With the popularization of CAD technology in the manufacturing industry, the number of 3D models in the industry is rapidly expanding and becoming increasingly complex. These models carry the intelligence of designers and are highly reusable in the product development process. Manufacturing practitioners realize that they can accelerate product development by reusing the structure of previously designed models. Today, many open-source and private model databases have emerged, but efficient and accurate retrieval of 3D models is still an urgent problem. Current mainstream 3D model classification and retrieval algorithms commonly use deep learning techniques, which can usually be subdivided into voxel-based, point cloud-based and view-based methods. However, they all have their own shortcomings. Therefore, in this paper, we propose a 3D model classification and retrieval algorithm based on graph convolutional neural network for feature extraction of STEP files, which improves the efficiency and accuracy of 3D model classification and retrieval.