Flow Measurement and Instrumentation 42 (2015) 16–25

Contents lists available at ScienceDirect

Flow Measurement and Instrumentation

journal homepage: www.elsevier.com/locate/flowmeasinst

Electromagnetic flow meters for open channels: Current

state and development prospects
Zbigniew Watral a, Jacek Jakubowski a, Andrzej Michalski a,b,n
a
Military University of Technology, Institute of Electronic Systems, S. Kaliskiego 2 Str, 00-908 Warsaw, Poland
b
Warsaw University of Technology, Institute of Theory of Electrical Engineering, Measurement and Information Systems, Koszykowa 75 Str. 00-662 Warsaw, Poland

art ic l e i nf o a b s t r a c t

Article history: Electromagnetic method to measure water flow has been known for nearly 180 years. This paper presents
Received 20 August 2014 how the method has undergone development and describes in detail the most common electromagnetic
Received in revised form flowmeters used for open channels. It also presents specific methods of signal processing used in this type
15 December 2014
of flowmeters. The analysis is associated directly with the relevant types of induced magnetic field. Some of
Accepted 6 January 2015
the analyzed types of magnetic fields considered are sine waves and bipolar rectangular fields. The concept
Available online 21 January 2015
of low-energy electromagnetic flow meter was introduced as part of deliberations concerning current
Keywords: development trends.
Electromagnetic flowmeter & 2015 Elsevier Ltd. All rights reserved.
Signal processing
Mathematical model
Exciting coil
Pulse flowmeter
Renewable energy sources

1. Introduction Some of the electric and magnetic properties of the environment

need to be assumed in order to describe how an electromagnetic
Electromagnetic flowmeters have been researched and subject of flowmeter operates [1]. Those assumptions are as follows:
analysis for over 180 years. Their indisputable advantages earned
them a permanent place on the flow measurement market, even  Flowing liquid is characterized by magnetic permeability
though competitive methods achieve comparable performance. In the μ ¼ μ0.
area of closed and open measurement channels, the main competitor  Water conductivity, γw, and conductivity of channel bed, γg, are
of the electromagnetic flowmeter is an ultrasonic flowmeter, which in isotropic and they are not functions of magnetic field and water
many aspects has comparable properties, especially with regards to velocity. The Hall effect is excluded.
 The type of flow and applied
measurements taken in pipelines. Since introduction of the Doppler !field excludes self-induction.
profile gauge, the ultrasonic method of measurements in open  Velocity vector of a liquid, ϑ has only one component (recti-
channels has become a serious alternative to electromagnetic meth- linear flow) in the area of magnetic field e.g. ϑy and ∂ϑy =∂y ¼ 0.
ods. By giving an overview of different solutions, this paper attempts
to systematize and structure the knowledge concerning application of
Analysis of distribution of magnetic field can be then limited to the
electromagnetic flow measurement method in open flow channels.
stationary state because its commutation frequency is low. All assu-
mptions were taken to simplify the analysis and do not affect practical
2. Mathematical model of electromagnetic flow meter application of electromagnetic methods. Shercliff [2] had described the
basic equation in his electromagnetic flowmeter theory,
Faraday’s law is the principle behind electromagnetic flow meters.  

 ! !
It associates electric potential φ induced
! in the measurement zone div γ gradφ ¼ divγ ϑ  B ; ð2:1Þ
!
with both liquid velocity vector, ϑ and magnetic flux density, B
(Fig. 2.1). where γ is electric conductivity.
Voltage, Ue, induced between two electrodes e1 and e2 of the
n
Corresponding author at: Warsaw University of Technology, Institute of Theory flowmeter, (Fig. 2.1) is given by
Z
of Electrical Engineering, Measurement and Information Systems, Koszykowa 75
! ! !
Str. 00-662 Warsaw, Poland. Tel.: þ 48 22 628 45 68. U e ¼ φe1  φe2 ¼ ϑ  B  J dv; ð2:2Þ
E-mail address: anmi@iem.pw.edu.pl (A. Michalski). V

http://dx.doi.org/10.1016/j.flowmeasinst.2015.01.003
0955-5986/& 2015 Elsevier Ltd. All rights reserved.
 Z. Watral et al. / Flow Measurement and Instrumentation 42 (2015) 16–25 17

long uniform magnetic field and two infinitely conducting long

electrodes spanning two opposite walls [4]. This configuration of
electrodes generates a uniform virtual current. Should magnetic flux
! !
density B be uniform, the weight vector W is also uniform and
consequently the output signal (electrode potential) is independent on
the shape of velocity profile. As shown in for instance [5], circular flow
meters with point electrodes could not be ideal as long as velocity
distribution is unspecified.

