In this paper, thermo-mechanical responses of shape memory polymers and their composites are investigated for improving the shape memory effect. First, a chain model composed of thermal and mechanical parts has been proposed. The properties of thermal and mechanical parts are considered by way of analogy with thermo-mechanical responses in macro-scope. Subsequently, the instantaneous elastic modulus of chain system under different temperatures is examined to calibrate the material's parameters by adjusting several variables: chain length, chain number, and distribution function. Next, numerical simulations of polyurethane composites are performed and compared with experimental data to validate the applicability of the model. Furthermore, the shape memory processes of polymer composites are optimized by changing the structure of the chain system: geometry, cross link, and internal filler. From the results, it can be concluded that the shape memory effect of polymer composite can be improved by increasing cross link density. (C) 2019 Elsevier Ltd. All rights reserved.