Signal de-noising method for vibration signal of flood discharge structure based on combined wavelet and EMD

A new de-noising method combining Wavelet threshold and empirical mode decomposition (EMD) (WTEMD for short) is proposed to improve the precision of de-noising performance for vibration signal of flood discharge structure in low signal to noise ratio (SNR). White noise is partially filtered out by decomposing the vibration signal with wavelet. Then conducting the further EMD on wavelet reconstructed signal to obtain Intrinsic Mode Function (IMF), through analyzing spectrum diagram of every IMF component, low-frequency waterflow noise and the rest of high-frequency white noise are filtered out, regarding SNR and root mean square error (RMSE) as evaluation index for noise reduction effect. The novelty of this method is that it can reduce the endpoint effect of EMD. By comparing the filtering effect of WTEMD with other methods on simulation signals, study shows that, WTEMD has a higher precision and a better de-noising effect. The dominant vibration information of dam structure is achieved by using WTEMD in Laxiwa arch dam hydro-elastic model and Three Gorges Dam, which can provide the basis for safe operation and on-line monitoring of the dam structure. This method can effectively solve the problem of dominant information extraction for large flood discharge structure.