Abstract
Chatter, a self-excited vibration phenomenon, remains a critical limiting factor in the high-speed milling of thin-walled titanium alloy structures, particularly Ti-6Al-4V, severely impacting surface quality, tool life, and process stability. Traditional chatter suppression techniques often rely on offline stability predictions or fixed-gain control strategies, which struggle to adapt to the highly dynamic and non-linear nature of the machining process and varying workpiece geometries. This research proposes a novel adaptive chatter suppression system leveraging Deep Reinforcement Learning (DRL) to enable real-time optimization of cutting parameters. A DRL agent, trained in a simulated environment and fine-tuned on an experimental milling setup, learns an optimal policy to adjust spindle speed and feed rate based on multi-sensor feedback (vibration, force, acoustic emission). Experimental validation demonstrates that the DRL-controlled system achieves a significant reduction in chatter amplitude (up to 70%), resulting in superior surface finish (Ra improved by 45%) and enabling higher material removal rates compared to conventional fixed-parameter and baseline adaptive control methods. The developed DRL framework exhibits robust adaptability to varying cutting conditions and workpiece deflections, marking a substantial advancement towards intelligent and autonomous machining of challenging materials.
