Mechanical and microstructural analysis of a AlSi-ZrO2 metal matrix composite using optimized artificial neural network and experimental data

Abstract Spray deposition methods has emerged as an alternative to powder metallurgy and ingot routes. This research aims to study the effect of flight distance as a potential key factor that changes the optimum percentage of aluminium silicon- zirconium oxide in terms of mechanical and microstructural properties. The alloy is sprayed at varying the flight distance from 320 mm to 480 mm. The composite were prepared by spray deposition technique and effects on microstructural properties were examined. The aluminium silicon- zirconium oxide composite was subjected to hot isostatic pressing for reducing the porosity of the deposit from 14.4% to 8.2%. A series of experimental study were carried out in the laboratory by varying the flight distance from 320 mm to 480 mm for aluminium silicon- zirconium oxide composite to characteristic loading. In this paper, an optimized artificial neural network using genetic algorithm are developed to predict the mechanical behaviour for aluminium silicon- zirconium oxide composites. Based on the experimental data, the ANN models were developed, trained and tested. The microstructure of the AlSi-ZrO2 composite consisted of finely divided globular shaped eutectic Si uniformly distributed in the Al matrix. With addition of zirconium oxide composition to AlSi alloy, the tensile strength and micro hardness increased from 123 MPa to 147 MPa and, 48 HV to 72 HV, respectively. The preferred flight distance for the current study is found to be 420 mm. Microstructural images obtained at flight distance consist of co-existing primary Si phase and needle like eutectic Si. The physical properties, such as tensile strength, compressive strength, yield strength, micro hardness and porosity of sprayed aluminium silicon- zirconium oxide can hence be adjusted by setting the optimized flight distance. The developed ANN-GA method proved to be accurate, reduced time and efficient to predict the numerous samples and it will help materials designers to design their future experiments effectively.