Application of Modelling for Ultrasonic Non-Destructive
Estimation of Cracks Caused by
Intergranular Stress Corrosion
L. Mazeika, R. Kazys, R.Sliteris, R. Raisutis
(Ultrasound Institute, Kaunas University of
Technology, LITHUANIA) (email@example.com)
Paper presented at the 8th ECNDT, Barcelona, June 2002
Ultrasonic non-destructive testing of components made of austenitic steel is complicated
by a structural noise and shape of cracks. For improvement of detection and
characterisation of such cracks the ultrasonic synthesised aperture focussing technique
(SAFT) may be applied, however efficiency of the SAFT depends on accuracy of the input
parameters. In order to optimise the processing algorithms and to estimate the potential
possibilities of various NDT methods, the numerical model has been developed. The model
enables simulate propagation of ultrasonic waves in materials possessing a granular
structure. For computer simulation of granular materials a novel method based on
application of Voronoi diagrams was developed. For simulation of cracks fractal theory
was applied. Propagation of ultrasonic waves in granular medium was simulated using the
ray tracing approach. The performance of the proposed model is illustrated both by the
simulated and experimental data.
Development of cracks in austenitic steel induced by intergranular stress corrosion is
essential factor reducing reliability of pipes used in nuclear power plants. Propagation of
ultrasonic waves in such materials is accompanied by strong backscattering, which causes
so called structural noise (1,4). Therefore, ultrasonic non-destructive testing of such objects
is complicated both by the structural noise and sophisticated shape of cracks developing in
this kind of materials (2,3). For improvement of detection and characterisation of such
cracks various testing methods and signal processing procedures are used, including split
spectrum technique (8), non-coherent filters (6), Wavelet transform (9), neural networks
(5), the SAFT (7). However, classical noise reducing techniques, like SAFT, are not
efficient in the case of a correlated noise (7).
Development of novel NDT methods and optimisation of signal processing algorithms may
be simplified using adequate numerical models of the medium under a test and models of
the testing method. However, the main attention until now was devoted to development of
novel signal processing procedures, but not to adequate models of testing procedures,
which are applied in the case of granular materials.
In this paper authors describe a numerical model, which allows simulate propagation of
ultrasonic waves in granular material and to investigate and estimate different ultrasonic
testing procedures and signal processing methods. The essential feature of this model is.2
that is based on physical considerations, like simulation of random granular structure of the
material and simulation of propagation of ultrasonic waves in such a medium. Such an
approach enables easy to modify parameters of the medium, including ultrasound velocity
in individual grains, reflectivity of grain boundaries, which are major factors influencing
formation of structural noise.
Mathematical model of granular material
The objective of this investigation is to develop the model, which enables to simulate
ultrasound propagation in a random granular medium with stress corrosion cracks of
different types. The model should enable to predict a structural noise in the medium also.
The structural noise is caused by reflection of ultrasonic waves by grains in the material,
however the exact mechanism of formation of these reflections is not clear yet. It is
possible to guess that mainly it is due to difference of the ultrasound velocities in different
granules, caused by anisotropy of the material and due to different elastic properties at the
boundaries between granules. Which phenomenon is dominating in formation of a
structural noise is not clear, however the proper model could help to give an answer to this
For solution of this task it is necessary to develop five independent models: model of a
granular medium, crack model, sound propagation model, model of ultrasonic transducer
and model of NDT process.
The first step is the generation of the virtual medium with randomly distributed anisotropic
granules, possessing different acoustic properties. The complexity of the task is caused by
the fact that each granule has its own form cross-section, which is close to some irregular
polygon with a random number of corners. The dimensions of these granules are also
random values. For generation of such a virtual medium the method based on the
calculation of Voronoi diagram was developed (13,14). The Voronoi diagram is the
partition of n-dimensional space with randomly distributed set of m points into m
polyhedral regions, which are called Voronoi cells. Each Voronoi cell satisfies the
condition that it contains all points that are closer to its data point than any other data point
in the set. This feature of the Voronoi diagram makes it especially attractive for simulation
of a granular medium.
In general, generation of virtual medium consists of the following steps:
- Generation of the randomly distributed set of points, which correspond to the
position of granules. The parameters of this set define the average size of the
- Calculation of the Voronoi diagram;
- Adjustment of the boundaries of the simulated region to the geometry of the
object under investigation.
The generated in such a way virtual medium is presented in Fig. 1. The presented example
illustrates that the developed approach enables to simulate granular material with a high
degree of similarity.
Acoustic properties of the material, such as an acoustic impedance and ultrasound velocity,
may be set separately for each granule. The reflection coefficients may be selected
separately for each boundary of granules also.
