Summary: | The work described in this thesis details the investigation into flash butt weld failure at Corus, Port Talbot, where the process is used to join coils of steel, permitting cold rolling without interruption. The aim of this work was to examine the factors which constitute weld failure and to determine whether failure could be eliminated. It was shown that weld failure cannot be predicted with appreciable levels of confidence with the use of neural networks based on inputs such as steel composition and the thickness and width of the coils to be joined. 817 weld breaks were recorded during 2006, of which all but two were observed during cold rolling. Analysis of the samples provided indicate that in some case the term weld break was incorrectly used to describe what was in fact failure in the base material, away from the joint. Prior to this work there were no reported data on the residual stress levels generated after flash butt welding of sheet steel with a thickness of 2.5 mm. It is now clear that in addition to the expected tensile stress in the weld, there is also a compressive stress of up to 200 MPa in the base material which extends far beyond the heat-affected zone. This result may go someway to explain why failure in the base material is observed outside the heat-affected zone during cold rolling. Finally, a post-weld heat treatment for a high hardening boron-steel was examined. The heat treatment was intended to supersede previous attempts to reduce the post-weld cooling rate by applying a voltage across the weld to locally heat the region to a desired temperature.
|