Prediction and Validation of Mechanical Properties of <i>Areca catechu/Tamarindus indica</i> Fruit Fiber with Nano Coconut Shell Powder Reinforced Hybrid Composites

Machine learning models can predict material properties quickly and accurately at a low computational cost. This study generated novel hybridized nanocomposites with unsaturated polyester resin as the matrix and Areca fruit husk fiber (AFHF), tamarind fruit fiber (TFF), and nano-sized coconut shell powder (NCSP). It is challenging to determine the optimal proportion of raw materials in this composite to achieve maximum mechanical properties. This task was accomplished with the help of ML techniques in this study. The tensile strength of the hybridized nanocomposite was increased by 134.06% compared to the neat unsaturated polyester resin at a 10:5:2 wt.% ratio, AFHF:TFF:NCSP. The stiffness and impact behavior of hybridized nanocomposites were similar. The scanning electron microscope showed homogeneous reinforcement and nanofiller distribution in the matrix. However, the hybridized nanocomposite with a 20:5:0 wt.% combination ratio had the highest strain at break of 5.98%, AFHF:TFF:NCSP. The effectiveness of recurrent neural networks and recurrent neural networks with Levenberg’s algorithm was assessed using R2, mean absolute errors, and minimum squared errors. Tensile and impact strength of hybridized nanocomposites were well predicted by the recurrent neural network with Levenberg’s model with 2 and 3 hidden layers, 80 neurons and 80 neurons, respectively. A recurrent neural network model with 4 hidden layers, 60 neurons, and 2 hidden layers, 100 neurons predicted hybridized nanocomposites’ Young’s modulus and elongation at break with maximum R2 values. The mean absolute errors and minimum squared errors were evaluated to ensure the reliability of the machine learning algorithms. The models optimize hybridized nanocomposites’ mechanical properties, saving time and money during experimental characterization.