3. Historical overview

The very first attempts to measure water flow using the electro-
magnetic method had been carried out in the nineteenth century by
Fig. 2.1. The idea of the electromagnetic flow measurement method in open flow Faraday, who tried to measure flow of the river Thames. Another
channel.
attempt at flow measurement was made nearly 100 years later. In
1915, Smith and Slepian were granted a patent for an electromagnetic
!
where V denotes the volume of measurement zone, J denotes a velocity meter used by vessels [6] and Young et al. [7] tried to measure
virtual density current vector, and φe denotes electric potential of flow of a river. The operating principle of a flowmeter is based on
an electrode. analysis of magnetic and electric fields carried out by Williams [8].
!
Here J is a “virtual current” i.e. current density set up in the Kolin [9] was probably first to suggest use of such a system in
liquid by driving an imaginary, unit current into electrode 1 and measuring flow in closed channels. The result of efforts initiated in
!
out of electrode 2. Bevir [3] named J the “virtual current” in order 1936 was a patent for an electromagnetic flowmeter granted in 1939.
to distinguish it from currents existing in working flowmeters. The That system used a constant magnetic field generated by a permanent
!
magnitude and distribution of virtual current J depends on magnet or an electromagnet powered by direct current. One of the
geometry of flow channel, the shape of electrodes, and conductiv- very first industrial flow meters was Tobiflux developed in the
ity of the liquid and channel walls. If the excitation coil of the Netherlands. It uses a coil of special shape and an iron core to increase
flowmeter generates a field such that induction of magnetic circuit. This made it possible to reduce dimen-
! !  ! sions of exciting coil-based solutions and improve stability of metro-
rot B  J ¼ rot W ¼ 0; ð2:3Þ logical parameters of the flowmeter. A company named Foxboro filed
! ! ! in 1952 for a patent, where they proposed the use of air coils and a
where W ¼ B  J is known as the weight vector [3] over the
simplified generation of magnetic field. The first flowmeter based on
entire domain V, then the voltage between electrodes
! is propor-
this approach was used in 1954 and since then electromagnetic
tional to mean velocity of unspecified flow ϑ m . This kind of
flowmeters technology has underwent steady development. The
flowmeter is considered as “ideal”.
! demand generated by the processing industry drove rapid develop-
The final potential difference (Ue), using weight vector ( W ),
ment of electromagnetic flowmeters solutions. Flow measurement
which is perceived as an extension of Shercliff’s weight function
was a serious problem in the paper industry, due to the density of the
[2], is expressed as follows:
Z medium. The solution came in form of an electromagnetic flowmeter
! ! which does not generate additional increase of hydraulic resistance in
Ue ¼ W  ϑ dv ð2:4Þ
V  the system, ensures correct measurement of the flow of suspensions
! and is relatively resistant to environmental factors.
According to our assumptions, flow is rectilinear ϑ ¼ ½0; ϑy ðx;
! First attempts to apply the electromagnetic method to measure
zÞ; 0Þ and vector W may be written as a scalar W. Then Eq. (2.2) may
be reduced to (2.5) currents in rivers, as mentioned earlier were unsuccessful due to
Z imperfections of measurement techniques. Between 1851 and
U e ¼ ϑy Wðx; zÞds: ð2:5Þ 1918, Young and Gerrard [7] made further attempts after Faraday’s
S work in the English Channel. That point on, there are no studies
here S is the cross-section of flow channel and W(x, z) is the rectilinear that would prove advancement of methods for electromagnetic
weight function [34,35], given by flow measurement in rivers. Thurlemann [10] mentioned in his
Z þ1 analysis of a rectangular flowmeter that potentially it could be
W ðx; zÞ ¼ W y dy; ð2:6Þ used in measuring flow of rivers. Development of modern electro-
1
nics (during the 60s) sparked a renewed interest in this method. In
where 1967, Engel published his paper [11] where he described how to
use electromagnetic method of flow measurement in open chan-
W y ðx; y; zÞ ¼ BZ J X  BX J Z : ð2:7Þ
nels. Engel in his deliberations considered a channel with a
rectangular cross section and placed electrodes on opposite banks
of the channel. He established a uniform magnetic field over the
1. Wy is the component of W. W(x,z) is a measure of the response entire area where measurements were taken. In order to measure
to rectilinear flow at a given point (x, z) in cross section of the flow, both the velocity and water level had to be measured. Engel
flow channel. proposed a measurement method utilizing an interesting relation-
2. The condition for an “ideal” flowmeter (2.3) may be reduced to ship between amplitude of the voltage induced in the conductor
Z þ1 loop and a water level. The loop consists of a liquid and electrodes.
W ðx; zÞ ¼ W y dy ¼ const: In 1970, Gils in [12] introduced an electromagnetic flow meter
1
system installed in a drainage channel in Bad Rehburg (Germany).
In an ideal flowmeter, the measurement signal does not depend on Characteristic feature of the system was the use of cross- coil
water level in neither flow channel nor velocity vector distribution. An system to excite an artificial magnetic field. Hence a variable
ideal flowmeter features a rectangular cross-section, insulating walls, magnetic field could be generated in two perpendicular directions,
18 Z. Watral et al. / Flow Measurement and Instrumentation 42 (2015) 16–25
thus enabling sequential measurement of velocity and water level
in the channel. The system proposed by Engel and Gils used so-
called “long electrodes” located on opposite banks of the channel
bottom-to-surface and excitation coils providing magnetic field
with uniform distribution. Rolff and Starke [13,14] used point
electrodes placed in corners of the flow channel. The work by
Herschy and Newman [15–17] was the main, go-to body of work
back in that period, which dealt with the subject comprehensively
for the very first time. It discussed the possibility of using the
electromagnetic method for flow measurement in rivers. They
used a simplified relationship between magnetic field, average
velocity of the water and the width of the river and treated
potential difference measured by the electrodes as an output
signal. This relationship is valid only for electrically insulated
rectangular flow channels. Environmental influences on the mea-
surement signal (e.g. variable bottom of the channel and water
conductivity, uneven distribution of magnetic field) were taken
into account by coefficients of proportionality, empirically esti-
mated during calibration phase. From a purely economic and
practical feasibility point-of-view, the authors proposed flat coil
as a source of magnetic field placed under the bottom of the river.
In the 70s Great Britain was the only country manufacturing
electromagnetic flow meters designed to measure average velocity
in natural watercourses (Plessey Radar Ltd). The maximum width
of the river, whose flow rate could be measured, was 100 m. The
error of 2% and the sensitivity of 5 mm/s were obtained.
The structures of primary transducers for electromagnetic flow
meters used in open channels, currently existing on the market, may
be divided into three basic solutions (Fig. 3.1). They mainly differ in
the way the artificial magnetic field is generated. In case of the
electrically insulated flow channel, the exciting coil may be an integral
part of the flow channel (Fig. 3.1b) thus making it relatively easy to
change the installation place. This construction ensures uniform
distribution of induced magnetic field thus enabling to measure
virtually any flow velocity. Flowmeter using this type of coil was
developed at the Warsaw University of Technology in the 1990s and
was first installed in the Łęgnowo polder near Bydgoszcz [18,19].
In the 1980s and early 1990s a system was built with a separate
exciting coil (e.g. Sarasota Instrumentation [20]). In a man-made
flow channel the coil was placed above the channel in a specially
prepared rectangular frame (Fig. 3.1a). Obtaining a uniform mag-
netic field distribution proved challenging with this solution. Thus
it became necessary to introduce correction factors taking into
account variable water level. The velocity distribution symmetry of
the measured flows has to be axial.
For installations in natural flow channels, the exciting coil could
also be buried beneath the channel bottom (Fig. 3.1c). However, this
requires an additional electrical insulation (plastic film) between the
bottom of the channel and the watercourse. The biggest disadvantage
of this construction is a non-uniform distribution of excited magnetic
field and the occurrence of the short-circuit effects at the edges of the
Fig. 3.1. Structures of exciting coils for electromagnetic flow meters for open channels, (a) excitation coil located above the water, (b) excitation coil as an integral part of the
flow channel, (c) excitation coil placed on or under the bottom of the flow channel.
insulation film. Both of those structures were based on research
carried out in the 1970s by Plessey Radar Instrumentation (UK).
Additional measurements of water level in the flow channel
were necessary to be taken in case of all of applications shown in
Fig. 3.1. In the structure shown in Fig. 3.1b, the measurement was
taken using the same pair of slightly inclined electrodes using
electromagnetic effects [21]. The other two solutions, utilized a
separate ultrasonic level meter to take the measurement.
Electromagnetic flow meters for open channels are quite expensive
devices but offer a number of significant advantages. They are suitable
for taking a wide dynamic range of measurements, they don’t feature
elements in the measurement section that could disrupt the flow, the
signal is proportional to the actual average flow velocity and they are
not sensitive to back flows. An average measurement error using an
electromagnetic flowmeter generally does not exceed 3%. They are
installed mainly in rivers (canals) several meters wide, which could
not be described by a stable relationship between water level and the
flow. A typical place where electromagnetic flowmeters could be
installed should meet several conditions. Flowmeter should be
installed on a straight section of the channel not less than 3 channel
widths long. The purpose behind this condition is to obtain a flow
profile similar to axially symmetric. In addition, one should avoid
proximity of power lines, electric railways, and large parts made of
steel, which could distort the induced magnetic field. Recently,
ongoing research focused on the use of electromagnetic methods in
measuring flow in polder areas and wastewater treatment plants. That
research effort was carried out in two centers in Poland—Warsaw
University of Technology and Military University of Technology.
Several structures of electromagnetic flow meters for channels whose
width does not exceed several meters were designed. Coil winding
technology and very specific signal processing algorithms were
developed [22,23]. Due to reported demand from companies mana-
ging water resources in areas undergoing extensive land reclamation,
a flowmeter with extremely low energy consumption was required,
which would operate fully autonomously in the field [24,25].
4. Dedicated method of signals processing in electromagnetic
flow meters for open channels
Processing algorithms have always aimed to achieve lowest
possible measurement uncertainty on a given stage of development
of measuring hardware and actuators. In the case of electromagnetic
flowmeters, the uncertainty is caused mainly by undesirable noise
voltages of various origin [26–29]. The most important one is caused
by electrochemical reactions between metal electrodes (electron
conduction) and liquid (ionic conduction). Many different materials
are available to make the electrodes, each with certain value of the
electrode potential relative to the standard hydrogen electrode SHE
[30]. Electric potentials exist on each of the electrodes and as such
 Z. Watral et al. / Flow Measurement and Instrumentation 42 (2015) 16–25 19

