18th World Conference on Nondestructive Testing, 16-20 April 2012, Durban, South Africa

Non-Destructive Testing of Continuously Cast Billets by Means

of the Laser Triangulation Method

Lukas PINDOR
1
, Stepan HEFNER
2
, Jiri CIBULKA
1
, Rostislav TURON
1


1
Department of technology and research, TRINECKE ZELEZARNY, a.s.
Trinec, Czech Republic, Phone: +420 558 535 439; e-mail: Lukas.Pindor@trz.cz,
Jiri.Cibulka@trz.cz, Rostislav.Turon@trz.cz
2
Department of quality assurance, TRINECKE ZELEZARNY, a.s.
Trinec, Czech Republic, Phone: +420 558 533 697; e-mail: Stepan.Hefner@trz.cz


Abstract
Nowadays, there have been increasing demands for the surface quality of hot rolled bars and wires evoked by
customers. Among other quality parameters is the presence of miscellaneous defects and other heterogeneities on
the surface of billets which serve as the input material for the production of the above mentioned rods and wires
in Tineck elezrny, a. s. In this paper we assume and predict a correlation between the surface defects found
on billets and the frequency distribution of various defects on the surface of rolled products. Therefore, it is
important to map the whole billet surface and for this purpose we have tested the laser triangulation method. The
results indicate a good applicability of this method for surface testing.
Keywords: Casted Billet, Laser Triangulation Method, NDT, surface quality

1. Introduction
Customer demands for surface quality of hot rolled bars and wires are currently increasing.
This trend reduces the labour productivity and burdens the metallurgical enterprises by
reducing the maximum permissible depth of surface defects on rolled product. This fact leads
to the fact that the main effort is directed at finding the cause or origin of the defects on rods
and / or wires, whereas the defects can occur during the rolling process, but can also be
transmitted from the billet, that is, the input material for the production of rods and wires.
Metallurgical enterprises react to this reality by trying to develop an automatic control of
continuously cast billets. Such an inspection line is equipped to detect surface but also internal
defects and makes usually part of the equipment for continuous casting of billets, where the
warm, solidifying billet surface is monitored, or the line is included in front of the rolling mill,
where the billet surface temperature does not exceed 100 C.
Defects on continuously cast blanks can be classified as surface, subsurface and internal.
Surface defects are roll marks, scratches, scales, and especially cracks. The internal defects
can include internal cracks, inclusions, segregation, etc. (see diagram in figure 1) [1].



Figure 1. Defects occurring on the surface and inside a continuously cast billet [1]

Classification of defects is one thing, the other thing is their precise detection of the billet
surface and further transfer of these defects on the surface of the rolled product surface (rods
and wires).
We experimentally tested several methods of detection of defects and evaluation of quality of
oscilation wrinkles, such as: visual, magnetic powder, magnetic field permanence, inductive
heating with thermography, whirling currents, infra scanner (in hot state) and ultrasound
testing shoe for subsurface defects. With regard to the continuous cast blank condition we
can say that each method has its advantages as well as limitations. For this reason it is obvious
that the future device for defect detection on continuously cast billets will have to combine
several methods.
The aim of this paper was to verify in the conditions of Tineck elezrny, a.s. laser
triangulation method of detection of surface defects, while every effort has been directed
mainly to processing of data obtained from the scanned surface of the continuously cast billet.

2. Experiment
Experiment regarding the detection of surface defects on continuously cast blanks was
performed at the technical university of Brno on samples delivered to this end, of
continuously cast billets with longitudinal section and dimensions of 500 x 150 x 5 mm
(Figure 2) presenting visually detected surface defects. After multiple testing of different
methods of detection of surface inhomogeneities laser triangulation method was chosen for
the experiment.




Figure 2. Experimental samples

2.1 Laser triangulation method
Laser triangulation method is based on using a laser profilometer, type MICRO-EPSILON
[2], whose principle of operation, use and method of measurement is shown in the block
diagram in Figure 3.

matice sensor sensor matrix pijma receiver
zdroj svtla light emission source mic rozsah osa-z
(vka profilu)
measuring range, z-
axis (section height)
mic rozsah osa x
(ka profilu)
measuring range, x-
axis (section width)


Figure 3. Principle of laser profilometer function [2]


