The increasing customized demand requires a manufacturing system that can quickly adapt its physical layout and corresponding production strategies to meet changes. The advancements of digital twin in Industry 4.0 provide a promising approach for dynamic optimizing a manufacturing system. This paper presents a digital twin-based optimization approach that provides a virtual simulation environment for dynamic configuration of physical layout and corresponding strategies of the Production-logistics systems. The proposed approach is encapsulated in software, providing a robust solution for modeling and optimizing manufacturing systems by incorporating randomness and dynamics between components. The software leverages Python’s Mersenne Twister algorithm to model the stochastic nature of job interarrival times, job types, and service times, ensuring accurate and reliable system performance analysis. Its modular design allows for reconfiguration of Production-logistics systems, by incorporating effective layout design and task allocation strategies, making it adaptable to diverse manufacturing environments with varying levels of dynamics and randomness. Statistical analysis from the simulation tests within a manufacturing system with six workstations and autonomous mobile robots (AMR) demonstrates the effectiveness of the proposed approach in optimizing layout configuration and task allocation under various dynamics and randomness. This work provides insights for designing and optimizing manufacturing systems through digital twin-based simulation and optimization.