Two-phase extreme learning machines integrated with the complete ensemble empirical mode decomposition with adaptive noise algorithm for multi-scale runoff prediction problems

Abstract The prediction of annual runoff in the Lower Yellow River can provide an important theoretical basis for effective reservoir management, flood control and disaster reduction, river and beach management, rational utilization of regional water and sediment resources. To solve this problem and improve the prediction accuracy, permutation entropy (PE) was used to extract the pseudo-components of modified ensemble empirical mode decomposition (MEEMD) to decompose time series to reduce the non-stationarity of time series. However, the pseudo-component was disordered and difficult to predict, therefore, the pseudo-component was decomposed by ensemble empirical mode decomposition (EEMD). Then, intrinsic mode functions (IMFs) and trend were predicted by autoregressive integrated moving average (ARIMA) which has strong ability of approximation to stationary series. A new coupling model based on MEEMD-ARIMA was constructed and applied to runoff prediction in the Lower Yellow River. The results showed that the model had higher accuracy and was superior to the CEEMD-ARIMA model or EEMD-ARIMA model. Therefore, it can provide a new idea and method for annual runoff prediction.

Download Full-text

Fault feature extraction of rotating machinery using a reweighted complete ensemble empirical mode decomposition with adaptive noise and demodulation analysis

Mechanical Systems and Signal Processing ◽

10.1016/j.ymssp.2019.106545 ◽

2020 ◽

Vol 138 ◽

pp. 106545 ◽

Cited By ~ 20

Author(s):

Liming Wang ◽

Yimin Shao

Keyword(s):

Feature Extraction ◽

Empirical Mode Decomposition ◽

Ensemble Empirical Mode Decomposition ◽

Rotating Machinery ◽

Mode Decomposition ◽

Adaptive Noise ◽

Fault Feature Extraction

Download Full-text

Adaptive Ultrasound Tissue Harmonic Imaging Based on an Improved Ensemble Empirical Mode Decomposition Algorithm

Ultrasonic Imaging ◽

10.1177/0161734619900147 ◽

2020 ◽

Vol 42 (2) ◽

pp. 57-73

Author(s):

Suya Han ◽

Yufeng Zhang ◽

Keyan Wu ◽

Bingbing He ◽

Kexin Zhang ◽

...

Keyword(s):

Empirical Mode Decomposition ◽

Second Harmonic ◽

Ensemble Empirical Mode Decomposition ◽

Harmonic Imaging ◽

Motion Artifacts ◽

Intrinsic Mode Functions ◽

Tissue Harmonic Imaging ◽

Mode Decomposition ◽

Adaptive Noise ◽

Harmonic Components

Complete and accurate separation of harmonic components from the ultrasonic radio frequency (RF) echo signals is essential to improve the quality of harmonic imaging. There are limitations in the existing two commonly used separation methods, that is, the subjectivity for the high-pass filtering (S_HPF) method and motion artifacts for the pulse inversion (S_PI) method. A novel separation method called S_CEEMDAN, based on the complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) algorithm, is proposed to adaptively separate the second harmonic components for ultrasound tissue harmonic imaging. First, the ensemble size of the CEEMDAN algorithm is calculated adaptively according to the standard deviation of the added white noise. A set of intrinsic mode functions (IMFs) is then obtained by the CEEMDAN algorithm from the ultrasonic RF echo signals. According to the IMF spectra, the IMFs that contain both fundamental and harmonic components are further decomposed. The separation process is performed until all the obtained IMFs have been divided into either fundamental or harmonic categories. Finally, the fundamental and harmonic RF echo signals are obtained from the accumulations of signals from these two categories, respectively. In simulation experiments based on CREANUIS, the S_CEEMDAN-based results are similar to the S_HPF-based results, but better than the S_PI-based results. For the dynamic carotid artery measurements, the contrasts, contrast-to-noise ratios (CNRs), and tissue-to-clutter ratios (TCRs) of the harmonic images based on the S_CEEMDAN are averagely increased by 31.43% and 50.82%, 18.96% and 10.83%, as well as 34.23% and 44.18%, respectively, compared with those based on the S_HPF and S_PI methods. In conclusion, the S_CEEMDAN method provides improved harmonic images owing to its good adaptivity and lower motion artifacts, and is thus a potential alternative to the current methods for ultrasonic harmonic imaging.

Download Full-text