STUDY ON FRICTION STIR WELDING OF ALUMINIUM PLATES USING AN ARTIFICIAL NEURAL NETWORK

Abstract

For attaching solid materials, friction stir welding (FSW) is a relatively novel
method recently developed. Compared to fusion welding processes, it has many benefits,
such as reduced distortion, porosity, shrinkage, and cracking. FSW was first used to link
aluminum alloys with limited weldability, but it has since been used to join other metallic
alloys and other dissimilar alloys. It is possible to fuse two plates using FSW by inserting a
non-consumable rotating tool with a specifically designed pin between them and moving it
along the welding line. Multiple applications in the aerospace and shipbuilding industries and
the automobile sector have seen success with this approach owing to its many benefits.
Computer-aided artificial neural network (ANN) modelling may be used in material science
and engineering to improve the FSW process. In the same manner, as the brain processes
information, ANN is a computer processing paradigm inspired by the brain's workings. There
are many nerve cells in the system. ANNs, like humans, are taught by examples and maybe
both a teaching and a forecasting tool. Well-trained neural networks are excellent prediction
tools and can predict results for inputs it has never seen. It may therefore be considered as an
approach to automating FSW.
A wide range of variables influences the FSW process. To better understand the relationship
between welded material's mechanical characteristics, such as ultimate tensile strength (UTS)
and hardness, this study considers three parameters: tool rotation speed, welding speed, and
axial force. An artificial neural network (ANN) is developed and then evaluated to determine
the mechanical characteristics of welded materials.