Ultrasonics 41 (2003) 319–322

www.elsevier.com/locate/ultras

Liquid level sensor using ultrasonic Lamb waves q

V.E. Sakharov a, S.A. Kuznetsov a, B.D. Zaitsev b, I.E. Kuznetsova b, S.G. Joshi c,*

a
Company ‘‘Gazavtomatika’’, Moscow 119435, Russia
b
Institute of Radio Engineering and Electronics, Saratov 410019, Russia
c
Department of Electrical and Computer Engineering, Marquette University, P.O. Box 1881, Milwaukee, WI 53201, USA

Abstract

This paper describes a novel, noninvasive method for measurement of liquid level in closed metal tanks that are under high
pressure. It is based on the use of ultrasonic Lamb waves propagating along the tank wall. Contact with liquid substantially changes
the characteristics of these waves and this can be used as an indicator of liquid presence. Theoretical analysis shows that the
symmetric and antisymmetric Lamb wave modes, both fundamental and higher order, are sensitive to presence of the liquid. The
optimal wave frequency depends on the thickness of the tank wall and wall material. A prototype level sensor based on this principle
has been developed. It uses two pairs of wedge transducers to generate and detect Lamb waves propagating along the circumference
of the gas tank. An operating frequency of 100 kHz is found to be optimal for use with tanks having a wall thickness of 30–50 mm.
Prototype sensors developed under this program have been used successfully in oil fields in the far northern region of Russia.
Ó 2002 Elsevier Science B.V. All rights reserved.

Keywords: Noninvasive sensing methods; Liquid level sensor; Ultrasonic lamb waves; Lamb wave transducers

The measurement of liquid level in closed metal tanks
that are under high pressure is an important problem in
the gas industry. The introduction of sensors inside the
tank disrupts its integrity, significantly complicates its
construction, and is highly undesirable due to possible
gas leakage and attendant fire hazard. Noninvasive
methods based on acoustic waves, which may be gen-
erated and detected outside the tank, are therefore best
suited for this application.
There are many methods of liquid level sensing based
on acoustic waves. Some of them use acoustic pulses
reflected from the liquid–gas interface to obtain infor-
mation about position of liquid level [1,2]. In this
method acoustic wave is generated by a transducer
placed on the bottom of the tank, passes through tank
wall and liquid, reflects from the liquid–gas interface,
and comes back to the launching transducer. The delay
time of the reflected signal gives the necessary infor-
mation. Other methods are based on transmitting an
q
This article is based on a presentation given at the Ultrasonics
International 2001.
*
Corresponding author. Tel.: +1-414-288-5579; fax: +1-414-288-
1584/5579.
E-mail addresses: zaitsev@ire.san.ru (B.D. Zaitsev), shri.joshi@
marquette.edu (S.G. Joshi).
acoustic wave through the tank wall and liquid in the
horizontal direction [1,2]. In this case the output signal
appears if the liquid level is higher than the spacing
between input and output transducers. These well-
known acoustic methods however are unsuitable due to
one or more of the following factors: presence of gas
bubbles in the liquid, absence of ideal reflecting surfaces,
and/or presence of acoustic wave scattering structures in
the tanks.
An acoustic method, which overcomes these limita-
tions, and on the first glance is suitable for use in gas
industry, is based on the reverberation effect of acoustic
pulses in a tank wall [3]. But attempts made to develop
this method show that its suitability is restricted for one
or more of the following reasons.
(1) The minimum usable value of acoustic wave fre-
quency is determined by thickness of the tank wall. The
width of the RF pulse used to excite the acoustic
transducer should be less than the time duration of one
cycle of reverberation, which is defined by thickness of
the tank wall. For typical wall thickness of 30–50 mm
the above dictates that an acoustic wave frequency
greater than 1–2 MHz should be used.
(2) This method requires the use of large acoustic
powers. In order to obtain high sensitivity, that is, long
time of reverberation, ultrasonic generator/receiver should

