3D CAD model retrieval in early design based on sketch and machining features

In the early design stages, before the construction of the three-dimensional (3D) model, the lack of a 3D model makes it impossible to retrieve expectation models from a 3D computer-aided design (CAD) model base. Furthermore, the high-dimensional analysis of 3D models increases calculations, reducing retrieval efficiency. The above issues have been studied and combined with previous research on 3D CAD model shape classification and machining feature recognition, and a framework of 3D CAD model retrieval in early design based on a 3D sketch and machining features is proposed in this study. In this framework, a shape classification method based on 3D Convolutional Neural Networks (CNN) is adopted to determine the retrieval scope to improve retrieval efficiency. A core issue of this framework is investigated, focusing on similarity calculation: a quaternion shape feature descriptor is established to extract the transformed point cloud, and then the shape feature is clustered and analyzed to calculate shape similarity. Furthermore, the 3D sketch shape and machining features are fused to calculate the similarity of the 3D CAD model. The F1 score calculated by this method is 98.21%, indicating that the proposed method has excellent retrieval accuracy. The validation and comparison of the experimental results indicate that this framework improves retrieval precision and efficiency in early design.