Application of an Artificial Neural Network Model to Predict Parameter of Friction Stir Spot Welding on Aluminum Sheet
Abstract
This research was conducted to predict the maximum load in the Friction Stir Spot Welding process using Aluminum Alloy 1050 material. There are 4 variations of process parameters, namely tool pin diameter, tool rotation speed, welding speed which has 3 levels for each, and plunge depth which has 2 levels. The experimental design in this research used the Taguchi method with 54 experiments. The results of the Backpropagation Neural Network training have a 4-8-8-1 network architecture consisting of 4 input layers, 2 hidden layers with 8 neurons, and 1 neuron in the output layer. The activation function used is "logsig" and the training function is "trainrp". With this network architecture, the MSE is 0.00918 and the average error is 6.99%.