Research On Feature Extraction And Prediction Of Deformed Signal Based On Local Mean Decomposition

Deformation monitoring and analysis is an important component of safety monitoring system, the reliability and precision of the analysis method plays an important role in maintaining the safe operation of the deformation. It is the key problem in deformation monitoring to treatment deformation data, extract characteristic vector and conduct deformation prediction, and it is also one of researching hot-spots for the disciplines of surveying and mapping. In 2005, Local mean decomposition, a new adaptive non-stationary signal processing method, was proposed by Jonathan s. Smith. The characteristics of this method is that it can adaptive to more effective resolve the complex non-stationary signal into several instantaneous frequency and the sum of PF components with physical significance, and it provided a new research method for forecasting and extracting the deformation feature. This paper is ground on the item of National Natural Science Fund, prediction model and extract deformation feature based on local mean decomposition was studied, and combined with the simulation signal and the dam data of deformation monitoring to analyzes and discusses the related problems.This paper is based on local mean decomposition principle and method. In-depth analyzed the LMD method, and carried on significantly improve in the algorithm, with the size of the average distance of adjacent a third as a moving average of the step length as improvements-LMD decomposition method. Comparison and analysis LMD and EMD method, and it is easy to found that the LMD method is less number of iterations and better than the EMD method in overcome the energy leakage. Research shows that, as the improvements-LMD decomposition method, the step length in the two aspects of computational efficiency and precision of decomposition is better than traditional LMD, CBI-LMD and other decomposition method. And the method of PF component spectrum distribution is more fully ready to extract the feature information, provided technical support for greatly the setted forecast model.Finally, established the combination of LMD-based BP neural network forecast model, feature extraction and the analysis of data deformation forecast model,the effect of model prediction is more better; the other one MLMD fast approximate entropy- LSSVM forecasting model, based on the displacement of dam deformation data, and compared the results with single LSSVM model, the SVM and BP neural network model, it indicated that the MLMD-fast approximate entropy- LSSVM model, the mean absolute error of MAE, mean absolute percentage error MAPE, RMSe and root mean square error, significantly higher than the precision of using a single LSSVM model, error of SVM and BP neural network model. It shows that the deformation characteristics of the displacement of the dam could be described well by the prediction model, not only provides effective technical support to predict departments, also provides an effective thinking deformation prediction for later research.