Explicit and implicit viscoplastic models for polymeric composite

The viscoplastic behavior of a carbon-fiber/polymer matrix composite was investigated through two different modeling efforts. The first model is phenomenological in nature and utilizes the tensile and stress relaxation experiments to predict the creep strain. The phenomenological model was constructed based on the overstress viscoplastic model. In the second model, the composite viscoplastic behavior is captured via neural networks formulation. The neural networks model was constructed directly from the experimental results obtained via creep tests performed at various stress–temperature conditions. The neural network was trained to predict the creep strain based on the stress–temperature–time values. The performance of the neural model is evaluated through the mean squared error between the neural network prediction and the experimental creep strain results. To minimize this error, several optimization techniques were examined. The minimization of the error when carried out by the Truncated Newton method outperforms the standard back-propagation and the conjugate gradient method in terms of convergence rate and accuracy. Using neural network with truncated Newton training algorithm, the prediction of the creep strain was very satisfactory compared to the phenomenological model.