Reinforcement Learning for Statistical Process Control in Manufacturing

The main concept of the authors is to place Reinforcement Learning (RL) into various fields manufacturing. As one first implementations, RL for Statistical Process Control (SPC) in production introduced paper; it a promising approach owing its adaptability and continuous ability perform. widely used Q-Table method was applied get more stable, predictable, easy overview results. Therefore, quantization values time series stripes inside control chart introduced. Detailed elements environment simulation are described interaction with reinforcement learning agent detailed. Beyond working adapting SPC manufacturing, some novel extensions also described, like epsilon self-control exploration–exploitation ratio, Reusing Window (RW) Measurement (MW). In related transformation, aim optimize cost while keeping ratio good products on high level as well. Finally, industrial testing validation that proved applicability proposed concept.