Graph Regression in CAD Data: A Graph Neural Network Architecture for Manufacturing Time Estimation

Manufacturing time estimation is crucial for many businesses. Especially small and medium-sized enterprises of-fering customized products need help as they often rely on traditional, non-digital methods and use heterogeneous machinery. This paper, therefore, explores the application of Graph Neural Networks for estimating manufacturing time based on CAD models. We propose a methodology for generating and preprocessing CAD models with multiple machining features. We then compare different Graph Neural Network architectures, such as Graph Convolutional Networks, Dynamic Graph Convolutional Neural Networks, and k-dimensional Graph Neural Networks, for estimating the manufacturing time of the generated CAD models. Our evaluation results indicate that an optimized Dynamic Graph Convolutional Neural Networks architecture best estimates manufacturing time, especially in CAD models with multiple features. The results of this work present small and medium-sized enterprises with a promising solution for accurate and efficient manufacturing time estimation to enhance their competitiveness.