form a common mode voltage that can be theoretically easily nullified The output signal could not be separated from this kind of noise
with the use of instrumentation amplifier of sufficient CMR coefficient. using phase sensitive devices. In order to do so, a zeroing procedure is
However, in real flow measurements surface conditions are change- necessary. Zeroing is done by measuring output voltage under
able and physical friction occurs between electrodes and the liquid. As conditions of no flow in the channel and then adjusting achieved
a result there is a mismatch causing differential noise. The noise offset level to keep it at zero. The procedure should be repeated as the
resembles a low frequency random AC voltage signal of the 1/f type time lapses since the noise changes over time. An additional dis-
spectrum as depicted in Fig. 4.1. advantage of the AC flowmeter is the fact that ac line voltage is not
The second kind of undesirable voltages are noises generated due regulated and requires taking into account some corrections. Another
to the fact that the magnetic field is time-variant. The voltages are solution developed to eliminate above drawbacks of the AC flowmeter
known as induced noises or the transformer effect. They are caused by (i.e. the necessity to take into account the quadrature component, the
an electric loop formed in magnetic field that consists of electrodes, offset voltage as well as the correction of changes in ac line voltage) is
liquid and lead wires. The voltage induced inside the loop is propor- based on excitation that uses bipolar rectangular waveform. The
tional to the rate of change of magnetic field. Moreover, it may be excitation offers alternating time intervals of quasi-steady magnetic
modified by magnetic field generated by eddy currents inside the induction as depicted in Fig. 4.3.
liquid alongside the effect of magnetic viscosity. The intensity of the The level of the induced noises was assumed to be neglected when
eddy currents depends on the temperature, local conductivity and the period of the excitation waveform is long enough. Consequently a
adhesion processes taking place in the channel. The effect of magnetic
viscosity is caused by the influence of electric and magnetic field on
the process of domains’ rotation inside areas neighboring with
measurement zone. As a result, a delay in environment’s reaction to
changing magnetic field is observed. This effect disappears after a
certain decay time. The resulting noise voltage could cause problems
with zero as it is not stable. Besides, there are noise voltages caused by
electromagnetic coupling with power lines of 50/60 Hz, other power
devices or noises evoked by stray currents.
All those undesirable voltages can effectively hinder, disrupt or
disturb relatively weak flow signals. In order to maximize the
probability of taking successful measurements in such conditions,
various approaches have been adopted over past decades to excite
magnetic field. Those approaches put emphasis on dedicated proces-
sing of the resulting output signals. One of the solutions uses a
common power line of 50/60 Hz to excite the field as the undesirable
electrochemical noises occupy lower frequency region than that of the
output flow signal achieved in this way (see Fig. 4.1).
Additional motivation is that the energy transited by the power
line is sufficient to get useful flow signals. A transducer, which
produces excitation in that manner, is referred to as AC flowmeter
and is renowned for its relatively quick response to changes in the
flow. The direct result of the transformer effect in AC flowmeters is
an additional voltage that is ideally 901 out of phase with the flow
signal, recognized as the quadrature voltage. The signal may be
taken into account and separated from the flow signal by phase
detection circuits employing synchronous demodulation [31,32].
The in-phase noises that take the same sinusoidal waveforms as
the AC excitation are much more problematic. Such noise is made
by the flowmeter magnet coils themselves and other devices e.g.
electric motors or welding equipment—Fig. 4.2. Fig. 4.2. The problem of the offset voltage in the AC electromagnetic flow meter.

