Experimental investigation and machine learning prediction of mechanical properties of graphene nanoplatelets based triaxial braided composites

This paper investigates the effect of graphene nanoplatelets (GNP) and braid angle on the mechanical properties of braided composites. Triaxial braided composite (TBC) specimens of three braid angles, 30 degrees, 45 degrees and 60 degrees, were fabricated with addition of different weight percentages (wt%) of GNP. Compared to the pristine TBC samples, braided filler samples showed improved tensile strength and flexural strength varying from 15% to 30% based on the braid angle and loading of GNP. The interlaminar shear strength (ILSS) of the TBC was improved by approximately 15% with the braid angle, but the GNP addition did not show a significant effect. Vickers' Hardness values were found to be high for 60 degrees braid angle TBC. The experimental results were used to build predictive models for the estimation of stress-strain curves for samples with intermediate wt% and braid angles using Machine learning techniques like Artificial Neural Networks (ANN) and Gaussian Process Regression (GPR). The ANN and GPR models have shown good correlation and prediction ability, with the coefficient of determination obtained as high as one and the root mean square error value as low as 0.37. The estimated results by the best performing models complied with the experimental results with an average difference of 10.67%. A similar methodology and model can be used to predict other mechanical properties and can be extended for building a robust predictive model to find the optimal combination of variables for an application of TBC.