Flow Measurement and Instrumentation 15 (2004) 179185

www.elsevier.com/locate/owmeasinst

Ultrasonic ow metering errors due to pulsating ow

J. Berrebi a,
, P.-E. Martinsson a, M. Willatzen b, J. Delsing a
a
EISLAB, Department of Computer Science and Electrical Engineering, Lulea University of Technology, SE-971 87 Lulea, Sweden
b
Mads Clausen Institute for Product Innovation, University of Southern Denmark, Grundtvigs Alle 150, DK-6400 Snderborg, Denmark

Abstract

Transit-time ultrasonic ow meters present some advantages over other ow meters for district heating industries. They are
both accurate and non-intrusive. It is well-known that ultrasonic ow meters are sensitive to installation eects. Installation
eects could be static or dynamic. Among the possible dynamic installation eects is pulsating ow. The inuence of pulsating
ow on the prediction and the zero-crossing operations is investigated. Expressions are found for the prediction error and the
zero-crossing error. The relative errors due to the prediction and the zero-crossing are plotted. The prediction error can reach
dramatic values while the zero-crossing operation is hardly inuenced by ow pulsations.
# 2004 Elsevier Ltd. All rights reserved.

Keywords: Ultrasonic ow meter; Pulsating ow; Zero-crossing; Installation eect; Prediction error

1. Introduction tions in the velocity prole of the ow. A pulsating

ow is a dynamic installation eect. In a pulsating ow
The perturbations due to devices like pumps, com- system, the velocity prole undergoes a dramatic
pressors and fast acting valves in a ow system will change during one cycle of pulsation compared to a
impose so called dynamic installations eects to ow stationary system [2]. This eect generates two errors
meters. This problem is known from the beginning (Fig. 2) that make the choice of an appropriate cali-
of the 20th century but is still a major source of error bration factor complicated [1,4]. The inappropriate
to ow measurements. Pulsating ows generate calibration factor modulates the value of the two pre-
errors in ow meters such as response-time errors, cedent errors in an uncontrollable way (Fig. 3). Sam-
resonance errors, velocity prole errors, sampling errors, pling and zero-crossing methods are analysed here in
etc. [13]. The errors generated depend on the type of
ow meter used, e.g. Pitot tubes, vane-type anem-
ometers, inductive ow meters, etc. In district heating
and gas applications, the use of ultrasonic ow meters
has been dramatically increasing for the last decade.
Around 30% in Sweden and almost 100% in Denmark
of the ow meters currently used in the district heating
industry are of ultrasonic type. Ultrasonic ow meters
have a high accuracy in stationary ow conditions [4].
Their maximum error does not exceed 2% or 3% at
turbulent ow rate (Re > 4000), and 5% at laminar
(Re < 2000) or transient (2000 < Re < 4000) ow rates
(Fig. 1). However, their main drawback is their great
sensitivity to installations eects that generate varia-

Fig. 1. The ow meter accuracy when there is no pulsating ow.

Corresponding author. Tel.: +46(0)920-491000; fax: +46(0) The Error from the meter is compared to the European standard EN
920-492082 1434-1 [5] showing the maximal permissible error (MPE). The stan-
E-mail address: jonathan.berrebi@sm.luth.se (J. Berrebi). dards requirements are respected.

