits. If it is incorrectly
installed, installed too close to valves or
used on water supply with high sediment content,
it’s performance may shift from the as-new
calibration. Usually, though not always, it
under-records and over a period due to component
wear, this under-registration may increase.
There may be a steady but noticeable fall in
accuracy of the meter with time. This trend
is not just for these types of meters. ALL flowmeters,
whatever the type from any source of supply,
show these time dependent effects to varying
degrees. The key to successful network management
is to estimate the rate of degradation (if present)
with time.
It is actually quite difficult
to calibrate flowmeters to much better than
0.2% total uncertainty, so this represents the
baseline when any meter leaves a manufacturing
facility. When the user installs the meter,
this 0.2% lab figure almost certainly changes
(increases). The outlet meters from a water
treatment plant for example should have the
highest accuracy and the best installation practice.
This is because they are handling large volumes
of treated water being put into supply. Any
flowmeters of 250mm and above should be carefully
selected and even more carefully installed.
Poor installation causes the greatest source
of error. Standards give guidance on the effect
of single fittings but little data exists on
the effect of multiple fittings close to flowmeters
(ref. 7) and no tabulated data exists for pipes
of 1500mm diameter and above. This may be the
first such technical study. The main problem
is that even accepted standards currently in
use may not be totally correct (ref. 8), usually
being out of date due to the rapid developments
currently taking place in metering. This is
why site verification is vital to ensure the
meter readings are valid. In addition both manufacturer’s
recommendations and those in standards state
that the upstream and downstream lengths are
fixed independent of pipe diameter. After many
years of measurements in large pipelines, we
contend this cannot be correct and the recommendations
for the upstream length should increase as line
size increases.
All flowmeters are affected by
lack of attention at the installation stage
to details such as protruding gaskets, pipe
bore alignment, the proximity of valves, presence
of pipe branches etc. (ref. 7). In addition
to these hydraulic considerations, close attention
must be given to environmental aspects on secondary
equipment to avoid excessive vibration, flooding,
ambient temperature swings and other effects.
These can greatly affect the actual measurement
uncertainty that can be achieved. References
9 and 10 recently discussed such effects.
The overall recommendation, taking
account of the many installation factors listed
above, is that it is often difficult to demonstrate
total installed errors of much better than 2%
anyway. It is suggested that in order to achieve
this long-term figure, it is necessary to specify
instruments with an intrinsic uncertainty of
around 0.5% or better and this then allows more
than 1.5% for all the other unquantified effects.
More attention should be paid to the nature
of the meter chosen and the application needs.
If water companies would standardise on this
2% value, it is suggested that leakage figures
would become much more understandable. It is
essential that all meters are site verified
to ensure that this 2% value is maintained,
or at least any deviation is estimated.
It should be remembered that if
each meter has a total installation within the
network of 2% or better than summing all these
meter readings may still give total metered
system imbalances of 10% of more. The performance
depends critically on the number of meters,
the type of meter, the design of each installation
and the degree to which maintenance and field
verifications are performed. To this is added
the uncertainties due to unmetered consumption,
losses due to undetected leaks and any assumptions
on inferred losses. These may all combine to
give system imbalance uncertainties greater
than 25%.
Magnetic meters vs. ultrasonic
meters in large pipes
The theoretical and experimental
work presented herein shows that fluid dynamics
effects are present in all piping systems, but
above 800mm such effects increase in a non-linear
manner. These are termed ‘scale effects’
and very few results have been reported in the
open literature. In order to correctly report
unaccounted for water figures, it is necessary
for any type of meter to have the lowest installed
uncertainty. The lowest installed uncertainty
is a combination of fluid dynamic, mechanical,
environmental, and time dependant effects. It
has been proven in many studies that electromagnetic
meters are less susceptible to all these influences.
However when compared to ultrasonic meters they
are more expensive and users are always focused
on price. However reported experiences in large
pipes are not favourable and several attempts
in Indian cities using both single beam and
multi-beam systems have not lived up to expectations.
