Water Resources Management VII 313

Water flow meter measurement uncertainties

R. P. Dias1, J. G. Dalfré Filho1 & Y. F. L. de Lucca2
1
College of Civil Engineering, Architecture and Urbanism – UNICAMP,
Brazil
2
Hydraulics Technological Center – CTH-DAEE-USP and UNICAMP,
Brazil

Abstract
In water supply systems, managers need to get water from a source and supply
an entire population. Knowledge regarding the flow of water is a critical
parameter within these systems, thus the need for flow meters. When flow
meters are placed near pipeline singularities, the velocity profile changes and can
cause measurement errors. Due to these potential errors, measurement norms and
standards are created to require flow meters to be placed in rectilinear-shaped
pipelines for a more accurate measurement. However, managers claim that in
field measurement situations, it is not always possible to meet these standards,
and consequently, adjusting the error from the acquired data is necessary.
Therefore, this study investigates how the error in the flow rate varies with the
distance from singularities when using an ultrasonic flow meter, which, as a non-
intrusive measurement method, is easily installed on the outside walls of the
pipeline. This flow meter was installed at different distances from an elbow and
the measurements obtained were compared to those from an electromagnetic
flow meter, which accumulate accurate data when installed at standard
conditions. Comparing the data from those two flow meters, it was possible to
obtain the variation of error as a function of flow and distance from the
singularity.
Keywords: uncertainties, ultrasonic flow meter, water supply systems.

1 Introduction
The water flow is an essential technical parameter that needs to be known by the
management staff of every water supply plant. This information allows the staff
to capture the water from the source, treat it and supply it to the population. In an

WIT Transactions on Ecology and The Environment, Vol 171, © 2013 WIT Press
[Link], ISSN 1743-3541 (on-line)
doi:10.2495/WRM130281
314 Water Resources Management VII

environment where water is scarce and demand is high, a precise control of water
flow is extremely important. In order for this control to be achieved, flow meters
must be employed. However, those flow meters usually work in continuous
periods, and from time to time they need to be calibrated. That calibration
becomes difficult when they are installed inside pipelines because they need to
be removed and transported to a certified lab for a precise calibration. This
represents an inconvenience and, moreover, requires the system to be completely
shut down to accommodate the maintenance of the device. With that being said,
the employment of meters that are external to pipelines is both an interesting and
potentially valuable option.
For a good precision in data acquired by the flow meter, a straight length
upstream from where it is installed is needed to avoid changes in the velocity
profile caused by singularities like curves and valves. ISO-9104 [1] requires a
straight length in the order of 10 diameters upstream of the flow meter, however
in many cases, it is not possible to meet this criteria. One example is the case
where water reuse systems are attached to pre-existing installations.
Because water flow control is related to waste and spills, it directly impacts
the budget of companies and the balance of the natural environment. Seeking to
avoid negative economic and environmental impacts, this experimental research
addresses water flow measurements in a hydraulic system, employing an
ultrasonic flow meter, which presents effective performance and can be installed
externally to pipelines. The ultrasonic meter was installed at different positions
from one singularity. An electromagnetic meter was employed to calibrate the
ultrasonic meter and to determine measurement error. The electromagnetic
device was installed according to all criteria required by technical standard
measurements. The goal of this research was to determine the uncertainties
derived from water flow measurement when an ultrasonic meter was employed
in straight lengths that are smaller than those required by normal standards.

2 Literature review
Flow meters are to be selected based on the following: type of fluid to be
measured, loss of charge, precision, available room for installation, temperature,
operational pressure, reliability and cost. All the flow meters have uncertainties
in measurements, but, prior to determining those uncertainties, it is necessary to
describe the types of errors that can occur.
According to ISO-5168 [2] and Vuolo [3], three types of errors exist: random
errors, which can be due to a large number of variables and cannot be controlled
by the researcher, and that cause measurements to differ from the true value in a
constant manner; systematic errors, which occur when all measures differ from
the true value in a constant manner; and spurious errors, caused by human error
or problems within the device. Systematic errors can be minimized with
instrument calibration. Additionally, even when environmental factors do not
interfere with measurement, it is customary to take note of the highest possible
number of these factors, such as temperature, pressure, and humidity.

