Reinforcement learning-enabled design of topological interlocking materials for sustainable multi-material additive manufacturing

Additive manufacturing (AM) has significantly expanded the possibilities to design sophisticated shapes and structures with unique properties and materials to achieve unprecedented functionalities. A notable trend in AM is the integration of multiple materials within a single structure to achieve multifunctionality while minimizing part count. However, multi-material AM presents inherent challenges, particularly in terms of printability constraints and environmental considerations, such as the recyclability of composite structures. Although the current effort in hybrid AM offers a partial solution to address some of these challenges, material versatility and sustainable disassembly remain major barriers. This research aims to introduce a computational interlocking design strategy for multi-material AM on a voxel basis, thus enabling controlled material disassembly and reuse. Reinforcement learning, especially Q-learning, is employed to optimize and explore the spatial arrangement of topological interlocking materials in the three-dimensional design space, which facilitates modularity while maintaining structural stability. Implemented via a Python-based computational framework interfaced with a computer-aided design environment, this approach is validated across various structural configurations, including cubic, beam, and irregular shapes. Our findings demonstrate a path toward sustainable, reusable, and recyclable multi-material AM, offering new possibilities for circular manufacturing and resource-efficient design.