A Closed Loop Reinforcement Learning Framework for Producing Geometry Optimized Robotic Toolpaths for 3D Printing with Cold Spray

Cold spray additive manufacturing (CSAM) holds significant potential for advanced material manufacturing and repair, yet it lacks control strategies to achieve near net parts. Here we present a novel reinforcement learning (RL) based simulation framework designed to optimize the cold spray additive manufacturing process. Our proposed framework leverages reinforcement learning to intelligently optimize the robot trajectory as per the user defined process parameters. Our framework's novelty lies in its ability to generate an optimal trajectory via combinatorial optimization of trajectory parameters, dynamically adapting to the complexities of the cold spray process. Through extensive simulation studies and physical sprays, we demonstrate our framework's effectiveness in producing complex geometries while minimizing over-deposition, validating the accuracy and reliability of the simulation environment for optimal use of cold spray for producing near net-shaped 3D printed parts. This research contributes to advances in the modeling and optimization of cold spray processes by introducing an RL framework for trajectory optimization. This framework enables intelligent decision-making in real-time by optimizing spray parameters such as velocity, stand-off, and angle to achieve desired surface geometries. By integrating this data-driven approach with CSAM, we offer a scalable solution for enhancing deposition accuracy, material efficiency, and overall process performance.