Prediction of weight change of glass fiber reinforced polymer matrix composites with SiC nanoparticles after artificial aging by artificial neural network-based model

In this study, the weights of SiC (silicon carbide) nanoparticle-filled and unfilled glass fiber reinforced polymer matrix composites (PMC) after artificial aging were estimated using an artificial neural network (ANN) model. Composite samples with different SiC nanoparticle weight fractions (0%, 0.5%, 1%, 1.5%, 2%) were produced by vacuum infusion method and subjected to artificial aging at 70 degrees C and 85% relative humidity for 0, 250, 500, 750, 1000, 1250, and 1500 h. The weights of the samples were measured and recorded periodically during the aging process. The developed ANN model was trained to estimate the sample weight using SiC nanoparticle weight fraction and aging time as input parameters. The network with four neurons in a single hidden layer was trained with the Levenberg-Marquardt feedforward backpropagation algorithm, and a total of 35 datasets were used for training, testing, and validation. The weights predicted by the model overlapped with the experimentally obtained data with high accuracy. The mean square error (MSE) value calculated to evaluate the accuracy and adequacy of the model was determined as 0.001225 in the 256th iteration. It was concluded that the trained artificial neural network model was able to predict the weights of SiC nanoparticle-filled and unfilled glass fiber reinforced PMCs with high accuracy and efficiency.