Imaging Molten Steel Flow Profiles: A Lancaster University Corus Uk LTD

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Imaging Molten Steel Flow Profiles

R BjflflSa, A.R.A. LyOnSL, A.J. Peytona, and W.D.N. Pritchardc

a Lancaster University
b Metal Process Control
CORUS UK Ltd

ABSTRACT

Control of molten steel delivery in continuous casting is critical to ensure stability of the meniscus and satisfactory mould
flow patterns, which in turn are determinants of steel cleanness and surface quality. Considerable effort has been expended
over the last 10 years in optimizing the design of the metal delivery system, particularly the pouring nozzle, to enable the
consistent production of high quality steel at high throughput.
This paper looks forward to possible systems that are capable of topographically imaging the distribution of molten steel
flows in these applications. The paper will concentrate on the feasibility of using electromagnetic methods. The paper will
present some initial results and an overview of the image reconstruction process used will also be included.

The paper will conclude with a discussion of possible future developments, such as the use of a tomographic or multi-
frequency approach, future research on the reconstruction image procedures and the potential for visualisation and flow
measurement. There is a need for further research in this area and some priority areas for future work will be suggested.

1. INTRODUCTION
In the continuous casting of steel, control of molten steel delivery through the pouring nozzle, between the tundish and the
mould, is critical in order to ensure the stability of the meniscus and create the optimum laminar flow patterns. These two
factors influence the surface quality and cleanliness of steel, product quality is also influenced both by the flow of steel and
the amount of 'clogging' within the pouring nozzle. Nozzle clogging is especially problematic when casting low carbon,
aluminum steels due to the deposition mechanism [1, 2] of non-magnetic elements upon the nozzle wall; this results in
quality problems due to asymmetric flow and inhomogeneous heat transfer within the mould.

Ladle ShrOUd JLtopper


Tundish

Water -cooled JPouring nozzle


copper mould

Spray chamber—' Torch cutter

Support roll
o
Straightener Runout table

Figure 1. Overview of a continuous casting process

Process Imaging for Automatic Control, Hugh McCann, David M. Scott, Editors,
284 Proceedings of SPIE Vol. 4188 (2001) © 2001 SPIE · 0277-786X/00/$15.00

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 08/24/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx


The position of the pouring nozzle in the casting process is shown in Figure 1 . Liquid steel from basic oxygen steelmaking
(BOS) is poured into the ladle, which supplies, via a refractory shroud, the tundish with molten steel. The tundish both
distributes the steel and removes inclusions. Steel delivery to the mould (through the pouring nozzle) is controlled via the
'stopper' rod; employment of the pouring nozzle prevents oxidization and hence maintains steel cleanliness. Argon gas may
also be injected at the pouring nozzle in order to exclude air [3]. Casting powder is used on the surface of the oscillating
mould to assist with lubrication. Once in the copper mould, the outer layer of the steel cools rapidly to form a partly
solidified shell containing liquid steel, this is supported through a series of water-cooled rollers; further uniform cooling is
achieved via banks of water sprays. The resulting cooled and solidified steel is then divided by cutting torches.

Although, there are several methods for sensing flows, such as for example X-rays. Due to inherent safety considerations
X-rays are yet to be used on plant for this application. Flow control is usually achieved using electromagnetic or
radioisotopes metal level sensors in the mould. Optimum flow profiles in the mould are disrupted by alumina deposition on
the stopper and in the pouring nozzle. This may cause powder entrainment and inclusions leading to subsequent quality
problems [4- 1 0] . A flow visualization approach based on a rugged and inherently safe sensor would be desirable. The
ultimate aim of such a sensor will be to visualize steel flows and detect clogging within the pouring nozzle. Flow anomalies
create differing sensing problems, for example one flow region may be bubbly (argon gas within the steel) and another may
be annular. Figure 2 illustrates several possible flow regimes that could exist within a submerged entry nozzle.
Considerable advances have been made in the modeling and simulation of the metal delivery system within continuous
casting processes [11-15], however there remains a need for on-line verification of the actual flow regime. The sensor would
thus be required to distinguish between such regimes and characterize performance of the pouring nozzle / 'stopper' rod
system. From this characterization and optimisation of the nozzle, better control over the flow rates could then be achieved.
Consequently, the purpose of this study is to assess the potential for using electromagnetic methods to address this problem.

