Combining static and temporal process data in the modelling of FSW weld quality and mechanical properties using Computational Intelligence

摘要

Establishing efficient operational windows in Friction Stir Welding (FSW) is a common but not trivial task. One has to consider a number of process and material parameters including the weld quality, final mechanical properties, tool wear, weldability as well as economical and weld efficiency factors. Considering the plethora of material types and thicknesses the task of establishing good operational windows becomes a difficult multidimensional problem especially when a relatively new material is used for welding. From a systems modelling perspective one can consider existing welding performance data as well as data specifically designed for modelling towards the development of hybrid data driven models aiming at establishing operational windows through parallel model simulations and optimisation. In this research work the process is studied in both the static and temporal domain towards the elicitation of such models, which aim at adding to the fundamental understanding of the process by simulating process variables and investigating the effect of various process parameters on the mechanical properties of the welded material. Granular Computing (GrC) and Neural-Fuzzy (NF) modelling are used to extract knowledge from of the data sets in a transparent and simple way but at the same time preserving the vital process dynamics. The linguistic format of the produced rule-base allows interaction with process experts and a better understanding of the models. The elicited models are also used to linguistically describe the process by inferring quantitative measures of weldability and weld quality.