Fig. 4.1. Power spectral densities of noise voltages acquired in two different conditions of electro-kinetic processes [24].
20 Z. Watral et al. / Flow Measurement and Instrumentation 42 (2015) 16–25
Fig. 4.3. Standard signal processing in the DC pulse flow meter. Case A corresponds to
the excitation frequency of 15 Hz when the average value within sampling interval
fluctuates with the induced noise. Case B corresponds to the excitation frequency of
5 Hz when the sampling interval is almost unaffected by the induced noise.
possibility to measure the voltage flow signal with only the electrolytic
noise superimposed arises within the first half of the period. Voltage
flow signal with opposite polarization but also with added noise is
measured during the second half of the period. If the electrolytic noise
doesn’t change much, a simple subtraction of those two above
voltages efficiently reduces the influence of that kind of noise. This
process is performed several times a second eliminating the need to
stop the flow in order to proceed with zeroing. This solution is
referred to as pulsed DC flowmeter. It is additionally immune to
power line noises of 50/60 Hz if the intervals of sampling cover an
integer number of their periods.
The main disadvantage is a relatively slow response to flow
changes if compared to the AC solution. The reason for that is low
excitation frequency that usually doesn’t exceed several hertz.
Moreover, if the electrolytic noises are not particularly slow, the
fundamental assumption of the measuring algorithm is ceases to
be valid and the DC flowmeter generates results that are much
more uncertain than their AC counterparts. All the aforementioned
factors explain why both solutions are available on the market
today as commercial transducers. Furthermore the complemen-
tariness of their advantages gave inception to development of a
third, more complex method of excitation that is nowadays
applied in magmeters. The method in question is referred to as
dual frequency excitation [33]. The general idea is to use a more
combined current driving the coil containing high and low
frequency components occurring concurrently. They both provide
good noise properties and zero stability, respectively.
An alternative solution that uses methods of statistical proces-
sing of the output signal is presented in [24]. The idea under-
pinning above pulse DC excitation is to calculate the mean of the
signal over limited time intervals within the period where the
transformer effect tends to be zero. The new concept represents a
generalization of the idea and takes the entire period to perform
the calculations including the intervals, where the transformer
effect exists. General motivation behind such an approach is the
fact that experimental flow signal captured by electrodes may be
modeled by a linear combination of signals’ waveforms formed by
magnetic field and its rate of change. The former one is modulated
by flow velocity according to the Faraday’s law, whereas the latter
one is caused by the transformer effect. Because the magnetic field
is proportional to the coil current, the equation describing flow
signal u(t) may be formulated as
diðtÞ
uðtÞ ¼ c1  iðtÞ þ c2  ; ð4:1Þ
dt
where unknown parameters c1 and c2 give the flow velocity and
intensity of the transformer effect, respectively. The model (4.1) is
linearly dependent on the two parameters and as such may
become a subject of statistical processing methodology that offers
minimum variance unbiased estimators to find them under given
noisy conditions. The methodology assumes that at the first step
the signal is affected by white Gaussian noise described by the
diagonal covariance matrix C ¼ σ2I. The signal is then treated as a
vector u of samples and expressed using a matrix product and an
additive influence of the noise vector ε. The product includes
matrix H formed by the samples of the current i and its derivative
according to (4.1) as well as a vector Θ ¼ ½c1 c2 T of the two
parameters to be estimated
u ¼ HΘ þ ε: ð4:2Þ
For such a model, the methodology based on maximum like-
^ for Θ
lihood offers the solution producing the estimator Θ
^ ¼ ðHT HÞ  1 HT u:
Θ ð4:3Þ
However, because disturbances affecting flow measurements
rather resemble colored noises, the proposal to treat them as white
must be rejected and thus the minimum variance is no longer
available. The second step is to reformulate the estimator (4.3) in
order to transform it into the minimum variance estimator—even in
the case of noises different from white. In order to achieve that, a
procedure referred to as whitening should be applied. It uses the
covariance matrix of the real noise Ca σ2I that must be calculated in
advance. Then the Cholesky factorization of the matrix C  1 is
performed
C  1 ¼ DT D; ð4:4Þ
to whiten the noise in (4.2):
Du ¼ DHθ þ Dε-unew ¼ Hnew Θ þ εnew : ð4:5Þ
With relatively little effort it may be shown that the covariance
matrix of the new, transformed noises εnew becomes an identity
matrix (white noise), making feasible to use again the formula (4.3)
but with a new, modified matrix Hnew and vector unew:
^ ¼ ðH T H Þ  1 HT u
Θ T T 1 T T
new new new new ¼ ðH D DHÞ H D
Du ¼ ðHT C  1 HÞ  1 HT C  1 u: ð4:6Þ
The final estimator guarantees minimum variance estimation in
given noisy conditions described by the covariance matrix C. The
covariance matrix depends strongly on intensity of the electro-
kinetic phenomena. It may assume the form of the identity-like
matrix for laminar flows and the form of a matrix showing
significant correlation when the flow is turbulent as depicted
in Fig. 4.4.
Practical application of that approach requires establishing the
concept of a two-stage measurement. At the first stage the covariance
matrix of actual noises registered by the electrodes before acquisition
of the flow signal should be estimated. Exemplary matrices were
 Z. Watral et al. / Flow Measurement and Instrumentation 42 (2015) 16–25 21

