Advanced Inspection of Surface Quality in Continuously Cast Products by Online Monitoring

(1)

Advanced Inspection of Surface Quality in Continuously Cast Products by Online

Monitoring

Alejandra Slagter

Mechanical Engineering, master's level (120 credits) 2018

Luleå University of Technology

Department of Engineering Sciences and Mathematics

(2)

1

Acknowledgements

I would like to express my gratitude to the European School of Materials (EUSMAT) for the financial support to carry out this Master Programme, and to the AMASE Master Program Secretary for their constant support during these two years.

I would like especially thank my supervisors Esa Vuorinen and Pavel Ramirez Lopez for their guidance and support during this semester. In addition, I would like to give my sincere thanks to Rosa Pineda and Pooria Jalali that have always been there to help, for their endless support.

In addition, I would like to thank Swerea MEFOS for the opportunity to carry out my master thesis in this research institute and the support during the project.

Finally, to everyone that has contributed to the development of this project, technicians at Swerea MEFOS and friends and colleagues both at MEFOS and LTU, whom not only contributed with productive discussions but also have been there to support me along the way.

(3)

2

(4)

3

Abstract

The present Master’s Thesis is dedicated to the study of a laser scanning system and its applicability for the detection of surface defects at the end of a casting machine in a steel production plant. Both room and high temperature trials were carried out on different carbon and stainless steel billets and slabs. For the high temperature tests, the samples were heated until a maximum temperature of 1200 C. For all trials, the surfaces were scanned with a blue laser sensor in order to generate a 3D representation of the as-cast product surface.

The applicability of a blue laser sensor was proven for carbon and stainless steel surfaces, both at room and high temperature. Defects such as depressions and oscillation marks were detected, as well as some small corner cracks. Furthermore, the full transversal section of billets and slabs was reconstructed from different scans of the faces of the products.

The effect of different scanning parameters on the resolution of the scans and the final results were analyzed and discussed with special focus on the scanning strategies that would be optimal for the industrial application of the sensor.

A microstructural analysis was carried out in order to correlate the subsurface microstructure with the presence of depressions in the edges of a duplex stainless steel slab. The as-cast structure of columnar and equiaxed grains was clearly observed, and some special features were analyzed and discussed. Nevertheless, no clear correlation between the subsurface microstructure and the presence of the defect was found.

(5)

4

(6)

5

List of Figures

Figure 1 Schematic representation of a continuous casting machine. After (4). ... 15 Figure 2 Typical possible process routes in the steel production. ... 17 Figure 3 Typical process route for the stainless steel production. Adapted from (9). ... 19 Figure 4 High-temperature region of a pseudobinary phase diagram for duplex stainless steel compositions. The shaded region represents the composition of commercial alloys. Extracted from (19). ... 21 Figure 5 Typical solidification structure observed in the transversal section of an as-cast steel. Chill zone, columnar zone, and equiaxed center region. Extracted from study material from the master program course “Tecnología Metalúrgica” at Universitat Polytecnica de Catalunya 2017. ... 23 Figure 6 As cast microstructures of a stainless steel (a) 316 (austenitic) and (b) 430 (ferritic). The presence of the equiaxed central region can be observed in the ferritic as-cast structure while it is not observed in the austenitic. Extracted from (23). ... 23 Figure 7 Schematic representation of the laser scanning equipment. Adapted from (31). ... 27 Figure 8 ScanCONTROL 3D-View 3.0 software interface (Micro-Epsilon). ... 28 Figure 9 Schematic representation of the ”z coordinate” and ”moment 0” color codings. It is possible to note how the ”moment 0” representation provides a realistic image of the object. Examples provided by Senso Test. ... 29 Figure 10 ScanCONTROL 2960-100/BL measuring range. Dimensions in mm. Extracted from (31). . 30 Figure 11 Examples of measuring fields defined in the sensor ScanCONTROL 2960-100/BL manual. 30 Figure 12 Maximum collection frequency for some of the defined measuring fields. Note that the smaller the measuring field, the higher the allowed collection frequency. ... 32 Figure 13 Set up for the small scale tests. The equipment is placed in a cooling jacket (prepared for high temperature tests) and installed in a milling table arm. Test sample is placed in the milling drill table. ... 34 Figure 14 Schematic representation of the grinding machine used for the full scale trials. Provided by Swerea MEFOS... 36 Figure 15 Top and lateral schematic view of the elements during the scanning of the top surfaces of the samples. ... 36 Figure 16 Schematic view of the elements during the lateral scanning of the samples. ... 37 Figure 17 Image of the grinding table were the blue laser is installed. ... 37 Figure 18 Images of some of the defects that were present on the surface of the steel products. Images in the right side show a depression in the inner bow (wide face) of a stainless steel slab, with a sharp edge at approximately 120 mm from the border (from the narrow face). Images in the left show a depression in the outer bow (wide face) of the same stainless steel slab, where the sharp edge is not present, but a more continuous curvature can be seen. In the interior of both depressions some oscillation marks are visible. ... 38 Figure 19 Image of the slab from which samples were taken for metallographic analysis. ... 41 Figure 20 Schematic representation of the area from which the samples were taken for metallographic analysis. ... 42 Figure 21 Results of a variation in the frequency of data collection. Both images show the same region on the surface where a small piece of scale can be seen. Both images were collected at a scanning speed

(7)

6

of 1 m/s and with 240 mm working distance; (a) 400 profiles per second, the image in the scanning direction is constructed with approximately 280 points; (b) 50 profiles per second, the image in the scanning direction is constructed with approximately 35 points... 43 Figure 22 Example of how different scanning parameters can lead to the same resolution in the scanning direction. (a) Collected with 1 m/s and 400 profiles per second; (b) 0.2 m/s and 80 profiles per second. ... 44 Figure 23 Resolution in the direction of the laser line remains constant even with a change in scanning parameters. Both images were collected at a scanning speed of 1 m/s and with 240 mm working distance; (a) 400 profiles per second; (b) 50 profiles per second. ... 44 Figure 24 (a) 2D representation of results for the smalls cale trials; (b) augmentation of a small area in (a). The color coding represents the height of the sample. Dark areas correspond to valleys while of brigth areas are associated with peaks. Oscillation marks are easily observed. ... 45 Figure 25 Area of the sample with corner cracks; color coding indicates in dark regions of the sample with low z values and in brigth regions of the surface with high z value. ... 45 Figure 26 Area of the sample with corner cracks; color coding ”Moment 0” provides a realistic representation of the surface. ... 46 Figure 27 Image of the grinding machine during the high temperature trials (left) and image of the blue laser on the surface of a high temperature sample during the trials (right). ... 47 Figure 28 Results from high temperature full scale trials. Area around a depression in the inner bow of 2101 steel slab. ... 47 Figure 29 Results from full scale high temperature trials. Complete scan of a 2101 steel slab with depressions. The surface corresponds to the inner bow of the slab and the width of the scan is equivalent to the width of the blue laser line... 48 Figure 30 Results for room temperature full scale trials. Region of the surface of a 304 stainless steel slab in which oscillation marks are visible. ... 48 Figure 31 Results for high temperature full scale trials. Region of the surface of a carbon steel billet.

