Optimisation in Friction Stir Welding: Modelling, Monitoring and Design

Abstract

The friction stir welding process involves a highly complex microstructure evolution. This makes it very difficult to derive intricate relationships among operating conditions, in-process variables and characteristics of welds, and utilise the relationships into modelling, monitoring and optimal design of operating conditions. In this research, a heuristic optimisation paradigm Reduced Space Searching Algorithm, combined with soft-computing-based modelling and data analysis techniques, is employed to solve the problem. The research investigates an aluminium alloy AA5083 and includes three facets of research: first, developing a weld quality indicator that can provide a reliable indication of ‘as-welded’ defects for an online monitoring system; second, generating accurate and interpretable prediction models for both internal process attributes and post-weld properties; third, finding optimal operating conditions to enhance welding productivity, process reliability and cost efficiency.