Fig. 4.4. Graphical presentation of covariance matrices obtained for disturbances occurring on the electrodes of a electromagnetic sensor for two different flow velocities.

depicted in Fig. 4.4 for two different flow velocities available in the
authors’ laboratory set. This stage involves knowledge collection and
requires nothing but little energy for data acquisition devices and
signal processing. The second stage is energy intensive as it is intended
to excite the output signal. As the acquisition of the output signal at
this stage is ready, the computations taking into account actual noises
described by the covariance matrix could be performed. The quanti-
tative effects of such an approach can be evaluated with the use of
Figs. 4.5 and 4.6, where the results of authors’ research are shown. The
research was aimed at both: finding the way to decrease measure-
ment uncertainty as the power supplied to the coil decreases and
finding best signal processing method in given noisy conditions. The
algorithm used in pulsed DC approach was taken as the reference
algorithm while the waveform of the current driving the coil was Fig. 4.5. Comparison of estimation results when two different processing approaches
strictly trapezoidal at the same time. That shape provides intervals of were applied in the conditions of strong magnetic field and turbulent flow.
constant magnitude and its transitions occur at slower slew rate as
compared to the square waveform. To generate trapezoidal waveform,
the voltage controlled current source was used [24]. The test channel
made of PVC had a 156 cm2 cross-section and was equipped with a
coil of 35 mH inductance.
A water pump enabled to control the flow velocity up to 14 cm/s.
Figs. 4.5 and 4.6 represent the values of flow velocities estimated
every 0.6 s (three periods of 5 Hz excitation) for velocity 13.7 cm/s
and for the power applied to the coil 45 V A and 1.8 V A, respec-
tively. When the power is strong enough (Fig. 4.5), there is no
difference in measurement uncertainty between the classic algo-
rithm and the algorithm that uses the covariance matrix of noises.
However, when only low power is available (Fig. 4.6) in low energy
applications, using the latter provides much better results.

Fig. 4.6. Comparison of estimation results when two different processing approaches
5. Possible direction of development
were applied in the conditions of weak magnetic field and turbulent flow.

5.1. The concept of pulse excitation

The rapid development of wireless sensor networks has awoken a
noticeable interest in electromagnetic flowmeters. When balancing
flow in areas of uncertain relationship between level and flow, it
appears that only electromagnetic flow meters are capable of perform-
ing the correct measurement. Those meters have unique metrological
properties but are rarely used in field systems. The fundamental
obstacle to do that in distributed systems is that the meters require
high power to operate. Problems related to minimizing power con-
sumption of the flowmeter are associated with a complete change of
the measurement algorithm and the necessity to optimize alternative
and mobile renewable power sources, especially if they are planned to
be used in wireless networks for environmental monitoring.
In the authors’ works partly presented above and concerning
research on electromagnetic flow meters [22–25], the exciting coil
was powered by a current source with a quasi-continuous stimu-
lation. It would appear that for a typical size flow channel, where
the peak value of magnetic field induction in the measuring zone
is about 4 mT, the exciting coil requires about 30 V A of power.
Under those conditions a relative standard uncertainty is expected
at a maximum of about 2% with the measuring phase lasting
0.6 ms. However, if the actual power absorbed by the source from
the mains was taken into account (measurement results showed
that it was approximately 90 V A at the phase of magnetic field
generation) as was the need for an initial excitation reserved for
transient decay during the time of 2 s corresponding to at least 10
periods of frequency 5 Hz, the energy that was required to perform
a single measurement reached the level of 230 J ([0.6 s þ2 s] 
90 V A E 230 J). To reduce energy consumption, the power supply
chain should involve as few intermediary devices as possible.
Furthermore, it should operate for a limited time to provide short
exposure of the output voltage signal to electrolytic noises. A
power system, designed by the authors to investigate the pro-
spects of the meter’s evolution in such aspects, is shown in Fig. 5.1.
Electric energy from a DC source UB (in this case it was a lead
acid battery) is transferred through the keying transistor T to drive
the coil in short intervals of time in which the measurement is
performed. Given the voltage was treated as a unit step function
22 Z. Watral et al. / Flow Measurement and Instrumentation 42 (2015) 16–25

