Sustainable Engineering: Predictive Modeling of Aluminum Matrix Composites using Deep Learning

In this research, the “Composite Property Prediction Neural Network” (CPPNN) is introduced as a deep learning algorithm designed for predicting the mechanical properties of aluminum matrix composites (AMCs) reinforced with coconut shell particles (CSP). Trained on the CSP-AMC Mechanical Properties Dataset, it comprises various CSP fraction values, composition condition details, and corresponding mechanical test results. Considered CSP fraction values range from 0% to $\mathbf{8 \%}$, alongside diverse compocasting conditions including temperature, pressure, and stirring time. Mechanical test results include microhardness and ultimate tensile strength (UTS) measurements from experimental testing of the fabricated composites. Through extensive training iterations, the CPPNN learns intricate relationships between input parameters (CSP fraction, compocasting conditions, Compocasting Pressure, Stirring Time) and output properties (Microhardness, UTS), enabling accurate prediction of mechanical performance for new composite formulations. Inspections of the model’s recall, accuracy, precision, and F1 score offer insight on the manner in which it performs. Leveraging the predictive capabilities of the CPPNN can expedite the materials development process, reducing the need for extensive experimental testing while providing valuable insights into the influence of composition and processing parameters on composite properties. The CPPNN stands as a significant advancement in predictive modeling for composite materials, offering a data-driven approach to enhance efficiency and sustainability in materials design and optimization.