0955-5986/$ - see front matter # 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.owmeasinst.2003.12.003
180 J. Berrebi et al. / Flow Measurement and Instrumentation 15 (2004) 179185
Fig. 2. The ow meter accuracy when pulsations are involved. The
Error from the meter is compared to the European standard EN
1434-1 [5] showing the maximal permissible error (MPE). The stan-
dards requirements are not respected for all ows.
Fig. 3. Amplication of the error by the k-curve. The k-curve indi-
^ for the estimated ow rate
cates the suitable calibration factor KQ
^
Q. The rst estimation of the ow rate is then modulated by the
^ Q
^ . A small error on the
k-factor. The nal estimation becomes: KQ
rst estimation can induce a large error on the determination of the
k-factor. By transitivity, it induces a large error on the nal esti-
mation of the ow rate.
order to study the eects of pulsations on ultrasonic
ow meter performance.
2. Theory
2.1. Description and principle of the ultrasonic ow
meter
An ultrasonic ow meter is described with the help
of Fig. 4. For convenience, the ow meter congur-
ation investigated is longitudinal. A similar analysis
could be done for diagonal ultrasonic ow meters. The
radius of the ow meter body is R and the radius of
the ultrasonic transducers is Rtd. The length between
Fig. 4. Axial transit-time ultrasonic ow meter. In the present com-
putations, R 0:5 and L 10 cm.
the two transducers is L. The estimation of the mean
ow velocity is divided into two steps. In the rst step,
the upstream transducer sends an ultrasonic pulse
received by the downstream transducer after time Tdw.
In a rst approximation, the pulse velocity is the sum
of the speed of sound in water c and the mean ow
velocity
v. In the second step, the roles of the transdu-
cers are inverted and the upstream transducer receives
a pulse travelling with the velocity c 
v after time Tup.
A system of two equations is then obtained. Its sol-
ution gives an estimation of the mean ow velocity [4]:
!
L 1 1
^
vt 
2 T ^ dw t  Ts T ^ up t
L dT
1
2T ^ up tT
^ dw t  Ts
^ up t  T
dT T ^ dw t  Ts 2
where T ^ dw t  Ts and T
^ up t are the estimations of the
downstream transit-time and the upstream transit-time,
respectively. The transit-times Tdw and Tup can be mea-
sured using dierent techniques, among which are the
cross-correlation technique and the zero-crossing tech-
nique. The cross-correlation technique is a digital tech-
nique that requires the sampling of the pulses and
heavy calculations. It consequently becomes too
expensive for industrial use. The zero-crossing tech-
nique is an analogue technique for determining the
transit-times. The zero-crossing technique has a low
cost compared to the cross-correlation technique and is
therefore widely used in the industry. In the following,
the estimator of the transit-times is dened by:
TNup
^ up
T ; ^ dw TNdw
T 3
N N
where TNup and TNdw are the times during which N pul-
ses propagate upstream and downstream, respectively.
Usually, transit-time technique is performed with
N 1. When N > 1, the technique is called Sing-
Around [6]. Tup and Tdw are sampled alternately and
periodically. Their sampling period is 2Ts, where Ts is
slightly longer than TNup (and consequently longer than
TNdw ). The estimation of
v made in (1) at time t then
 J. Berrebi et al. / Flow Measurement and Instrumentation 15 (2004) 179185 181

becomes at time t Ts : to give a description of the zero-crossing technique, the

! transformation between electric and pressure signals is
L 1 1 assumed to be linear with gain one. This assumption
^

vt Ts  4
2 ^ ^
T dw t Ts T up t will not lead to qualitatively dierent results in terms of
1
Each estimation of Tup and Tdw is used twice. They ow measurement properties. A pulse pu is sent by the
are alternately introduced in relations (1) and (4). emitting transducer and received by the receiving trans-
Thereby
v can be estimated at rate 1/Ts. The fact that ducer at time t0 0. This pulse is composed of ve per-
the estimations of Tup and Tdw are not simultaneous iods of a sinusoid of frequency fu 1=Tu damped by
has an inuence on the estimation of
v [4]. This inu- an exponential window. The damping in water is
ence is studied in the next paragraph. neglected. Diraction is not taken into account for
convenience. As the propagation model studied in the
2.2. The prediction error precedent paragraph is linear and non-dissipative, the
received signal can be written as a translation in time
The sampling of the mean ow velocity is aected by of the pulse sent.
the ow pulsations especially when the time delay Ts 8t 2 t0 ;t0 5Tu ;
between the measurement of the upstream transit-time  
Tup and the measurement of the downstream transit- t  t0
p1
u t p 0
u sin2pf u t  t0 exp  12
time Tdw is in the same order of magnitude as the per- Tu
iod Tp of the ow pulsation. The error Es is caused by 8t 62 t0 ;t0 5Tu ; p1
u t 0 13
the prediction error on the ow rate. The dierence in
transit-times at time t can be written as: Assuming that the rst pulse is sent independently of
the ow pulsations, the ow pulsations have a phase
dTt Tup t  Tdw t 5 2pfuh that is a free parameter since t0 0 is dened by
However, as it is technically impossible to estimate 1
pu :
both Tup and Tdw at time t, the estimation of dT is per-
formed as follows: 8t 2 R; pp t p0
p sin2pfu t  h 14
^ Tup t  Tdw t  Ts
dT 6 The value of the total pressure at time t0 pp t0
1
The prediction error on dT is then: pu t0 is memorized as the threshold for the zero-
1
crossing. This is in fact simply pp(t0) since pu t0 0.
EdT Tdw t  Tdw t  Ts 7 1
Simultaneoulsy as pu
is being received, a second pulse
Denoting by Td L=c the zero-ow transit-time
is sent from the same emitter. This process is repeated
between the two transducers and by
vp t the mean
N1 times.
ow velocity at time t when pulsations are involved,
the expression of EdT can be developed as follows by 8j 2 1; N  1;
using (1) and (4) in (7): 8
>
> s 0
> 0
>
L L >
> L
EdT     8 >
> j 6 0 ) sj 2
vp tj1