Figure 2. Pouring nozzle flow regimes

Proc. SPIE Vol. 4188 285

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 08/24/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx


2. OVERVIEW OF THE IMAGING TECHNIQUE

Over the past decade there has been considerable interest in the application of tomographic systems to industrial processes
in order to improve performance and productivity [16] . A wide variety of tomographic techniques have been researched for
process applications and electrical techniques in particular, despite their modest image resolution, have been considered
mainly because of their potentially high imaging speeds, relatively low cost, non-intrusive and non-hazardous nature. The
combination of these properties makes electrical techniques attractive for a number of process applications and especially
on-line monitoring and control. EMT [17,18] uses magnetic coupling between inductive sensors in order to tomographically
image electrical conductivity profiles (by detecting the eddy current loss) and magnetic permeability profiles (detection of
inductance variances). Recently eddy current techniques have been researched for the similar problem of imaging metal
solidification JI 19,20].

A typical EMT system has three major parts: sensor array, data acquisition electronics and host computer. A typical system
is arranged as shown in Figure 3. The sensor array is energized with a sinusoidally varying AC magnetic field created by
one or more of the excitation coils. Electrically conductive or ferromagnetic objects within the space cause the applied
magnetic field to be modified; the resultant field changes are measured with an array of detection coils. A set of such
measurements is taken for a variety of energizing magnetic field distributions; from the acquired data, the host computer,
via reconstruction software, is able to produce an image of the material distribution.

Sensor array Conditioning Host computer

Reconstruction algorithm
= C C21. CM1 D1
'2 C12 C22. CM2 D2
13 C13 C23. . CM3 D3
Field control Data
& &
Measured signals Control
'N C2NC2N I I
AC magnetic DM
field

Figure 3. Block diagram of a typical EMT system

3. EXPERIMENTAL RESULTS

The aims of the experimental study were to gain insight to whether this application of EMT and to assess the potential of
using electromagnetic sensors to distinguish between different types of flow regimes. Specifically, this involved:
U
Demonstrating that the correct characteristics can be visualized, and that simple object distributions could be
determined in their correct positions.
U
Gaining experience on the design of sensor arrays for this application in order to propose a sensor design that could be
used on subsequent tests involving molten steel.
• Formulating the forward problem in order to predict the outputs from the sensor coils given a known distribution of
electrical conductivity.
• Applying an image reconstruction algorithm to solve the inverse problem and thus determine the electrical conductivity
distribution within the pouring nozzle from the measured coil outputs.

286 Proc. SPIE Vol. 4188

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 08/24/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx


3.1. OPERATING FREQUENCY
The optimum excitation frequency is determined by several factors including sensitivity of the detectors, the affects of the
electrically conducting pouring nozzle wall (i.e. induced eddy currents within the wall) and the decrease in resonating
frequency incurred in a real system due to the extra capacitance associated with the connection cables. The pouring nozzle
was fabricated from an alumina and graphite based composite (electrical conductivity of graphite o 70k S/m, compared
with a target steel conductivity 2. 1M S/m and copper 58M S/m), thus the electromagnetic skin depth, 8, for frequency,f,
given by equation (1) for the pouring nozzle wall must be large enough to cause negligible effect.

1
(1)
£7

The affect of the pouring nozzle wall on the flux linkage is illustrated in Figure 4, which shows the measured mutual
impedance between two coils (100 turns, 50 mm diameter) placed either side of the nozzle (in this case a submerged entry
nozzle, SEN). At relatively low frequencies the pouring nozzle has little effect, however above 10kHz, the pouring nozzle
starts to attenuate the coupling and finally resonance occurs at 500kHz. This graph indicated that the maximum operating
frequency should be kept below 10 kHz, to avoid excessive eddy current losses in the nozzle walls.

1.E+05 Test set-up


1.E+04 Impedance analyser
—Air
1.E+03
—SEN
E
-C
0 1.E+02
0)
C Pouring nozzle with
0. 1.E+O1 two coils attached
0
1.E+OO
1E+04 1.E+05
1.E-O1

1.E-02
Frequency (Hz)

Figure 4. Coupling through the pouring nozzle wall

3.2. EFFECT OF THE FLOW PROFILE ON THE IMPEDANCE SPECTRA


A series of experiments were performed in order to determine the effects of different types of flow profile on the impedance
spectra of the sensor coils. Figure 5(a) shows a block diagram of the test set-up. In this case, a null coil was introduced in
order to reduce the background to a minimum. Figure 5(b) shows the three flow regimes tested. Models of each regime
were constructed from Woods metal (Bi 50%, Pb 25%, Cd 12.5%, Sn 12.5%) and glass beads to represent the argon
bubbles. At room temperature, Woods metal has an electrical conductivity of approximately three times that of molten
steel. Model A had an outer diameter of 83mm, model had a central rod of diameter 36mm, and annulus in model C was 83
mm OD and 71 mmID.

