Towards Industry 5.0: digital twin-enhanced approach for dynamic supply chain rescheduling with real-time order arrival and acceptance
Xing Zhu, Baoyu Liao, Yexing Shen, Min Kong
Abstract
In the Industry 5.0 era, digital twin (DT) technology enables a real-time connection between physical factory systems and virtual scheduling models, enhancing resilience and adaptability in response to market fluctuations. This paper introduces a resilient, human-centric, and sustainable DT-based dynamic scheduling framework tailored for supply chain rescheduling problems, particularly under dynamic order arrivals and acceptance. The static scheduling model focuses on minimising the total weighted tardiness of existing orders, while the dynamic model extends this by balancing disruption costs with potential penalties for rejecting orders. Within the DT-based framework, we integrate deep reinforcement learning (DRL) with a genetic algorithm (GA), utilising an Actor-Critic mechanism to select genetic operators dynamically. Extensive computational experiments demonstrate that the proposed DT-based framework substantially enhances supply chain resilience, offering manufacturers a sustainable and human-centric solution aligned with Industry 5.0 goals in the face of volatile demands.