In the current context of Industry 5.0, sustainable scheduling has emerged as an evolution of classical scheduling, now integrating environmental and human-centric considerations. The objective is to strike a balance between economic, environmental, and societal concerns. Additionally, there is a growing need to enhance the resilience in the Industry 5.0 era, necessitating dynamic systems capable of reacting to unforeseen disruptive events. This paper introduces a multi-objective dynamic scheduling model designed to simultaneously minimize the makespan, the energy consumption, and a standardized ergonomic risk factor. The initial schedule is generated using a non-dominated sorting algorithm (NSGA-III), and in the event of worker absence during production, a rescheduling process is initiated. The choice of rescheduling strategy is determined using a Q-learning algorithm, allowing continuous improvement in the selection of the optimal strategy depending on the scenario. Results, derived from experiments conducted on literature instances, demonstrate the model's effectiveness in swiftly generating new efficient schedules. The motivation for this research stems from the increasing demand for sustainable industrial practices that not only enhance productivity but also reduce environmental impact and improve worker well-being, thereby contributing to the development of more resilient and sustainable industrial systems.