exciting the coil of inductance L and overall resistance R including function. As the signal is linearly processed, however, by the
the resistance of the coil itself as well as the resistance of the amplifier using parameters of the transfer function known in
battery and the electronic switch, an increasingly exponential advance, the resulting waveform and its components may be
waveform of the current should be expected easily derived with the use of the Laplace transform and subjected
 
to the same processing methodology. The results of experiments
UB L
iðt Þ ¼ 1  exp  t : ð5:1Þ conducted with equipment shown in Fig. 5.1 prove that the
R R
acquired pulse signals conform to the postulated model (4.2) i.e.
The current is responsible for the following induced signal they can be synthesized by a linear combination of waveforms
voltage formed by differentiation according to the law described determined by circuit parameters just like time constant L/R and
by (4.1) parameters of the amplifier. Fig. 5.2B presents real output voltage
    signals acquired under conditions of the same flow velocity for
UB UB L L
uin ðt Þ ¼  c1 þ  c2  c1  exp  t ¼ A þ B three different durations of current pulses transferring energy to
R R R R
  the coil. The pulses were depicted in Fig. 5.3A and correspond to
L
exp  t : ð5:2Þ the “on” phase of the switch that lasted 50 ms, 20 ms and 10 ms.
R
The corresponding energy consumption per one pulse was 1.4 J,
The voltage is of a transient nature but it is also linearly 0.43 J and 0.15 J, respectively. By taking into account the rising
dependent on the coefficients A and B containing information on slope of current pulses, linear combination of known waveforms
sought parameters c1 and c2. This kind of dependence makes it may be performed to obtain approximations denoted in Fig. 5.2B
clear that the voltage may be processed by the methodology by gray thick lines.
described again by (4.3) given that the time constant L/R is known. To investigate the possibility to estimate flow velocity with use
To find the constant, some off-line and preliminary experiments of the coefficient of above combination, an experiment aimed at
should be carried out. Close attention is also advised as the signal the input–output characteristics of the flowmeter with this kind of
is further modified by the conditioning amplifier intended to processing was carried out. The repetition rate necessary to obtain
minimize differential noises through its high pass transfer descriptive statistics was one measurement per second. One
hundred measurements were taken to determine the mean and
standard deviation just like in the case of the quasi-continuous
stimulation to perform a comparison study. The results were
presented in Fig. 5.3 where two different levels of energy were
used to find the quantitative results.
When pulse duration is shortened to reduce energy consump-
tion, the measurement uncertainty obviously deteriorates. How-
ever, it still remains comparable to the uncertainty achieved in the
case of the quasi-continuous stimulation (2%) but with consider-
ably energy savings (230 J reduced to hundreds of mJ). Moreover,
the uncertainty may be even greater as the measurement phase in
the quasi-continuous stimulation lasts several seconds for transi-
ent decay and one can perform additional pulse excitations in this
time using the proposed approach. Falling slope of the current
pulse may also be applied provided that an additional condenser is
Fig. 5.1. Laboratory set used to investigate the directions of development in the used. The condenser recovers the energy stored in magnetic field
field of the electromagnetic flow meters for open channels. and changes the pure exponent shape of the slope. When the

