Objectives  It is difficult for the single-point deformation monitoring model for concrete arch dam to reasonably characterize the cooperative response characteristics of the arch dam spatial deformation field, and the traditional regression method has obvious limitations to interpret the complex function relationship between environmental quantity and dam deformation.

  Methods  On the basis of the analysis of traditional single-point deformation monitoring model construction method, a multi-point hybrid model based on support vector machine-particle swarm optimization(PSO-SVM) for concrete arch dam deformation monitoring is proposed by introducing spatial coordinates, and using the finite element method to calculate the water pressure component. SVM optimized by PSO is used for analysis and prediction of dam deformation sequence with the aid of PSO-SVM's good nonlinear processing capabilities. Thus, a multi-point hybrid model based on PSO-SVM for concrete arch dam deformation monitoring is established.

  Results  To verify the effectiveness of the proposed PSO-SVM-based multi-point hybrid model, a single-point statistical model, and an SVM-based multi-point hybrid model are established at the same time. The mean absolute error and root mean square error values of the established model are lower than those of other models, and the determination coefficient R² is closer to 1.

  Conclusions  The engineering example analysis shows that the proposed model has better ability to analyze the deformation performance of multi-point, and has better fitting and prediction accuracy than the statistical model of single-point, and can effectively reflect the overall behavior of the dam. The proposed theory and method can be generalized and applied to other dam behavior safety monitoring models.