In a multi-stage production system, a systematic optimization of stock levels can have a significant influence on the improvement of various logistical objectives. However, a complex production system and interdependencies between the logistical objectives make it difficult to determine an ideal stock situation. This thesis particularly focuses on the inventory optimization within a multi-stage production system with divergent material flow and a bottleneck control. Moreover, we consider capacity restrictions, stochastic customer ordering behaviour, working time fluctuations, machine malfunctions and supply uncertainties. In the empirical part, a simulation-based inventory optimization of a multi-stage production system is carried out, based on a practical case study. For this purpose, a generic simulation model is conceptualized, validated and verified by using the simulation tool DESMO-J. For this purpose, multi-objective genetic algorithms are used to determine an optimized inventory configuration. Furthermore, the constant adaptation of the input data enables a continuous optimisation of the current stock situation.