Some scale at the bottom of the scan can be recognized and variations in the surface profile are also detected... 49 Figure 32 Slab cross section constructed from multiple profiles of different faces for one of the grade 2101 slab sample. The height variations in the profiles are exaggerated to generate a better perspective.

Dimensions in mm. ... 49 Figure 33 Billet cross section constructed for multiple profiles from different faces. The height variations in the profiles are exaggerated to generate a better perspective. Dimensions in mm. ... 50 Figure 34 Slab cross section constructed for multiple profiles from different faces for the grade 304 slab sample. The height variations in the profiles are exaggerated to generate a better perspective.

Dimensions in mm. ... 50 Figure 35 Macro etching results for a 200x200 mm section of a stainless steel slab (left); augmentation of an area in the narrow face in which a transition inside the columnar zone can be distinguished (right). Marble’s reagent etching... 51 Figure 36 Area of the cross section presented in Fig.30 in which the angle between columnar grains and surface of the slab can be clearly seen. Marble’s reagent etching. ... 51 Figure 37 Microstructure of an area of the sample close to the surface (a); and microstructure in the equiaxed region of the material (b). Etching with Beraha’s: Ferrite dark, austenite white. ... 53

(8)

7

Figure 38 Microstructure of the material in a large region (aprox. 30 mm from the surface). The surface of the material is on the left while the equiaxed region can be noted on the right. Beraha’s etching: Ferrite dark, austenite white. ... 53 Figure 39 Microstructure of the material in a large region including the surface of the material in the wide face. Beraha’s etching: Ferrite dark, austenite white. ... 53 Figure 40 Results from a Thermocalc simulation of the amount of phases against temperature for a steel within the compositional ranges presented in Table 5... 54

(9)

8

(10)

9

List of Symbols

dy Distance between consecutive collected profiles in the scanning direction vs Scanning speed

pcf Profiles collected per second Cr Chromium weight percent Mo Molybdenum weight percent Nb Niobium weight percent Ni Nickel weight percent C Carbon weight percent N Nitrogen weight percent

Cu Copper weight percent Creq Chromium equivalent number Nieq Nickel equivalent number

δ Delta ferrite phase γ Austenite phase

V Velocity at which the solid phase grows from the liquid K Thermal gradient

(11)

10

(12)

11

Abbreviations

CC Continuous Casting

AOD Argon Oxygen Decarburization VOD Vacuum Oxygen Decarburization OM Oscillation Marks

AISI American Iron and Steel Institute

ASTM American Society for Testing and Materials FEPA Federation of European Producers of Abrasives CCD Charged Coupled Device

(13)

12

(14)

13

1. Introduction

Continuous casting (CC) is an established technology for the production of steel and is responsible for the solidification of most of the millions of tons of steel that are produced in the world every year (1). In the continuous casting process, the liquid steel is poured into a tundish and from the tundish it flows to a bottomless copper mold, Figure 1. Once in the mold, the molten steel solidifies against the water-cooled copper walls and forms a solid shell (1). The solidifying metal is withdrawn from the mold at a given casting speed, which also matches the liquid metal flow into the mold (1). The tundish holds the molten steel and provides a control flow of metal into the mold (2), and while in slab casting usually one mold is served by one tundish, several billet molds can be supplied by the same tundish. The mold level is a key parameter in the casting process since it exerts a large influence on the liquid flow to the mold and especially in the formation of vortex in the tundish, which could incorporate air or slag in the melt (2). The liquid steel is transported to the mold trough pouring nozzles located along the bottom of the tundish. The design of these nozzles controls the volume and flow of the steel to the mold and also plays a key role in the fluid control. Nevertheless, the mold is the most important component in the continuous casting process (2). It has the primary function of extracting heat from the molten steel as efficiently as possible. It is also continuously oscillating in the vertical direction in order to avoid the adherence of the solidified metal to the copper surface, with oscillation frequencies that can be in the order of 100-200 cycles per minute (2).

The frequency of mold oscillation together with the oscillation characteristics (mold velocity, mold displacement, and time for upward and downward movement) has a strong influence on the formation of surface defects, such as oscillation marks (3).

Figure 1 Schematic representation of a continuous casting machine. After (4).

The continuous casting makes possible not only high productivity but also lower energy consumption, better labor efficiency and quality assurance when compared

Ladle

Tundish Mold

Inner bow Outer bow

Cut-off point

(17)

16

with earlier ingot casting techniques (1-3). Since its first appearance in the last century, an extensive amount of work has been devoted to improve the productivity as well as the quality and cost of steel products. From the first mold oscillation introduced by Junghans (7) (with the purpose of overcoming the sticking of the initially solidified shell) to the more recent introduction of continuous casting modeling to improving the yield and quality of steel production (8), the list of technological improvements is vast.

The development of refractories with better performance, the invention of refining methods such as argon oxygen decarburization (AOD) and the vacuum oxygen decarburization (VOD) are milestones in the development of continuous casting technologies; as have also been the electromagnetic stirring, the better understanding and design of the secondary cooling zone and, nowadays, the implementation of advanced modelling to control and predict the quality of the steel produced (9).

Nevertheless, there are plenty of defects that still plague the industry and that affect both the quality of the final product and productivity (9).

