Application of Computational Fluid Dynamics to Steelmaking Processes

Transcription

1 Application of Computational Fluid Dynamics to Steelmaking Processes J. Schlüter, J. Kempken, H.-J. Odenthal, M. Reifferscheid, N. Vogl (SMS Demag AG) INTRODUCTION The processes in steel production are so complex that it is hardly possible to obtain any deeper insights into the flow related effects and chemical reactions. In the operational test, chemical and physical processes can be reproduced under certain preconditions. However, most process variables can not be measured due to the extreme boundary conditions and the lack of measuring technologies on the one hand and for safety reasons on the other. If the focus is to be on aspects of flow technology, the physical simulation is performed on cold water models providing important information on the flow phenomena. It is not possible to simultaneously fulfil all similarity conditions in the models which are mostly on a reduced scale. Thermal effects are also difficult to take into account. Due to the tremendous increase in computing power, metallurgical processes are currently be analysed with the help of the Computational Fluid Dynamics (CFD) and Finite Element Method (FEM). It is possible to simulated three-dimensional, transient, non-isothermal multiphase flows including particle or bubble interaction, phase changes, electromagnetic forces in fluid flows and chemical reactions such as combustion, reduction and oxidation. The gained insight into the flow phenomena makes the process more transparent and supports the design activities of new products. Metallurgical applications of CFD/FEM are widespread, e. g. direct reduction and electric arc furnaces, converters, dust collecting facilities, continuous and thin strip casters, strip processing lines as well as rolling and metal plants. At SMS Demag, applicationoriented numerical simulations are carried out. For this purpose, well-known and already validated submodels are used, by means of which practical solutions for the design and operating performance of plants for the steel and NF-metal industries are prepared. Likewise causes for disturbances in operation can be found. In the following, several examples for successful CFD/FEM solutions are described. BOF CONVERTER SMS Demag has developed a CFD model for the combined blowing BOF process. The model is able to rate the influence of the melt domain width-to-heightratio, the oxygen blowing and argon stirring rate, the top lance position as well as the number, type and location of the bottom tuyeres on the melt flow behaviour. Due to computing time reasons, the mixing time and homogenization time respectively, can only be simulated for pure bottom stirring. The results for a 335 t BOF converter with six-hole top lance and 14 bottom tuyeres show a varying penetration depth of the oxygen blowing cavities up to 0.5 m, Fig. 1. Due to a low lance position of one meter, the cavities overlap and interfere 1,2,6. Figure 1: 335-ton BOF converter with six-hole top lance, V & ox = 1100 m 3 /min (STP), and bottom stirring, V & ar = 650 m 3 /h (STP); velocity distribution inside the melt domain (yellow), slag layer (red) Towards the end of blowing, when decarburization progresses slowly and the formation of CO bubbles required for thorough mixing decreases, the bottom tuyeres play the major role for the homogenization and speed of reaction. If the tuyeres are put into operation additionally, the cavities will change as a result of the rising gas bubbles. When the oxygen jets come into contact with the plumes on the free surface, the depth of the cavities is reduced and the oscillation of the melt funnels is increased. Thus, top lance and porous plugs must always be aligned to each other in such a way that they support the melt movements induced by them 8. This ensures stable and reproducible blowing patterns 7,9,10. However, the BOF-CFD model has got some restrictions, i. e. missing chemical and foaming processes, which will be reworked in the near future. Concerning the blowing process, it is known that each top lance is designed as a function of the process data (blowing rate, pressure and temperature at the lance entry, number of nozzles, converter back-