WIT Transactions on Ecology and The Environment, Vol 171, © 2013 WIT Press
[Link], ISSN 1743-3541 (on-line)
 Water Resources Management VII 315

The goal of the measurement is to determine the specific measure of the target
variable. According to Kamal [4], in general, this process will only be completed
when the uncertainties are determined. The number of significant digits must
also be calculated and should be used to represent both the experimental standard
deviation and the mean standard deviation. According to ISO-5168 [2] and
Miller [5], if the systematic error is equal to zero, the flow Q given by eqn. (1)
will include the true mean approximately 95% of the time:

∙ (1)

where: = Mean flow; t95 = Student T value, α = 95%, dependent on degrees of

freedom (df = n - 1); = mean standard deviation of the flow.
Additionally, according to Miller [5], the systematic error Bi can be calculated
through the relationship between the mean and arbitrary value as the reference
value. In the present study, the systematic error was calculated through the
comparison between the mean flow from the ultrasonic meter and the mean flow
from the electromagnetic meter (reference value), according to eqn. (2):

, ,
% . 100 (2)
,

where: Bi = systematic error of situation i (in percentage); Qi,u = mean flow in

the ultrasonic meter in situation i; Qi,e = mean flow in the electromagnetic meter
in situation i.
Commonly, the correction factor FB is used to adjust readings when a
systematic error is calculated using eqn. (2). The mean value can be corrected
when it is multiplied by the value of FB, given by eqn. (3):

1 (3)

Uncertainty, U, in a measurement can be calculated by combining the random

and systematic errors in order to determine a confidence interval. It’s considered
a standard deviation in the systematic error and a standard deviation in the
random error. Therefore, eqn. (4) is obtained.

(4)
√

where: U = uncertainty; B = systematic error; t95 = Student T value; n = sample

size; sxi = experimental standard deviation.
If the value of the systematic error is known, each measurement needs to be
corrected with the correction factor FB, in which case the uncertainty, U, is given
by eqn. (5) below:
1 (5)

WIT Transactions on Ecology and The Environment, Vol 171, © 2013 WIT Press
[Link], ISSN 1743-3541 (on-line)
316 Water Resources Management VII

In this experiment, the variations of the systematic and random errors were
determined in the case when an ultrasonic meter was installed near singularities.

3 Experimental facility
This research was performed at the Laboratory of Hydraulics and Fluid
Mechanics at the College of Civil Engineering, Architecture and Urbanism at
UNICAMP, located in Campinas, Brazil. The test bench was constructed using
two straight pipeline pieces which interconnect through two 90-degree curves.
The flow meters were installed within the straight pieces. Two valves were
employed to control the flow, as demonstrated in fig. 1.

Figure 1: Testing Bench diagram.

The electromagnetic flow meter employed in the study is a 4” Model 570TM

by Fisher-Rosemount from Brazil, with uncertainty equal to 1% of the measured
value. This meter was installed in the straight length section as demonstrated in
fig. 1, in accordance with the norms, criteria and suggestions by ISO-9104 [1],
Delmée [6], Spink [7], Troskolanski [8] and Souza and Bortoni [9]. The
ultrasonic flow meter is a TransPortTM Model PT868 Portable Flowmeter by
Panametrics. According to the manufacturer’s manual, its uncertainty value
depends on the pipeline diameter and on the velocity of the fluid, whose values
are presented in table 1:

Table 1: Typical error values for the Ultrasonic Flowmeter model PT868 by
Panametrics.