In a conventional steel plant, the continuous casting of slabs is usually followed by the hot rolling, which not only reduces its width to approach the dimensions of the final product but also breaks the casting structure and promotes chemical and microstructural homogenization. After the liquid steel with the proper composition has been obtained, the liquid is continuously cast to form billets, blooms or slabs. The following step is the hot rolling of the steel product and different process routes between the casting and the hot rolling are possible, Figure 2. The most energetically efficient routes are the direct rolling, which consists the hot rolling of the material directly after the casting, and the direct rolling, which involves an intermediate stage in a furnace in order to control the temperature and add time and flexibility to the processing. Other possible routes include the storage of the material either for a short or a long term. Since the presence of severe defects in the cast products can lead to problems during the hot rolling, both the direct rolling and the direct charging require a cast product with a high surface quality (5). As a consequence, it is very common that the billets and slabs after the casting are cooled down to room temperature, inspected, ground if necessary, and then reheated to be hot rolled with an evident increase in energy consumption (10).

(18)

17

Figure 2 Typical possible process routes in the steel production.

An extensive effort has been done in the past years to better understand and predict the formation of surface defects, which are usually generated in the mold or in the secondary cooling zone (11, 12). The evolution of computer systems and the invaluable work of technologists and researchers have made it possible to gain understanding in the complexity of the high temperature solid-liquid interaction of slag, molten steel and copper mold.

More than two decades ago, Brimacombe dreamed with the idea of the “intelligent mold”, capable of “thinking” and taking process decisions based on the temperature

“feelings” and the “observation” of mold level and other process parameters (13).

Motivated by the necessity to empower the workforce with the knowledge of the fundamental rules that govern the solidification of the metal and determine the quality of the product, he had worked and developed the expert system CRAC/X. This computer program, focused on both internal and external cracks, condensed a large amount of information regarding billet defects, and linked the presence of one (or more) of them with the casting parameters to help the machine operator in correcting operational problems and improve the product quality (14). Naturally, the crack length(s), and type(s) are user’s inputs in the program.

Further work has been done in this direction in the past years. The general concept is that, with the help of computers, quality predictions can be me made considering the real, on-time, casting parameters (15). The result is that not only decisions can be made with respect to the process downstream (hot rolling, direct charging, grinding, discard) but it is also possible to control the system to eradicate future defects.

Nevertheless, in order to correlate defects with operating conditions, the model has to be trained to associate a specific defect and its appearance with different casting parameters, and nowadays this can only be done with the invaluable help of experienced operators (15).

(19)

18

2. Aim and Objectives

At the present time, there are no commercially available systems capable of assessing the as cast surface quality in a reproducible and reliable way, and this task is usually performed in the plant by experienced operators. Nevertheless, even with the best- trained personnel, it is not possible to inspect the product online, since the visual inspection can only be performed at room temperature after the as-cast products have been cooled down, which necessary means time and additional cost. Even more, different defect severity criteria may still be present as they depend on the operator and in some cases the severity of the defect is not assessed. For example, a common practice is to divide the slab surface into 3 areas, left edge, center and right edge. A typical report contains the slab number (from the cast sequence) and a statement about the presence of defects in the different areas of the slab, such as “longitudinal cracks”, “inclusions”, and “deep oscillation marks”, “depressions” “transversal cracks”, etc., always associated with the position on the slab in which they are observed (15). If a severity criterion is also included, and depending on the plant or industry standards; a classification into severe, medium and light can be established for cracks based on the opening of the mouth, the crack density or the crack length, or for example based on the depth for oscillation marks. It is important to note that the fact that the severity is assessed by visual inspection may lead to differences between the results when different persons are performing the inspection.

All these inconveniences can be overcome with the use of adequate inspection techniques in the form of systems installed on-line in the plant. Although some surface inspection systems have been developed for their use in rolling mills or mapping samples offline (16-18) transferring them to online monitoring of continuous casting is difficult because of the higher topography variations on the as- cast surface compared with the more defined surface after rolling. In addition, the high temperature of the solidified slab, the on-line process variability (e.g. speed, width and grade changes) as well as the sensibility of the detection system to all the different defects (cracks, depressions, oscillation marks, etc.) are the main challenges that still remain to make the on-line installation of such equipment a reality.

In this context, the main purpose of this project is to support the design and construction of new sensors for the steel industry, with focus on online surface measurements at high temperature in the casting machine.

The specific objectives of the project are:

 To determine the range of defects that can be detected by using the ScanCONTROL 2960-100/BL Blue laser.

 To determine the optimum scanning parameters and the scanning system to be implemented at a steel plant (Otokumpu Stainless, Avesta).

(20)

19

 For one particular steel grade (LDX 2101), to analyze the correlation between surface defects and internal structure

3. Background

3.1. The Continuous Casting of Stainless Steel

Even though the continuous casting of stainless steels shares a common principle with the continuous casting of carbon steels, a number of differences stems from the high amount of alloying elements that are present in the stainless steels composition and from the numerous interactions that can take place either in the solidifying metal or between the latter and the environment.

Stainless steel is nowadays manufactured following the processing route that is shown in Figure 3. The process starts with the melting of a mixture of stainless steel scrap, carbon steel scrap, and other minor additions, and it continues with the refining of the melt, either via AOD or VOD, to achieve molten steel with the adequate composition and temperature (9). Then, the liquid is poured into the tundish to be continuously cast in the form of a slab, billet or strip (9).

The reactivity of certain elements (especially chromium) and their affinity to oxygen and nitrogen make the protection of the molten metal a critical issue. The oxidation of the bath can lead both to operational problems such as nozzle blockage and to a product with a deteriorated quality. In addition, the mold-flux technology necessary to cast stainless steel is highly complex due to the variety of compounds that can be formed. As a consequence, it is usually impossible to cast all the stainless steel grades with the same mold flux and different formulations must be used to handle the different compositions (9).

Figure 3 Typical process route for the stainless steel production. Adapted from (9).

(21)

20

The occurrence of superficial and internal defects during continuous casting is affected both by the intrinsic solidification behavior according to the steel composition and by extrinsic variables that include casting parameters such as the casting speed, mold oscillation characteristics, mold flux, secondary cooling design, mold taper, strand dimensions, etc. A brief review of the most important considerations of both intrinsic behavior and extrinsic variables influencing the defect occurrence is presented in the following sections.

