The present study is devoted to developing a digital twin for a composite overwrapped pressure vessel (COPV) used in electric propulsion engines of spacecraft. The digital twin is used to predict the future behavior and performance of a real physical object based on the currently available information without carrying out expensive and time-consuming full-scale prototyping and testing. Multiscale approach is employed to link the macroscopic stiffness degradation and failure with a progressive damage evolution at the microlevel of composite. The computational models for the stress state and failure analysis at different scale levels are presented. Based on a comparative analysis of the traditional approach for assessing the load-bearing capacity of the COPV and its digital analogue, the advantages of the latter are shown as the predicted burst pressure is in good agreement with the experimental results.