Abstract: To address the challenges of high computational resource consumption and uncontrollable convergence time commonly encountered in optimization problems, a multi-fidelity surrogate model-based sequential sampling method is proposed, which incorporates the time cost of computational models into the optimization process. By dynamically adjusting the dependence on high-fidelity models, the method achieves convergence to an improved solution within a predefined time frame, meeting the practical engineering demand for obtaining superior solutions under strict time constraints. The effectiveness of the proposed method is validated through various numerical case studies. To enhance the waverider's aerodynamic performance across a wide speed range, a parameterization method for the waverider's rear fairing is developed, and the proposed optimization approach is applied to the wide-speed-range optimization of the rear fairing. The optimization results demonstrate that the rear fairing significantly improves the aerodynamic performance of the waverider at both supersonic (Ma = 2, 4) and hypersonic (Ma = 6, 8) speeds. Within a small angle of attack range, the lift-to-drag ratio is increased by 5%–25%. The aerodynamic performance improvement as well as the detailed flow field analysis both indicates the fairing's contribution across a wide speed range and the proposed method's effectiveness for engineering applications.