Applied of ACS Meta-heuristic Algorithm and PSO

Heuristic Algorithm in Ultrasonic Signals

Farid CHIBANE a,b,* , Abdessalem BENAMMAR a and Redouane DRAI a

a
Reserch Center in Industrial Technologies CRTI
PO Box 64, Cheraga 16014, Algiers, Algeria
b
Laboratoire Electrification des Entreprises Industrielles
University of Boumerdes – Algeria
*
[email protected]
{f.chibane, a.benammar, r.drai}@crti.dz

Abstract—The accurate estimation of the ultrasonic process. Equation (3) is an altered version to describe the
backscattered echoes pattern is essential in ultrasonic non- multiple echoes from the reflector [1],
destructive evaluation. In this paper, a generalized parametric
M
y (t ) =  s ( m ; t ) +  (t )
ultrasonic echo mode was presented. It is influenced by a set of
parameters: echo bandwidth, arrival time, center frequency, (3)
m =1
amplitude and phase. The Adaptive Cuckoo Search (ACS) and
Particle Swarm Optimization (PSO) algorithms are used to Here, M denotes the number of superimposed Gaussian
estimate these parameters and there performances are
echoes; echo vector m indicates the shape and position of each
compared. In first, simulations are carried out to assess the
echo. The summation term represents a signal model of
performance of the two algorithms, then these algorithms were
applied on experimental ultrasonic signal. The ACS algorithm
multiple reflections while the number of reflections is
outperforms PSO. unknown. In order to estimate M, we can use the information
theoretic criteria such us the minimum description length
Keywords—Ultrasonic Signal, Echo Parameter Estimation, (MDL) [3].
Adaptive Cuckoo Search Algorithm, Particle Swarm Optimization Cuckoo search is a metaheuristic search algorithm which
I. INTRODUCTION has been proposed by Yang and Deb [4]. The algorithm is
inspired by the reproduction strategy of cuckoos. At the most
In ultrasonic Non Destructive Evaluation (NDE) basic level, cuckoos lay their eggs in the nests of other host
applications, the pattern of ultrasound echo signal reveals birds which may be of different species. The host bird may
important physical information such as the location, size and discover that the eggs are not its own and either destroy the
orientation of defects, as well as attenuation and dispersion eggs or abandon the nest all to gather. This has resulted in the
characteristics of the propagation path. It has been shown that evolution of cuckoo eggs which mimic the eggs of local host
model-based signal analysis is robust in echo evaluation and birds [5]. To apply this as an optimization tool, Yang and Deb
parameter estimation [1-2], especially when the adopted [4] used three idealized rules: (i) Each cuckoo lays one egg,
model has parameters strongly pertinent to the physical which represents a set of solution co-ordinates at a time and
information of reflectors in material and structures. Estimated dumps it in a random nest; (ii) A fraction of the nests
parameters of echoes can be used to quantitatively evaluate containing the best eggs, or solution, will carry over to the nest
structures and characterize material. Nevertheless, signal generation; (iii) The number of the nests is fixed and there is
processing and analysis is still challenging due to the a probability that a host can discover an alien egg. If this
nonstationary nature of ultrasound NDE signals. happens, the host can either discard the egg or the nest and this
The ultrasonic backscattered echo from a flat surface result in building a new nest in a new location. The CS
reflector can be written as [1]: algorithm has been applied successfully to variety of
optimization problems [6-7]. The ACS algorithm is made
y (t ) = s ( ; t ) +  (t ) (1) without using the Levy distribution and it does not require any
initial parameter to be defined [9]. The PSO algorithm has a
Here, s(;t) is a Gaussian echo model, and swarm composed of multi-particles. Each particle has its own
position and velocity (transfer vector). The search by multi-
particles distinguishes PSO from many optimization methods
− (t − )
(
cos 2f c (t −  ) + ϕ )
2
s( ; t) = e (2) proposed so far. Each particle shares global information and
interaction among particles makes their search efficient.
 = [  f c ϕ  ] . Although only simple operations compose the PSO algorithm,
the PSO method can solve nonlinear and multi-peaked
 is bandwidth factor that determines the bandwidth of the optimization problems efficiently [14]. In this study, we
echo or the pulse duration in the time domain;  is arrival time, propose to apply the ACS and PSO algorithms for the
indicates the location of the reflector; fc is center frequency, estimation of ultrasonic echoes parameters. The performances
governed by the transducer center frequency;  is phase and  of these algorithms is tested and compared with simulated and
is amplitude. (t) is the additive white Gaussian noise (WGN) experimental ultrasonic echoes. The remainder of this paper is

