Neural network application for prediction mechanical properties of Mg-Al-Zn alloys

Purpose: The paper presents results of the research connected with the development of new approach based on the neural network to predict chemical composition and cooling rate for the mechanical properties of the Mg-Al-Zn cast alloys. The independent variables on the model are chemical composition of Mg-Al-Zn cast alloys and cooling rate. The dependent parameters are hardness, ultimate compressive strength and grain size. Design/methodology/approach: The experimental magnesium alloy used for training of neural network was prepared in cooperation with the Faculty of Metallurgy and Materials Engineering of the Technical University of Ostrava and the CKD Motory plant, Hradec Kralove in the Czech Republic. The alloy was cooled with three different cooling rates in UMSA Technology Platform. Compression test were conducted at room temperature using a Zwick universal testing machine. Compression specimens were tested corresponding to each of three cooling rates. Rockwell F-scale hardness tests were carried out using a Zwick HR hardness testing machine. Findings: The results of this investigation show that there is a good correlation between experimental and predicted dates and the neural network has a great potential in mechanical behaviour modelling of Mg-Al-Zn alloys. Practical implications: The presented model can be applied in computer system of Mg-Al-Zn casting alloys, selection and designing for Mg-Al-Zn casting parts. Originality/value: Original value of the work is applied the artificial intelligence as a tools for designing the required mechanical properties for Mg-Al-Zn castings.