Several attempts by the first named author in
UK systems have shown results no better than
3-4%. Usage in Mumbai has shown imbalances of
10% or more in pipeline sections that are above
ground and where we can see that no leakage
occurs. The recent work in Brazil appears to
confirm the variations in velocity both radially
and axially even though the average flowrate
is constant over time.
The predictions from this work
qualitatively match the reported results from
references 3 and 4 for both ultrasonic and electromagnetic
meters. The theory presented here also indicates
greater errors in ultrasonic meters as Reynolds
number increases. This is certainly the case
for the Mumbai transmission systems, and for
single beam meters installed in other parts
of Maharastra. The data from the single beam
meters in Brazil seems to indicate that the
current mounting position may not be correct
to record to true average velocity within the
pipe. Further as the swirl is flowrate dependent,
the conclusion may be drawn that the position
should change as the flowrate changes to give
a true representation of the average flow across
the pipe.
It is always difficult to directly
compare two different flow technologies. Parameters
could be chosen to bias the reader in either
direction. The purpose here is to look at the
true basic characteristics and not to make a
selection. We have tried to choose parameters
that a customer would consider when making such
a technical judgement. The last but one line
below is a suggestion as to more correct lengths
in larger pipes. The performance characteristics
of the two technologies can be briefly compared
as follows:
The table shows that inherently
magnetic meters have lower uncertainty. The
performance of transit time ultrasonic meters
is generally inferior in all respects to magnetic
meters for large diameter pipelines. The other
major difference is the area of the flow actually
sensed. As the area sensed by even multiple
beams is a fraction of the total area, this
for us, is the major deficiency with ultrasonic
meters. We feel it is one reason why their performance
will not be as good as an electromagnetic meter,
especially in swirling flow situations. The
major advantage of ultrasonic meters is the
retrofitting of existing pipes. Here there is
simply no alternative and the higher uncertainty
will have to be accepted. Ultrasonic meters
have a purchase price advantage and can be used
on very large pipes (>3000mm).
Implications for balancing
The balancing of water supply
lines is carried out all over the world. The
majority of recent systems have been through
the use of transit time ultrasonic meters. Some
cities (Johannesburg, London) have used magnetic
type meters and reports show better success
in balancing the daily totalised flows. The
work reported in this paper has major implications
for the balancing of pipelines. We can state
the following is implied from this preliminary
study and from experience in previous work:
- Recommended upstream and downstream lengths
should increase for larger diameters
- The lengths given in accepted standards
may not correct for larger diameter lines
- Increasing pipe roughness will lengthen
the distance over which swirl will decay
- Errors in ultrasonic meter behaviour in
swirling flow are greater than for magnetic
meters
- The effect of swirl on ultrasonic meter
errors will increase as flowrate increases
- The effect of swirl on magnetic meters
will decrease as flowrate increases
- Single beam transit time ultrasonic meters
will generally under-read in swirling flow
Further work is planned in large
pipes to try to assess how the upstream length
should vary with pipe diameter, but it is inconceivable
that swirl will decay away in 10D in pipe sizes
above 1000mm. In systems where the inside surface
roughness has not been examined for ten years
or more, the effects of roughness will be unknown.
Data from reference 9 below shows a Venturi
meter with tuberculation typical of hard water
systems showing errors of between 3 and 19%
depending on the flowrate.
Similar errors have been measured
by the first named Author in pumping stations
in Southern Africa for DP-type meters installed
for more than 40 years. If the inside surface
conditions are not taken into account, we would
contend that roughness effects could swamp all
the other sources of uncertainties. A recent
report undertaken by NEL on behalf of the DTI
(ref. 12) has shown that uncertainties from
a variety of sources can add up to sizeable
metering errors. However in Appendix C of that
report, the worked example assumes no ageing
or installation effects. Such data can only
be obtained from internal inspection of the
pipeline and would be an estimate only. In our
experience in pipes of 10 years age and older,
the roughness/fluid dynamic effect would be
considerable. Many water pipes we have examined
are made with mortar lining from new. Once tuberculation
begins, then the roughness will increase well
outside expected limits. We have found roughness
of 25mm and greater in some of the older pipelines
and tuberculation of 10-15mm in pipes around
10 years old. Such influences may help to explain
why it appears difficult to balance transmission
lines for 1200mm and greater to better than
3-4% over 10kms or more. This has been our experience
over many years of site work in several parts
of the world. Using established CFD methods
might therefore tend to cloud the real effects
that we contend are present.
