Designing a more efficient traffic network system by building new roads or expanding existing roads is a direct and effective approach to alleviate traffic congestion. The corresponding network design problem (NDP) is commonly formulated as a bi-level optimization problem, incorporating the user equilibrium as the lower-level objective function and the system optimal as the upper-level objective function. In this paper, the NDP is simplified into a single-level optimization problem by assuming that vehicles in the traffic network will follow the provided guidance and obey the overall arrangement. Consequently, we only need to consider the upper-level system optimal objective, taking the maximum flow of the reconstructed traffic network and the total reconstruction cost as objectives. To solve the proposed simplified NDP model, four different multi-objective evolutionary computation algorithms (MOECs) are adopted as solvers, including both classic and recently proposed algorithms. We analyze the performance of these algorithms based on experiments conducted on generated traffic networks.