Abstract: The accuracy of traditional classification methods for classifying damage of CFRP-PMI sandwich structure materials is relatively low. A vision transformer (ViT) model based on Gramian angular field (GASF) enhanced time series features, combined with an improved adaptive feature fusion module (HFF), referred to as GASF-ViT-HFF, was proposed to improve classification accuracy. To address the issue of inconspicuous features in one-dimensional time series data, GASF was employed to enhance the feature representation of the time series data. The enhanced features were then used as inputs to the ViT model for classification. Additionally, the HFF module was integrated into the ViT model to further improve feature fusion. Experimental results demonstrated that the proposed model exhibited superior performance, high classification accuracy, and good stability, providing an effective approach for damage classification of CFRP-PMI sandwich structure materials.