Proc. SPIE Vol. 4188 287

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 08/24/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx


A B

0
00000
0 Woods metal O Woods
metal

0-
000 Glass
—bead
0c0
00

oco

(a) (b)
Figure 5. Test set-up to determine the effects of flow regime

Figure 6 contains plots of the impedance spectra, in relative units, for the empty nozzle and for the three flow models. A
frequency range of approximately 100 Hz to 1 kHz was used, so the effects of the conducting nozzle walls can be largely
ignored, as described in section 3 . 1 . All the curves increase with frequency due to the differential nature of Farday's law.
The plots can be split into two pairs. The curves for the empty core and for case B (central stream) are similar, but case B
gave a slightly higher output than for no metal. In case B, only a small cross section is available for the flow of induced
eddy currents within the central stream of metal and therefore the signal change is relatively small. The curves for cases A
(bubbly) and C (annular) are also similar to each other. In both these cases, a large bulk of metal is present for the flow of
induced eddy currents and hence much larger signal changes. However, these two curves do display some significant
differences, with for example a cross over point at approximately 800 Hz. This is because the bubbles impede the flow of
induced eddy currents more at higher frequencies. It indicates that the impedance spectra could be used to differentiate
between these two flow regimes.

N
=
0
0

C
C)
U)
0

0 200 400 600 800 1000 1200


Frequency (Hz)

Figure 6. Graph of relative signal output versus frequency for the flow models

288 Proc. SPIE Vol. 4188

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 08/24/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx


3.3. TOMOGRAPHIC IMAGING
The test arrangement in Figure 7 was employed to assess the potential for tomography. The number of coils chosen was six,
setting the compromise between system complexity, signal-to noise-ratio and resolution. All the coils were matched and
arranged at 60° intervals to preserve symmetry.

GPIB SEN with sensor array Arrangement of


coils around SEN
link

4
Impedance analyser
1
Host computer
6

Figure 7. Tomographic test arrangement

The physical principle for this form of EMT can be expressed as a diffusion equation [21] for magnetic flux density B as
given in complex phasor notation by,
= jw/i0 (2)

Naturally the boundary conditions necessary for the excitation coil and interfaces between the different materials must also
be included. There are a number of established procedures for the solution of such eddy current problems including direct
analytical techniques and finite element methods (FEM). The approach adopted here was a standard sensitivity method [22]
widely used for electrical tomography problems, which involves determining how each sensor coil combination responds to
pixel perturbations within the object space. In this case, the solution was obtained by direct measurement using a small test
sample (a 12.5 mm diameter copper rod) to create the pixel perturbation. To ensure that the entire object space was
recorded, the sample was moved upon a 5mm grid, thus 100 sample positions were taken over the 3000mm2 object space.
To provide the full information about the object space, the resulting impedance changes for every pixel position and for
each coil pair was recorded. This test procedure created a series of sensitivity maps, which relate impedance, coil pairs and
sample position. Figure 8(a) shows the changes in the imaginary component of the mutual impedance between opposite
coils 1 & 4 and shows a distinctive saddle shape common with opposite elements. Figure 8(b) shows the sensitivity
between coils 1 and 3, and as expected this shows a reduction in coupling as the perturbation approaches the coils.

SENSI11VITY FOR EXCITE 1 - SENSE 4 averaged resuIts) SENSITIVITY FOR EXCITE 1 - SENSE 3 ( averaged results )

40

X Co-ord X co-ord

(a) (b)
Figure 8. Sensitivity maps when coil 1 is excited

Proc. SPIE Vol. 4188 289

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 08/24/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx


Symmetry can be exploited so that only the maps for coils 1 to 4 (inclusive) were required. The projection matrix, A, in
equation (3), describes the system response. The projection matrix (A in equation 3) relates how impedance values for each
coil pair is dependent upon the contents of each pixel and is created from an amalgamation of all the sensitivity maps.
Vectorf contains the image pixel values arranged as a column and the corresponding measurement values are arranged in a
vector p. The system is taken to be linear, such that pixel interactions are ignored, and therefore superposition is assumed.
of course, this is a significant simplification and does not accurately represent the actual non-linear response of the system.

pA.f (3)

For the image reconstruction problem, the coupling impedances in p are known, but the pixel conductivities in I are not.
Unfortunately the A matrix cannot be inverted directly and consequently iterative techniques are often used to obtain an
approximate solution. A Simultaneous Increment Reconstruction Technique (SIRT) was employed, with an iterative step
given by:

f(K+1)f(K)+2(PA.f(K)).D (4)

In the above equation D is a pseudo-inverse matrix of A and 2 is a relaxation factor, which dictates the speed and accuracy
of algorithm convergence. Vector/" is the previous value of the state of the pixels. The maximum value of the pixels is 1,
corresponding to a full element and values between 0 and 1 correspond to partially full elements. Some constraining was
implemented, so the value of each pixel was forced to lie between 0 and 1 ,as only conductive media is present.

