Based on the classic uncapacitated fixed charge model, and considering simultaneously the system operation cost and reliability, this paper proposed a nonlinear mixed integer programming model taking into account the failure probability and the customer multilevel redistribution. The linear processing model was solved by using the optimized Lagrangian relaxation algorithm. The data case shows that Lagrangian relaxation algorithm is more effective for midsize network location problem, and relative to multi-level reassignment of customers, the failure probability  of facilities has more influence on it.