Hybrid finite elements method-artificial neural network approach for hardness prediction of AA6082 friction stir welded joints. (8th August 2022)

• Record Type: Journal Article
• Authors: Quarto, Mariangela; Bocchi, Sara; D'Urso, Gianluca; Giardini, Claudio
• Is Part Of: International journal of mechatronics and manufacturing systems. Volume 15:Number 2/3(2022)
• Page Start: 149
• Page End: 166
• Publication Date: 2022-08-08
• Subjects: hybrid approach -- FEM -- finite element analysis -- neural network -- process sustainability -- FSW -- friction stir welding -- aluminium alloy -- artificial intelligence -- forecasting model -- hardness prediction; Mechatronics -- Periodicals; 629.89
• Languages: English
• ISSNs: 1753-1039

Abstract

One of the main important aspect of friction stir welded parts is the different hardness values reached in the characteristic welding zone, as a function of the maximum temperature derived from the welding process. Indeed, these differences affect the mechanical properties and the service quality of component. For these reasons, a hybrid model for predicting the final hardness of the single points of the welding as a function of the maximum reached temperature is developed. Specifically, the hybrid approach takes into account the finite element method (FEM) and the artificial neural network (ANN). The FEM model was set-up and the temperature map output was introduced into the ANN together with experimental results for the ANN training. The hybrid approach FEM-ANN provides a robust framework for forecasting aluminium hardness after the FSW process without experimentally investigating each welding.