A pneumatic servo-system has essentially non-linear elements such as compressibility of air, friction between cylinder and piston. The control scheme with fuzzy inference is effective for such non-linear plants. However, fuzzy rules in the fuzzy controller should be adjusted optimally to realize excellent control performance. 0n the other hand, the genetic algorithm can easily obtain the optimum solution in optimum problems. In this paper, we propose a fuzzy control for the vertical type pneumatic servo system in which the antecedent clause in the fuzzy controller is optimized by the genetic algorithm. In the genetic algorithm, a penalty function which regulates an excessive overshoot and a vibration of the plant output is proposed. Further, the effectiveness of the optimized scheme was confirmed by experiments using the existent vertical type pneumatic servo-system.