In this paper, we present a novel surrogate-assisted evolutionary algorithm CSMOEA for multi-objective optimization problems (MOPs) with computationally expensive objectives. Considering most surrogate-assisted evolutionary algorithms do not make full use of the population information, which only use population information in the objective space or design space independently for tackling expensive MOPs, we propose a new strategy to comprehensive utilization of population information of objective and design space. The proposed algorithm adopts an adaptive clustering strategy to divide the current population into ‘good’ and ‘bad’ groups, and the clustering centers in the design space are obtained respectively. Then, a bi-level sampling strategy is proposed to select the best samples in the design space and objective space, where new samples are screened out according to their distance to the clustering centers and approximated objective values of radial basis functions. After the comparison with five state-of-the-art algorithms on 21 widely-used benchmark problems, CSMOEA shows high efficiency and good balance between convergence and diversity. Finally, CSMOEA is applied on the shape optimization of blend-wing-body underwater glider with 14 decision variables and two objectives, and the results show its effectiveness on the engineering problem.