Taguchi and ANN approaches for predicting the FSW parameters to analyse the mechanical properties

Taguchi design utilized to execute mathematical analysis to assess friction stir welding factors on the mechanical properties like ultimate tensile strength (UTS) and Hardness (BHN). Tool rotational speed (TO), pin depth of tool (PI) and welding speed (WE) were selected. The collected experimental data were used to train ANN model using the LM backpropagation algorithm, aiming to achieve accurate predictions of the maximize the mechanical properties. The optimized ANN was structured as a multilayer perceptron with a feed-forward architecture, consisting of 3 input nodes, 10 neurons in the hidden layer, and 2 output node (3-10-3 structure). The developed model accurately which predicted UTS and BHN values, with an overall R of 0.99335 which indicates a high level of accuracy. The optimal settings for UTS at the highest and lowest levels in Exp. No 8 and Exp. No 3 are identified as TO3-WE2-PI1(1600 rpm, 200 mm/min, and 0.3 mm) and TO1-WE3-PI3 (1200 rpm, 250 mm/min, and 0.9 mm).