抄録
Thermoplastic matrix of the composite has time-temperature dependent properties. For nonlinear properties of the matrix, a neural network model is used to represent those complex properties. The time-temperature superposition principle is known in viscoelastic properties, it is used to verify a neural network model. A homogenization theory with two-scale asymptotic expansion is used to homogenize the creep properties of the composite. Effective constitutive equations and microscopic disturbed displacements are derived from the homogenization theory. In numerical calculations, the effective creep compliance of the composite is determined by using the homogenization theory and predicted creep property of the matrix by neural network.
