Abstract

Manufacturing environments are increasingly characterized by high complexity, low production volumes, and a growing diversity of product variants. Simulation is a powerful tool for analyzing such systems. However, simulations are still expert tools, offering limited accessibility for less experienced planners. This paper presents a hybrid framework that integrates numerical optimization, discrete-event simulation, and large language models to support production planners. The approach explores the solution space for optimizing work-force allocation in semi-automated work cells. A large language model-based approach is introduced to support simulation studies by reducing manual configuration efforts and enabling systematic experimentation. The method is applied in two automotive aftermarket plants with varying levels of automation to identify optimal worker-machine allocations. By integrating large language model capabilities into established simulation-optimization workflows, the approach contributes to the gradual development of autonomous simulation studies.