Fundamental modelling of friction during the hot rolling of steel.

Friction is one of the most significant physical phenomena influencing metal
forming, yet in comparison with metallurgy, heat transfer and mechanics it remains the
least understood. The goal of this project was to develop, on as fundamental a level as
possible, a friction model based upon the physics of the process to be applied to the hot
rolling of steel.
A fundamental friction model was developed based upon the simplified
approach to the adhesion theory by Straffelini (Wear, 249, 79-85, 2001), which is an
extension of Bowden and Tabor's original adhesion theory. In this work, the simplified
approach's dependence on the thermodynamic work of adhesion was exploited to apply
it over a wide range of temperatures. The thermodynamic work of adhesion describes
the work required to form a new surface and is a function of the surface energy of the
contacting materials was estimated using two approaches: Rabinowicz's and the
geometric mean rule. Since high temperature surface energy data is not generally
available the relative change in Young's modulus with temperature was used to estimate
a material's surface energy at a desired temperature. Reciprocating friction
experiments, which provided a controlled environment in which to investigate friction,
were conducted to verify the application of this theory to high temperature conditions
and metal-oxide contacting materials.
The fundamental model describing friction was applied to the hot rolling of steel
via a friction algorithm using the commercial finite element (FE) code MARC. Simply
described the friction algorithm calculated a friction coefficient using material
properties, defined by the user, and contact temperatures, taken from the rolling model.
This resulted in the friction coefficient predicted throughout the roll bite, compared to
an average friction coefficient typically employed in rolling models. The combined
friction algorithm-rolling model was validated against laboratory rolling experiments.
One of the assumptions of the finite element rolling model is the presence of a
thin, continuous and adherent scale layer. To achieve this in the laboratory a two pass
rolling schedule was employed; the first pass to remove the furnace scale and the second pass to input the desired deformation. The success of the friction algorithm was
determined by comparing the experimental torques and loads to the predictions of the
finite element model. The FE model with the friction algorithm predicted the friction
coefficient to vary in the roll gap between approximately 0.25 and 0.35 and was able to
predict the measured rolling torque with an average error of 15%, which was considered
acceptable and the accuracy was increased after the bearing torque was considered. The
error in the load predictions compared to the measured loads was 13.5% on average,
which was also acceptable.