.png){kind=icon}
.png){kind=icon}
Return
Deep reinforcement learning-enabled closed-loop process optimization for aerosol jet printing
.png){kind=icon}
PRE
AI
DOI：10.1088/2058-8585/ae5830.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
To address process instability in aerosol jet printing (AJP), this study proposes AeroOptima-DRL, a deep reinforcement learning-based closed-loop process optimization system. A data-driven virtual environment, termed NeuroPrintMapper, is developed to model the relationship between process parameters and printed line quality, enabling efficient agent training without extensive physical experiments. Based on this environment, a reinforcement learning agent is trained to autonomously generate optimal process parameters in real time. Experiments on silver nanoparticle ink deposition on polyimide substrates demonstrate that the proposed system achieves accurate line width control within a target range of 10-30 mu m, with a mean absolute deviation of 0.19 mu m and an average error rate of 1.09%. During a 5 h continuous printing test, the closed-loop system effectively suppresses process drift, limiting resistance fluctuations to 2.5 times the initial value, compared to approximately 30-fold fluctuations under open-loop control. These results verify the effectiveness of DRL-based closed-loop optimization for improving the stability and consistency of high-resolution AJP processes.
Keywords:
aerosol jet printing
flexible electronics
deep reinforcement learning
closed-loop control
Journal
IF:
3.2
Papers:
124
Citations:
1.5K