To develop a high-precision surrogate model of the combustion physical parameter field on the combustor longitudinal section

this study uses the fuel flow rate and swirl blade installation angle as design variables. The surrogate model is constructed by using proper orthogonal decomposition and Kriging methods. Subsequently

we analyze key factors influencing model accuracy based on the distribution characteristics of the combustion physical parameter field and propose a method to enhance the prediction accuracy of the combustion field surrogate model. The study finds that except for turbulence kinetic energy and radial velocity

the field errors of other physical field parameters are less than 10%. The error of the surrogate model is inversely proportional to the similarity of the samples. The difference between the axial positions of the peak value of the turbulent kinetic energy field and radial velocity field of each sample in the flame zone is the main reason for the significant error of the surrogate model. Therefore

the paper proposes the similarity-based sample processing method

including the sample clustering method prioritizing the global error and the upstream and downstream partitioning method to reduce the model error by increasing sample similarity. The proposed method elevates model precision by reducing the average error of the turbulent kinetic energy field and radial velocity field to less than 10%.