3.2. Stainless Steel Solidification during Continuous Casting

The solidification of metals is in itself a complex process, affected not only by external parameters but also by the intrinsic solidification behavior of the metal or alloy under consideration. It is also of relevant importance because the final microstructure and the different properties of the material are largely affected by the unavoidable solidification process.

Compositional variations can be found on a solidified metal as a result of the solute distribution between the solid and liquid phases which usually accompanies solidification. The composition of the liquid and solid in contact at the solidification front changes continuously during the solidification process as predicted by the equilibrium phase diagrams. In addition, the non-equilibrium conditions during solidification (finite-time process and incomplete diffusion of one or more elements in the alloy) are responsible for the chemical segregation observed in the final solid material (6).

Different solidification behavior can be observed in different stainless steel grades depending on the compositional balance. The possible solidification modes that can be found on stainless steels are (6, 19):

Ferritic (F), the first solid phase that appears is δ ferrite, and δ—>γ transformation only takes place below the solidification temperature.

Ferritic-austenitic (FA), the first solid forming is δ ferrite but austenite also begins to form before the liquid has been completely consumed.

Austenitic-ferritic (AF), austenite is the first phase solidifying from the liquid but ferrite begins to form before the liquid has been completely consumed.

Austenitic (A), the first solid phase is austenite and it continues to be present after subsequent cooling.

The occurrence of any of these modes is determined by the Creq/ Nieq ratio, which represents the relative presence of elements that are ferrite stabilizers (Cr, Si, Mo, Nb, Ti, Al) or austenite stabilizers (Ni, C, N, Mn). Many equations have been developed to estimate the effect of the alloying elements, especially in the context of solidification during welding processes. One of the most renowned equations was developed by the Kotecki and Stewart (20):

(22)

21 Creq= Cr + Mo + 0.7Nb

Nieq= Ni + 35C + 20N + 0.25Cu

Where Cr, Mo, Nb, Ni, C, N, Cu represent the weight percent of each element respectively.

Figure 4 presents the high-temperature region of a pseudobinary phase diagram in which the compositional axes contain the Creq/ Nieq ratio for stainless steels. The shaded area represents the compositional ranges in which commercially available duplex stainless steels are located. It is possible to observe that the solidification mode proceeds as fully ferritic. Other austenitic and ferritic steels would have positions in the diagram that could correspond to austenitic, ferritic-austenitic, austenitic-ferritic or fully ferritic modes.

Figure 4 High-temperature region of a pseudobinary phase diagram for duplex stainless steel compositions. The shaded region represents the composition of commercial alloys. Extracted from (19).

Since the solidification mode determines the presence of one or more phases at the earlier stages of solidification; it can exert a large influence on the occurrence of certain defects such as cracks, depressions, etc. (21). For example, the high- temperature strength of the austenitic steels has been found to be higher than for ferritic stainless steels; and hence, the side bulging problem tends to be more prevalent in ferritic than in austenitic stainless steels (21). Furthermore, small additions of gamma stabilizer elements have been proposed in ferritic stainless steel (within the compositional range of the specific grade) in order to increase the δ—>γ transformation temperature and promote the formation of a stronger shell earlier in

(23)

22

the casting process (21). On the other hand, a small amount of ferrite phase (2-6%) during the solidification of austenitic stainless steel is usually desired to diminish the hot cracking tendency, since the ferrite phase has a larger solubility for tramp elements such as P and S and can prevent the formation of low melting point films between austenitic grains (22). It has also been proposed that a small amount of ferrite phase on the solidification of austenitic stainless steels can refine the size of the columnar grains during solidification and improve the hot workability of austenitic steels (9).

3.2.1. Solidification Structure

Three distinctive zones are generally observed in the as-cast structure of a metal or alloy and are schematically represented in Figure 5. The first solidifying metal that is next to the mold wall usually suffers a large cooling rate, which results in a high rate of nucleation. Consequently, the so-called chill zone is constituted by small equiaxed grains with random orientations (Figure 5 b) (6). As the solidification progresses, the grains become larger in shape and tend to orientate in the direction of the heat flow.

This area of elongated grains is usually referred as columnar zone (Figure 5 c) and also tends to present a determined crystallographic orientation (6). In pure metals, these grains may continue to grow until they meet in the center of the slab.

Nevertheless, a final central region of equiaxed grains is usually observed in the casting of alloys (Figure 5 d). The growth of the columnar grains ceases when they meet other crystals growing from the opposite direction, or when the nucleation and growth rate of new grains in the center of the slab is higher than the growth rate of columnar grains. This can be favored by the concentration of solute elements in the center of the slab (23) which lowers the actual solidification temperature of the melt, or by the breaking-off of the tips of the crystals in the columnar region (6) which provide nuclei for the growth of new grains.

The chemical composition of the stainless steels strongly affects the solidification structure. It has been reported that austenitic stainless steels solidify with a macrostructure which is essentially columnar to the center of the cast (Figure 6 a), while fully ferritic stainless steels tend to present the classic chill zone followed by the columnar and then by the equiaxed central region (Figure 6 b) (5). If the solidification mode proceeds as columnar growth, the concentration of solute elements in the center of the slab tends to be high and centerline segregation is usually formed (this is known as “zone refining action”). On the other hand, the presence of an equiaxed zone in the center of the slab tends to stop the advance of the columnar grains and allows the distribution of the solute elements in a wider region in the center of the slab, diminishing the macrosegregation severity (20). The columnar to equiaxed transition is mainly dependent on the velocity at which the solid phase grows (V) and the thermal gradient (K) (21, 22). The velocity V is influenced by the undercooling, the chemical composition of the alloy and is affected by the presence of electromagnetic stirring. On the other hand, the thermal gradient

(24)

23

K is controlled by the overheating in the tundish, the casting speed, and cooling parameters both in the primary and secondary zone.

Figure 5 Typical solidification structure observed in the transversal section of an as-cast steel. Chill zone, columnar zone, and equiaxed center region. Extracted from study material from the master program course “Tecnología Metalúrgica” at Universitat Polytecnica de Catalunya 2017.

Figure 6 As cast microstructures of a stainless steel (a) 316 (austenitic) and (b) 430 (ferritic). The presence of the equiaxed central region can be observed in the ferritic as-cast structure while it is not observed in the austenitic. Extracted from (23).

(a) (b)

(25)

24

3.2.2. Surface Quality

The surface quality of the slabs after the continuous casting process is the result of the highly complex conditions under which the solidification process takes place.

