Abstract: Considering the whole loading-failure process of a steel mesh, the mechanical behavior of the steel mesh in the bolt-mesh-shotcrete structure can be verified, which can guide the selection and parameter design of the steel mesh practically. The existing experimental results were used to verify the rationality of the calculation model and parameters. On this basis, the mechanical behavior analysis of the steel mesh in the whole process of loading-failure was completed. The maximum stress and vertical displacement of the steel mesh were predicted and analyzed by multivariate nonlinear regression and BP neural network method. It is found that: The stress distribution of the steel mesh under a single load presents a zoning phenomenon. The steel stress in the specific range of area 1 and area 2 is basically in the uniaxial tensile state, and the stress in the other areas of area 2 is in a tensile-shear composite state. The mechanical behavior of the steel mesh is significantly affected by the steel type such as the steel mesh made of HPB300, HPB400 and HPB500 bars has plastic flow range phenomenon. There is no plastic flow range in the steel mesh made of 1650 grade and of 1770 grade high strength stainless steel wire. The prediction accuracy of BP neural network model is better than that of multivariate nonlinear regression method obviously. The average relative errors of maximum stress and vertical displacement predicted by BP neural network are 1.31% and 1.67%, respectively. The prediction errors of multivariate nonlinear regression method are 7.39% and 8.19%.