Process-Centric Simulation To Optimize Complex Production Lines Considering Multi-Class Parts Combination

Modern manufacturing systems have become increasingly complex due to growing product customization, fluctuating demand, and inherent variability in processing times. These factors create significant challenges in effectively allocating the workforce within mixed-model assembly lines, where multiple product types share the same production resources. Addressing these challenges requires a systematic and adaptable decision-support framework capable of balancing production targets and labor utilization under uncertain conditions. This thesis introduces a Simulation-Guided Mathematical (SiGMa) approach that integrates mathematical optimization with discrete event simulation to solve the Mixed-Model Assembly Line Balancing Problem (MALBP). The proposed method aims to determine the optimal workforce size and allocation strategy while ensuring that production objectives are met. The SiGMa framework can accommodate complex system layouts involving both series and parallel workstations, stochastic task times, and varied product combinations and ratios. A mathematical programming model was formulated to represent the operational structure and practical constraints of such hybrid assembly systems. The model was implemented and solved using the PuLP optimization package in Python, which provided an optimized workforce configuration under diverse production conditions. To verify and validate the optimization outcomes, a Discrete Event Simulation (DES) model was developed in FlexSim, replicating the dynamic interactions and uncertainties present in real operations. The SiGMa approach was applied to a real-world case study conducted at an electrical structure manufacturing company. The study demonstrated that the proposed framework significantly improves workforce utilization and production efficiency while maintaining target output levels. Extensive simulation-based experiments, performed through FlexSim's experimenter tools, further confirmed the accuracy, reliability, and adaptability of the model across multiple production scenarios.