Abstract: Aiming at the problem of vibration suppression to large and nonlinear structures, the control strategy of adaptive wavelet algorithm was proposed based on on-line system identification method. Combining the wavelets and least mean square (LMS) algorithm, this paper focused on adaptive vibration control to body structure by using decomposition LMS algorithm. Experiments research on active vibration control was carried out with the piezoelectric elements as sensors and actuators. Firstly, adaptive LMS wavelet control strategy was analyzed and adopted. Input signals were decomposed into a series of different frequency bands through a set of band pass filters. By using LMS algorithm to deal with each component of frequency bands, parameter matrix equation of the adaptive controller was obtained. Secondly, experimental modal analysis and system identification of the car body were carried out, and the mathematical model of the system was obtained by experimental method. A car body was hung on the bracket, and it was arranged with 106 test points. In the center of the wheel, a vibration exciter was arranged at the selected frequency to stimulate the car body to vibrate. Vibration signals to each measuring point were digitized by data acquisition and analysis system. Modal parameters and frequency response function of the structure were obtained through parameter identification. Using the modal analysis software, the experimental function curves were fitted to the modal vibration mode. According to the input and output to the measuring points, the system identification toolbox of MATLAB was used to obtain the structural parameters matrix, and to establish the mathematical model of the body structure. The mathematical model would be used for vibration control experiments as the control object in the adaptive wavelet control system. Thirdly, parallel on-line system identification method was applied. Another adaptive digital filter was introduced into the traditional system identification method, and the two filters were all performed according to the LMS algorithm. The on-line system identification method had the characteristics of good real-time performance, high identification accuracy, easy implementation and simple structure. It could greatly improve the adaptability of the adaptive wavelet control system. MATLAB software was applied to establish adaptive wavelet control system based on parallel on-line system identification method. Identification was integrated into the control behavior, and the functions of identification and control were automatically accomplished in the control process. Finally, experiments based on the adaptive vibration control system were carried out. The experimental platform to vibration control was built by using the piezoelectric elements as the sensors and actuators. Vibration exciter with random signal motivated the car body to vibrate. The vibration signals were detected by piezoelectric sensors, and the signals were filtered and amplified by charge amplifiers to send into computer through data acquisition card. The data were changed into the control signals by adaptive wavelet control program to drive the piezoelectric actuators to produce deformation. The body panels of nested together were driven in a synchronous deformation to realize the goal of adaptive response to the vibration and adaptive adjustment to the vibration deformation. Comparing the vibration signals before controlled with the signals after controlled, it can be seen that the vibration amplitude to the body panels was reduced about 60% because of the application of the adaptive wavelet control. In the main range of vibration frequency, the control system had obvious vibration suppression effect. Especially in the low frequency region, where the vibration amplitude was relatively large, and the control effect was very good. The effectiveness of control system is verified to show that the adaptive wavelet control system, for uncertain vibration on large and nonlinear structures such as car body, can achieve good control results.