vp t
vp t  Ts >
> c
c 1 c 1 >
>
c c >
> X j
>
>
< tj jTd sk
Since
vp 5 c, the Taylor expansion of (8) gives:
  k0 15
>
>
L L > 8t 2 tj ;tj 5Tu ;
>
EdT Td  2
vp t  Td  2
vp t  Ts 9 >
>  
c c >
> t  tj
>
> p
j1
t p 0
sin2pf t  t exp 
>
> u u j
L  >
>
u
Tu
EdT 2
vp t  Ts 
vp t 10 >
>
c : 8t 62 t ;t 5T ; pj1 t 0
j j u u
The prediction error EdT made on the estimation of
dT leads then to the relative prediction error Es on the where sj denote the dierence between the current tran-
mean ow velocity estimation: sit-time (when the ow velocity is
vp tj1 ) and the

vp t  Ts 
vp t zero-ow transit-time Td (Fig. 5). From time tj 1 , a
Es 11 trigger is looking for the rst time when the received

vp t j
pressure pp t pu t reaches the threshold pp(t0). The
trigger is looking for the rst zero of the signal
2.3. Zero-crossing error j
pp t pu t  pp t0 .
The upstream and downstream transit-times are very
pp t pj1
u t  pp t0 0; t 2 tj ;tj Tu  16
often measured by zero-crossing techniques. In order
182 J. Berrebi et al. / Flow Measurement and Instrumentation 15 (2004) 179185

(Fig. 6):
pj
u zj ej pp zj ej  pp t0 0 24
Assuming that ej 5 zj (assumption H1) and applying
Taylors formula to (24) yields:
j
dpu dpp
pj
u zj zj ej pp zj zj ej  pp t0 0 25
dt dt
j
From (17), we have pu zj 0. Moreover, if
j
jdpp =dtzj j5 jdpu =dtzj j (assumption H2), ej
becomes:
pp t0  pp zj
ej j
26
dpu
zj
dt
Inserting (14) and (15) into Eq. (26) yields:
Fig. 5. Comparison between zero-crossing at zero ow and at non-
zero ows. On the upper part of the gure, a pulse is received every 0
pp sin2pfp t0  h  sin2pfp zj  h
tj jL=c since the mean ow velocity is zero. In the lower part of the ej 0
    
gure, ow pulsations are involved. They induce positive (resp. nega- pu Tu Tu
fu 2pcos 2pfu  sin 2pfu exp  1=2
tive) extensions sj of the upstream (resp. downstream) transit-time. 2 2
27
That is in fact:
2.3.1. Constant mean ow velocity 0
pp sin2pfp  h  sin2pfp zj  h
If the ow velocity is constant, then pp is also con- ej 0
28
pu fu  2pexp  1=2
stant, equal to the threshold pp(t0), and sj s. The trig-
ger simply determines the rst zero zj of each pressure The transit-time and the average transit-time then
j become:
pulse pu . Then zj 1 is the solution of:
8j 2 2;N; Tj zj ej  zj1  ej1
pj1
u t 0; t 2 tj ;tj Tu  17
Td sj1 ej  ej1 29
Solving for (17) using (15) leads to:
 1 N X1
1
^ Td
T si
e
e eN  e1 30
sin2pfu t  tj 0 N  1 i1 N 1
17 () 18
0 < t  tj < T u
Tu
zj1 tj 19
2
The transit-time Tj is then determined by the follow-
ing formula:
8j 2 2;N; Tj zj  zj1 tj1  tj2 20
8j 2 2;N; Tj Td s 21
A reduction of the noise present in the measurement
is performed by taking the average of all Tj. The esti-
mation of the transit-time is then:

