Cascade-Correlation (CC) network is a new architecture and supervised learning algorithm for artificial neural networks. The learning algorithm of CC network and its network structure are described in this paper, the CC network with an excellent fitting ability is applied to fitting vibration characteristics of hydraulic turbine units according to different parts under three water heads. Compared with the BP network, the simulation experiments demonstrates that the CC network has a faster convergence speed and a higher accuracy, it is much closer to true to describe the vibration characteristics of hydraulic turbine units under different working conditions for their parts than its counterpart.