This paper proposes a novel adaptive genetic algorithm (AGA) for the multi-objective optimization design of a PID controller and applies it to the control of a real active magnetic bearing (AMB) system. The performances of the AGA are compared with that of the simple genetic algorithm (SGA) in optimizing dynamic responses of the controlled AMB. It shows that because of the proposed AGA can adjust the parameters adaptively according to the value of individual fitness and dispersion degree of population, this algorithm realizes the goals of maintaining diversity in the population and sustaining the convergence capacity of the genetic algorithm. The problems of convergence and prematurity occurred in SGA are then solved. The dynamic model of AMB system for axial motion is also presented, together with experimental and simulation results to verify its availability and good dynamic response.