Based on the radial basis function (RBF) neural network, a mathematical model of vibration load for active controlled flap (ACF) is established. In the model, the trailing-edge-flap angle is as input and the rotor hub oscillating loads are as output. An orthogonal experimental method is used for the selection of the training set, then the training samples are calculated in CAMRAD II software and the RBF network is trained off-line by the orthogonal samples. With the aid of RBF network, a multicyclic controller is given for the active control of rotor vibration loads. In order to verify the effectiveness of the network and the controller, the vibration loads of a two-blade ACF rotor with diameter 4 m are analyzed. Numerical results indicate that the vertical 2Ω load is reduced to almost 50%, and the 2Ω load in the other directions shows different degrees of reduction.