Search

End effects of EMD (empirical mode decomposition) make a serious distortion of the decomposition result. In order to reduce the end effects in the process of decomposition, support vector machine (SVM) which is a kind of intelligent algorithm is combined with EMD, then a solution to the end effects problem during the course of decomposition using SVM model is proposed. Firstly, one or more extreme values are obtained by extending two endpoints of the original data with SVM. Moreover, in order to get more reasonable extension at endpoint, SVM algorithm is combined with particle swarm algorithm (PSO) to optimize the parameters, and the extension of two endpoints will be more accurate, then the end-points of cubic spline curve will not have large swing so as to achieve that intrinsic mode functions (IMF) of EMD are more accurate and reliable. Simulation results indicate that the extension method for data based on PSO-SVM method can restrain the end effects effectively. [ABSTRACT FROM AUTHOR]

A technique to predict the gas emission from the coalface is presented which is realized by the intrinsic mode SVM modeling on the basis of the intrinsic mode components drawn out from the observed dada by means of the EMD (empirical mode decomposition) method. The intrinsic mode functions, that is, IMF components, are obtained by the EMD analysis of the historical recording dada of gas emission, and after the prediction of each intrinsic mode is carried out by the extrapolation of its regression function determined by the SVM flmction regression approach, then the prediction result of gas emission is derived through the reconstruction summing all prediction results corresponding to different intrinsic modes. From an application example related to the monitoring data it can be seen that the introduction of EMD method into ther theoretical modeling to predict the gas emission from the coalface obviously improves the accuracy in comparison with the conventional SVM method, to have the prediction results agreement with the monitoring data. The theoretical analysis shows that in modeling the gas emission from the coalface, the extraction of intrinsic modes and the operation of SVM method make full use of the sampling data driven adaptive performances, and hence provide better theoretical flmdamentals for guarding perfect prediction efficiency. [ABSTRACT FROM AUTHOR]