978-1-7281-0003-6/19/$31.00 ©2019 IEEE

organized as follows. The flowcharts of the two ACS and PSO random numbers with a uniform probability density function
algorithms are presented respectively in sections II and III. [0,1].
Section IV addresses the application of these algorithms on a
6. Evaluate the particles positions.
simulated ultrasonic signal. Section V presents experimental
results. Section VI concludes the paper. JG JG
7. Update pbest i and g best i :
II. FLOWCHART OF THE ADAPTIVE CUCKOO SEARCH JG JG JG
JG pbest i ( l ), if f ( x i ( l + 1)) ≥ pbest i ( l )
ALGORITHM
pbest i ( l + 1) =  JG
A. Initialization  x i (l + 1), else if (9)
(09)
Randomly initialize the N number of host nests JG JG
X i = ( x i , ..., x i , ..., x i ) for i = 1, 2,..., N for n dimensional
1 d n gbest (l + 1) = arg min pbest i f ( pbest (l +1) ) , 1  i  N (10)
problem and define the fitness function fit(X). Initially take 8. End
l = 1 and evaluate the fitness function of the hosts nests fit(Xi)
for i = 1,2,...,N for the first time. IV. SIMULATED RESULTS
B. Iterative algorithm
To study the ACS and PSO algorithms performance, we
1. Find the bestfit and worstfit of the current generation tested them on superimposed ultrasonic echoes with additive
among the host nests. WGN. The signal is made up of two echoes.
2. Calculate the step size using the equation: For simulation, two ultrasonic Gaussian echoes (2) are
bestfit ( l ) −fit i ( l ) simulated in terms of bandwidth factor  (MHz)2, arrival time
1 bestfit ( l ) −worstfit ( l )  (s), center frequency fc (MHz), phase  (rad), and amplitude
step i ( l + 1) =   (4) . The signal is shown in Fig.1. (a). These echoes are sampled
l at the sampling frequency fs = 200 MHz. A zero mean WGN
3. Then calculate the new position of Cuckoo nests using is added to this signal [1] with variance corresponding to a
the equation: Xi (l +1) = Xi (l ) + randn × stepi (l +1) (5) signal to noise ratio (SNR) of 12,5 dB. The actual parameters
used in the simulation are listed in Table I. Then, the ACS and
4. Evaluate the objective function of the host nests fit(Xi) for
PSO algorithms are applied to estimate the echoes parameters.
i = 1,2,...,N. For the ACS algorithm setup, the number of nest N is 100,
5. Then choose randomly a nest, j, among N pa = 0,25 and the maximum number of generation is 5000. For
If (fiti > fitj) the PSO algorithm setup, the population size N is 200,
Update jth nest by the new solution. C1 = C2 = 1,5 and w = 0.729. The choice of these parameter is
End made empirically [14].
6. The worst nests are abandoned with a probability
(pa ∈ [0, 1]) and new one is built. In order to convert the parameter estimation problem into
optimization problem, the following objective fitness function
7. l = l + 1.
is defined [12]:
8. Verify (l <= lmax), if yes then go to 1; otherwise end.
T
 ( y − yˆ )
III. FLOWCHART OF THE PARTICLE SWARM 1 2
F= , (11)
OPTIMIZATION ALGORITHM T
i =1
1. Random initialization of N particles (position and velocity)
where i = 1, 2, ... , T is the sampling time point and T denotes
2. Evaluation of the positions of the particles by the objective the length of data. For comparison, we use best estimate
function f. (‘Best’), mean (‘Mean’) and standard deviation (‘Std’) to get
JG JG the best results which are shown in Table I. The estimated
3. For each particle i, calculate pbest i = x i the best past echoes are plotted in Fig. 1 (b) and Fig. 1 (c) respectively with
position. ACS and PSO algorithms.
JG The parameters estimation tabulated in Table I, is robust,
4. Calculate gbest by : regardless of a significant noise level and echo time location.
JG JG
gbest (l + 1) = arg min pbest i f ( pbest (l +1) ) (6) 1