Parameters such as the superheat, casting speed, casting powder, mold tapper, and mold oscillation have a strong effect on the slab quality and particularly on the characteristics of the surface (5, 9-11, 24). Even more, the steel composition, which affects the solidification process, determines the temperature at which each phase appears and hence the evolution of the volume with temperature (shrinkage) (25).

This, in turn, affects the development of strains-stresses during the solidification and, in combination with the thermal history in its way down in the casting machine;

determines for example, weather solidification cracks will appear or not on the steel surface.

Other usual defects associated either with the shape and/or surface characteristics include oscillation marks (OM) and depressions. The OM are present on the steel surface as a consequence of the continuous movement of the mold during the casting process, but its severity and characteristics such as pitch and depth are dependent on other numerous parameters, which include casting speed, casting powder and type of mold oscillation (24). Because of the complexity of the process, the reader is referred to other references for further information on the mechanisms of OM formation (24, 26). Nevertheless, it is important to note that the presence of deep oscillation marks provides a site in which strains are prone to concentrate; and thereby, providing a potential site for crack nucleation.

Defects such as depressions and bulging are more related to the overall shape of the slab than to the surface itself but are also considered surface defects because they do not affect the internal quality of the material. Depressions are generally associated with a lack of planarity or lack of material in the form of gutters near the corners of the strand either in one or both of the wide faces (27). On the other hand, the bulging effect is usually seen on the narrow phase of the strand because the unconstrained faces are unable to withstand the ferrostatic pressure of the molten metal. The defect is then the result of a weak metal shell, being the thickness (and hence the strength) of the shell a function of the casting speed, mold tapper, overheating, cooling rate, mold flux, etc. (17). Ferritic stainless steels have been reported to be more prone to bulging (due to a lower hot strength when compared with austenitic), and hence the chemical composition, which controls the ratio of gamma and alpha phases during solidification, also plays a role in the occurrence of the bulging effect (17).

Depressions can also be seen near the edges on the wide faces of the slab associated with the bulging in the narrow sides.

(26)

25

3.3. Surface Inspection Techniques for Cast Products 3.3.1. Background and Inventory of Existing Techniques

The importance of reliable surface inspection techniques in the steel industry started to gain importance since the 1980s, primarily due to quality demands from car manufacturers (28). In the beginning, the interest was mainly focused on cold-rolled steel because of its direct visual impact on the product but; with time, hot strip surface quality and surface quality of structural products like rods/bars have also gained importance. Nowadays, online quality control of production is becoming a key factor to ensure quality at a minimum cost. In a cold rolling mill, it allows the earliest defect detection and the chance to take corrective actions faster. On continuous casting it also allows energy savings, making possible the direct charging or hot rolling billets or slabs completely free from surface defects.

Surface inspection techniques are commercially available since the 1970s but it was not until 1990s that started to be used in the inspection of steel surfaces, mainly in cold rolling mills (28). The real-time detection of surface defects presents many challenges. For instance, the variety of surface defects is wide both in cause and in topographical features such as size, location and depth, without mentioning the severity and frequency of the defect. As a result, the selection of a technique able to detect all of them is not straightforward. Additionally, the operating speed during production is usually high, especially at the end of the rolling where typical speeds are in the order of 20 m/s (28).

Visual-based automated surface inspection systems have been widely developed to collect online information of the surface quality of steel products. The system usually consists of one or more light sources, one or more cameras (bright and/or dark field), a fast image processor, server, and the operator interface. For visual-based systems, the place of installation of the equipment is also a key issue, particularly for hot rolling mills. The presence of high ambient temperature, vibration, dust and oil must be considered to ensure optimal operating conditions. In addition, the presence of surface scale on plain carbon and micro-alloyed steels is usually an inconvenience. As stated by Obeso et. al. (14), the presence of oxide scale makes it almost impossible to produce an algorithm that classifies defects and differentiates cracks from inclusions, etc. For this reason, Obeso et. al. (14) developed a system that combines imaging techniques with conoscopic holography in order to detect both cracks and inclusions in hot slabs (700°C or more) even in the presence of oxide scale. The advantage of the conoscopic holography is that it is able to differentiate a crack from scale by means of the topographic characteristics of each one (positive or negative deviation from the surface plane). Nevertheless, other defects such as depressions, bulging and oscillation marks are not considered (14).

(27)

26

An experimental study on the ability of a laser scanning combined with CCD (Charged Coupled Device) imaging system to detect defects in high-temperature slabs has been presented by Ouyang et. al. (29). The system, based on the triangulation distance measurement, proved to be useful in determining the topographical characteristics of slabs and differentiate between defects and scale. In the previously mentioned study, a laser source projects a line on the slab which is reflected by the hot surface and captured by an array CCD. A working distance of 600 mm was successfully used in the study.

Another “tried and tested” technique to measure the surface quality of slabs is the Oscillation Mark Depth Measurement with a laser (30). This measuring technique is applicable to cold slabs (offline measurement), where evaluations are carried out at the narrow face due to the flattening effect of the supporting rolls on the wide faces.

The data acquired is only for a single line along the product surface by use of a relative oscillation mark method which requires manual identification of the mark’s deepest point. The main disadvantage of this technique is its offline nature, which hinders any quick online response to avoid defect formation. Therefore, there is still a need for reliable information of the as-cast surface topography (e.g. not only on a single line) as a precondition for diagnosis of surface defects.

It is clear that the assessment of the surface quality of steel products at high temperature is still a challenge, especially considering the hazardous conditions in the casting machine and the variety of surface defects that can be observed. Then, one interesting strategy to achieve a complete representation of the surface condition of steel products is to combine different inspections techniques such as imaging, temperature measurements and topography measurements. The development of these techniques in order to be able to implement them online and the combination of them is a challenging task that can lead to considerable improvements.

3.3.2. Laser Scanning

The focus in this project is in the assessing of the surface quality of steel products by means of a blue laser scanner (Figure 7). The scanner (Micro-Epsilon) projects a blue laser line over the target surface via a linear optical system and a sensor array collects the diffusively reflected light. Both the distance from the surface to the scanner (Z- axis) and the distance in the laser line (X-axis) are evaluated by the sensor (Fig. 7), which operates according to the optical triangulation principle. The licensed software (ScanCONTROL 3D-View 3.0, Micro-Epsilon) controls the scanning parameters, collects data and plots both 2D and 3D views of the scanned surfaces. The collection frequency can be as high as 2000 Hz and a reference resolution of 12 µm can be achieved (31). A discussion on the parameters that influence the operation of the scanning system; and particularly the resolution, will be presented in this section.

