The invention provides a method for predicting the effective stress relaxation coefficient of a polymer matrix composite.The method includes the steps that firstly, a linear viscoelatic material constitutive equation in a time intra-domain integral form is established based on the Boltzmann superposition principle; then the effective stress relaxation rigidity of the polymer matrix composite formed by periodic unit cells is defined; then a mesomechanics model for solving an effective stress relaxation rigidity matrix of the polymer matrix composite is established based on the variation asymptotic homogenization theory; finally the effective stress relaxation rigidity of the polymer matrix composite is predicted based on the established mesomechanics model.According to the method, attributes of different materials in different directions can be obtained in the one-time solving process, and compared with a method in which rerunning is conducted under different loading conditions, the method is simpler, more convenient, more efficient and rapider.The calculated effective stress relaxation coefficient is consistent with the precision of a wave function, no more approximation relations need to be introduced, calculation precision is higher, and the strength and service life of the polymer matrix composite are predicted more accurately.