Factory Layout Design and Optimization using Production Simulation and Reinforcement Learning

Designing the factory layout is important process that has a significant impact on productivity, logistics and manufacturing costs. The design of factory layout require careful consideration of the various factors mentioned above, and the relationships between them become more complex as the scale of the factory expands Implementing a new production line or making layout changes due to process redesign, such as adjustments in equipment count, process, and logistics, demands considerable investment of time and resources for design. However, it is difficult to know if the layout is optimal. It is necessary to find methods to reduce the costs and discover the optimal layout. This study explores the method of determining optimal factory layouts, considering productivity and logistics, using digital twin and reinforcement learning algorithms. Through this methodology, the research deviates from conventional rules traditionally considered optimal in existing layouts. The study focuses on finding an optimal layout that maximizes performance metrics(productivity, logistics) while satisfying given conditions. Utilizing design data containing information about the equipment, a digital twin model is automatically generated and the results are predicted through simulation. The predicted outcomes are evaluated and updated by the reinforcement learning algorithm, iterating the process of creating a digital twin model and predicting results based on the updated design data. Through this iterative process, the study gradually identifies layout with better performance metrics. This study demonstrates the possibility of deriving optimal layouts under complex layout design conditions. The proposed methodology is expected to the identifying optimal layouts for the varying conditions.