0041-624X/02/$ - see front matter Ó 2002 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0041-624X(02)00459-6
320 V.E. Sakharov et al. / Ultrasonics 41 (2003) 319–322
be acoustically mismatched with the tank wall material.
Moreover, under real conditions the number of rever-
beration cycles are defined not only by reflection coef-
ficients, but also by attenuation of acoustic waves, their
scattering by rough surfaces, excitation of spurious
Lamb waves, etc. The losses due to most attenuation
sources increase with increase of wave frequency. Thus,
in order to detect and measure all reverberated signals, it
is necessary to use high frequency acoustic generator of
high power. Development of such generators with reli-
able operation is a complicated problem.
(3) During use the inside surface of tank may be
covered by viscous layer (paraffin or grease) that leads to
decrease of reflection coefficient from the inner tank
wall. This fact significantly weakens the reverberation
effect and even after increase of generator power the
efficiency of this method cannot be fully recovered.
So, for practical realization of the reverberation
method, one needs a high frequency ultrasonic generator
operating at a high power level, and a high sensitivity,
low noise receiver. Besides, the presence of viscous lay-
ers decreases the reliability of the reverberation method.
This paper describes a new method of level sensing
based on ultrasonic Lamb waves, which overcomes
limitations of the earlier methods. Theoretical analysis
of Lamb wave propagation in steel plates allows one to
find an optimal value of operating frequency and type of
Lamb wave mode most sensitive to presence of liquid.
The method of analysis has been described previously
[4–6]. Fig. 1 shows plots of group velocity versus the
Fig. 1. Group velocities of various symmetric (S) and antisymmetric
(A) modes as a function of the product hf (h ¼ plate thickness,
f ¼ acoustic wave frequency) for Lamb waves propagating in a steel
plate.
Fig. 2. Attenuation in dB=k (k ¼ acoustic wavelength) as a function of
product hf for Lamb waves propagating in a steel plate in the presence
of a liquid.
product hf (h ¼ plate thickness, f ¼ acoustic wave fre-
quency) for a number of symmetric (S) and antisym-
metric (A) Lamb waves propagating in a steel plate.
Analysis shows that these velocities are negligibly af-
fected by presence of the liquid. On the other hand, the
liquid causes significant attenuation of the waves. This is
due to the radiation of bulk acoustic waves into the
liquid. Fig. 2 shows dependence of attenuation in dB=k
(k ¼ acoustic wavelength) as a function of hf for vari-
ous Lamb wave modes. One can see that there are
ranges of hf values where symmetric and/or antisym-
metric Lamb waves of both fundamental and higher
order are sensitive to presence of liquid contacting with
tank wall. Analysis shows that the optimal wave fre-
quency depends on the tank wall thickness and wall
material. An operating frequency of 100–120 kHz is
found to be optimal for use with tanks having wall
thickness of 30–50 mm. This is lower by more than one
order of magnitude than the optimal frequency of 1–2
MHz for use in the reverberation method. Since acoustic
wave attenuation varies as f 2 , this results in decrease in
wave attenuation by more than a factor of 100.
First, the characteristics of acoustic Lamb waves
propagating in a steel plate contacting with a liquid were
investigated experimentally. The experiments were car-
ried out on steel plates of thickness about 50 mm. The
operating wave frequency was 100–120 kHz. For gen-
eration and detection of Lamb waves we used wedge
transducers consisting of bulk acoustic wave transducers
(1) made of lead zirconate titanate (PZT) ceramic and
acoustic wedges (2) made of polyurethane (see Fig. 3).
Electric field applied to bulk wave transducer (1) gen-
erated bulk longitudinal wave, which was transformed
into Lamb waves by the polyurethane wedge. The op-
 V.E. Sakharov et al. / Ultrasonics 41 (2003) 319–322 321

Fig. 3. Block diagram showing generation, propagation, and reception

of Lamb waves along circumference of tank wall: 1––bulk wave PZT
transducers, 2––polyurethane wedges, 3––tank wall.

