A computational model of the depolymerization of composites during solvolysis of wind turbine blades is developed. The model is based on a phenomenological approach, describing the dissolution of an epoxy matrix as a local phase transition influenced by temperature, solvent diffusion, and local microstructures. The model is implemented in the finite element code Abaqus, using the user‐defined field and heat flux subroutines. Parametric studies are carried out to study the influence of defects and heterogeneities on the depolymerization of composite materials. The results show that variations of the solvent diffusivity in the vicinity of fiber/matrix lead to nonhomogeneity of depolymerization, and smaller diffusivity may explain matrix residues remaining on the fibers. Fiber volume density and distribution influence the polymer dissolution rate because the resistance to solvent diffusion arises between fibers. Further, voids in the polymer may lead to a local acceleration of the polymer dissolution. This work also looks into the contributions of diffusion and reaction to depolymerization, and the former dominates. The rate and homogeneity of depolymerization, therefore, depend significantly on the manufacturing quality of composites.