Production scheduling optimization for manufacturing cells in smart factory

Factories worldwide are increasingly focused on sustainable and resilient production, with Industry 5.0 driving the development of self-organized and self-adapted intelligent decision-making brains. This study introduces a comprehensive model to achieve self-organization and self-adaptation in smart factories. The model addresses complex challenges, including multi-type products, multi-stage operations, limited durations, and varying cell capacities. By integrating assignment and scheduling into a mixed integer programming framework, the model aims to minimize makespan. A column generation-based method is employed to solve the problem, achieving zero optimality gap in experiments. The algorithm can solve large-scale cases within 831 s on average. The method improved factory efficiency by 0.8% and 1.1 % across two industrial datasets, with each 1% efficiency gain potentially adding up to $4.6 million in annual revenue. The findings suggest that optimizing batch size based on product complexity and quantity is crucial, as overly homogeneous or excessively diverse batches should be avoided. Additionally, versatile cells do not always result in higher efficiency, as this can depend on factors such as order size and process time variability. This method not only enhances productivity but also provides a solid foundation for production decision-making and the development of intelligent systems.