<p>The concept of archives has received significant attention within the realm of multi-objective optimization. However, there exists a substantial gap in exploring archives within surrogate-assisted multi-objective evolutionary algorithms (SAMOEAs). In this paper, a novel framework called surrogate archive assisted multi-objective evolutionary algorithm (SAAMOEA) is introduced. The framework uses a surrogate archive to guide the evolution of the algorithm. This archive serves as an external database to store individuals that perform better during evolution, reducing the randomness of the individuals. Firstly, the surrogate archive is initialized using multiple MOEAs, and the surrogate constructed by the individuals in this initialization archive are used to approximate the exact function. Secondly, a cluster-based model management strategy is proposed that employs K-means method to fit the Pareto front and identify infilled individuals. Simultaneously, an environment selection method is proposed. The method determines whether infilled individuals are used to update the surrogate by calculating a convergence quality metric. The quality of the individuals in the surrogate archive is maintained inversely through the two strategies described above to provide conditions for accurate model construction. Furthermore, the paper presents three hypotheses about the relationship between the Pareto front of the surrogate model and the Pareto front of the exact function evaluation. The SAAMOEA was compared to ten state-of-the-art (SOTA) algorithms, and simulation experiments were conducted on 12 benchmark functions to obtain non-parametric estimation results and distribution plots of the true PF. Additionally, SAAMOEA was validated on an airfoil design optimization problem. The HV results and optimized parameter results demonstrate the performance advantages of the algorithm.</p>