2 pressure) to fit a single operating condition. If the process data deviate from this condition, the lance works inefficiently. Pressure and velocity disturbances in form of shock waves (over-expanding jet) or expansion waves (under-expanding jet) arise inside and outside the Laval nozzle. The jet radially spreads and the supersonic jet length decreases. These phenomena lead to a high wear rate of the lance. Fig. 2a is a cutaway view of a six-hole blowing lance after approx. 200 converter heats. The original outline of the nozzle with the outlet diameter d 1 is shown. Variables d 1 and d* denote the inlet and throat of the nozzle. More severe wear can be clearly seen towards the centreline of the blowing lance. A line of unsteadiness is formed at the diffuser outlet where the flow is separated. a) Six-hole top lance after 200 converter heats detail A detail B b) Simulated influence of the lance wear Ma recirculation new nozzle worn nozzle Figure 2: a) Cut through the centre plane of a six-hole top lance; b) Simulated influence of the lance wear on the Ma number distribution at the design point Fig. 2b shows results for the design condition (p 0 = 8 bar, V & ox = 650 m 3 /min STP ) both of the new and worn nozzle. While in the first case the flow leaves the diffuser barely causing any disturbance, it is disrupted at the separation line in the second case. A local recirculation area is formed in the diffuser outlet. The outlet diameter d 1 is now located inside the nozzle and no longer matches the theoretical outlet pressure p 1. Hot converter gas enters the nozzle and reaches the copper wall via the recirculation area. The cooling effect is reduced and local wear as well as the risk of a water breakthrough rise. At the same time the length of the supersonic jet decreases. Once wear has started at a certain location inside the diffuser, this region will be increasingly charged by hot converter gas in the further course of blowing operation. While underexpanding nozzles are uncritical in terms of wear, the service life of over-expanding nozzles, i. e. those which are operated at low supply pressures, decreases rapidly due to the absence of cooling. Operational tests show that more severe wear is found towards the centre axis of the top lance. Currently, SMS Demag is working on design solutions to increase the service lives of oxygen top lances. AOD CONVERTER A further application of CFD to steelmaking processes is the AOD technology. The melt flow conditions, i. e. the melt level, process gas type and flow rate, design, number and arrangement of submerged nozzles and converter tilting angle, play an important role for the vibration level. It is the main objective to detect vibration reasons, predict vibration levels and determine safe and efficient operating conditions. The optimal/ maximum metal level as well as the process gas flow rate has to be known for each AOD process just as the best nozzle design and converter position. Thus, SMS Demag carries out operational tests, physical and numerical simulation. For example, the one-fourth scaled model of a 135 t AOD converter with up to seven sidewall nozzles has been set up at the Institute for Industrial Furnace and Heat Engineering of the RWTH Aachen University. Innovative measurement methods such as the laser-light-sheet, Digital Particle Image Velocimetry (DPIV) and Residence Time Distribution (RTD) techniques in combination with wave height and vibration measurements are used. As an example, Fig. 3 shows results of laser-lightsheet visualisations in the centre plane of the water model with varying filling level. According to the Froude criterion, the air flow rate of V& 3 air = m 3 /s (STP) is corresponding to a total oxygen blowing rate of V& ox = 117 m 3 /min (STP). Both, water model and CFD simulation emphasize that the penetration depth of an air jet into the water domain is low 5. The bubbles rise on the wall of the sidewall nozzles, inducing a large-space vortex with clockwise rotation. The size and position of the vortex that ensures intensive mixing remain stable in terms of time. On the free surface, high amplitudes arise in the region of the plumes. The highest amplitudes are found immediately after the start of air injection. The plumes move back and forth creating a transient water level with an intensive low-frequency. Close to the wall of the nozzles, high pressure and velocity gradients are the

