局部均值分解的端点效应优化及应用

发布时间：2018-11-11 07:56

【摘要】：局部均值分解(Local Mean Decomposition,LMD)是一种新的能自适应处理非平稳信号的方法。LMD方法作为新提出的一种数据分解方法,在算法上还存在着一些不足之处,如端点效应问题。端点效应使数据分解的分量在两端发生了形变,在使用LMD方法对变形数据进行特征提取与变形预测分析时,端点效应的影响将被放大。文中将使用新的方法抑制LMD的端点效应,并结合实例进行分析。对LMD方法的理论进行研究和分析,总结LMD方法存在的问题和探讨方向。对LMD产生端点效应的原因进行分析,并用仿真信号说明端点效应出现的原因。使用一种新的延拓方法,即基于极值点和距离的SVM延拓方法,以实际工程信号数据为例证进行验证,说明了本延拓方法能够在一定程度上抑制LMD的端点效应问题。使用另一种抑制LMD端点效应的方法,即抑制LMD端点效应的自适应选权迭代法。并使用本优化的LMD方法对大坝的变形数据进行多尺度分解和分析,并与其它常用的抑制LMD端点效应的方法进行比较,验证了优化的LMD方法能够有效的对变形数据进行特征提取。同样,采用优化的LMD方法对一百多年海平面变化数据进行趋势项提取,也与其它常用的抑制LMD端点效应的方法进行比较,验证了优化的LMD方法相比于其他方法,能够更加有效的提取出海平面变化的趋势项。利用优化的LMD方法(抑制LMD端点效应的自适应选权迭代法)与SVM模型和GM(1,1)模型组成优化的LMD-SVM-GM(1,1)模型,用两组不同的实例进行验证,分别是蓄能水电厂的大坝监测数据和鄱阳湖入湖总水量数据。对监测数据进行分解,将分解的余量采用GM(1,1)模型进行预测,其他的分量采用SVM模型进行预测,将各自的预测结果进行重新融合,得到最后的预测结果。通过实例验证,文中优化的LMD-SVM-GM(1,1)对多尺度的变形监测数据具有适用性,值得推广。
[Abstract]:Local mean decomposition (Local Mean Decomposition,LMD) is a new method for adaptive processing of non-stationary signals. As a new data decomposition method, LMD method has some shortcomings, such as the endpoint effect problem. The end point effect causes the component of data decomposition to deform at both ends. The effect of the end point effect will be magnified when the feature extraction and deformation prediction analysis of the deformation data are carried out by using LMD method. In this paper, a new method is used to suppress the endpoint effect of LMD, and an example is given to analyze it. The theory of LMD method is studied and analyzed, and the problems and direction of LMD method are summarized. The cause of endpoint effect in LMD is analyzed, and the reason of endpoint effect is explained by simulation signal. A new continuation method, the SVM extension method based on extreme point and distance, is used to verify the actual engineering signal data. It is shown that the method can suppress the endpoint effect of LMD to some extent. Another method to suppress the end point effect of LMD is adaptive weight selection iteration method, which is used to suppress the end point effect of LMD. The optimized LMD method is used to decompose and analyze the deformation data of the dam at multiple scales, and compared with other commonly used methods to suppress the LMD endpoint effect. It is verified that the optimized LMD method can extract the deformation data effectively. Similarly, the optimized LMD method is used to extract the trend term of sea level change data for more than one hundred years. It is also compared with other commonly used methods to suppress the endpoint effect of LMD. The results show that the optimized LMD method is compared with other methods. It can extract the trend of sea level change more effectively. Using the optimized LMD method (adaptive weight selection iteration method to suppress the LMD endpoint effect) and the SVM model and the GM (1K1) model to form the optimized LMD-SVM-GM (1K1) model, two different examples are used to verify the optimization. The dam monitoring data of power storage power plant and the total water volume of Poyang Lake entering the lake are respectively. After decomposing the monitoring data, the residual of the decomposition is predicted by GM (1 / 1) model, the other components are predicted by SVM model, and their prediction results are re-fused to obtain the final prediction results. It is proved by an example that the optimized LMD-SVM-GM (1 / 1) is applicable to the multi-scale deformation monitoring data and is worth popularizing.
【学位授予单位】：东华理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN911.7

【相似文献】