Prediction Of Shear Force Characteristics Of Dissimilar Friction Stir Spot Welded Joints Using Neural Network Model

Hanafy, Amr Mohamed; Gaafer, Ahmed; Mansour, Sayed Hamza; Abdel-Aleem, Hamed A.

doi:10.21608/erjsh.2024.294971.1320

Prediction Of Shear Force Characteristics Of Dissimilar Friction Stir Spot Welded Joints Using Neural Network Model

Document Type : Research articles

Authors

¹ Higher technological institute

² Department of Mechanical Engineering, Faculty of Engineering at Shoubra, Benha University, Cairo, Egypt.

³ Welding Technology and NDT Department, Central Metallurgical Research and Development Institute (CMRDI), Cairo, Egypt

10.21608/erjsh.2024.294971.1320

Abstract

An artificial neural network (ANN) system was developed and implemented for analyzing and simulating the process parameters concerning -aluminum alloy 6061 and pure copper - dissimilar welded joints for their mechanical properties. In the present study, 2.2 mm thick Aluminum alloy 6061 and 1.4 mm thick pure copper are welded using the friction stir spot welding process. The process parameters involved in welding are the tool rotation speed, plunge depth, and dwelling time. There exists an optimized level of the parameters of friction stir spot welding (FSSW) for the highest shear load of AA 6061 and pure copper lap-welded joints, predicted as 20 seconds with a plunge depth of 0.2 mm at 2000 rpm. The shear load increases with a further increase in the plunge depth for a 15 seconds dwell time and 2000 rpm. The network of 10 neurons achieves the best performance with the highest validation and test correlation coefficients, whereas the network of 20 neurons may be overfitting or underfitting, as suggested by its lower training correlation coefficient.

Keywords