Based on flexure hinge and piezoelectric actuator of two-axis fast steering mirror is a complex system with time varying,uncertain and strong coupling.It is extremely difficult to achieve high precision decoupling control with the traditional PID control method.The feedback error learning method was established an inverse hysteresis model which was based inner product dynamic neural network nonlinear and no-smooth for piezo-ceramic.In order to improve the actuator high precision,a method was proposed,which was based piezo-ceramic inverse model of two dynamic neural network adaptive control.