^ 1 X N
T Tj 22
N  1 j2 Fig. 6. Error induced by a pulsating ow on the zero-crossing pro-
cess. The threshold level is given by the value of the pressure at time
^ Td s
T 23 t t0 . From that moment, the detector is looking for the rst time
when the pulses cross the threshold. The measured zero is then given
by the zj. As the ow pulsations are due to variation of the pressure
pp(t), they induce an error denoted by e. When the pressure is increas-
2.3.2. Sinusoidal mean ow velocity
ing (case of the gure), the threshold becomes lower than it should be
If
vp is slowly varying over the duration of N pulses, relatively to the pulse. Then, e is negative. If the pressure is decreas-
the zero-crossing found in (19) is slightly delayed by ej ing, e is positive.
 J. Berrebi et al. / Flow Measurement and Instrumentation 15 (2004) 179185 183

With help of (28),
e becomes: Taking the logarithmic dierentiate of (1) and focus-
ing only on the zero-crossing terms leads to:
0
pp sin2pfp z1  h  sin2pfp zN  h D
v
eH h 
eH h  Ts
eH h

e 0
31 Ez 
pu N  1fu  2pexp  1=2
v dT Tup h


z z 
eH h  Ts
1 N  38
0
pp 2cos 2pfp  h sin2pfp zN  z1 Tdw h  Ts

e 2
0
pu N  1fu  2pexp  1=2 By neglecting the two latter terms, the relative error
on the ow velocity becomes:
32

eH h 
eH h  Ts
As h is a free parameter, it can be changed to Ez 39
dT
z1 zN L
v h
2  h. Moreover, assuming that fp zN  z1 As dT cp2 ; Ez becomes:
fp N  1Td 5 1 leads to: h
c2
Ez
eH h 
eH h  Ts 40
0 L
vp h
cos2pfp h pp f p

e 0
Td
e0 cos2pfp h 33 Developing the latter expression leads to:
exp  1=2 pu f u
c2
e0H cos2pfp h  cos2pfp h  Ts
Notice that the latter is valid if and only if there is Ez 41
L
v0p sin2pfp h
solution for the zero-crossing. If the threshold does not
0
cross the pulse (pu too small), formula (33) has
0 3. Results and discussion
no meaning. The problem has a solution if pu >
 
dpp =dt NTd 2NTd pfp p0
p . The expressions of Es and Ez, respectively, found in
Max
(11) and in (41) are suitable for both liquids and gases.
Both Es and Ez cannot be overestimated since the sinu-
2.3.3. Case 3: ltering the ow pulsations
soids present in the quotient of (11) and of (41) can be
By applying a high-pass lter on the total received
in principle equal to zero. In practice, the probability
pressure (before zero-crossing), it is easier to have a
that h will be equal to (or in the neighbourhood of) 0
solution of form (33) to the problem posed in (24). By T
mod 2p is small. The sinusoid present in the quotient of
using a rst order lter:
(11) and of (41) will not be equal to zero.
ix=xc
Hx H0 ; 34
1 ix=xc 3.1. Prediction error
where H 0 1, the pressure from the ow pulsations is A typical value for the ow pulsations amplitude is
damped by a factor H(xp) xp 2pfp . Then replacing
v0p 1 m=s. By choosing dierent values for the num-
0 0 0
pp in (33) by H(xp)pp gives for pu > 2NTd pfp Hxp ber N of loops and for the pulsations frequency fp, Es
0
pp (assumption H3): can vary considerably. Figs. 7 and Fig. 8 are, respect-
ively, plotted the prediction error for N 2 and for
cos2pfp h Hxp pp fp
0 N 100 for fp 10 Hz. These gures show that Es can

eH 0
Td
e0H cos2pfp h 35 easily reach around 10% when there is no averaging of
exp  1=2 pu fu
the transit-times (N 2). The same error can reach
dramatic values when averaging the transit-times
2.3.4. Case 4: error on the ow measurement (N 100). This is due to the fact that when N increa-
Assuming now that a downstream transit-time is ses, the sampling interval also increases leaving a
measured at time h and that the following upstream longer time interval to the variations of the ow velo-
transit-time is measured at time h Ts , the estimation city. There is a compromise to nd between N and
fp. When fp reaches high values, it is wiser to take N
of the mean ow velocity according to (1) then reads:
! smaller.
L 1 1
^

