• 목록

[Abstract]

Maximizing product quality while sustaining high productivity remains a key challenge in multistage manufacturing, where delays or scrap can rapidly escalate costs. Semiconductor fabrication exemplifies this difficulty, as each product passes through multiple sequential stages with parallel machines operating under varying conditions. This study presents a reinforcement learning-based scheduling framework that jointly optimizes productivity and quality. The proposed framework integrates systematic analysis of production data, machine-level quality characterization, and learning-based scheduling optimization. By combining predictive modeling with adaptive decision-making, it enables dynamic trade-offs between productivity and quality, thereby enhancing overall manufacturing performance.