Fig. 5.2. Exemplary three different pulses of excitation (A) corresponding to output flow signals (B) together with approximations of the latter based on known parameters
of the coil and the conditioning amplifier.
 Z. Watral et al. / Flow Measurement and Instrumentation 42 (2015) 16–25
Fig. 5.3. Input–output characteristics of the laboratory flowmeter for two energy levels used to excite the magnetic field in the proposed mode of operation ((A) energy
0.42 J, (B) energy 0.15 J).
current waveform is strictly exponent, the components forming
the output voltage according to (4.1) are of the same shape and
they cannot be distinguished by means of processing.
The analysis and preliminary experiments formulated the founda-
tions to propose the concept of controlled injection of energy into a
distributed measuring system operating in off-the-grid conditions
supplied by power stations equipped with sources of renewable
energy. They demonstrate that significant optimization of energy
consumption may be achieved compared to existing solutions.
5.2. Discussion on the usage of renewable sources
Renewable energy sources should be recognized for very small
unit density and therefore power stations of this type are generally
large, sometimes much larger than the flowmeter itself. Therefore,
on one hand there is a great need to reduce electricity demands of
the measuring station in the way presented above and on the
other—the need for optimizing power supply and the method of
supplying energy to the transducer. Given typical installation
locations of the electromagnetic flowmeters working in areas with
difficult access to AC power, the renewable energy sources, just
like micro-hydro-power stations, photovoltaic, small wind tur-
bines and fuel cells should be considered first. They are able to
provide sufficient levels of electricity when combined with a
buffer battery and a suitable driver.
The flowmeter may be installed on watercourses to use their
energy in the process of electricity production. Hydraulic power of
flowing water is equal to its kinetic energy in the time
ρ  A  ν3
Ph ¼ ; ð5:3Þ
2 cell itself. In case of a chemical battery, its size and weight should
where A is the cross-section of flow channel filled with water [m2];
v is the velocity of water flow [m/s]; ρ is the water density [kg/m3].
After substituting into Eq. (5.3) the typical value of water
density (1000 kg/m3) an expression for the theoretical hydraulic
power of the flowing water is obtained
P h ¼ 500  A  ν3 : ð5:4Þ
In practice, the actual power consumption is much less because the
velocity of water before and after passing the turbine decreases. Due
to low speed of flowing water – in most rivers (1–2 m/s) – the power
obtained is not considerable. Commercially produced submersible
23
micro power plants operating at a speed of the order of 1 m/s to
2 m/s provide a power of 10 W to 50 W.
In case of photovoltaic power plants location is absolutely critical.
For instance, in Poland sun exposure during the year amounts to on
average about 1000 kW h/m2/y over about 1600 h available for
practical use (an average of 5.5 h/d). However, the main problem with
solar energy is not its low density, but its high volatility on a daily,
monthly and annual basis. With an average sun exposure of about
1000 kW h/m2/y, its rate of change varies from about 18 kW h/m2 in
the month of December to around 155 kW h/m2 in the month of July.
Taking into consideration such high volatility as well as the efficiency
of solar photovoltaic panels and its dependence on operating condi-
tions, the panels with capacity of approximately 40 W peak power
should be used to obtain a constant power level of 1 W throughout
the year, regardless of the sun exposure. In turn, as far as the wind
energy is concerned, the problem is not just low density above the
surface (W/m2), but also its high variability in space and time. For
instance, again in Poland, the average wind velocity is about several
m/s but it depends on the region and the altitude where the meas-
urement is taken. Fortunately, wind velocity is usually larger on
cloudy days when sun exposure is reduced.
Despite possibilities of using the energy of water, sun and wind
as indicated above, it seems to be inevitable to consider fuel cells
as a reserve source of battery charging. The advantage of a fuel cell
as compared to a battery is that during charging the energy is not
supplied by an electric current reversing the chemical reaction as
it happens in the battery, but is charged by filling up the cell with
methanol or hydrogen fuel. In such construction the increase of
energy capacity is achieved by supplying the cell with the required
amount of fuel without the necessity to increase dimensions of the
be adequately increased.
According to above considerations, one could propose an autono-
mous power supply system based on the hybrid structure as depicted
in Fig. 5.4.
The proposed structure includes all available sources of renew-
able energy. Those ensuring that necessary measurements are
carried out inexpensively and with required reliability under all
conditions should be selected. If, for example, the speed of the
watercourse in which the meter is installed exceeds 1 m/s, an
underwater micro turbine could be considered, such as UW100 by
Ampair Energy Ltd. The power generated by this turbine highly
24 Z. Watral et al. / Flow Measurement and Instrumentation 42 (2015) 16–25
depends on water velocity, but it is also a source providing the
station with energy all day long, even at night. Second, the
possibility to preserve all or part of the required power by the
solar panels should be considered. The power of panels depends
on a day and varies across seasons but small wind turbines are
even more unpredictable as they require wind speed of a few m/s
to start. Consequently, they should be installed only in pla-
ces where wind conditions are good enough throughout the all
year round.
5.3. Exemplary balance of power
Given the above, installation of an electromagnetic flowmeter
should be preceded with necessary calculations identifying the
balance of required energy. With respect to the proposed method of
pulse excitation let us again assume that the peak magnetic induction
of about 4 mT is necessary. Establishing 5 measurements per second
with pulse duration of 50 ms consuming 1.4 J requires 7 W h (420 J) of
energy per hour. Assuming the system is planned to operate 24 h a
day, its projected energy demand is 168 W h. Provided a 12 V batteries
were used, that equates to required capacity of 14 A h. Assuming that
the flowmeter should be supplied the entire year by a battery charged
with solely renewable sources, one has to select the most reliable one
providing the cheapest energy at the same time. Due to the cost to
power ratio, the most suitable solution is based on photovoltaic panels
as opposed to the fuel cell that proves the most expensive. On the
other hand, due to the reliability required to maintain charge of the
battery, the best result is obtained with fuel cells while the worst with
micro wind and water turbines. However, the final decision to choose