The laser scanner, known as a laser profilometer uses the principle of triangulation for two-
dimensional detection of profiles on measured surfaces. Choppered laser beam is extended by
means of special lenses to static line laser beam, which falls onto the examined area. Diffusely
reflected radiation from the surface imperfections returns to a very sensitive sensor array and
the high quality optical system. The output signal depends on the frequency of the laser start
and is demodulated using an internal synchronous detector. In addition to distance
information (z-axis), parameters from sensitive sensor array are used to determine the
positions in x and y axes. These measured values form an output in two-dimensional
coordinate system, which is determined with regard to the position of the measured surface
towards the sensor. In the case of the measured sample or by sensor shift it is therefore
possible to obtain a 3D image of the measured values [2].
2.2 The measurement
During 2011, experiments were performed with a laser profilometer on selected samples of
billets (Figure 2). This method showed excellent results in terms of billet surface sensing and
its main advantage is that the recognition depth of defects is about 10 micrometers. Other
positives can be found in its applicability in dynamic mode without external lighting, balance
setting of sharpness and wavelength of light, and it is a less demanding method in terms of
maintenance.
However, it appears that the technique of scanning the sample surface is only a part of the
overall evaluation of surface quality. Considerable role is further played here by the speed of
movement of the sample to the sensing device, but especially the procedures and evaluation of
the scanned data.
For this purpose, the FEKT faculty of the technical university of Brno created a program for
data processing in MATLAB and the measured data were processed by two procedures.
The first procedure is based on digital filtering of surface using a modified STA / LTA
method. This is the filtering of data and does not address image processing. Data obtained
from the profilometer are filtered by moving average with matrix of the bigger dimension
(LTA) and a matrix of smaller dimension (STA). This filtered data is then processed to obtain
information that contain only significant changes from the original measured data (Fig. 4).



Figure 4. Filtering by STA / LTA method

In other words, this highlights the surface inhomogeneities with defects and the effects of
transverse mounting of laser profilometer or sample are suppressed. In the next step it is
necessary to separate defects and other inhomogeneities and thresholding is performed for this
particular purpose. It can be clearly described on the example of oscillatory wrinkles (Figure
4)
Thresholding value is based on the consideration that defects are found in greater depth than
the oscillating wrinkles, by means of which the information of defects location and oscillatory
wrinkles is obtained. Despite the probability distribution of black dots (oscillatory wrinkles)
on the x-axis, an average centre of each oscillation wrinkles is received.
If we split the binary image with oscillation wrinkles into multiple segments, and again we
find a probability distribution of oscillatory wrinkles on the x-axis, we obtain a more accurate
shape of the original oscillation wrinkle. Thus we are able to describe both the shape of
oscillatory wrinkles, their average spacing, the average depth (Figure 5), presence of
discontinuities between oscillation wrinkles, and we can also see display defects and measure
their depth (Figure 6, 7) [3].



etnost frequency rozhodovac rove
defektu
defect resolution
level
rozhodovac rove
vrsek
wrinkle resolution
level
pepotan z
souadnice
recalculated z
coordinate
pouze vrsky only wrinkles

Figure 5. Conversion to binary image containing defects and oscillatory wrinkles


vskyt v binrnm
obraze
occurrence in binary
image
poet segment number of segments

Figure 6. Finding the oscillatory wrinkles





prmrn hloubka
vrsky
average wrinkle
depth
oblast nespojitosti
vrsky
wrinkle discontinuity
area
oblast defektu defect area prmrn vzdlenost
vrsek
average wrinkle
spcing
Figure 7. Resulting display

The second procedure is based on the torque description of areas. It is a method of image
processing here. Through central torques, standardized central moments it is possible to
calculate seven torque characteristics for each object in the binary image.
At the beginning it is necessary to again adjust the measured data from the profilometer and
eliminate slow-growing trends and then convert the modified data to a binary image. In this
binary image it is then necessary to find individual objects and to calculate seven torque
characteristics for them. Figure 8 shows the value of calculation of the first torque of the
seven torque characteristics for each object in the image. It can be noted that objects that
resemble the oscillating wrinkles, have a high value of this torque.
On the other hand, defects are characterized by low value of the first torque. In this way it is
necessary to define the other torques as well and thus to differentiate the oscillatory wrinkles
from defects.
When connecting multiple oscillating wrinkles together, such as by a longitudinal defect, it is
necessary to check the objects, whether it is an oscillation wrinkle or not, in a narrow window.
The width of this window is selected again by the probability of distribution of black points
on the x-axis. In this way it is possible to retrieve information about the oscillatory wrinkles
and defects.



Figure 8. Finding the defects and oscillation wrinkles using torque description of areas
in the image

3. Conclusion
Laser triangulation method was used to evaluate the possibility to detect surface defects on
continuously cast blanks. It turned out that the used laser profilometer is able to scan the
surface quality of billet, but also requires a sophisticated method of processing of the data
obtained. In this way signal processing was made by digital surface filtration using STA/LTA
method and torque description of areas, which has brought about promising results.
The paper was prepared with the contribution of the FR-TI 2/536 project, within the TI P
2010 program of the Ministry of I ndustry and Trade of the Czech Republic.

References
1. www.ndt.net/search/docs.php3?id=4544
2. http://www.micro-epsilon.cz/download/products/cat--scanCONTROL--en.pdf
3. L Grmela, J ikula, P Koktav, P Sedlk, J Majzner and P Tofel `Dl zprva o een
projektu v programu TIP v roce: 2011 internal document, FEKT, VUT Brno,
prosinec 2011