(28)

27

Figure 7 Schematic representation of the laser scanning equipment. Adapted from (31).

3.3.2.1. Software Interface

The software used for the collection of the surface information is ScanCONTROL 3D- View 3.0 (Micro-Epsilon). Figure 8 presents the software interface. The control setting panes can be seen on the left side of the screen while windows for preview, 2D, and 3D views are seen on the middle and right side of the screen. From the control settings pane, it is possible to control different scanning parameters such as the exposure time, number of profiles per second and also the buffered profiles. A more detailed explanation of the significance of all these factors will be presented in following sections. In addition; in the 3D view pane, the preview and color coding can be selected according to the user preferences, as well as other display settings.

Z direction

Y direction X direction

(29)

28

Figure 8 ScanCONTROL 3D-View 3.0 software interface (Micro-Epsilon).

3.3.2.2. Preview Mode and Color Coding

The software used for the collection of 3D information (ScanCONTROL 3D-View 3.0, Micro-Epsilon) can work either in a continuous mode or in an intermittent mode ("preview mode" in the "3D mode" settings in the control pane, Figure 8). In the continuous mode; a certain number of profiles are displayed every few seconds, and the data is not automatically saved. On the other hand; in the intermittent mode, the user defines the number of profiles to be collected (“buffered profiles”) and once the data has been collected the user can save the scanning or export the information either as a 2D or 3D views of the results. The maximum amount of buffered profiles is limited by the RAM/paging file of the computer that is used and the operating system/architecture (32 bit/64bit). For this project, a Lenovo ThinkPad L470 with a 7th Gen Intel® Core™ i7-7600U processor and 16GB RAM were utilized and the maximum buffered profiles were 12500.

Different color coding modes are available in the previously mentioned software. A color coding is basically a specific way to represent the surface of an object taking into account a determined aspect of it; for example paying attention to the topography, or paying attention to the reflectivity of the surface. Each color coding provides information on different features that allow for interpretation of different aspects of the scanned surface. As a common example, the “Moment 0” mode collects and plots with different gray intensities the value of the zeroth moment of the reflection, which provides a realistic image of the scanned object (Figure 9). On the other side, the “z - coordinate” mode plots each point in the 2D or 3D view according to its distance to the scanner device, providing a topographical image of the surface (Figure 9). Other

(30)

29

available color coding includes Moment 1, Intensity, Width, and Threshold, but none of these modes is analyzed during this project.

Figure 9 Schematic representation of the ”z coordinate” and ”moment 0” color codings. It is possible to note how the ”moment 0” representation provides a realistic image of the object. Examples provided by Senso Test.

3.3.2.3. Measuring Range and Measuring Fields

The measuring range of the scanning device, namely the spatial range in which the scanner is able to detect a surface, allows the detection of objects at a distance from 190 mm to 290 mm from the scanner. Additionally, an extended measuring range allows also for detection of objects as close as 125 mm from the scanner and as far as 390 mm.

The projection of the laser line from the scanner is done with an opening angle of 21.4° (Figure 10), which causes the measuring range to be trapezoidal. As a consequence, the length of the blue laser line changes with the distance of the scan to the scanned object.

In the processing of the signal, the measuring range is mapped on a rectangular matrix with a grid spacing of 1/8 of the matrix. The laser settings can be selected in a way in which not all the measuring range is “active” and only the selected portion of it is actually “reading” or “recording”. This selection of a specific region of the measuring range is carried out by selecting a measuring field, i.e. the portion of the matrix that will be active on the scanning. Figure 11 shows the way in which some of these measuring fields are defined. A total of 127 different measuring fields can be selected. For example, if a measuring field of 4 is selected (Figure 11), the scanner will not record any information of this part, and the error “No points in region of interest”

will be displayed even if there is some object in the lower part of the measuring range.

The selection of a smaller measuring range allows to a higher frequency for data collection, which in turns means a higher amount of profiles per second. For example, with a standard measuring field of 2 a maximum of 190 profiles can be collected per second, while when selecting a measuring field of 100, a maximum of 1030 profiles can be collected per second.

(31)

30

Figure 10 ScanCONTROL 2960-100/BL measuring range. Dimensions in mm. Extracted from (31).

Figure 11 Examples of measuring fields defined in the sensor ScanCONTROL 2960-100/BL manual.

The selection of a proper measuring field is an important parameter in the design of the set-up for scanning in the plant because it does not only influences the data collection frequency (and hence the maximum scanning speed) but also modifies the actual measuring range, making invisible portions of the measuring range that would be visible with other measuring fields.

(32)

31

3.3.2.4. Resolution

In the manual of the scanning device, a reference resolution of 12 µm is mentioned.

Nevertheless, a simple analysis of the scanning conditions can easily show that the resolution both in the direction of the laser line (X direction) and in the scanning direction (Y direction) are strongly affected by the scanning parameters.

In the direction of the laser line, the length of the laser projected over the surface will vary according to the distance of the surface to the scanner. At the same time, the blue laser line is divided, for the analysis of the recorded information, on a maximum of 1280 points (80, 160, 320, 640, or 1280). As a consequence, the distance between two consecutive points in the x direction will vary depending on the distance of the object to the scanner (different distance divided into the same amount of points).

With respect to the resolution in the scanning direction, a combination of parameters has to be considered. The distance between consecutively collected profiles (resolution) in the y-direction (dy) is determined by the scanning speed (vs) and the number of profiles collected per second (pcf). Different combinations of scanning speed and profiles collected per second can lead to the same resolution. On the other hand, a constant scanning speed with a higher amount of profiles collected per second invariably means a higher resolution in the same way in which a constant amount of profiles per second with a smaller scanning speed leads to a smaller distance between consecutive profiles.

𝑑_𝑦 = 𝑣_𝑠 𝑝_𝑐𝑓

The maximum profile collection frequency (pcf) for the ScanCONTROL 29xx-100/BL has a maximum value of 2000 Hz. Nevertheless, the actual collection frequency that can be selected during the operation of the equipment is also dependent on the measuring field. The larger the measuring field, the smaller is the maximum allowed profile collection frequency. Figure 12 shows some examples of maximum operative collection frequencies for different measuring fields.