3 reason for high wall shear stresses. In general, the filling level affects the flow structure as well as the vibration level. The higher the water level the more intensive the vortex and the vibration become. a) t 0, γ = 0 b) t 1, γ = 5 (tilted converter) Figure 3: Laser-light-sheet visualisation in the centre plane of the one-fourth scaled AOD water model with varying filling level; V & air = m 3 /s (STP) y Position of rest Fig. 4a shows the calculated velocity distribution and the melt movement for the 135 t converter filled with melt. The converter axis (black line) and the position of the sub-lance (blue line) are marked. The pressure colour map does not show individual bubbles, but an iso-volume covering the respective bubble plume and containing the melt phase in which the velocity is in the range of 0 u 1 m/s. The velocity vectors for the melt and gas phases and the corresponding colour map are shown in the central cross-section of the AOD converter. At t 0 the converter is in vertical position, and at t 1, as is common practice, it is inclined by γ = 5 in clockwise direction. The intention is to move the rising argon bubbles away from the refractory wall and thus to reduce wear. In steady-state operation the amplitude of the melt level fluctuations around the static position of rest is approx. y = ± m. The bubbles adhere to the converter wall and move back and forth. If the free surface moves upward, also upward-directed velocity components are induced in the melt metal and vice versa. In the case of the converter which is inclined by γ = 5, Fig. 4b, the bubbles move slightly away from the wall where the nozzles are located. Yet the surface movement becomes more intensive, since the plumes have more space for moving back and forth. The mean melt-movement amplitude of the free surface around the static position is y = ± 0.15 m and is thus twice as high as in the upright converter. First CFD results reveal that the wall shear stress rises in the case of the inclined converter. axis Figure 4: 135-ton AOD converter C: Velocity distribution and melt movement for V& ar = 117 m 3 /min (STP), n tuy = 7 CASTING TECHNOLOGY In the following the focus will be put on the casting of steel. The processes that take place during the solidification in a continuous casting mould are highly complex. The main issues are (Fig. 5): Transient fluid-dynamic effects of several phases. Multidimensional thermal conduction. Diffusion of dissolved substances in steel. Various chemical reactions. Phase changes in the casting flux and steel. All these effects occur at the same time and mutually affect each other. Additionally, an attempt is made to influence the flow through electromagnetic fields. All processes cannot within a foreseeable period of time be dealt at the same time, as generally accepted models for sub-processes do not yet exist. On the other hand, connecting the numerous sub-models would require disproportionately higher computing power, thus making it expedient to only use submodels for the questions that arise in plant design and construction.

4 Figure 6: Vortex formation in the meniscus of a standard submerged entry nozzle Figure 5: Conventional/CSP caster-flow phenomena Currently, the flow at the meniscus is primarily being taken into consideration for the design and development of submerged entry nozzles. Decisive for a reliable casting operation and a good quality is a steadystate inlet flow in the mould within a defined speed range, with strong mould level fluctuations to be avoided. This partial aspect of all the complex operations taking place in the mould level and during solidification can be reproduced in a full-scale water model. This type of test has been carried out for many years. The velocity inside the mould is either measured by laser-optical processes or determined by measuring the surface camber. Today it has become possible to determine these effects by CFD because, previously, calculations within an acceptable cost frame could only be made for steady-state flows. The fluctuations in the mould level could therefore not be assessed. About five years ago, CFD simulations of the instabilities in the mould level became possible as well within an acceptable time frame and thanks to more sophisticated computing procedures and powerful computers. Based on a conventional mould, the customeroriented process optimization of a SEN is explained here as an example. In Fig. 7, the velocity vectors in the symmetry plane, in two planes parallel to the narrow sides and in the free melt surface level are shown for two different SEN designs. The strand dimension in this simulation is 2150 mm x 220 mm with a casting speed of u cast = 2.2 m/min. For these casting conditions the standard SEN shows surface velocities higher than 0.5 m/s, which is the maximum allowed value. The improved design shows much lower velocities at the surface resulting also in much smaller surface fluctuations. The mould flow is inherently instable so that is impossible to eliminate the fluctuations. They can only be minimized for a certain combination of casting speed and strand geometry. Conventional mould Focussing at the free surface of the liquid steel in the mould, the fluid properties of melt are used but solidification is not taken into account. Moreover, the slag and casting flux in the mould level have been neglected. In Fig. 6, undesired surface swirls that transport the slag from the phase boundary surface into the strand can be clearly recognized. Figure 7: Customer-oriented process optimization of a submerged entry nozzle, 2150 mm x 220 mm, u cast = 2.2 m/min CSP caster The benefit of a CFD simulation can be seen when in a more detailed model of the strand shell is included

