Prediction of static strength properties of carbon fiber-reinforced composite using artificial neural network

Abstract In this paper, an artificial neural network (ANN) based model is developed considering the significant parameters affecting the strength properties of the fiber-reinforced composite. The model utilizes the experimental data obtained from Composite Materials Handbook, Volume 2—Polymer Matrix composites material properties (Military Handbook 17-1F). The data is extracted for unidirectional carbon fiber reinforced composite (CFRP) which represents the mean data obtained from experimentally tested specimens in batches. The dataset consists of 74 samples with eight input parameters: fiber strength, matrix strength, number of plies, loading axis, temperature, volume fraction, void percentage and thickness of ply. The output of the ANN model is the strength of the composite. The hyper-parameter of the ANN model is tuned and selected optimally. The network architecture arrived at is 8-[4]-1 with training function as Levenberg–Marquardt and activation function as tan-sigmoid in the hidden layer and pure-linear in the output layer. The agreement between the prediction from the developed model and experimental data is satisfactory, indicating the model’s applicability and efficacy. The trend analysis with respect to the input parameters is also carried out to verify that the model captures the mechanics-based behavior of CFRP.