Use of artificial neural network for prediction of mechanical properties of Al-Si alloys synthesized by stir casting

Abstract

Mechanical testing plays an important role in evaluating the fundamental properties of engineering materials as well as, in developing new materials. The use of conventional mathematical models in analytical calculating of the mechanical properties in most materials is very complex. In the current study Al-Si alloys were synthesized using the stir casting method. The mechanical properties as the tensile strength, Brinell hardness and wear property for the produced Al-Si alloys were investigated. Then, the obtained experimental results were used to train the artificial neural network (ANN). The neural network model is used to predict the effect of silicon content on the tensile strength, the hardness test, and wear loss for the prepared Al-Si alloys. Three neural networks were used in this study and the percent of silicon content variable was used as the ANN's input for each. Tensile test is used as ANN's output and the training function used is (traincgp) in first neural network. Also, hardness test is used as ANN's output and the training function used is (traincgf) in second neural network and wear loss test is used as ANN's output and the training function used is (traincgf) in third neural network. The obtained outcomes showed that predictions data in the applied neural networks were closer to the experimental results. The optimum mean square error (MSE) for ANNs during the tensile test, the hardness test and the wear loss test equal to 0.0335, 0.0023, 0.014 respectively and these results were satisfactory.