The conventional measurement methods, constrained by the limited spatial capacity, predominantly rely on finite sensors, consequently exhibiting a deficiency in global sensing capacity for structural field responses. Meanwhile, existing reduced-order model (ROM) algorithms, while capable of achieving full-field reconstruction through finite sensors, exhibit limitations in addressing complex structural mechanical loads. To enhance the global sensing capacity for structural field responses, this paper proposes a structural field reconstruction method based on residual super-resolution, which establishes a mapping relationship between finite inputs and a high-fidelity simulated strain field. When applied to blended-wing-body underwater glider (BWBUG) structural field prediction, validation using finite element analysis datasets demonstrates that the proposed method maintains superior prediction accuracy compared to conventional data-driven ROMs and state-of-the-art network-based reconstruction methods, showing promising potential for further BWBUG digital twin applications.