Prediction of slab quality through metallurgical modeling.

Careful monitoring of the process parameters influencing quality during continuous casting is indispensable for finding the origins of slab defects and for continuous quality improvement.

Additional parameters influencing quality can be derived by applying computational metallurgical

models. These parameters describe the casting process more precisely and show good correlations with the actual defects. Such computational modeling is likely to play an increasing role in generating deeper process understanding as advances in both computer hardware and software are making the modeling tools more powerful [1].

Work carried out on computational modeling in a joint project between Austrian steelmaker voestalpine and plant builder VAI tested the computer models for continuous casting and were used online for quality prediction.

APPLICATION 1 ALUMINUM NITRIDE PRECIPITATION

Transverse cracking is a serious problem in the continuous casting of steel, and has a detrimentally affect the final product quality. These surface defects are thought to form mainly as the strand is straightened [2] from the vertical plane as it exits the mould to the horizontal run out table Fig 1 shows typical pictures for this kind of defect.

[FIGURE 1 OMITTED]

The cracks are intergranular and meander along the prior austenite grain boundaries. The straightening operation is performed in the temperature range between 700 to 1000[degrees]C, which coincides with the interval in which steel exhibits a minimum in ductility as demonstrated in laboratory hot tensile tests which shows a fall in reduction of area (RA) in this temperature range. Increasing the N or AI content of the steel extends both the depth and width of the low ductility trough in such tests due to increasing precipitation of aluminum nitride (AIN) which supports microvoid coalescence.

More detailed investigations have shown a strong influence of nitride--and carbonitride precipitation kinetics on the hot ductility behaviour of steels. Especially:

--particle density;

--particle size; and

--amount of precipitation.

The shape and appearance of these ductility curves are determined by the chemical composition of the steel and its cooling history [3]. For example, coarse precipitation is less harmful than fine precipitation of an equal volume fraction.

The target of the project was to develop a computer model capable of describing precipitation kinetics during casting steels containing Al, V, Nb nitrides and carbonitrides [4]. This model provides quantitative information on the nitride and carbonitride precipitation on each individual cast. Consequently the prediction of the surface quality of the strand is more reliable.

By means of a nucleation/growth model [5,6], density, size and total amount of particles per volume are predicted as a function of the concentration of micro-alloying elements and the temperature profile. The model introduces a thermodynamic approach to determine precipitation parameters taking into account the mutual interaction between the alloying elements and dissolution of precipitates as a consequence of strand reheating. Zener's steady state approximation was employed for the description of the growth of spherical particles [7].

MODELING PRECIPITATION PARAMETERS

The main aim of the model is to improve product quality and quality prediction. Therefore correlations between quality and precipitation parameters have to be established. Data recorded by VAI on caster start-ups was used to reconstruct thermal histories of slabs where surface defects had been observed and correlated to A1N precipitation. From this, the behaviour and kinetics of precipitation were determined. Low carbon aluminum killed peritectic grade steels were selected based on the assumption these offered stable casting conditions. By employing VAI's in-house developed computer software the surface temperature history was calculated as well as the precipitation parameters:

--particle density;

--particle size; and

--number of particles.

RESULTS

A typical result of such a simulation of precipitation kinetics for a continuously cast slab containing 60ppm N and 450ppm Al is shown in Fig 2 in the form of a calculated Precipitation-Time-Temperature (PTT) diagram. The surface temperature of the slab is superimposed on the precipitation curve as well as the region where straightening of the slab takes place. The straightening operation coincides with a maximum precipitation rate. Grain boundaries weakened by precipitates are exposed to a mechanical deformation processes and, as a consequence, surface cracks occur.

[FIGURE 2 OMITTED]

The results of a statistical evaluation of calculated precipitation and quality data is presented in Fig 3. Mean values and standard deviation of precipitated A1N are indicated for good and cracked samples. Whereas defective slabs had precipitated approximately 260ppm A1N as a mean value, slabs free of defects had precipitated only a mean of 140ppm. A strong correlation is evident.

[FIGURE 3 OMITTED]

This model was implemented at the caster process computers and integrated with the existing computer aided quality control system [8]. The benefit was 9% less scarfing of slab ends compared to the previous rate.

It should be noted that prediction of surface cracks by this model is limited to cracks caused by precipitation. There are also several other metallurgical phenomena leading to this kind of defect, one of which is pro-eutectic ferrite precipitation [9].

In conclusion:

--Ductility of Al-killed steels is affect ed by AIN precipitation.

--Additional extended models have been developed to including the precipitation of Niobium carbonitrides and Vanadium nitride for the analysis of microalloyed steels.

--Surface quality shows a dependency on the precipitation parameters, density, size and quantity.

--Crack prediction seems possible and more effective using the physical A1N precipitation model.

--Pro-eutecfic ferrite precipitation is not covered by the present model and will be the next step of development.

APPLICATION 2 MOULD POWDER ENTRAPMENT

One of the quality problems encountered for low carbon steel and IF steel originating from the continuous caster is entrapment of mould powder slag where it is trapped near the surface of the slab and in the vicinity of hooks [10] formed by the strand shell due to mould oscillation. Fig 4 shows such an inclusion on an etched sample of slab [11].

