Mechatronic systems are conventionally modeled by integer order state equations for simplicity. Based on these integer-order models, integer-order PID or other classical model-based controllers are designed. However, often complex mechatronic systems require fractional-order state equations for a more adequate description. In this paper, a Steward platform (ball and plate system) with different balls is used to demonstrate this. A fractional-order PD controller is designed and tuned by a genetic algorithm based global optimizer to obtain the suitable parameter values in time domain. The experimental results show that the fractional-order PD controllers can provide better performances than conventional integer-order PD controllers.