NUMERICAL-SOLUTION TO THE SOLIDIFICATION OF ALUMINUM IN THE PRESENCE OF VARIOUS FIBERS

In an attempt to understand the experimentally observed solidification microstructures in metal matrix composites, the influence of SiC, gra phite and alumina fibres on the solidification of aluminium has been s tudied numerically. Irregular geometries of the composite material wer e mapped into simple rectangles through numerical conformal mapping te chniques to analyse the influence of a single fibre or a row of fibres on a unidirectionally advancing planar solid-liquid interface. The fi bres were assumed to be circular in cross-section and the direction of the interface movement was perpendicular to the length of the fibres. The study showed that for fibres with lower thermal conductivity than aluminium, the interface first goes through acceleration as it approa ches and ascends the fibre and then deceleration as it descends the fi bre. The acceleration and deceleration phenomena of the interface incr eases as the thermal conductivity ratio of fibre to liquid aluminium d ecreases. With low thermal conductivity ratios (K-f/K-L much less than 1), the interface is orthogonal to the fibre surface. When the conduc tivity of the fibre is lower than that of the melt, the interface beco mes convex facing the fibre; this mode would lead to pushing of the fi bre ahead if it was free to move, as has been experimentally observed in cast microstructures of metal matrix composites. The temperature ve rsus solidification time plots of two points, one in the fibre and the other in aluminium, show that the fibre with a conductivity lower tha n the matrix is at a temperature higher than the melt; the temperature difference between the two points increases with increasing solidific ation rate for all the positions of the interface before it touches th e fibre. The three-fibre study shows that as the number of fibres incr eases, the curvature of the interface increases upon approaching the s ubsequent fibres. The relationship between these numerical computation s and experimental observations has been discussed.