Knowledge discovery for friction stir welding via data driven approaches

Abstract

For a comprehensive understanding towards friction stir welding (FSW) which would lead to a unified approach that embodies materials other than aluminium, such as titanium and steel, it is crucial to identify the intricate correlations between the controllable process conditions, the observable internal process variables, and the characterisations of the post-weld materials. In Part 1 of this paper, multiple correlation analyses techniques have been developed to detect new and previously unknown correlations between the internal process variables and weld quality of aluminium alloy AA 5083. Furthermore, a new exploitable weld quality indicator has, for the first time, been successfully extracted, which can provide an accurate and reliable indication of the as welded defects. All results relating to this work have been validated using real data obtained from a series of welding trials that utilised a new revolutionary sensory platform called ARTEMIS developed by TWI Ltd, the original inventors of the FSW process.