Use this URL to cite or link to this record in EThOS:
Title: Modelling of the liquid slag behaviour in the continuous casting mould
Author: Kountouriotis, Zacharias
Awarding Body: University of Greenwich
Current Institution: University of Greenwich
Date of Award: 2011
Availability of Full Text:
Access through EThOS:
Access through Institution:
This work presents a fluid dynamics model of a continuous caster mould region, including the transient behaviour of the steel/slag interface. The research was carried out in collaboration with ArcelorMittal Research (AMR), based in Maizieres-les-Metz in France. The industrial objective of the thesis was to understand the factors affecting the transient behaviour of the liquid slag layer covering the steel and its interaction with the Submerged Entry Nozzle (SEN) jet supplying the steel from the tundish to the Continuous Casting (CC) mould. The study includes the very complex phenomenon of argon bubble transport which also affects the behaviour of the slag layer and SEN jet. The model developed in this study is based on the finite volume method with the liquid regions (steel, slag) solved in an Eulerian scheme on a fixed unstructured mesh. The interface behaviour is modelled using a number of VOF type schemes, including the time-efficient Counter Diffusion Method (CDM). A coupled Lagrangian particle tracking scheme is used to represent the presence of argon bubbles and their influence on the flowfield (mainly due to buoyancy) in conjunction with the fluctuating surface dynamics. The bulk of the research concerns comparisons against flowfield and interface data obtained from an experimental water/oil study of the process. However, the model is extended to include predictions of heat transfer and phase change in the steel and flux powder regions in an industrial CC unit and validation against available data. The three-phase model is developed making use of the unstructured mesh multi-physics finite volume code PHYSICA [1]. As stated, the main goal of this particular work has to do with the study of the dynamic behaviour of the steel/slag interface, including the effects of casting speed and injected gas. Because of the great difficulty in physical experiments with a real caster, the research is supported with water model experiments and mathematical simulation. Comparisons of observed interface profiles, measured and predicted mean and fluctuating velocities, gave an insight to the degree of coupling between interface behaviour and the fluid dynamics within the mould region. In particular, a spectral analysis of the dominant fluctuation frequencies in the water/oil experiment suggests a strong link between the upper and lower recirculation loops generated by the SEN jet as it splits after contact with the narrow face for the mould. The presence of gas bubbles alters the spectral picture, since the buoyancy induced in the flow affects the behaviour of the jet, leading to the one/two loop behaviour known from experiments. Good qualitative and quantitative agreement was achieved between the numerical results and water-model experimental data. The main observations drawn from the water model simulation and experiment are as follows: - An increase in casting speed, which is equivalent to an increase in SEN velocity leads to an increase in the amplitude of interfacial fluctuations. - At the highest SEN velocities, the oil layer is pushed away from the narrow ends of the mould, exposing the water surface to air. - When there is no oil on top of the water surface, the surface remains for all practical purpose flat. - Air entering through the SEN influences the flowfield in the mould and also disturbs the oil/water interface when it passes through it. - The ratio of water to air flow rate seems to be the most important parameter, with high air/water flow ratios leading to a change in flowfield at the top of the mould as the gas buoyancy lifts the SEN jet towards the surface. To achieve a good correlation between the experiments and the simulations a number of factors in the numerics were found to be important. These are: - The quality of the mesh used, especially in the complex transition from the SEN geometry – essentially a cylinder with two outlets set at a specific angle of 20o to the vertical, to the thin rectangular geometry of the mould which is designed to cast flat products. - The turbulence model, which was initially found to suppress interface oscillations whenever an oil layer was introduced. Various approaches were followed to overcome this problem, (a)removing the turbulence model from the oil layer, (b) using a low frequency filter to remove resolved turbulence kinetic energy from the k-ε model, (c)opting for the high order SMART numerical scheme in preference to the default Hybrid. - The interface tracking algorithm used as a default in the code PHYSICA is essentially a VOF technique with options for a Van-Leer (TVD) scheme [2] or alternative the popular Donor Acceptor scheme [3,4], both options work well but they are explicit and therefore extremely expensive computationally. Due to the size of the mesh and the CFL limit for stability, timesteps as small as 10-3s become necessary, meaning a 600s simulation could take up to 8 weeks! To overcome this, the implicit CDM scheme [5] was used, which allows the interface to spread by diffusion but then pushed back against the gradient to re-sharpen at the end of each timestep. With this scheme, timesteps up to 2 orders of magnitude larger become possible, the limit then governed by the frequency range to be resolved. A non-standard approach to the Lagrangian particle tracking scheme was adopted in the simulation with the following characteristics: - The amount of gas entering was divided into packets of equal bubble diameter and then each packet was further divided into individual tracks. The transport of 1000 and more particles tracks was used to ensure a realistic dispersion. - Tracks were updated at regular time intervals (but not necessarily at each Eulerian timestep) and then followed until they exited the calculation domain. - The residence time of particles in each cell provided information for the gas content of the cell and therefore its density. With this approach, the Navier-Stokes equations then solve for mixture (gas and liquid) and lighter cells are influenced by buoyancy. - The bubble tracks are affected by the mean velocity of the surrounding fluid and also by a stochastic component derived from the turbulence model. However, there is no direct feedback to fluid turbulence from the bubbles. To extend simulations to a real caster, heat transfer and phase change were introduced in the model, in addition to the property changes (water to steel, oil to slag, air to argon). Of importance here was the development of a solidified skin of steel on the water-cooled mould walls and also the melting of flux powder into a liquid layer on the top surface. This last component of the research was introduced to enable comparisons against plant data obtained by AMR. Of particular interest in this study was the transition from a double to a single roll recirculation in the top section of the mould, as a function of the relative quality of argon entering the SEN. The model was able to reproduce this behaviour for the cases studied. Although much has been done in developing this model of the continuous casting process it is evident that much more research is needed, especially in the case of a real caster. For example, the thermophysical property variations in the slag due to temperature, composition and mass transfer were ignored. A very simple approach was used for the phase change in steel and flux powder, although since the PHYSICA framework is modular, more sophisticated alternatives can be easily introduced. The boundary conditions for heat transfer remain uncertain and the values used in this study were obtained from the industrial partner from earlier experiments. In spite the aforementioned limitations, the model is very useful, especially in understanding the dynamic interactions between the SEN jets, and the slag/steel interface and in this respect in advance of other models used by industry.
Supervisor: Pericleous, Kyriacos A. ; Patel, Mayur ; Djambazov, Georgi Sponsor: ArcelorMittal Research (AMR) ; University of Greenwich
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: QC Physics