Prediction of mechanical properties of aluminium metal matrix hybrid composites synthesized using Stir casting process by Machine learning

Abstract

In recent days, aluminium based hybrid metal matrix composites (Al-HMMC) is heavily demanding in aerospace and marine industry due to its enriched mechanical properties. Al7075 and Al6061 aluminium alloy are most widely used as a base matrix in Al-HMMC and stir casting method is an efficient process to manufacture them. The mechanical properties of Al-HMMC depend on material parameters and fabrication process parameters. Effecting parameters were identified based on a thorough literature review, and they include material parameters such as different types of matrix, reinforcements, and their weight percentages and sizes, as well as fabrication process (Stir Casting) parameters such as stirring speed, stirring time, and reinforcement mixing temperature. To achieve desirable results of hybrid composite materials from stir casting method, it requires high experimental cost and time. To overcome these limitations, machine learning and artificial intelligence are emerging tools, which utilize probabilistic and statistical methods to learn from the past experience based upon the experimental data to predict more accurate outcomes. In this work an attempt has been made to develop a machine learning model that can predict mechanical properties accurately. Experimental data of mechanical properties of aluminium HMMC has been collected and then used for training and testing of different machine learning models. Decision tree regression model showed maximum accuracy of 92.029 % on dataset for prediction of Ultimate Tensile strength (UTS) of Al-HMMC. The UTS of two samples of Al-HMMC with Al7075 and Al6061 as base matrix were predicted from decision tree model and compared with actual experimental values with an error < 10%.