[FIGURE 4 OMITTED]

It is assumed that hook depth as well as the number of entrained slag particles both influence the frequency of the defect occurring on the coil, together with other casting parameters.

The project therefore focussed on:

--empirically estimating the hook depth on process parameters; and

--modeling mould steel flow using the commercially available CFD code 'FLUENT', verified by physical water modelling in a 1:1 scale mould to estimate the probability of slag entrainment.

MODELING OF HOOK DEPTH

Based on experiments carried out by voestalpine and others earlier [11,12], an empirical hook depth model could be set up according to the oscillation parameters, carbon content and casting speed. This was in good agreement with the experimental results (Fig 5).

[FIGURE 5 OMITTED]

The model is calibrated for sinusoidal oscillation, but future work will adapt the model for non-sinusoidal oscillation, a feature installed in recent casters and achieved by means of hydraulic oscillators.

MOULD FLOW BY CFD

Modeling the flow of metal in the mould considers the main parameters; mould width, casting speed, SEN immersion depth and the flow of argon applied to the stopper. Since these parameters can change during casting a study was carried out for different casting situations. The steel flow velocity near the meniscus as well as the probability of vortex formation were calculated. These parameters may be significant for slag entrainment and have been investigated in detail [13,14].

Fig 6 shows a typical CFD result for the casting parameters: speed 1.4m/min, width 1300mm, thickness 215mm and SEN immersion depth 150mm.

[FIGURE 6 OMITTED]

The CFD calculation for the different casting situations were approximated by regression analysis to allow the calculation of these parameters on-line during casting.

WATER MODELING

Using a 1:1 scale water model of the mould, the steel flow patterns in the mould for the various conditions modeled by CFD were compared. Particles with a specific density close to that of water were added to the water to demarcate the flow. Four ultrasonic sensors were positioned on top of the mould to measure level fluctuations as well as the average mould levels under various flow rates. The levels recorded were compared with the calculated values derived from the CFD analysis.

The experiments were documented by video camcorder and afterwards the frequency at which vortexes occurred were counted for each trial. The vortex index derived from the CFD calculations showed good agreement with that observed as well as the location of the vortex.

The modeled parameters were correlated with the defects found by strip inspection. As an example, Fig 7 shows good correlation of the CFD vortex index and the frequency of actual defects on the coils inspected.

[FIGURE 7 OMITTED]

Due to the statistical nature of the turbulent steel flow and transient flow phenomena [15], an exact prediction of the location of the defects on the coil proved impossible. However, the frequency of the defects can be predicted and casting situations where there is a high probability of forming defects can be avoided using the model results. Further, the know-how gained is useful for rating new SEN designs.

In conclusion:

--A prediction model for hook depth was established and fitted to the measured results.

--The steel flow in the mould could be characterised by model parameters such as steel flow velocity near the meniscus, vortex index and steel temperature at the meniscus.

--Using the modeled results the casting practice was adjusted to avoid unfavourable operating parameters.

--Future investigations will focus on finding new methods of measuring flow in the mould to better characterise the mould steel flow as well as mould level fluctuation.

APPLICATION 3 LADLE SLAG CARRY OVER

Slabs cast during a ladle exchange have an inferior quality in regard to steel cleanliness than the bulk of the slabs. In earlier work it has been shown that the inclusions causing defects in deep drawn products are often particles consisting of ladle slag, approximately 0.1mm in diameter.

The task of the project was to evaluate different ladle exchange conditions and to determine the strand length showing this inferior quality.

Three evaluation methods were applied:

--Physical water modeling of the tundish flow;

--Plant trial and slab sampling; and

--Combined CFD model of tundish and strand.

WATER MODELING OF TUNDISH

Using a 1:1 scale water model of the voestalpine tundish at caster CC5, a pulse of particles was injected into the shroud. The tundish water content as well as the filling rate and casting rate were varied for the different experiments. The particles arriving in the mould were collected periodically, dried and weighed.

Fig 8 shows a cloud of white particles propagating through the tundish after injection into the shroud.

[FIGURE 8 OMITTED]

PLANT TRIAL AND SLAB SAMPLING

From a slab cast during a ladle exchange, 48 samples were cut and rolled using a laboratory rolling stand. The rolled samples were assessed in the laboratories of voestalpine by ultrasonic inspection. Fig 9 shows the measured cleanliness index versus strand length.

[FIGURE 9 OMITTED]

The cleanliness of the steel begins to decrease before the current ladle is closed. The decrease of cleanliness continues for more than 10 meters of strand length after the new ladle is opened for the next heat.

CFD MODEL

A CFD model of the steel flow in the tundish and strand, and a water model of the same region using particle floatation were compared to calibrate the calculated results with those of the physical water model. Fig 10 shows a typical result of the CFD calculations carried out with using the proprietory CFD code FLUENT [16]. The particle size for the calculation was ll0mm. The contour plot shows the concentration of particles related to the injected concentration.

