Modelling of the prediction of tensile and density properties in particle reinforced metal matrix composites by using neural networks

In this study, density and tensile strengths properties of Al2O3/SiC particle reinforced metal matrix composites (MMCs), which is produced by using stirr casting process, are predicted by designing a back-propagation neural network that uses a gradient descent learning algorithm. Firstly, to prepare a training set some results has been experimentally obtained. In the experiments, Al2O3/SiC powder mix has been prepared by reacting of aqueous solution of aluminium sulphate, ammonium sulphate and water containing SiC particles at 1200 °C. Ten percent vol. of this dual ceramic powder with different SiC particle size ranges was added into liquid matrix alloy (A332) during mechanical stirring between solidus and liqudus under inert condition [Altinkok N, Demir A, Ozsert İ. Composite Part A 2003;34:577–8. [1]]. Density and tensile strengths of dual ceramic reinforced aluminium matrix composites have been investigated at room temperature. In the neural networks training module, it were used different SiC (μm) particle size ranges as input and density and tensile strengths in produced MMCs. Then, neural network is trained using the data obtained in experimental process. In this paper, density and tensile strengths in produced MMCs have been estimated for different SiC (μm) particles size range by using neural network efficiently instead of time consuming experimental processes.