The actual engineering structure is usually large-scale and has various local damages. Different damage has different impact on the overall performance of the structure. Small damages in critical regions can threaten the safety of the overall structure.Therefore，it is necessary to identify the critical regions of the structure，so that the monitoring，evaluation，and management of the structure have a clear priority and focus. This paper proposes a method for discriminating structural critical regions based on weighted sensitivity analysis of dynamic eigensolutions. The proposed method deduces the sensitivity of eigenvalues and eigenvectors with respect to element stiffness parameters. Firstly，the number of modes for analysis and the weight coefficient of each retained mode are determined based on modal mass participation coefficient，and a weighted sensitivity index that combines multi-order normalized eigenvalues and eigenvector sensitivity is established. The sensitivity index seeks the second-order norm to obtain the critical region index，which is used as the basis for determining the critical regions. This method is a scientific and quantitative judgment basis，and is applied to a frame and an arch bridge. The results show that the critical regions of the frame structure are the column area near the beam-column node and the support，and the critical regions of the arch bridge are the arch toe area of the arch ring. In addition，the critical region is not static，and local damage can cause the originally uncritical region to become a critical region. The proposed critical region identification method improves the efficiency of health monitoring and evaluation，and to save manpower and material resources.