(33)

32

Figure 12 Maximum collection frequency for some of the defined measuring fields. Note that the smaller the measuring field, the higher the allowed collection frequency.

It is also important to note that the equipment assumes a constant scanning speed for the analysis of the data. As a consequence, any local variation on the scanning speed or vibration will generate a region of the point cloud that is not accurately representing the reality. That is to say, a momentarily faster speed will result in points that are more separated in the virtual representation than in the reality. At the same time, a vibration in the z or x-direction will also affect the results and will have an effect on the resolution of the technique. In order to achieve the best results, the sample or the equipment that is moving must be as free as possible of vibrations and have a steady movement with a controlled speed.

Another important matter related to the resolution is the maximum amount of profiles that are buffered and the size of the sample that is being scanned. If the full size of one sample is to be scanned in only one file, the relation between the scanned length and the maximum buffered profiles (limited by the computer used in the scanning, see section 3.3.2.2 Preview Mode and Color Coding) determines a maximum resolution. In this sense, the scanning strategy is of primary importance, since it can strongly influence the resolution.

3.3.2.5. Exposure Time

The exposure time is a relatively simple parameter that controls the amount of time that the sensor is exposed to light. If the exposure time is too short some areas of the surface will not be visible for the sensor while a long exposure time limits the number of profiles that can be collected per second. A good balance can be achieved by using the "auto exposure" tool in the advanced settings, but the indirect effect of the profiles per second must be recognized. For new (not previously evaluated) surfaces, the testing of different exposure times is recommended.

190 280 550

(34)

33

3.4. Metallographic Techniques

Metallography is the most widely used technique to reveal and study the internal structure of metals and alloys (32). It basically consists of the etching of a metal surface with an appropriate reagent to reveal microstructural features that can be observed either with the use of a microscope or with the naked eye. It is a usual procedure to control the internal quality of steels and alloys in any production line as well as a tool to analyze and understand the solidification structure.

Generally speaking, the first step of the process is the cutting of a sample from an area that is significant for the analysis of the structure. Following steps include mounting (optional), grinding, and polishing until a mirror-like surface is obtained.

The final stage before the observation in the microscope is usually the use of a chemical or electrochemical reagent to reveal the microstructure. Although the reagents that are used to reveal the microstructure are usually called etchants, they do not always refer to the selective chemical dissolution of various structural features.

Another frequent case is found, for example, with the color etching, which in fact consists of the use of a chemical mixture to deposit an interference layer on the metal surface (33). Other contrast techniques that may not involve etching include the use of dark field illumination or polarized light (33).

Another general distinction must be made between macro and micro-etching. A micro-etching is carried out when the microstructural features such as precipitates, phases, phase morphology, phase fraction, grain size, etc., are to be observed or quantified (32). On the other hand, macro etchants are used to reveal the structure on a macro scale when features such as the chill zone depth, the size of the columnar or dendritic grains, and the size of the equiaxed zone, are of prime interest. In addition, the presence of macrosegregation and macroporosity can also be observed (32).

In this project, both macro and microstructural analysis were carried out in order to correlate the presence of surface defects with the subsurface structure. In particular, the presence of depressions in the surface of one particular stainless steel grade was analyzed and the focus was placed on the subsurface grain structure and characteristics of the phases that are present.

4. Experimental Procedures

In order to prove the applicability of the scanning system in industrial conditions, pilot trials both at small and full scale were performed. The experimental work can be divided in 2 different stages:

1) Small scale test – as a first approximation to analyse the effect of scanning parameters in the scanning results.

(35)

34

2) Full scale test – in order to prove the applicability of the scanning technique to full size slabs and billets

2-a) Room temperature trials 2-b) High temperature trials

4.1. Small Scale Tests

4.1.1. Apparatus and Material

The main purpose of the small scale trials was to determine the effect of the scanning parameters on the resolution of the equipment. For this purpose, the scanner device was mounted in a milling table (Figure 13), which offers a precise control of the displacement of the sample and a constant speed movement. During the tests, the equipment remained stationary while the sample was moved at a constant speed with the milling table.

The material used for the small scale trials was a piece of slab, namely the narrow face of a high strength low alloy steel slab, of approximately 200 x 200 x 50 mm (Figure 13). It has various superficial defects such as scale, oscillation marks and small cracks. According to SEM images of a previous work in which a similar material was analyzed (34), the mouth opening of the corner cracks is in the order of ~200 μm. For this reason, and because the small size facilitates handling of the material, this piece was selected as sample for the small scale trials.

Figure 13 Set up for the small scale tests. The equipment is placed in a cooling jacket (prepared for high temperature tests) and installed in a milling table arm. Test sample is placed in the milling drill table.

Laser scanner

Sample

(36)

35

4.1.2. Procedure

The scanning procedure includes variations in the scanning speed and in the amount of profiles collected per second in such a way that the resolution in the scanning direction can be controlled. Table 1 presents the summary of scanning parameters that were used during the small scale trials. It can be noted that the variations on scanning parameters effectively lead to a variation in the resolution (distance between profiles). In addition, the use of higher amount of profiles collected per second determines the need for different measuring fields (see section 3.2.2 Measuring Field and Measuring Range).

