As the requirements for industrial operation and military work, the frequency characteristics should be changed artificially sometimes. Active control is a good choice, but the current active control mainly focuses on time domain for vibration control. In this paper, the structural active control on frequency domain is studied through theory and experiment. Firstly, multivariable wavelet finite element method with two kinds of variables (TWFEM) which is suitable for modeling of great and complex structures with high efficiency and precision is used to construct the mathematical model for the controlled structure and do static and dynamic analysis. Then the control algorithm based on neural network including two parts, identification implement and controller is constructed. The present study takes frequency response as control objective, and can not only do vibration control but also change the vibration frequency characteristics, providing a new perspective for active control.