eration of this transducer is similar to that of the well

known wedge transducer used for the generation and
detection of surface acoustic waves [5,6]. The wedge
angle h is chosen so as to satisfy the condition cos h ¼
vB =vL , where vB is the velocity of bulk longitudinal
waves in the wedge, and vL the velocity of Lamb waves Fig. 4. Schematic pattern of delayed signals observed on the output
in the plate. transducers. The distance between the input and output transducers
The liquid level sensor operated in pulse regime at a was 2 m. The amplitudes of pulses are not to scale.
frequency of 100 kHz. The thickness of tank wall was 50
The theoretical and experimental results obtained
mm, so that the operating point was at hf ¼ 5000 m/s.
above were used for development of a prototype level
Symmetric and antisymmetric Lamb waves of various
sensor. The sensor consisted of two pairs of wedge
orders were excited by the input transducer. The sche-
transducers (generating and receiving) placed on the
matic pattern of delayed signals observed on the output
tank wall. The disposition of Lamb wave transducers is
transducer is shown in Fig. 4. The signals corresponding
shown in Fig. 5. The propagation of Lamb waves was
to the A0 , S0 , and A2 modes occur very close together in
along the circumference of the tank. The first pair of
the time domain and cannot be easily distinguished from
Lamb wave transducers (1 and 2) defined the lower
each other. This is due to the fact that at the operating
admissible liquid level, while the second pair (3 and 4)
value of hf , the velocities of these modes lie within 300
m/s of each other. One can see that pulses corresponding
to S1 , A1 and S2 modes may be clearly distinguished from
other pulses as well as from each other. Amongst these
modes, at hf ¼ 5000 m/s, only the A1 mode is sensitive to Trans.4
Trans.3
presence of the liquid (see Fig. 2). This mode was Upper admissible
level
therefore chosen as the operating mode. The A0 mode,
although more sensitive to presence of the liquid, could
not be used since its signal cannot be easily distinguished
from those of the S0 and S2 modes. The measured values
of group velocity and attenuation of A1 wave in presence
of the liquid were 2000 m/s and 0.2 dB=k respectively.
The group velocity was calculated from measured value
of delay time of pulse and known distance between the Trans.1 Trans.2
Lower admissible
input and output transducers. The attenuation was de- level
fined from change in amplitude of the pulse due to
presence of liquid. Experimentally measured values of
velocity and attenuation are in good agreement with
calculated values of 2268 m/s and 0.2 dB=k, respectively.
The slight discrepancy between theoretical and experi- Fig. 5. Block diagram showing arrangement of wedge transducers in
mental results may be explained by lack of exact data on the prototype sensor. Transducer numbers 1 and 2 define the lower
properties of steel plates and the contacting fluid. liquid level, while 3 and 4 define the upper liquid level.
322 V.E. Sakharov et al. / Ultrasonics 41 (2003) 319–322
defined the upper admissible level. The distance between
generating and receiving transducers was about 2 m.
This sensor was connected with special electronic cir-
cuits including control circuits of liquid pumps and al-
lowed one to maintain the liquid level in admissible
range defined by distance between pairs of Lamb wave
transducers. Prototype sensors developed under this
program have been used successfully in oil fields in the
far northern region of Russia in a wide operating tem-
perature range of )40 to þ60 °C.
Acknowledgements
The authors thank A.Yu. Kozlov for his experimental
assistance.
References
[1] O.I. Babikov, Ultrasonics and its industrial applications, Consul-
tants Bureau, New York, 1960.
[2] L.C. Lynnworth, Ultrasonic measurements for process control,
Academic Press, New York, 1989.
[3] N.N. Bakurskii, V.V. Ahremenko, Liquid level sensor using
outside acoustical transducers, Gas Industry (5) (1997) 25–26 (in
Russian).
[4] S.G. Joshi, Y. Jin, Propagation of ultrasonic Lamb waves in
piezoelectric plates, J. Appl. Phys. 70 (8) (1991) 4113–4120.
[5] I.A. Viktorov, Rayleigh and Lamb waves: physical theory and
applications, Plenum Press, New York, 1967.
[6] J.L. Rose, Ultrasonic waves in solid media, Cambridge University
Press, New York, 1999.