Autonomous Layer Time Optimization in Large-format Additive Manufacturing through Machine Vision and Feedback Control

Abstract

Proper thermal management in large-format additive manufacturing (LFAM) to maintain geometric control and mechanical integrity requires, in part, maintaining the substrate, or the previously deposited layer temperature, within a defined process window. Temperatures that are too low result in weak interlayer bonds; conversely, elevated temperatures cause geometric instability. This work presents an autonomous closed-loop control system that modulates feedrate to maintain a setpoint substrate temperature using on-gantry thermal sensing. Geometry-independent monitoring of material deposition was enabled by six radiometric cameras mounted around the deposition nozzle. An edge-deployed computer vision pipeline enabled automatic identification of layer boundaries and extraction of substrate temperature without manual region-of-interest assignment or offline workstation process, both of which preclude closed-loop control. A proportional-integral (PI) controller with Internal Model Control (IMC) tuning adjusted the feedrate based on the measured temperature error. The layer-synchronous architecture triggered control actions at layer completion events rather than fixed time intervals. The autonomous control system was validated on a Big Area Additive Manufacturing platform during printing of wood-filled polylactic acid. The experiment demonstrated that the controller successfully recovered from substrate temperatures ~20°C below the setpoint to achieve steady-state regulation of 124.3°C ± 1.8°C against a 125°C setpoint. Notably, the controller, without recalibration, was able to adapt to increasing part thermal mass by automatically reducing steady-state feedrate from 57% to 52% between two controller activation periods, early and late in the build, to maintain the same 125°C setpoint.