National High-tech R&D Program of China (863 Program) (2013AA040403).
In order to remove noise from grain flow signal, this paper investigated a new method using empirical mode decomposition (EMD). Grain flow signal with noise is decomposed adaptively into intrinsic mode functions (IMFs) that contain specific frequency through sifting. Minimum energy criterion (MEC) is proposed to determine the partial IMFs (low-frequency components) that are used to reconstruct a new grain flow signal. The other IMFs (high-frequency components) are removed as noise. A test rig was built to investigate performance of the proposed method. The test rig was equipped with impact plate sensor and weighting sensor. Data from both impact plate sensor and weighting sensor were analyzed to verify the improvement of accuracy using the proposed method. The results indicate that the proposed method have a higher accuracy of grain flow signal measurement. Compared to other reduction noise method, the proposed method can effectively suppress noise improve noise-signal ratio (SNR). The relative error of total grain mass was less than 1.6 %, which is also better than other reduction noise method. The proposed method provides a new way for preprocessing of grain flow signal.