Speed < 0.3 m/s Speed > 0.3 m/s

Diameter > 150 mm (6”) ± 0.01 m/s 2% (1% if calibrated)
Diameter ≤ 150 mm (6”) ± 0.05 m/s 2% to 5%

The ultrasonic meter was installed at 10 different positions from valve V1

upstream, as shown in fig. 1. For each position of the flow meter, tests were
conducted for 20 different flows, at a flow interval between 0 and 1,000 1/min.
For each test, a reading of the flows was performed in the ultrasonic and

WIT Transactions on Ecology and The Environment, Vol 171, © 2013 WIT Press
[Link], ISSN 1743-3541 (on-line)
 Water Resources Management VII 317

magnetic meters, and the results were recorded for future analysis. The
measuring distances from the valve were as follows: 1Dint = 98mm, 2Dint =
196mm, 3Dint = 294mm, 4Dint = 392mm, 5Dint = 490mm, 6Dint = 588mm, 7Dint =
686mm, 8Dint = 784mm, 9Dint = 882mm and 10Dint = 980mm.
Per the manufacturer’s recommendation, the ultrasonic meter was installed
with the two transducers placed on the same horizontal line, at the upper side of
the pipeline. Water temperature was considered to be at 20ºC. Testing began
once all relevant parameters for the test were established.

4 Results and analysis

When repeated flow readings are performed in a lab, the obtained values are
subject to random errors and remain at a certain average for a permanent flow
regimen. Maximum and minimum values were obtained for each flow range
established during 30 seconds for each position of the ultrasonic meter. The
mean flow Qi was then calculated, creating the arithmetic mean of the maximum
and minimum flows at the established flow range.
For the distance of 1Dint, it was observed that, even for distances near to valve
V1 upstream, the ultrasonic meter generated reasonable results as when
compared to the flow results generated by the electromagnetic meter used as a
reference. However, for this specific distance and flow of 1,000 1/min, two
equipment errors occurred, E1 (Low Signal) and E2 (Sound Speed Error), and as
a result no readings were recorded. When the flow was decreased through valve
V2 downstream to approximately 950 1/min, the equipment no longer presented
reading errors, and the readings were performed. It is probable that the
equipment presented this error because four traverses were used, which caused
the transducers to be located further apart. It is important to remember that a
traverse represents each crossing made by the ultrasonic wave, from one side to
the other side of the pipeline. If the transducers are installed on the same side of
the pipeline, there will always be an even number of traverse. Another factor to
be considered is that the fluid speed in the 1,000 1/min flow was higher, which
caused more turbulence and interfered with flow measurement.

Figure 2: Systematic flow error variation at a distance of 10 Dint.

WIT Transactions on Ecology and The Environment, Vol 171, © 2013 WIT Press
[Link], ISSN 1743-3541 (on-line)
318 Water Resources Management VII

Having the mean flow of each meter, and establishing that the mean obtained
from the electromagnetic meter is the flows reference value, it is possible to
determine the systematic error present in each measurement by using eqn. 2. On
calculating the systematic error for each distance and different flows, graphs
were prepared to represent how the systematic error varies with the flow. The
graph in fig. 2 presents this variation at a distance of 10 Dint from the valve.
The tendency for the systematic error to increase or decrease as the flow
increases was not present. The same was found true as the distance decreases, the
tendency for the systematic error to increase or decrease was not present. Fig. 3
presents the maximum, minimum and mean values between distances for each
flow. Fig. 4 presents the maximum, minimum and mean values between flows
for each distance.

Figure 3: Maximum, minimum and mean values of the systematic error

between distances according to flow.

Figure 4: Maximum, minimum and mean values of the systematic error

between the 20 measured flows according to distance.

In fig. 3, a flow of 50 1/min generated the highest systematic error which also
corresponds to the distance of 1Dint. This plot is out of the range of mean values,

WIT Transactions on Ecology and The Environment, Vol 171, © 2013 WIT Press
[Link], ISSN 1743-3541 (on-line)
 Water Resources Management VII 319

and this situation could have been caused either by a measurement error or as a
consequence of the low speed along with the short distance from the valve. It
should also be noted that for low flow values, a small difference in values causes
a relatively large systematic error.
Because the systematic error is independent from both the distance to the
valve and from the flow, it is possible to obtain a mean, a mean standard
deviation, the number of significant digits and the calibration factor that can be
employed in the equipment, according to ISO-5168 [2]. These referred values are
described in table 2. There were a total of 400 flow points which were obtained
in the lab. Therefore, one can apply the Student T-value of 2.0, according to
Miller [5], in order to calculate a standard deviation with 95% confidence.

