Experimental modeling and the effective analysis of friction stir welding parameters in 5456 aluminum alloy using artificial neural network

Friction stir welding is a solid state process developed by the welding institute TWI in 1991¡ as a technique for welding aluminum alloys. In this research¡ lap joint welding of 5456 aluminum alloy was carried¡ so that the hard working H321 sheet with the thickness of 5 mm was placed on an annealed sheet with the thickness of 2.5 mm and has been investigating the simultaneous effects of parameters such as rotational speed¡ feed rate¡ and tool angle¡ on failure force parameters and the hook disadvantage height in friction stir welding were investigated. After evaluation and analysis of the welded samples; the modelling of multilayer perceptron neural network was carried out using experimental results¡ and for the purpose of the trained model verification¡ several verification models was performed where the results of failure force in training model show 2.017 percent average error with respect to the experimental results. Next¡ the multilayer perceptron neural network was used for investigating the effect of welding parameters as in single or group mode on welding mode variations¡ where rotational speed parameter has the greatest effect and tool angle has the least effect than the other two parameters in lap joint friction stir welding.