In order to test that the system could distinguish between differing areas of conductivity, copper bars were used to test three
main events, as follows: small samples (pixel sized), areas much greater than the test sample, and multiple events, i.e. two
small samples within the object space. It should be noted that these tests are not truly representative of the real problem,
i.e., in reality small areas of low conductivity will occur in a highly conductive background material, and the conductivity of
copper is also much higher than molten steel. However, despite these limitations, the tests do provide confirmation that the
system could work and also provide valuable insight into the design of the sensor array and image reconstruction algorithm.
A selection of results from these tests is shown in Figures 9 to 11.

SMALL BAR AT(3.3) SMALL BAR AT (33)


-40

-30

-20
LU
-10
•0
>

10 0
20

30

40 30 20 10 0 -10 -20 -30 -40


X Co-ord

Figure 9. Small bar at x, y position (3,3) mm

Figure 9 shows the response to a bar (12mm diameter) placed in the centre of the pouring nozzle, and this position suffers
from the least spatial and contrast resolutions. It can be observed that the pixel value at the point of the target is much less
than 1 and the target is spread over several pixels due to the filtering effects of the image reconstruction algorithm. However
it can be seen that there is good accuracy between the real and estimated bar positions.

290 Proc. SPIE Vol. 4188

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 08/24/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx


A large bar (25 mm dia.) is shown in Figure 10. Comparing Figure 9 and 10 shows the large difference in pixel values due
to the quantities of fully occupied pixels. It is intended to use this pixel intensity to estimate the actual void size; thus
providing a greater degree of information about the steel flow.
LARGE EAR AT(0,0)
LARGE BAR AT(00)

30

20

10

-10

-20
40
-30

Oo -30 -20 -10 0 10 20 30 '0


XCo-ord

Figure 10. Large bar at (0,0)

Visualization of dual streams is investigated by placing two 12 mm diameter rods within the object space, with the results
shown in Figure 11. This test is very important, as the ability to visualize complicated flows would be required for a real
system. The confidence values are approximately that for a single bar, however artifacts also occur. The resolution for
these events is within 5mm (for the lowest bar) while the upper bar's prediction is approximately that of the real position.

2 BARS IN THE OBJECT SPACE


2 BARS THE OBJECT SPACE

30

20
04 ----t
10

U
-10

-20

20 40 -30

YCo-ord - 10 -30 -20 -10 0 10 20 30 40


XCo-ord X Co-ord

Figure 9. Two bars

4. DISCUSSION
The results presented in this paper support the potential feasibility of using electromagnetic sensing techniques to monitor
and image the flow of molten metal through a pouring nozzle, such as a SEN. For the type of nozzle in this study, the
results show that:

• with suitable choice of excitation frequency, the conducting nozzle itself may have no effect

• spectral information may be used to distinguish between different types of flow profile

tomographic imaging may be a possibility.


There is a clear need for further research. In particular, tests need to be performed on hot metal, possibly using modified
nozzles to ensure known and consistent flow profiles are established. The use of spectroscopy indicates the potential do

Proc. SPIE Vol. 4188 291

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 08/24/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx


distinguish between bubbly and continuous metal phases, however inversion algorithms will be required to extract this
information. The use of coil arrays for tomographic imaging also shows some promise, even though the phantoms tested
are simple and do not have the same electrical conductivity as molten steel. Even with existing two-dimensional algorithms
the images show good accord with the real distributions. It is, however, likely that more sophisticated image reconstruction
algorithms will be required, which take into account the three dimensional geometry of the actual process. In particular the
effects of the close proximity of the tundish and mold should be taken into account.

5. CONCLUSION
From this investigation into the application electromagnetic techniques, some promising results have been produced that
support the feasibility of using these methods to visualize molten steel flow regimes. The paper has presented results from
laboratory tests on cold phantoms using both spectroscopy and tomography. From these results, some suggestions for future
research have been made.