Fig. 1. Example of the medium with a granular structure generated using Voronoi
Application of fractals for simulation of cracks
All cracks in granular materials like austenitic steel may be classified into stress and
intergranular corrosion cracks (7). Both of them have branched character and difference is
that intergranular corrosion cracks are developing usually along the boundaries of
neighbouring granules (2,3).
For simulation of cracks we have applied the fractal theory, which enables to represent the
crack by the random set of linear segments. Stress cracks were generated according to the
where x1k, y1k, x2k, y2k are the coordinates of the end points of the crack branches; Lk is the
length of the k-th segment of the crack; KL and KB are the coefficients defining linearity
and branchiness of the crack; L new= Rand (KBL LK) is the length of the new branch of the
crack; KBL is the coefficient limiting the length of the new branches, anew= Rand (aBL aK)is the direction angle of the crack branch, ak is the crack segment direction angle; a BL
defines the limits of the branch angle. Examples of a few cracks generated using fractals
are presented in Fig. 2.
Fig. 2. Examples of generated cracks: 1-crack with short branches, 2-crack with long
Intergranular corrosion cracks were generated along arbitrary selected boundaries of
granules (Fig. 3).
Fig. 3. Simulated intergranular corrosion crack
Simulation of propagation of ultrasonic waves
The model of ultrasound propagation is based on the ray tracing approach. It was assumed
that reflections from boundaries of granules are dominating in generation of a structural
noise. Therefore, refraction phenomena inside granules were neglected assuming that the
ultrasound velocity is the same in all cells. In this case the ultrasonic beam splits at each
cell boundary into two beams: the first is propagating in the same direction as the incident
beam, the second is specularly reflected by the boundary. The beam propagation path is
generated as branched structure, each segment of which is defined by 5 parameters: the
coordinates xk, yk of the segment origin, the propagation direction ak, the signal delay time
tk and the signal amplitude Ak at the origin point, where k is the number of the segment. For
the selected segment k of the transmitted or reflected beam the parameters of new segments
are given correspondingly by:
where Lk is the propagation path length in the k cell, KT, KR are the transmission and
reflection coefficients at the k+1 boundary correspondingly.
The main problem in such an approach is that the number of segments is increasing 2N
times, where N is number of boundaries through which the ultrasonic beam propagates.
This leads to very essential increase of calculation time. Amount of calculations may be
reduced neglecting branches in signal propagation path, amplitudes of which are small
enough due to multiple reflections at the boundaries of granule and limiting the number of
reflections which are taken into account. Then the complete algorithm splits into a few
- Calculation of propagation of the direct ultrasonic beam. In this stage
propagation path from the transmitter till the back wall of the object and back
through a granular medium is calculated.
- The branching stage. At this stage each segment of the sound propagation path
is "branched" till the selected depth, that is, the complete propagation tree
including reflections and transmission at the boundaries is found.
- The "back to the receiver" stage. At this stage all segments, parameters of
which satisfy directivity angle and amplitude limits, are traced back to the
receiver. In this case only a transmitted signal is calculated at the boundaries
taking into account the diameter of the transducer.
The most essential parameters, which define the level of a structural noise in the model
presented, are reflection and transmission coefficients at the boundaries of cells. The
reflection coefficient was set as
where aR is the reflection angle at the boundary, a0 the reflection directivity pattern set as
parameter depending on the wave length, KR0 is the reflection coefficient selected the same
for all boundaries inside the object.
At the output of the model we obtain the set of amplitudes and the delay time of the
ultrasonic beams, which reach the receiver. In general this set may be treated as the
impulse response of the object under investigation. Convolution of this impulse response
with the transmitted signal gives the waveform of the received signal.
Such a modelling enables to simulate propagation of ultrasonic wave along a single ray,
including multiple branching of this ray at the boundaries of granules. The ultrasonic
transmitter /receiver with finite dimensions was simulated by the set of the acoustic rays
transmitted from the group of the points on the transducers surface and the reflected back
signals were integrated over the aperture of the ultrasonic transducer. Performing such
calculations for different positions of the ultrasonic transducer, the complete NDT process
is simulated and results are presented like typical B-scan images. The results of such
modelling are presented in Fig.4. The B-scan image was obtained simulating scanning of
the granular medium presented in the Fig.1, 3, using 4MHz, 45° degrees longitudinal wave.6
probe. For comparison experimental B-scans were acquisited on the 25x40x30mm steel
sample with a granular structure using a similar ultrasonic probe (Fig.5). The results
presented in Fig.4 and 5 demonstrate a very similar character of a structural noise.