5 in the calculation. In Fig. 8, the calculation of the melt flow during start of cast of a CSP caster is presented. The strand has a width of 1500 mm and a thickness of 60 mm. The casting speed is u cast = 5 m/min. The temperature distribution in the symmetry plane of the strand is shown at different process times. The highest temperature approx. 20 C above the liquidus temperature of the melt is marked red, the temperature lower than the solidus temperature is marked blue. The start of cast is shown in the left part of Fig. 8. At the beginning, the mould is filled with melt of a uniform temperature and the remaining domain is fully solidified with a temperature of 25 C. In the following, the solidified parts are moving downwards with the casting speed. In the right part, the behaviour of the flow at quasi-steady-state conditions is shown. It is notable, that the solidification front is fluctuating due to the mould flow instabilities. The position of the solidification front moves up and down over a length of 0.5 m. CFD calculations are the basis for the understanding of the interaction between solidification and melt flow. The time dependent geometry and the temperature distribution of the strand shell can be analyzed due to thermal stresses and conclusions on the segregation and bulging behaviour can be drawn. process which directly influences the design of the caster. This knowledge is a precondition for the design of an automation control system as well. The customers can be sure to have a state of the art SEN design and are not restricted to buy SEN s from one special manufacturer. Thin strip casting Process optimization of feeding system Fig 9 shows an application of CFD to the Belt Strip Technology (BESTT) is demonstrated. The melt is poured from above onto a belt that moves at casting speed via a distribution system. The heat is withdrawn via a spray cooling system arranged under the belt. Having passed a secondary cooling zone and a thermal balance zone, the solidified steel strip with a thickness of 8 mm to 15 mm is rolled right away and coiled after intensive cooling. A test unit has been in operation for a few years at the Institute of Metallurgy of the Technical University of Clausthal, Germany 11. In addition to the low cost of the equipment, this casting process features the following advantages: Casting without mould powder. The thin strip is not subjected to any bending stresses. The process is suitable for casting stainless steels and carbon steels. The caster allows a high flexibility in terms of casting thickness and speed Sufficient casting thickness may be obtained to allow hot deformation. Figure 8: CSP caster simulation of solidification The simulation of the mould flow and solidification process gives a better understanding of the entire Figure 9: Belt Strip Technology (BESTT); pilot plant at Technical University of Clausthal, Germany The pouring of the melt onto the casting belt is a typical flow-related problem. The entire casting width up

6 to the side limiters must be filled with melt. The melt flow behaviour has been calculated using the CFD software FLUENT including a solidification model and a special turbulence model to describe the effects of the low Reynolds number flow. The free surface of the melt was calculated time dependently. Since for design reasons the feed line must be narrower than the width of the casting belt, external forces have to ensure that the melt is filled in completely up to the side limiters. This can be accomplished by electromagnetic stirring. The electromagnetic fields of the stirrer move at a speed which depends on the stirring frequency and the distance between the poles of the linear motor. Since FLUENT is unable to describe electromagnetic moving fields, the FEM software ANSYS was used. Since ANSYS offers special elements which provide the calculation of the movement with a fluid, the moving melt can be