0.5
(a)

the best position reached -0.5

-1

5. Move the particles by :

0 0.5 1 1.5 2 2.5 3 3.5 4

v i (l + 1) = wv (l ) +C 1r1 (l ) × ( pbest i (l ) − x i (l )) +
(b)
0.5

i 0

-0.5

C 2 r2 (l ) × ( gbest i (l ) − x i (l )) (7)
(07) -1
0 0.5 1 1.5 2 2.5 3 3.5 4

1
(c)
0.5

x i ( l + 1) = x i ( l ) +v i ( l + 1)
-0.5
(8) -1
0 0.5 1 1.5 2 2.5 3 3.5 4
Time ( sec.)

with v velocity vector, x position vector, w inertia Fig. 1. a) Two echoes with a SNR of 12.5 dB; b) estimated echoes for (a)
coefficient, C1 and C2 coefficients acceleration, r1 and r2 two with ACS algorithm; c) estimated echoes for (a) with PSO algorithm
 TABLE I. PARAMETER ESTIMATION RESULTS FOR TWO ECHOES WITH SNR OF 12.5 DB USING ACS AND PSO ALGORITHMS

 α (MHz) 2 τ (μs) f (MHz) ϕ (rad)  α (MHz) 2 τ (μs) f (MHz) ϕ (rad)

First echo Second echo
1 15 1 8 0,52 0,9 12 2,5 7,5 1,04

ACS 1,0209 13,6684 1,0000 8,0205 0,4909 ACS 0,8847 11,6533 2,5061 7,4951 1,3072
Best Best
PSO 0.9887 15.2257 1.0014 7.9461 0.5706 PSO 0.9131 12.3701 2.5023 7.5086 1.1813

ACS 0,9197 15,3541 1,3374 7,2463 3,0888 ACS 1,0922 15,3897 4,0115 6,2106 3,0033
Mean Mean
PSO 0.9788 15.1913 0.9914 7.9666 0.5649 PSO 0.9137 12.3984 2.4873 7.5129 1.1763

ACS 0.4796 9.7771 1.3029 2.0869 2.6377 ACS 0.7432 14.8676 2.6078 2.6157 2.1789
Std. Std.
PSO 0.0986 0.3741 0.0999 0.2046 0.0569 PSO 0.0070 0.2828 0.1495 0.0438 0.0527

V. EXPERIMENTS where Vsample = 2830 m/s is the velocity of sound in the

material [13]. The data are obtained using a transducer
In this study, the ACS and PSO algorithms have been centered at 2,25 MHz. A thickness of CFRP is 3,3 mm [13].
tested on the experimental data. In the ultrasonic NDE Fig. 4 (b) and Fig. 4 (c) show the estimated echoes by
applications, the useful parameters of interest are ,  and fc respectively the ACS and PSO algorithms using M = 2.
[13]. Parameter estimated results are listed in Table II.
The sample used is a carbon fiber-reinforced polymer
multi-layered composite material (CFRP) achieved with one
delamination defect located at front surface [13]. The results show that both ACS and PSO algorithms are
Delaminations in thin composite laminates are usually robust and reliable for estimating the delamination depth.
detectable by an immersion transducer operating in pulse- ACS provides an accurate estimate of the position of the
echo mode. Fig. 2 shows a typical ultrasonic setup. Multiple delamination.
waves are reflected from the surfaces of the specimen as well
FS d
as from delaminations, as shown in Fig. 3.
1
Fig. 4 (a) shows closely-spaced echoes between 0.5
(a)

delamination and front surface echo. d represents the defect 0

-0.5
echo location of the delamination zone, FS represents the time -1

location of the front surface echo. The delamination depth in 0 1 2 3 4 5 6 7 8 9 10

the material can be given by using Eq. (11): 1

0.5
(b)
0

-0.5

d=
1
V sample
2
( τ d − τ FS ) 
 (11)
-1
0 1 2 3 4 5 6 7 8 9 10

0.5
(c)
0

Ultrasound -0.5

-1
Transmitter/Receiver
0 1 2 3 4 5 6 7 8 9 10
Time ( sec.)

