A Data‐Driven Co‐Optimization Framework to Concurrently Enhance Lightweight Performance and Micro‐Vibration‐Induced Imaging Stability of Space Optical Systems

ABSTRACT With the advancement of high‐resolution optical payloads, the impact of micro‐vibrations on imaging accuracy has become increasingly critical. Micro‐vibrations induce optical axis jitter, leading to the degradation of imaging performance. Optimizing the parameters of the load‐bearing structure helps suppress the transmission of micro‐vibrations, providing a feasible approach for improving imaging performance under micro‐vibration excitation. This paper proposes a novel parameter optimization framework based on a surrogate model to achieve collaborative optimization of lightweight design and minimal image displacement. First, a structural‐performance dataset was prepared, and a multi‐task learning (MTL) surrogate model was established to accurately capture the nonlinear mapping from structural characteristics to performance indicators. Second, an enhanced dynamic weight non‐dominated sorting genetic algorithm (DW‐NSGA), incorporating a dynamic weight mechanism and weighted crowding distance, was proposed to efficiently perform multi‐objective optimization. Results showed that the MTL model achieved high prediction accuracy across all indicators, with R 2 values exceeding 0.9. The DW‐NSGA converged more rapidly and obtained superior optimized solutions, demonstrating excellent performance in evaluation metrics such as hypervolume (HV) and generational distance (GD). Through collaborative optimization, the final design solution effectively balanced the trade‐offs between mass and image displacement. The mass was reduced by 29.2%, the root‐mean‐square vibration response decreased by 29.9%, and the maximum image displacement decreased by 49.10% and 22.78% along two directions. This optimization framework provides an efficient approach for intelligent design and imaging‐stability improvement of optical payloads under micro‐vibration excitation.