simulated as well. The current flow that excites the magnetic fields was defined in the area of the coils as volumetric-flow density with other elements. Any feedback of the induced magnetic fields on the current flow in the coils was neglected. Because of the strong interaction between melt flow and magnetic fields the calculation has to be performed coupled. The flow was calculated with FLUENT and the electromagnetic forces are calculated with ANSYS. Information between the programs was exchanged by files. The entire simulated domain consisting of the fluid regime and the electromagnetic stirrer can be seen in Fig. 10. application of external forces the casting with a defined width is not possible. With the help of the stirrer, it is possible to affect the melt flow on the casting belt such that the flow is distributed over the entire casting width, Fig 11b. The forces induced by the stirrer can be seen in the right part of the figure. Due to the influence of the copper side dams, no forces are induced in the middle part and near the side dams. Operational tests gained with the stirrer confirm these results. a) Melt flow conditions without stirrer b) Melt flow conditions with stirrer Figure 11: Belt Strip Technology (BEST); FEM/CFD simulation of the melt flow conditions Figure 10: Geometric boundary of the FEM/CFD domain The simulation results without the use of a stirrer are illustrated in Fig. 11a. The calculation was performed for a casting speed of u cast = 10 m/min, a casting thickness of 15 mm and a width of 300 mm. The free surface of the melt is coloured by the velocity. On the right side of the figure, velocity vectors are added to give a better indication of the flow direction. The abrupt expansion of the channel in the area of the ramp causes the flow to break away from the side wall and in addition, due to the surface tension and the rising speed, also restricts the flow. Thus, without the CONCLUSIONS Owing to the progress in computer technology, even complex metallurgical flow processes can be successfully analysed with the help of CFD and FEM. In this presentation the current potential of both methods was demonstrated by reference to converters, conventional and CSP casters as well as the belt strip technology. Selected numerical models of commercial software codes are used at SMS Demag to compute fundamental phenomenological flow processes and make them more transparent. These newly developed integrated models are capable of simulating timedependent, non-isothermal and multi-phase flows including electromagnetic forces in order to improve

7 the metallurgical results. Thus, further process optimization in terms of design and plant rating are derived by improving the operating conditions. A high plant safety and operational reliability can be guaranteed. Customized solutions are offered emphasizing the high competence of SMS Demag technologies. LITERATURE [1] Balkos, T., Batham, J., Russo, T., Fash, R., Howanski, B.: Cold shroud boosting converter performance, 6 th European Oxygen Steelmaking Conference, Aachen, Germany, 2006, p [2] Deo, B., Boom, R.: Fundamentals of steelmaking metallurgy,1 st ed., Prent. Hall Intern., [3] Fabritius, T.M.J., Kurkinen, P.T., Mure, P.T., Härkki, J.J.: Vibration of argon-oxygen decarburisation vessel during gas injection, Iron and Steelmaker 32 (2005) No. 2, p [4] FLUENT "FLUENT User s Guide", [5] Hoefele, A., Brimacombe, J.K.: Flow regimes in submerged gas injection, Metallurgical Transactions B, 10B (1979), p [6] Koria, S.C., Lange, K.W.: Penetrability of impinging gas jets in molten steel bath, steel research int. 58 (1987) No. 9, p [7] Lachmund, H., Bruckhaus, R., Fiedler, V., Xie, Y.: Optimisation of the BOF process after replacement of vessels with a different geometry, stahl u. eisen 123 (2003) No. 11, p [8] Luomala, M.J., Fabritius, T.M.J., Härkki, J.J.: The effect of bottom nozzle configuration on the bath behaviour in the BOF, ISIJ Intern. 44 (2004) No. 5, p [9] Odenthal, H.J., Falkenreck, U., Kempken, J., Schlüter, J., Uebber, N.: CFD simulation of melt flow mixing phenomena in combined blowing converters, 6 th Europ. Oxygen Steelmaking Conf., Aachen, Germany, 2006, p [10] Odenthal, H.-J.; Emling, W.H.; Kempken, J.; Schlüter, J.: Advantageous numerical simulation of the converter blowing process, Association for Iron & Steel Technology Conference, Indianapolis, USA, , p [11] Spitzer, K.H., Rüppel, F., Viscorova, R., Scholz, R., Kroos, J., Flaxa, V.: Direct Strip Casting (DSC) - an option for the production of new steel grades, steel research 74 (2003) No. 11/12, p