A Sugeno type fuzzy controller is proposed for the automatic control of the crossflow microfiltration process for raw cane sugar. This process is becoming a field of increasing importance in the food industry. In most cases, and especially for sugar products, widespread applications of crossflow microfiltration is limited by the low permeate fluxes usually obtained. This limitation is linked to the difficulty to reduce the degree of membrane fouling (settling of particles on the membrane), a phenomenon which is dependent on several variables and difficult to control. In the present work, human expertise of the process is used to set up the fuzzy controller. The fuzzy controller is validated through simulations using a neural network model of this process and by real time experimentation on a pilot plant. The results of simulations and pilot tests show that it becomes possible to impose dynamics to the process which permits to maintain the state variable close to a given reference and to limit membrane fouling considerably. An off-line optimization of the fuzzy controller is performed using genetic algorithms. The cost results obtained during a real experiment, after genetic optimization of the fuzzy controller are much better than those obtained previously. Furthermore the tuning of the controller through optimization is realized under constraints which lead, after optimization, to a heuristic structure completely understandable by microfiltration experts.