In order to solve the problem of multi-objective size optimization of truss structures with stress constraints, a multi-objective optimization algorithm with multi-neighborhood was proposed. Based on the Euclidean distance between individuals, the population was divided into multi-neighborhood to form several niche populations. A number of individuals were assigned to each cell as neighborhood by the proposed algorithm. The individuals were only allowed interacting with each other within its neighborhood and generating offspring. The influence of different sizes of neighbors on the performance was analyzed through simulation experiments. The test results on 18 benchmarks revealed that the proposed algorithm outperformed some state-of-the-art algorithm in terms of covered area and diversity, which showed good uniformity and diversity. The obtained Pareto front showed good uniformity and diversity when solving the classic multi-objective optimization problem of 25-bar truss structure. The algorithm showed good performance in program design, solution space and generality and so on, which was very simple, practical and suitable for engineering practice.