The approach described enables to investigate formation of a structural noise depending on
the reflectivity conditions at boundaries of cells taking into account influence of ultrasonic
Fig. 4. The simulated B-scan image of a granulated medium; average size of granules
is 1.7mm (the dimension of the region are given in mm).|
Fig. 5. The experimental B-scan (the dimension of the region are denoted in mm
The influence of structural noise is only one factor affecting detection of stress-corrosion
cracks. Another one is the sophisticated shape of cracks possessing a branched structure,
which causes a quite diffused reflection. For investigation of ultrasonic signals reflected by
such cracks, the approach based on the transducer diffraction model was applied. The
model developed enables to take into account direct reflection, specular reflections from
the bottom and mode conversion signals (10). For the sake of simplicity, simulation results
presented in this paper were performed neglecting structural noise. The simulation set-up is
presented in Fig. 6. The transducer model corresponds to the 45, 2.3 MHz, 10 mm
diameter shear wave probe. The simulation results are presented in Fig.7. In Fig.7a the
ordinary B-scan image is presented in which specularly reflected and crack edge
diffraction signals can be seen. The B-scan corrected according to the geometry of the
object (10) is shown in Fig.7b. The presented example illustrates possibilities of the
developed model to simulate reflection of ultrasonic signals from surface braking cracks
caused by stress corrosion.
Fig. 6. The simulation setup
Fig. 7. B-scan images: a-ordinary, b-corrected
The developed numerical model enables simulation of propagation of ultrasonic waves in
random granular materials with cracks caused by stress corrosion. For computer simulation
of a granular medium a novel method based on application of Voronoi diagrams was.8
developed. Various cracks were simulated by fractals, what enabled to generate cracks
possessing branchy character. Propagation of ultrasonic waves was modelled using the ray
tracing approach taking into account multiple reflections between granules. That enabled to
investigate formation of a structural noise. The performance of the proposed model was
verified by experiments.
- J. Moysan, G. Corneloup, "Ultrasounds back-scattering measurements for new
anisotropy indicator construction", 15 th WCNDT, Roma, 2000.
- S. Ahmed, R.B. Thompson, "Effect of preferred grain orientation and grain elongation
on ultrasonic wave propagation in stainless steel", in Review of Progress in
Quantitative Nondestructive Evaluation, Vol.11, edited by D.O. Thompson and D.E.
Chimenti (Plenum, New York, 1992), pp.1999-2006.
- S. Hirsekorn, "Directional dependence of ultrasonic propagation in textured
polycrystals", J.Acoust.Soc.Am. 79 (5), May 1986, pp.1269-1279.
- J. Saniie, T. Wang, N.M. Bilgutay, "Statistical evaluation backscattered ultrasonic
grain signals", J.Acoust.Soc.Am. 84(1), July 1988, pp.400-408.
- J. Spanner, "Neural networks for ultrasonic detection of intergranular stress corrosion
cracking", NDT.net, Vol.5, No.07, July 2000.
- L. Ericsson, M.G. Gustafsson, "Perception and entropy inspired ultrasonic grain noise
suppression using noncoherent detector statistics", Proc. 7 th ECNDT, Copenhagen,1998
- Nondestructive testing. Handbook. Ultrasonic testing. Vol.7. American Society for
Nondestructive Testing, 1991, pp. 580-756.
- V.L. Newhouse, N.M. Bilgutay, J. Saniie, "Flaw-to-grain echo enhancement by Split-Spectrum
processing", Ultrasonics, vol.20, Mar.1982, pp.59-68.
- M.C. Robini, I.E. Magnin, "Application of the Wavelet packet transform to flaw
detection in ultrasound B-scans", IEEE Ultrasonics Symposium, 1995.
- R. Raisutis, L .Mazeika, "The simulation of ultrasonic imaging in the case of the
objects with a complex geometry", Ultragarsas, No.1.(38), 2001, pp.46-53.
- F.E. Stanke, G.S. Kino, "A unified theory for elastic wave propagation in
polycrystalline materials", J.Acoust.Soc.Am. 75(31), March 1984, pp.665-681.
- A. Lhemery, "Impulse response method to predict echo responses from targets of
complex geometry. part I: Theory," J.Acoust.Soc.Am. 90(5), Nov. 1991, pp.2799-2807.
- C. B. Barber, D.P. Dobkin, H.T. Huhdanpaa, "The Quickhull Algorithm for Convex
Hulls," ACM Transactions on Mathematical Software, Vol. 22, No. 4, Dec. 1996, pp.
- F. Aurenhammer, "Voronoi diagrams- a survey of a fundamental geometric data
structure", ACM Computing Surveys 23, 1991, pp. 345-405.