v 
2 T ^ dw h  Ts T ^ up h 3.2. Zero-crossing in water applications
L dT
36 The ow pulsations in water appears mostly in a
2T^ up hT^ dw h  Ts
range of frequencies between 0.1 and 100 Hz. The
^ up h  T
dT T ^ dw h  Ts 37 pressure is of the order of 1 bar. The central frequency
fu of the pulse is between 0.5 and 4 MHz. The ultra-
184 J. Berrebi et al. / Flow Measurement and Instrumentation 15 (2004) 179185
Fig. 7. The variations of the prediction error Es over time when
N 2. The maximal error can easily reach 50% (2.2).
105 %.
sonic pressure varies in a range between 1 and 1000 Pa.
A typical example is given for ow pulsations of ampli-
tude 1 Pa and of central frequency 10 Hz. Taking
0
pu 1000 Pa, fu 4 MHz, fp 10 Hz and N 2,
leads to the graph plotted in Fig. 9. Calculating the
transit-times with N 100 gives the curve plotted in
Fig. 10. The frequency fu of the ultrasonic pulse is so
much higher than fp that the lter erase almost totally
the ow pulsations. That is why the zero-crossing
errors found have no signicance.
3.3. Zero-crossing in gas applications
Flow pulsations in gas can in practice reach higher
frequencies, up to 1 kHz. The pulse central frequency is
usually lower than in liquids. fu is then conned
between 100 and 1 MHz. One can imagine that by
Fig. 8. The variations of the prediction error Es over time when
N 100. The maximal error can easily exceed 100%. The error is
104 %. It shows that
increased by the high number of pulses sent. The sampling time is too
long, inducing a large prediction error (2.2).
Fig. 9. The variations of the zero-crossing error Ez over time when
N 2. The maximal error does not exceed 4
choosing a high ow pulsation frequency of 1 kHz and
a low pulse central frequency of 100 kHz, ow pulsa-
tions would still be present after ltering. Since the
speed of sound in air is only 330 m/s, the time of travel
Td is equal to 0, 3 ms, the ow can no more be con-
sidered as static during the time of propagation and it
is not valid to apply the model employed in the present
paper. In our study, ow pulsations cannot reach
higher frequencies than 100 Hz. Moreover, since the
damping is much stronger in gas than in liquids, the
0
amplitude of the ultrasonic pressure pu at the receiver
will be sensibly lower than at the emitter. All these con-
siderations make assumptions H2 and H3 more dicult
to satisfy for gases. However, these assumptions are
satised in practice by ltering more eciently ow
Fig. 10. The variations of the zero-crossing error Ez over time when
N 100. The maximal error does not exceed 2
the ow pulsations do not have any inuence on the zero-crossing
process.
 J. Berrebi et al. / Flow Measurement and Instrumentation 15 (2004) 179185
pulsations. It implies to use a high-pass lter of higher
order than H. This is usually the case in the industrial
applications. It is then possible to reach the same level
of error than in liquids.
4. Conclusion
The zero-crossing method used for measuring the
upstream and downstream transit-times is not sensitive
to ow pulsations. Indeed, an appropriate lter can
remove the ow pulsations before operating the zero-
crossing. However, the sampling of the ow velocity
cannot be perfect in the way that upstream and down-
stream transit-times cannot be measured simul-
taneously. A prediction error depending on the time
position, the pulsations frequency and the number of
loops used for averaging the transit-times can then
reach dramatic values. Moreover, the present work
does not deal with the possible inuence of the ow
pulsations on the ultrasonic propagation due to velo-
city prole variations [7]. So far, no propagation model
relevant for the eects caused by the pulsations has
185
been found. But one has to be aware that such errors
probably exist.
References
[1] A. Hayward, Flowmeters, The Macmillan Press Ltd, London,
1979, pp. 112116.
[2] E. Hakansson, J. Delsing, Eects of Pulsating Flow on an Ultra-
sonic Gas Flowmeter, Lund Institute of Technology, Lund,
Sweden, 1993.
[3] C. Carlander, Installation Eect and Self-Diagnostics for Ultra-
sonic Flow Measurement, Lulea University of Technology, Lulea,
Sweden, 2001.
[4] L. Lynnworth, A. Brown, Ultrasonic Flow meters. Flow
Measurement, second ed., D.W. Spitzer, ISA, Research Triangle
Park, 2001.
[5] J. Delsing, US patent 5 796 009, Method for meausing in a uid
with the aid of sing-around technique.
[6] S. Uchida, The Pulsating Viscous Flow Superposed on the
Steady Laminar Motion of an incompressible Fluid in a Circular
Pipe, Zamp, 1956, pp. 403421.
[7] European Committee for Standardisation (CEN) and Swedish
Standard Institution (SIS). Heat meters-part 1: General require-
ments. EN 1434-1 (1997).