In optimization design algorithms, considering uncertainty quantified by evidence theory can effectively ensure the reliability of the design outcomes. However, current evidence theory-based reliability optimization design (EBDO) algorithms still face issues such as low solution efficiency and complex model construction. To address these challenges, this study proposes an EBDO method based on the optimal sparse polynomial chaos expansion (OSPCE) technique. This method introduces the OSPCE technique to complex systems or structures, enabling efficient model construction under various complex uncertainty variables. Subsequently, this study further integrates a sequential sampling generation method into the OSPCE model, thereby improving the construction efficiency of the model. By replacing the original complex system functions with the constructed OSPCE models, the proposed method effectively resolves the nested solving problem in EBDO. The effectiveness of the proposed method is validated through two engineering case studies. The results show that, compared to other EBDO solution methods, the proposed approach offers significant advantages in terms of solution efficiency.