[FIGURE 10 OMITTED]

Further, a combined numerical 'mix-box' model of the tundish and the strand was set up and adjusted to fit the results of the water model trials, plant trial (Fig 9) and CFD simulations. This numerical mix-box model was then fed into the continuous casting process computer for on-line determination of the strand length showing inferior steel cleanliness according to the measured casting parameters during ladle exchange.

As an immediate result of the trials and simulations, the ladle exchange practice has been changed at voestalpine. Special care is now taken not to fill up the tundish too rapidly.

As a next step, the inclusions found in the slab samples will be analysed to determine their chemical composition. Future work will correlate the on-line mix-box model prediction results with quality feedback information from the final product.

In conclusion:

--Because the particles of interest are large, but rare, samples taken from the liquid steel or from hot-rolled coil could not be successfully evaluated whereas the ultrasonic inspection of slab samples was successful.

--Weighing the particles reaching the mould following injection into the tundish shroud in a full-scale water model was used to illustrate particle entrapment.

--Selection of the type of material and particle size used in the water model was difficult. Future work will try to further improve the particle size distribution and the method of measuring the quantities of particles reaching the mould.

--A combined CFD model for the tundish and strand could be set up. The CFD calculation has proven to be important in the investigation of multiple ladle changes because plant trial sampling is time consuming and costly.

--The ladle exchange practice was changed and a quality improvement due to this change can be expected.

ACKNOWLEDGEMENT

The authors wish to thank all project team members for their engaged work and valuable contributions.

REFERENCES

[1] B G Thomas: 'The importance of numerical simulations for further improvements of the CC process', Proceedings CCC2000, Linz. Austria, 2000

[2] B Mintz, R Abushsha: 'Influence of vanadium on hot ductility of steel', lronmaking and Steelmaking, Vol 20, 1993, 445

[3] Y Maehara, Y Ohmori: The precipitation of A1N and NbC and the hot ductility of low carbon Steels', Materials Science and Engineering, Vol 62, 1984, 109

[4] H Resch, C M Chimani, J Lanschutzer: 'Precipitation Modeling--A Tool for Predicting Cracks and Quality', Proceedings CCC2000, Linz, Austria, 2000

[5] A Samoilov, B Buchmayr, H Cerjak: 'A thermodynamic model for composition and chemical driving force for nucleation of complex carbonitrides in microalloyed steel', Steel research, Vol 65, 1994, 298

[6] W J Liu, J J Jonas: 'Nuclation kinetics of Ti carbonitride in microalloyed austenite', Metallurgical Transactions A, Vol 20A, 1989, 689

[7] J S Kirkaldy: 'Diffusion in the condensed state', The Universities Press, Belfast,1985

[8] T Fastner, H P Narzt, J Vlcek, W Marschal, H Prei61, N Hubner: Automatic Quality Control in Continuous Casting of Slabs for Hot-Ro0ing Use, MPT International, 1995

[9] E Gamsjager, C M Chimani, K Morwald, H J B6hm, F D Fischer: 'Influence of solid phase transformation on mechanical properties of steel during casting', Proceedings CCC2000, Linz, Austria, 2000

[10] T Fastner, C Furst, H P Narzt, G Xia, G Zuba: 'Investigation on slab surface quality by using a hydraulic oscillation unit at the new slab caster of Voest-Alpine Stahl Linz', 3rd European Conference of Continuous Casting, 1998

[11] M Nadif, A Solimine, P Ominetti, P Velluet M Blandel, M Roscini: 'influence of shell solidification on generation of slab defects use of hydraulic mold oscillation at Sollac Florange', 1st European Conference on Continuous Casting, Florence, Italy, 1991

[12] A W Cramb, F J Mannion: 'The measurement of meniscus marks at Bethlehem Steel's Burns Harbor Slab Caster', Steelmaking Proceedings, Vol 68, Detroit, Michigan. USA, 14-17 Apr 1985

[13] S L Feldbauer, A W Cramb: Slag entrainment in the mold of a continuous easter, thesis 1995. Carnegie Mellon University, Pittsburgh

[14] M Gebhard, Q I, He, J Herbertson: 'Vortexing Phenomena in Continuous Slab Casting Moulds', 761h Steelmaking Conference, Dallas, Texas, USA, 28-31 Mar 1993

[15] X Huang. B G Thomas: 'Modeling of transient flow phenomena in continuous casting of steel', Conference of Computational Fluid Dynamics and Heat/Mass Transfer Modeling in the Metallurgical Industry, Montreal, Quebec, Canada, 1996

[16] M Javurek, Ph Gittler. G Zuba, B Kaufmann: 'Particle separation in tundishes--some new general aspects', Steel Research 73, 2002

Parameters, which describe the causes of defects in continuous casting, have been modelled by computation and by physical water models. Results enabled slab end scarfing to remove transverse cracks caused by AIN precipitation to be reduced 9%; mould powder entrapment to be better understood, and showed that a slower filling rate of the tundish following ladle exchange would reduce ladle slag entrapment.