In this study, a fuzzy control system for the cleaning unit on a combine harvester was developed. In preceding studies considerable knowledge about the working principles of the cleaning unit has been gathered by means of databased models. Based on experimental data and fuzzy modelling techniques, optimum and non-optimum working conditions were identified. In this study, a fuzzy control system was developed which combines the knowledge of experienced operators with these databased models. The main goal of the controller for the cleaning section is defined as finding an optimum trade off between acceptable sieve losses and return volume for a desired level of MOG in the grain bin and a certain amount of incoming biomass. The performance and robustness of the control system was evaluated by field tests during wheat harvest. Promising results were obtained, demonstrating the benefits of an automatic control system for the cleaning section of a combine harvester when environmental conditions rapidly change over time.