In order to solve the difficulty of damage diagnosis for hydraulic structure during operation, a valid damage diagnosis method based on permutation entropy is proposed. Firstly, a kind of combined wavelet threshold and empirical mode decomposition (EMD) method is used to filter out the noises, which can reduce the disturbance of environmental excitation in structural damage characteristic information to improve the signal-to-noise ratio. Then the de-noised signals are analyzed by permutation entropy to extract their entropy values. Finally, the damage identification of hydraulic structure can be achieved by the contrast of entropy values under different working conditions. The proposed method is applied to study the vibration responses of cantilever beam model under environmental excitation. The results show that the permutation entropy of structural vibration signal is the largest in the normal state without damage. When there are certain damages, the permutation entropy decreases, the greater the damage degree, the smaller the permutation entropy. Permutation entropy is sensitive to structural initial damage. Permutation entropy is almost constant under different working conditions when there is no damage to structure. Therefore, the proposed method can effectively identify structural damage with high diagnostic accuracy.