Table 2: Final values of the systematic error for the Ultrasonic meter.

Mean Systematic Error B = - 3.236%

Systematic Error Standard Deviation SB = ± 0.159%
Mean Systematic Error Standard Deviation
t95*sB = ± 0.319%
with the Application of Student T
Significant Digits 0.003
Mean Systematic Error Final Value B = -3.236 ± 0.319%
Calibration Factor 1.03344

One can notice that the Systematic Error Value is close to the Maximum
Value of the uncertainty of the instrument, which is 5%, as presented in table 1.
Hence, the calibration factor can be used in the ultrasonic meter in order to
improve measurement precision.
Relating to random error, fig. 5 and fig. 6 were prepared and present a
comparison between the standard deviation in relation to the mean values
including flows and distances. Fig. 5 presents the maximum, minimum and mean
values of the standard deviation between the distances for each flow point. Fig. 6
presents the maximum, minimum and mean values of the standard deviation
between flows for each distance.

Figure 5: Maximum, minimum and mean values of the standard deviation

between distances for each flow point.

WIT Transactions on Ecology and The Environment, Vol 171, © 2013 WIT Press
[Link], ISSN 1743-3541 (on-line)
320 Water Resources Management VII

Figure 6: Maximum, minimum and mean values of the standard deviation

between flows for each distance.

The graph in fig. 5 shows a higher standard deviation for the low flows. This
can be explained due to the fact that even one digit that varies represents a
significant portion of the total. A constant random error of approximately 0.17%
was obtained between readings of 300-850 1/min, along with a tendency of
increase in the random error as the flow surpasses the value of 850 1/min.
On the contrary, the graph in fig. 6 demonstrates that the random error is
independent of the distance to a specific singularity, remaining at an average of
0.33%. Some maximum points appear to be discrepant, which could have been
caused by an installation problem, despite all the caution taken.
The worst-case scenario of the relative standard deviation occurred with a
flow around 1,000 1/min, for which 27 values were recorded. Hence, it has a T-
distribution value of 2.056, in accordance with Miller [5]. The results are
presented in table 3.

Table 3: Random error final values of the ultrasonic meter for the worst-case
scenario.

Relative Experimental Standard sxi = ± 1.3%

Deviation
Final Value of the Random Error after t95*sxi = ± 2.6%
Employment of the T-distribution
Significant Digits 0.2

By knowing the equipment’s random and systematic errors, it is possible to

calculate the total uncertainty, U, through the use of eqn. (4). The results found
in this case was an uncertainty, U, between -5.855% and -0.617% of the
measurement value. In case every measurement is calibrated with the calibration
factor, the uncertainty is then between -2.619 % and 2.619 % of the flow value,
for the critical case in which the flow equals 1,000 1/min. This value is within

WIT Transactions on Ecology and The Environment, Vol 171, © 2013 WIT Press
[Link], ISSN 1743-3541 (on-line)
 Water Resources Management VII 321

the uncertainty range of the equipment, described in the manufacturer’s manual

as being between ± 2%–5% of the flow value.
When analyzing the central range of the graph in fig. 5, with flow values
between 300~850 l/min, table 4 is obtained through the same method described
above. In this case, 12 flow intervals with 3 points each for 10 different distances
were adopted for a total of 360 points, with T-value equal to 2.0.

Table 4: Ultrasonic meter random error final values for flow range between
300 1/min and 850 1/min.

