An integrated data-driven adaptive architecture for intelligent robotic wire arc additive manufacturing

Abstract

This thesis focuses on the use of industrial robots for Wire Arc Additive Manufacturing (WAAM), a widely adopted directed energy deposition process for metal additive manufacturing. Compared to processes such as laser powder bed fusion, WAAM offers higher deposition rates and enables large-format fabrication when integrated with robotic systems. Despite these advantages, achieving high-quality robotic WAAM remains challenging due to kinematic inaccuracies, multi-robot coordination complexity, and strong coupling between robot motion, thermal behavior, and material properties. This work addresses key challenges in robotic WAAM optimization, including multi-robot calibration, motion planning, and in-process adaptive control. Conventional calibration approaches rely on static models that do not maintain accuracy across the workspace. To enable high-precision WAAM, a configuration-dependent kinematic calibration framework is developed to compensate for pose-dependent, non-geometric factors. Kinematic parameters are identified at multiple configurations and interpolated into a global model, achieving improved accuracy across the workspace compared to conventional methods. A direct dual-arm calibration framework is further introduced, relying solely on relative tool-to-tool pose measurements and eliminating the need for external metrology systems. Building upon accurate robot kinematics, the thesis develops a data-driven, sensor-integrated framework for WAAM process modeling and control. A motion planning strategy ensures both deposition requirements and sensing coverage. Using in-process data, data-driven input–output dynamical models with observer-inspired innovation feedback are developed to capture geometry evolution and are incorporated into a model predictive control (MPC) framework for adaptive regulation. In addition, a thermal dynamics model based on multi-sensor infrared measurements is developed to represent the temperature field, enabling predictive thermal regulation through a multi-step MPC strategy. Experimental validation on the WAAM testbed demonstrates improved geometric consistency and reduced reliance on manual parameter tuning. Overall, this thesis presents an integrated methodology combining robot calibration, coordinated motion planning, data-driven modeling, and adaptive feedback control to improve the precision and robustness of robotic WAAM systems.