Fig. 4. a) Experimental signal; b) Estimated signal for (a) with ACS

Transducer Water algorithm; c) Estimated signal for (a) with PSO algorithm

Material

TABLE II. DEPTH OF DEFECT

Fig. 2. Experimental setup for immersion pulse-echo testing FS d Position of the
Signal
(μs) (μs) delamination (mm)
Actual
− − 0,63
Reflected waves parameters
r1 r2 r3 ACS 3,32 3,77 0,64
Best
Delamination Composite PSO 3,31 3,79 0,67
laminate
ACS 3,41 4,19 1,23
Mean
PSO 3,24 3,79 0,78
ACS 0,70 1,65 1,70
Std.
Fig. 3. Reflected echoes PSO 0,37 0,38 0,02
 VI. CONCLUSION [6] I. H. Gandomi, X. S. Yang & A. H. Alavi, ‘‘ Cuckoo search algorithm:
a metaheuristic approach to solve structural optimization problems’’.
In this paper, we present two methods for the estimation Engineering with Computers, 29(1), pp. 17-35. 2013
of superimposed Gaussian ultrasonic echoes parameters using [7] X. S. Yang, & S. Deb, ‘‘ Engineering optimization by cuckoo search’’.
the two algorithms: Adaptive Cuckoo Search and Particle International Journal of Mathematical Modeling and Numerical
Swarm Optimization. The obtained results have a good Optimisation,1(4), pp. 330-343, 2010.
accuracy, and the performances are almost independent of the [8] S. Chakraverty, and A. Kumar, ‘‘Design optimization for reliable
noise. The ACS and PSO algorithms show a high robustness emgedded system using Cuckoo search,’’ Proc. International
Conference on Electronics, Computer Technology, vol. 1, pp. 164-268,
and reliability in the parameters estimation. The ACS has 2011.
better estimation precision than PSO. [9] M. Naik, MR Nath, A, Wunnava, S. Sahany and R. Panda ‘‘A
new Adaptive Cuckoo Search Algorithm, ’’ IEEE 2nd
REFERENCES International Conference on Recent Trends in Information Systems
[1] R. Demirli, and J. Saniie, “ Model-based estimation of ultrasonic (ReTIS), 2015, pp. 1-5.
echoes. Part I: analysis and algorithms,” IEEE Transactions on [10] P. Barthelemy, J. Bertolotti, and D. S. Wiersma, ‘‘A Lévy flight for
Ultrasonics, Ferroelectrics, and Frequency Control, vol. 48, pp. 787- light,’’ Nature, vol. 453, pp. 495-498, 2008.
802, May 2001. [11] H. Soneji, and R. C. Sanghvi, ‘‘Towards the improvements of Cuckoo
[2] R. Demirli, and J. Saniie, “ Model-based estimation of ultrasonic search algorithm,’’ Information and Communication Technologies
echoes. Part II: nondestructive evaluation applications,” IEEE (WICT), 2012 World Congress on, pp. 878-883.
Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, [12] J.Wang, B. Zhou and S. Zhou, “An Improved Cuckoo Search
vol. 48, pp. 803-811, May 2001. Optimization Algorithm for the Problem of Chaotic Systems Parameter
[3] M.Wax and I. Ziskind, “Detection of the number of coherent signals by Estimation,” Computational Intelligence and Neuroscience, Vol. 2016
the MDL principle,” IEEE Trans.A coust.Sp eech Signal Processing, (2016), pp. 1-8.
vol. 37, no. 8, pp. 1190–1196, Aug. 1989. [13] A. Benammar, R. Drai and A. Guessoum “Detection of delamination
[4] X. S. Yang, & S. Deb, “Cuckoo search via Lévy flights”. Proceedings defects in CFRP materials using ultrasonic signal processing,”
of World Congress on Nature & Biologically Inspired Computing Ultrasonics 48 (2008) 731–738.
(NaBIC 2009, India), IEEE Publications, USA 2009; p. 210-214. [14] Eberhart, R. C., and Kennedy, J. (1995). “A new optimizer using
[5] S. Walton, O. Hassan, K. Morgan, M.R. Brown, Modified cuckoo particle swarm theory”. Proceedings of the Sixth International
search: ‘‘ a new gradient free optimization algorithm’’ , Chaos, Symposium on Micromachine and Human Science, Nagoya, Japan. 39-
Solitons & Fractals, Vol 44, pp. 710-718, 2011. 43.