Abstract: A structural damage prediction method based on the Kriging surrogate model was proposed for a multi-storey frame structure of a museum in Luding. The method proposed was validated upon the seismic damage of the multi-storey frame structure in the Luding Ms6.8 earthquake that occurred on September 5, 2022. The specific process and method of structural seismic damage prediction based on the Kriging surrogate model are as follows: Establishing a finite element model of the structure; Selecting near-fault seismic motion records that can represent the characteristics of the site where the structure is located for elastic-plastic time history analysis; Training the Kriging surrogate model by the base shear and by displacement data obtained from time-history analysis to obtain the seismic capacity curve of the structure; Combining with LASSO regularization method and K-means clustering algorithm, and training the Kriging surrogate model by the input multiple seismic parameters, by the base shear and, by the displacement data from time-history analysis to obtain the seismic demand Kriging surrogate model of the structure; According to the seismic capacity curve, defining the seismic damage evaluation criteria for the structure, and estimating the seismic damage of the structure based on the seismic demand Kriging surrogate model and, on potential seismic motion parameters. The research results indicate that: When using 300 sets of training data, the earthquake damage prediction method based on the Kriging surrogate model can meet engineering requirements and serve as a reference for improving the HAZUS method; In the case of limited data, feature selection methods can improve the accuracy of the Kriging surrogate model, but the K-means clustering algorithm cannot effectively improve the accuracy of the results; The high cost of obtaining training data is a limiting factor for the application of Kriging surrogate models in the field of structural seismic analysis.