Table 1 Scanning parameters used during the small scale trials

Scan Number

Profiles per second

Exposure time (ms)

Scanning mode

Measuring Field

Scanned distance (mm)

Scanning speed (m/min)

Scanning speed mm/s

Distance between profiles (mm)

1 80 1 z 2 188 0.2 3.4 0.043

2 80 1 z 2 193 0.2 3.5 0.043

3 80 1 z 2 186 0.2 3.4 0.042

5 400 1 z 7 195 1 16.6 0.042

6 400 1 z 7 191 1 17.4 0.043

7 400 1 z 7 198 1 17.0 0.043

8 50 1 z 2 193 1 17.5 0.351

9 50 1 z 2 196 1 17.0 0.341

10 50 1 z 2 196 1 17.3 0.347

11 280 1 z 4 204 1 16.7 0.060

12 280 1 z 4 206 1 17.0 0.061

13 280 1 z 4 203 1 17.5 0.062

14 280 1 z 4 205 1 17.1 0.061

4.2. Full Scale Tests 4.2.1. Apparatus

The full scale trials were performed in order to analyze the surface characteristics of full size slabs as a step forward to the determination of the viability of the system to be installed in a steel plant, where full scale means that the samples analyzed have the

(37)

36

dimensions of real billets and slabs. Both room and high temperature trials were performed at Swerea MEFOS, the latter with the special purpose of proving the operability of the laser equipment in the vicinity of high temperature surfaces such as those encountered in the casting machine of a steel plant. For this purpose, the scanning system was installed in a grinding machine, usually used to perform grinding tests (Figure 14 and 17). This machine has the possibility to control the movement of the working table and also to control the movement of the arm in which the laser is installed. Figure 15 presents a top (a) and lateral (b) view of the elements during the scanning of the samples. For the top and bottom scans on the slabs and billets, the equipment was stationary while the sample was moving on the grinding table. On the other hand, for the scan of the lateral faces of the slabs, the sample was stationary while the sensor was moving (Figure 16).

Figure 14 Schematic representation of the grinding machine used for the full scale trials. Provided by Swerea MEFOS.

Figure 15 Top and lateral schematic view of the elements during the scanning of the top surfaces of the samples.

(38)

37

Figure 16 Schematic view of the elements during the lateral scanning of the samples.

Figure 17 Image of the grinding table were the blue laser is installed.

4.2.2. Materials

The materials used for the full scale trials include three stainless steel slabs cast at Outokumpu Stainless (Avesta) and two carbon steel billets cast at Sidenor. The steel grades and dimensions of all samples are summarized in Table 2, while Figure 18 presents some examples of surface defects in the samples.

Sensor

Moving arm

Grinding table

(39)

38

Table 2 Grades and dimensions of the steel slabs and billets used for the full scale trials

Sample Steel grade W (mm)

L (mm)

T (mm)

A-1 LDX 2101 1150 500 200

A-2 LDX 2101 950 300 200

A-3 304 2050 300 140

S-1 Microalloyed 240 1000 240

S-2 Microalloyed 240 1000 240

Figure 18 Images of some of the defects that were present on the surface of the steel products. Images in the right side show a depression in the inner bow (wide face) of a stainless steel slab, with a sharp edge at approximately 120 mm from the border (from the narrow face). Images in the left show a depression in the outer bow (wide face) of the same stainless steel slab, where the sharp edge is not present, but a more continuous curvature can be seen. In the interior of both depressions some oscillation marks are visible.

4.2.3. Procedure

4.2.3.1. Room Temperature Trials

Before any of the scans were performed, the surface of all the samples was cleaned using a steel brush and a vacuum cleaner to remove scale and dust. Once the samples were cleaned, each specimen was placed on the grinding table for the scanning. A working distance of 240 mm was used for all the scans and an overlapping of 25 to 50

(40)

39

mm was allowed between consecutive scans. Table 3 shows the general scanning parameters that were used during the room temperature full scale trials.

Table 3 Example of scanning parameters employed during the room temperature full scale trials

Working distance 240

Scanning speed in the control machine 1 m/min Y displacement respect to previous scan 50 mm

Profiles per second 200

Scanned profiles (buffered profiles) 12250

Exposure 1.75 ms

Scanning mode Moment 0

Measuring field 97

Distance between profiles ~0.08 mm

4.2.3.1.1. Billets

For the scanning of the billets, the sample was placed on the grinding table with the scanning direction parallel to the grinding table longitudinal axis. The scanning direction was then parallel to the casting direction and the scan was generated by the movement of the grinding table (Figure 15). Consecutive scans were performed every 50 to 75 mm, and the sample was rotated and re-scanned until the total surface of the billet was evaluated.

4.2.3.1.2. Slabs

For the scanning of the slabs, 2 different situations can be distinguished. For all slabs, the wide faces were scanned with the casting direction perpendicular to longitudinal axis of the grinding table and the scan was generated with the movement of the grinding table (Figure 15). For one duplex stainless steel sample, the narrow faces were also scanned in order to obtain a complete representation. In this case, the sample was stationary and the scan was generated by the movement of the laser senor in the arm of the grinding machine (Figure 16).

4.2.3.2. High Temperature Trials

In order to test the ability of the sensor to scan steel surfaces at high temperatures;

such as those encountered in the casting machine, the steel samples were heated up to 1200°C and scanned with the Blue laser device. The heating of the samples was performed in a bell furnace under nitrogen atmosphere, both for carbon and stainless

(41)

40

steels, in order to minimize the surface oxide formation. In addition, an air cooling jacket was built at Swerea MEFOS in order to protect the scanning system from the high temperatures and radiation from the hot surface.

The scanning procedure was similar to the room temperature trials in the sense that the samples were placed on the grinding table and scanned with the movement of the table, or with the movement of the arm for the lateral scan of one stainless steel sample (See sections 4.3.2.1.1 Billets and 4.3.2.1.2 Slabs). Nevertheless, in order to keep the temperature of the slabs as high as possible during the test, each sample had a reheating period in the furnace in between the scanning of each face. For example, a typical scanning routine at high temperature would include 2 samples in the furnace.

While the first one is being scanned, the second one remains in the furnace. After one of the sides of the first sample is scanned, sample 1 is taken back to reheating and the first side of sample 2 is scanned. When the scanning on the first side of sample 2 is finished, the material is taken to the furnace for reheating and the scanning continues in sample 1. Table 4 presents an example of the scanning parameters used during the high temperature trials.

Before the first scan was performed, all samples had a heating period of at least 16 hours (samples were placed in the furnace at least at 4:00 pm the previous day to the tests). This time is considered more than sufficient for the material to reach the furnace temperature. Considering that a general rule of thumb is 1 hour per inch and that the thicker section is 240 mm, 10 hours would have been the minimum soaking time.

Table 4 Example of scanning parameters employed during the high temperature full scale trials

Working distance 240

Scanning speed in the control machine 2 m/min Y displacement respect to previous scan 50 mm

Profiles per second 330

Scanned profiles (buffered profiles) 12250

Exposure 3 ms

Scanning mode Moment 0

Measuring field 97

Distance between profiles ~0.1 mm

Advanced Inspection of Surface Quality in Continuously Cast Products by Online Monitoring