Abstract Smart structures with build-in sensor/actuator that can actively and adaptively change their physical geometry and properties are shown to have significant potential in engineering applications. The implementation of neural networks to system identification and vibration suppression of smart structures is conducted in this paper. Three neural networks are developed, one for system identification, the second for online state estimation, and the third for vibration suppression. It is shown both in analysis and in experiment that these neural networks can identify, estimate, and suppress the vibration of a composite structure with embedded piezoelectric sensor and actuator. The controller is also shown to be robust to system parameter variations.