the proper power supply should be always taken with respect to the
conditions existing in the flow meter location. For watercourses with
significant flow velocities, the best choice is a submersible micro
power plant instead of other sources. To conclude considerations
concerning required energy for the flowmeter, the solution based on
photovoltaic panels supported with a fuel cell was selected. Using the
web off-grid calculator by Udomi, one of fuel cells manufacturers, the
most appropriate set can be easily selected. The fuel cell EFOY-PRO-
800 should be used (25 W, 12 V and 2.1 A charging current) for a 7 W
receiver when supplied from the battery charged by the photovoltaic
100 W panel. For such a set, a battery with a capacity of approximately
three times greater than the daily requirement 14 A h should be
selected, i.e. a battery with the capacity of 42 A h. This choice provides
complete coverage of the demand for electricity generated by PV
panels. The fuel cell will support battery charging during the winter
months when there are worse conditions of insolation, consuming
6.6 l of methanol and providing 6.5 kW h of energy per year. A lead-
acid battery or a lithium-ion battery can fulfil the function of energy
storage. The fuel cell acts as a failsafe energy source charging the
battery in extreme circumstances. Operating temperature range for
this type of cells is between  20 and þ451 C at theoretical capacity of
Fig. 5.4. The structure of power station for pulse electromagnetic flowmeter.
5 kW h is obtained from 1dm3 of methanol. The above studies
indicate that it is possible to take low-energy flow measurements
using a pulsed magnetic field in off-the-grid conditions.
Summary
This paper discussed both the history of developing electromag-
netic flowmeter structures and prospects for development of this
method dedicated to the measurement of flow in open channels.
Methods of measuring and processing measurement signals were
emphasized the most, as they are convoluted and challenging issues.
The encountered relationship – signal to noise ratio – is one of the
most demanding issues of them all. It necessitates using highly
complex measurement algorithms. Among the most important devel-
opment trends the problem of a low energy flowmeter was shown.
The concept of such a flowmeter and first data confirming effective-
ness of such solutions was described.
References
[1] Michalski A, Starzyński J, Wincenciak St. Electromagnetic flowmeters for open
channels—two dimensional approach to design procedures. IEEE Sens J 2001;1
(1):52–62 (June).
[2] Shercliff IA. The theory of electromagnetic flow measurement. Cambridge
University Press; 1962.
[3] Bevir MK. The theory of electromagnetic flowmeter. J Fluid Mech 1970(43):.
[4] Bevir MK. Long induced voltage electromagnetic flowmeters and the effect of
velocity profile. Q J Mech Appl Math 1971;24(3).
[5] Bevir MK, O’Sullivan VT, Wyatt DG. Computation of electromagnetic flow-
meter characteristics from magnetic field data. J Phys D: Appl Phys
1981;14:373–88.
[6] Smith CC, Slepian I Electromagnetic ships log, US Patent 1,249,530.
[7] Young FB, Gerrard H, Ievon W. On electrical disturbances due to tides and
waves. Philos Mag Ser. 1920(5):.
[8] Wiliams EI. The induction of electromotive forces in a moving liquid by a
magnetic field and its application to an investigation of the flow of liquids.
Proc Phys Soc London 1930.
[9] Kolin A. An electromagnetic flowmeter principles of the method and its as
location to blood flow measurements. Proc Soc Exp Biol Med 1936;35.
[10] Thurlemann B. Zur elektromagnetischen Geschwindigkeitsmessung von flus-
sigkeiten. Helv Phys Acta 1955;28.
[11] Engl L Uber ein induktives verfahren zur durchflussmessung in offenen
gerinnen, Acta Imeco; 1967.
[12] Gils M Discharge measurement in open water by means of magnetic induc-
tion. In: Symposium on hydrometry, Koblenz; 1970.
[13] Rolf J, Starke H Magnetische Abflussmessung in offenen Rechteck-gerinnen,
ATM Blat V, mai; 1973.
[14] Rolf J Ein neuartger induktiver Abflussmesser, messen þprufen/automatik;
Juli/August 1977.
[15] Herschy RW. Hydrometry: principles and practices. Department of the
Environment Water Data Unit, Reading; 1978.
[16] Herschy RW. Hydrometry. New York, NY: John Wiley & Sons; 1978.
[17] Herschy RW, Newman JD Electromagnetic river gauging. In: Symposium on
river gauging by ultrasonic and electromagnetic methods, Water Research
Centre & Department of Environmental Water Data Unit Reading, UK; 16
December, 1974.
[18] Michalski A. Flow measurements in irrigation channel. Instrum Meas Mag
2000;3(1):12–6 (March).
[19] Michalski A, Sikora J, Wincenciak St. Optimal shape design of an electro-
magnetic flow gauge. (January). IEEE Trans Magn 1988;24(1):565–8 (str.).
[20] Technical Bulletin. Electronic flow gauges for open channels, Sarasota; 1992.
[21] Bonfig KW New developments in magnetic flow measurement in partly filled
open channels, Acta Imeco, Berlin; 1982.
[22] Watral Zbigniew, Andrzej Michalski. Practical validation of a flow meter
design environment. IEEE Instrum Meas Mag 2011;14(2):42–7 (Apr).
[23] Michalski A, Watral Z, Sienkiewicz J Some practical aspects of excitation coil
design for electromagnetic flow meter. In: 19th IMEKO world congress, Lisbon,
Portugal; Sep 06–12, 2009. p. 1254–1259.
[24] Jakubowski J, Michalski A. An energy-saving algorithm for electromagnetic
flow measurement in open channels. IEEE Instrum Meas Mag 2011;14
(6):40–5.
[25] Jakubowski J, Watral Z, Michalski A. The problems of pulse excitation in
electromagnetic flowmeters. IEEE Instrum Meas Mag 2013;16(5):47–52.
[26] Bonfig KW Induktive Durchflussmessung mit getastetem Gleichfeld, ATM,
Blatt V—1249; 11, August 1971. p. 153–156.
[27] Bonfig KW. Einflusse auf die Nutzspanung bei der induktiven Durchflussmes-
sung. ATM 1970 (no 8).
 Z. Watral et al. / Flow Measurement and Instrumentation 42 (2015) 16–25
[28] Bonfig KW Elektrochemische Storspanungen an den Elektroden bei indukti-
ven Durchfussmessung, Teil I: Uberblick und Einflus auf die Polarisation, ATM,
Blatt V—1249; 9 Juni 1971. p. 113–114.
[29] Bonfig KW Elektrochemische Storspanungen an den Elektroden bei indukti-
ven Durchfussmessung, Teil II: Einflusse auf die Unsymmetriespannung, ATM,
Blatt V—1249; 10 Juli 1971. p. 135–138.
[30] Ramette RW. Outmoded terminology: the normal hydrogen electrode. J Chem
Educ 1987;64(10):885.
[31] Xu KJ, Wang XF. Signal modelling of electromagnetic flowmeter under sine
wave excitation using two-stage fitting method. Sens Actuators, A 2006;136
(2007):137–43.
[32] Middelhoek S, Yamasaki H. Handbook of sensors and actuators—intelligent
sensors. Oxford: Elsevier Science; 1996.
25
[33] Kuromori K, Gotoh S, Matunaga Y. An electromagnetic flowmeter with dual
frequency excitation. Trans Soc Instrum Control Eng 1994;30(9):1020–6.
[34] Baker Roger C. Flow measurement handbook, industrial designs, operating
principles, performance and application. Cambridge University Press; 2000. p.
308–310.
[35] O’Sullivan VT, Wyatt DG. Computation of electromagnetic flowmeter char-
acteristics from magnetic field data III, rectilinear weight functions. J Phys D:
Appl Phys 1983;16:1461–76.