Closure
This may be the first experimental
work reported in large transmission systems.
The study calls into question the recommendations
from manufacturers and standards alike, with
regard to the upstream lengths required to measure
correctly in large bore pipes. The study has
also highlighted the influence of tuberculation
and pipe roughness for the first time and here
much more work needs to be done to correctly
quantify the effects of swirl on reported UFW
figures. It is therefore possible that the true
UFW figures are not accurately known in any
municipal system and reported leakage figures
are estimates only.
Acknowledgments
The mathematical model comes from
a Cranfield University PhD thesis given below
as reference (1). It has been re-interpreted
by the Author in this case study. The Authors
gratefully acknowledge major help from Mr. Bhatia
and Mr. Diverker of Greater Mumbai Municipal
Corporation for making the facilities available
for this study for the Indian work and to Jomildo
da Silva from COPASA who freely gave of his
time information and facility in the operator’s
attempts to explain the results they have been
obtaining. The instrumentation and recording
equipment for the Indian tests was kindly loaned
by Endress + Hauser GmbH, Weil-am-Rhein, Germany.
For the work in Brazil, the equipment came from
E+H Flowtec in Reinach, Switzerland. The Author
is especially grateful to the E+H Group of Companies,
and the Consult Board in particular for the
on-going support to advance knowledge in this
area.
References
- Halsey D. M: Flowmeters in Swirling Flow.
Journal of Physics E, 20, 1987 (Also PhD thesis,
School of Mechanical Engineering (DFEI), Cranfield
University, Bedfordshire)
- Private communication from IMTA Cuernavaca,
July 2004.
- Deacon, J.E: Electromagnetic meter Installation
tests. Flowmeko Conference, Budapest Hungary
20-22 September 1983
- Baker R.C. et al: Installation effects
on electromagnetic and ultrasonic flowmeters.
Flowmic report No.5, Cranfield University
report, Department of Fluid Engineering, Bedfordshire
UK, November 1988
- Carman T and Furness R.A.: Installation
and maintenance, Chapter 7 of Flow Measurement,
(Spitzer D: Ed) 3rd Edition, ISA Raleigh USA.
ISBN Number 1 55617 736 4
- Private communication from IPT, Sao Paulo,
June 2004.
- ISO 5167; Measurement of Fluid Flow in
Closed Conduits – Differential Pressure
Devices. International measurement standard:
ISO Geneva 1991
- Furness R.A.: Contentious and outstanding
issues in flow measurement. Keynote speech,
Flowmeko 94. IMEKO conference on flow in the
mid 90s. East Kilbride, Scotland 1994
- Phair D: Errors occur often in municipal
flow metering. InTech Magazine, p54-57, ISA
Publications ISSN 0192-303X, September 1997
- Furness R.A.: Do fittings really affect
flow meter performance? Journal of Measurement
and Control, IMC, London 1996
- Brown, G.J, Barton, N.A. and Moore, P.I.
Installation effects on ultrasonic flowmeters.
North Sea Flow Measurement Workshop. NEL East
Kilbride, 1999.
- Windfal Club: Flowmeter audit guidelines
and uncertainty case study reports. Project
WM04: report290/2001, National Engineering
Laboratory, East Kilbride, 2001.
(See: www.windfal.co.uk) (See
also pages 250-272 of E+H Flow Handbook 2nd
edition, June 2004: ISBN 3-9520220-4-7)
Authors
Dr.RA Furness -Consultant
JD+F Associates
The Author has more than 35 years
combined experience in flow measurement and
water supply. He is a private consultant working
in metering and leakage analysis, with have
more than 100 publications and several public
awards to his name. He is an Instrumentation
Fellow in both Europe and North America, a Chartered
Professional Engineer and a UN Consultant (New
York). |