Relative Standard Deviation sxi = ±1.01%

Final Value of the Random Error after t95*sxi = ±2.02%
Applying the T-Distribution
Significant Digits 0.04

The results for uncertainty, U, in this case show an uncertainty between

-5.281% and -1.191% of the measurement value. In each case where the
measurement is calibrated with the calibration factor, the uncertainty lies
between -2.045 % and 2.045 % of the flow value for a flow range between 300
l/min and 850 l/min. By calibrating the meter, the uncertainty obtained is exactly
at the lower range of the uncertainty of the equipment, described in the
manufacturer’s manual. This value corresponds to ± 2% of the flow value.
Therefore, the ultrasonic meter does present a systematic error, but when the
obtained calibration factor is employed, measurement values become more
precise. It was also verified that the systematic error does not vary according to
the flow, nor does it vary according to the distance to the singularity, which in
this case is a valve. On the other hand, the random error does vary according to
the flow, although it does not vary according to the distance to the valve. There
is a range in which the random error is shown to be constant, and when a flow
value of 850 1/min was achieved, this error tended to increase. When applying a
calibration factor to the measurements, it is safe to say that the uncertainty value
will be ± 2.619 % in the most critical scenario where the flow is approximately
1,000 1/min. At the same time, for a flow range between 350 1/min and 850
1/min, the total uncertainty drops to a value of ± 2.045 %.

5 Conclusion
It can be concluded that flow meters present both systematic and random errors
that affect the total uncertainty of the collected data. In the case of the ultrasonic
meter, it was demonstrated that the systematic error does not vary according to
the distance to a valve, nor does it vary with the flow. It was also demonstrated
that the meter’s random errors are constant at a specific flow range, and increase
as the flow increases, as the random error showed to be dependent on fluid
velocity. However, the random error is not dependent on the distance variable.
It was also noted that the devices need to be adequately calibrated before they
are installed, if they are to be used in real-world, non-controlled environments.

WIT Transactions on Ecology and The Environment, Vol 171, © 2013 WIT Press
[Link], ISSN 1743-3541 (on-line)
322 Water Resources Management VII

With this calibration, the uncertainty range of measurements can be decreased,

which is made possible by calculating the systematic error in comparison to a
standard meter used as a reference. In the case of facilities where the length of
straight stretches does not meet the 10 diameters upstream and five diameters
downstream that are required by the standards for ultrasonic meters, it is still
possible to achieve adequate precision, given the conditions described by this
article.
However, it is not possible to affirm that a measurement yielded by an
ultrasonic meter will present systematic errors when installed near different
singularities. In order to investigate that, future studies need to be developed.

References
[1] International Organization For Standardization - ISO 9104. Measurement of
fluid flow closed conduit methods of evaluating the performance of
electromagnetic flow meter for liquids. Geneva, Switzerland, 18 p., 1991.
[2] International Organization For Standardization - ISO 5168. Measurement of
fluid flow: Procedures for the evaluation of uncertainties. Geneva,
Switzerland, 65 p., 2005.
[3] Vuolo, J. H. Fundaments of error theory. São Paulo: E. Blucher, 225 p.,
1992.
[4] Kamal, I.A.R. Measurement and Instrumentation Techniques. Ed. do autor,
369p., 2000.
[5] Miller, W.R., Flow measurement engineering handbook. 3rd edition. New
York: McGraw-Hill, 1168 p., 1996.
[6] Delmée, G. J., Manual of flow measurement. 2nd edition. São Paulo: Ed.
Edgard Blücher, 476 p., 1983.
[7] Spink, L.K. Principles and practice flow meter engineering. Foxborough,
Mass: Foxboro Company, 575 p., 1967.
[8] Troskolanski, A.T., Hydrometry: theory and practice of hydraulic
measurement. New York: Pergamon Press, 684 p., 1960.
[9] Souza, Z.; Bortoni, E.C., Instrumentation for Energetic and Industrial
Systems. 1st edition. Itajubá: Ed. Novo Mundo, 387 p, 2006.

WIT Transactions on Ecology and The Environment, Vol 171, © 2013 WIT Press
[Link], ISSN 1743-3541 (on-line)