Health monitoring plays an important role in guaranteeing the structural safety of a cable-stayed bridge throughout its life. However， the use of sensor data for health monitoring of cable-stayed bridges has some limitations. Combining the Benchmark model to monitor bridge structures can compare and evaluate different health monitoring methods criteria in a given state， which is valuable for bridge design， operation and management. This paper proposes a modified Benchmark model for health monitoring technology of cable-stayed bridges. Firstly， a simplified finite element model of the spine-beam model is established using Matlab software to conduct the dynamic analysis and model modification of the structure. In addition， a detailed finite element model is established as a simplified supplementary model and modal correction model. Then， an orthogonal test design is used to perform a significance analysis on the dynamic characteristic parameters of the cable-stayed bridge for the parameters to be modified in the modeling process， and the sensitive parameters are clarified. Finally， the model parameters are modified based on the genetic algorithm， and the dynamic adaptive technique and the method of parallel selection are proposed to optimize the structure of the genetic algorithm， which interferes with the genetic selection process and gives higher dispersion to the initial population. Compared with the GA function， the method with the improved genetic algorithm has higher computational efficiency and smaller error for model parameters.