Search for: All records

Toolpath design plays a significant role in determining the efficiency of Additive Manufacturing (AM) processes. Traditional toolpath optimization methods frequently depend on empirical methods, which may not adequately account for the complex dynamics of the printing process. This study introduces a novel reinforcement learning (RL) approach, leveraging Proximal Policy Optimization (PPO), to optimize toolpath generation with a particular focus on reducing energy consumption. A custom-built environment was created, simulating the toolpath planning scenario as a discrete grid space, where an RL agent representing the printing nozzle learns to navigate and optimize its path. The RL agent, implemented using Proximal Policy Optimization (PPO), was trained on grids of increasing complexity (10x10 and 25x25) using two reward systems: a default system and an energy-optimized system based on a custom energy model. The energy model penalizes energy-intensive vertical and diagonal movements while rewarding horizontal movements. Results from training showed that the energy-optimized model achieved a significant reduction in energy consumption without compromising toolpath efficiency. On the 10x10 grid, energy consumption decreased from 92.7 𝑘𝑊𝑚𝑠 to 83.5 𝑘𝑊𝑚𝑠, while on the 25x25 grid, it dropped from 400.2 𝑘𝑊𝑚𝑠 to 395.4 𝑘𝑊𝑚𝑠. Statistical analysis using paired t-tests confirmed these reductions with p-values of 0.00, demonstrating the effectiveness of incorporating energy constraints in RL training for AM. This research highlights the potential of RL in improving the sustainability and efficiency of AM processes through intelligent toolpath design.