Foundation models in reverse engineering, CAD reconstruction and additive manufacturing: A systematic review

Reverse engineering is undergoing a profound transformation driven by the emergence of foundation models. These models, together with their applications, are progressively extending the limits of artificial intelligence (AI) from geometrical reconstruction to a more sophisticated level engineering reasoning. In contrast to traditional methods, which usually use sequences of algorithms for specific 3D retrieval and reconstruction tasks, current foundation models train geometrical representations that are able to support semantic reasoning, computer-aided design (CAD) reconstruction, and intelligent manufacturing in a unified way. This work provides an overview of foundation models within the context of reverse engineering and their combination with CAD reconstruction and additive manufacturing. Starting from a systematic bibliometric analysis of the recent literature, the present review is mainly focused on the development of representation learning and reviews the key geometric forms that are being used by the current foundation models. The paper also explores the engineering abilities facilitated by such models in the reverse engineering process. Furthermore, current challenges, research obstacles, and future directions are critically explored, emphasizing the recent transition toward intelligent engineering systems. The analysis shows that foundation models are gradually making reverse engineering move from being an assortment of discrete computational activities to becoming an engineering process workflow that is made possible by means of geometry representations and multimodal reasoning as well as engineering AI. Finally, considerations on the importance of developing engineering conceptual frameworks to interpret the emerging role of foundation models in future intelligent engineering systems are drafted.