6. REFERENCES
1. Goldschmit M. B., Javier Principe R., and Koslowski M, "Numerical Modelling of Submerged Entry Nozzle", 3rd
European Conference on Continuous Casting, pp. 191-200, 1998.
2. Yin H., Shibata H., Emi T., and Suzuki M., "Characteristics of Agglomeration of Various Inclusion Particles on Molten
Steel Surface", ISIJ International, Vol. 37, pp. 946-955, 1997.
3. Sjostrom U. Burty M., Gaggioli A., and Radot J. P., "An Experimental Study of Argon Injection and Aspiration of Air
into Stopper Rods in Continuous Casters", Steelmaking Conf Proc., pp.63-71, 1998.
4. Wilson F. G., Heesom M .J., Nicholson A., and Hills A. W. D., "Effect of Fluid Flow Characteristics on Nozzle
Blockage in Aluminum-killed Steels", I&SM, Vol. 6, pp. 296-309, 1987.
5. Dawson S., "Tundish Nozzle Blockage during the Continuous Casting of Aluminum-killed Steel", I&SM , Vol. 4, pp.
33-42, 1990.
6. Robinson Q., Hanse E., and Dumas P., "Effect of Sub Entry Nozzle on Longitudinal Face Cracks of Continuously Cast
Slabs", Stahl und Eisen Special, Vol. 10, pp. 108-1 1 1, 1994.
7. Rackers K. G., and Thomas B. G., "Clogging in Continuous Casting Nozzles", Steelmaking Conf Proc., pp. 723-734,
1995.
8. Alavanja M., Gass R. T., Kittridge R. W., and Tsai H. T., "Continuous Improvement of Practices to Reduce Tundish
Nozzle Clogging", Steelmaking Conf Proc. ,pp. 415-425, 1995.
9. Tai H., Morishtia M., and Miyake T., "Prevention of Longitudinal Cracks due to Uneven Solidification at Slab Corner
by Optimization of Submerged Nozzle", 3rd European Conference on Continuous Casting, pp. 447-454, 1998.
10. Kemeny FL., "Tundish Nozzle Clogging - Measurement and Prevention", McLean Symposium Proceedings, pp. 103-
110, 1998.
1 1. Herbertson J., He Q.L., Flint P.J., and Mahapatra RB., "Modelling of Metal Delivery to Continuous Casting Moulds"

Steelmaking Conf Proc. , pp. 17 1- 1 85, 1991.


12. Heaslip L .J., and Schade J., "Physical Modelling and Visualisation of Liquid Steel Flow Behaviour during Continuous
Casting", McLean Symposium Proceedings, pp. 223-237, 1998.
13. Harvey A.D., Lucas T., and Winder E.A., "Optimisation of Steel Flow conditions in the Mould using Mathematical
and Physical Modeling", 3rd European Conference on Continuous Casting, pp. 769-778, 1998.
14. Thomas B.G., Huang X., and Sussman R.C. "Simulation of Argon Gas Flow effects in a Continuous Slab Caster"
Metallurgical and Materials Trans. B, Vol. 25B, pp. 527-547, 1994.
15. Burty M., Larrecq M., Pusse C., and Zbaczyniak Y., "Experimental and Theoretical Analysis of Gas and Metal Flows
in Submerged Entry Nozzles in Continuous Casting", La Revue de Metallurgie -CIT, pp. 1249 — 1255, 1996.
16. Williams, R.A., and Beck, MS. (Ed), "Process Tomography: Principles Techniques and Applications", Butterworth-
Heinemann, ISBN 0 7506 07440, 1995.
17. Peyton, A.J., Yu, Z.Z., Lyon, G., Al-Zeibak, S., Ferreira, J., Velez, J., Linhares, F., Borges, A.R., Xiong, H.L.,
Saunders, N.H. and Beck, MS., "An overview of electromagnetic inductance tomography: description of three
different systems", Meas. Sci. Technol., Vol. 7., pp. 261-271, 1996.

292 Proc. SPIE Vol. 4188

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 08/24/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx


18. Yu Z.Z., Peyton A.J., Xu L.A., and Beck MS., "Electromagnetic Inductance Tomography (EMT): Sensor, Electronics,
and Image Reconstruction for a System with a Rotatable Parallel Excitation ", lEE Proc. A, Science Measurement and
Technology, Vol. 145, pp. 20-25, 1998.
19. M.H. Pham, Y. Hua, and N. Gray, "Imaging the solidification of molten metal by eddy currents —Part 1", Inverse
Problems, Vol. 16, pp. 469-482, 2000.
20. M.H. Pham, Y. Hua, and N. Gray, "Imaging the solidification of molten metal by eddy currents — Part 2", Inverse
problems, Vol. 16, pp. 483-494, 2000.
21. Jackson J. D,. "Classical electrodynamics" 3rd edition, John Wiley & Sons, ISBN 0-471-30932-X, 1998.
22. Natterer F., "The mathematics of computerised tomography", Wiley, Chichester, 1986.

Proc. SPIE Vol. 4188 293

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 08/24/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

You might also like