surface of the melt or stick to the refractory walls of the containing vessel. The inclusions are brought there by the process of fluid flow. There are many factors which influence the rate of removal of inclusions from the melt, for example, the inclusion size, the density difference between the inclusion phase and the metal, the viscosity of the metal and the size of the
system17*79*80. However, stirring has perhaps the most significant effect on the removal rate17*66*74*81*82. Under industrial conditions metal stirring is achieved in a number of ways; by mechanical action, by using an
electromagnetic field, and by the injection of an inert gas83. Numerous
workers1 T49*50*51*58*60*63*66*79-*82*84 have studied the factors influencing the removal rate, and some have developed mathematical models for the
2.4.4.1 REMOVAL OF DEOXIDATION PRODUCTS FROM STIRRED MELTS
Several models for the removal of deoxidation products from stirred iron melts have been proposed58*60*79-81.
Miyashita et al.66 showed that the separation rate of inclusions in an agitated melt was generally larger than in a quiescent melt, concluding that Stokes' Law was not appropriate to describe the removal of deoxidation products from liquid steel under such circumstances.
Lindborg and Torssel58 made a quantitative, statistical analysis of the growth and separation of inclusions based on a collision theory. They postulated that an interparticle collision would result in the instantaneous coalescence of the particles into one large particle, which in turn collides with other particles until separation takes place at the top surface of the melt. It was suggested that stirring the melt has two effects on deoxidation:-
(i) the homogenisation of the melt with respect to the deoxidant and, (ii) increasing the number of interparticle collisions.
There are however, inconsistencies in their proposed growth mechanisms, when compared with other experimental observations82. Their calculations suggested that the rate of gradient collisions was less than the rate of
Stokes' collisions. The major weakness in Lindborg and Torssell's58 model
was that the trajectory of the moving particle was assumed not to be affected by an approaching particle, and that every collision resulted in coalescence. Three possible mechanisms for the removal of oxygen in deoxidised,
induction stirred steel, melts were postulated by Lindskog and Sandberg49; (a) the deoxidation products were nucleated directly on the surface of the crucible;
(b) the inclusions rose to the surface of the melt with a velocity given by Stokes' Law;
(c) the inclusions were adsorbed on the inner surface of the crucible wall due to the fluid flow within the system.
Total oxygen and dissolved oxygen measurements indicated that some deoxidation products were formed in the melt. If all the deoxidation products had been nucleated directly on the crucible wall the total oxygen content of the melt would equal the dissolved oxygen content of the melt, which was shown not to be the case, Figure 2.4. It was proposed that during the initial precipitation period, direct nucleation of the deoxidation products on the crucible wall, flotation according to Stokes’ Law and adsorption on the inner surface of the crucible could all be active in the removal of the deoxidation products. However, after precipitation had ceased, direct nucleation onto the crucible wall could no longer occur and deoxidation products could only be removed by flotation due to Stokes' Law or by adsorption onto the crucible wall.
Further evidence of the importance of inclusion adsorption onto crucible walls has been presented by Nakanishi et al.85, who detected an adsorbed
layer of AI2O3 in the inner surfaces of SiC>2 and CaO crucibles after the deoxidation of liquid iron with aluminium in a high frequency induction furnace.
Results obtained by deoxidising induction melted steel with radioactive 3^Si indicated that 60% of the oxygen removed from the melt, after precipitation
ceased, was removed by adsorption onto the crucible wall. The Si02
particles were adsorbed preferentially onto the upper part of the side wall just below the slag line and the centre part of the bottom of the crucible. Other
experiments have shown49-85 that AI2O3 deoxidation products can also be
absorbed in high concentrations in the centre part of the bottom floor and the upper part of the side wall of the crucibles. The high concentrations coincide with the points were the stream of liquid metal is thought to encounter the crucible at right angles in an induction stirred melt, Figure 2.5.
Linder79 considered the hypothetical behaviour of small spherical particles in a turbulent metallic melt for situations similar to those for the deoxidation of steel in stirred systems. It was stated that the most important mechanisms to achieve relative movement of two particles in a turbulent melt are associated with the rise of the particles due to the gravitational force and with turbulent velocity gradients. Differentiations were made between three possible situations when collisions could take place.
1. The rise velocity due to the gravitational force is small compared with the relative movements due to the velocity gradients in the melt. In this case the main collision mechanism is gradient collision.
2. The rise velocity due to gravitational forces is greater than the relative movement due to velocity gradients in the melt, in this case the particles can be considered as rising in a quiet viscous fluid. If a large particle approaches a smaller particle, Figure 2.6, at first it would appear that a collision would take place if the distance between the smaller particle and
the central axis of the large particle is smaller than L'+ L", but as the two particles come closer, the smaller particle will be pressed aside by the flow around the larger particle. Only particles coming from a distance
Lc < L'+ L" will collide, where L' and L" are the radii of the small and large particles respectively and Lc is the critical distance for a collision.
Linder79 suggested that the possibility for a collision by this mechanism for two deoxidation products is quite limited and "Stokes' Collisions" could probably be neglected when discussing the removal of deoxidation
products.
3. The velocity rise due to gravitational forces and the velocity due to velocity gradients are of the same order of magnitude, there might be a mutual interaction effect such that the number of collisions are not just a sum of (1) and (2) considered separately.
The method presented by Linder79 to predict the behaviour of the small fluid
spherical particles in a stirred metallic melt gave some indication of the parameters which when varied might affect a change in the rate of a
deoxidation operation, ie. the amount of stirring, the mean flow velocity, the size and shape of the system, the density difference between the deoxidation product and the liquid steel, and the viscosity of the liquid steel. However, much of the Linders' work was based on the use of hydrodynamic parameters for which values are generally not known, and to overcome this, many of the hydrodynamic parameters were estimated.
Engh and Lindskog81 produced a fluid mechanical model for inclusion
removal, where the separation rate of inclusions was given as a function of the stirring power. The model was based on the eddy diffusivity concept,
which was used to calculate the rate of removal of inclusions to the walls. However, the values tended to give too high removal rates, which may have been due to the fact that not all inclusions which contact the walls of the containing vessel will adhere to it.
2.4.4.2 REMOVAL FROM MELTS STIRRED BY NATURAL CONVECTION
It is generally considered58*60*66-81*82 that the rate of removal of liquid deoxidation products is slower under conditions of natural convection, than with stirred melts.
Miyashita et al66 investigated the rate of removal of primary deoxidation products from a quiescent iron melt and reported that experimentally the results lay between values calculated for Stokes' Law and values calculated from the following equation for the case where slip occurring at the
particle/liquid interface was used to express the rising velocity of the particles:-