Experimental Study and Modeling of Friction Stir Welding Process of Aluminum 1100 Alloys, using Artificial Neural Network with Taguchi Method

In this paper, the temperature distribution in workpiece and microstructure of welded zone in friction stir welding of aluminum 1100 alloys and the effect of the tool rotational speed on these parameters have investigated experimentally. Also feed forward back propagation neural network has been used to predict the temperature of the workpiece during the welding process by considering the process time and tool rotational speed as input parameters of the neural network. For this purpose, the Taguchi design of experiments has been used and the network with minimum mean squared error was selected. This way of neural network selection is very formal and effective than the existing methods. The selected network mean squared error with this approach is 0.000388, its most differences with experimental inputs is 0.770997ºC and its regression R values is 0.99113. Also according to experimental results, increasing tool rotational speed leads to higher plastic deformation in materials and also causes increasing the friction between tool and workpiece which leads to higher workpiece temperature.