移动平均和经验模态分解法在大坝位移分离中的应用

Application of Moving Average Method and Empirical Mode Decomposition Method in Separation of Dam Displacement

经验模态分解方法由于缺少数据预处理,分解结果具有冗余性、与自变量数量不能对应、模函数无法进行物理解释等问题,因此在经验模态分解前增加移动平均数据预处理过程,以减少原始数据量,保留趋势过程,并以某双曲拱坝位移分离为例,对比了经过数据预处理与未经数据预处理的经验模态分解结果。结果表明,预处理后位移序列趋势性不变,但数据量和分解模函数数量均有所减少;为了使分解得到的模函数与水位和温度的物理解释相吻合,选择合适的移动平均参数M,使模函数数量为2,避免了分解的冗余性。

Due to the lack of data preprocessing, empirical mode decomposition＇s results have the problem of redun- dancy, the number of arguments, and not corresponding to the physical interpretation, adding the moving average data pretreatment process in the empirical mode decomposition method can reduce the amount of raw data and retain the trend process. Taking displacement separation of a certain double-curvature arch dam for an example, empirical mode decomposition＇s results are compared with and without data preprocessing. The results show that after the pretreatment of displacement, time series trend remains unchanged, but the amount of data and the number of decomposition mode function are reduced. In order to make the model function correspond to the variables of water level and temperature, it can choose appropriate parameters M to make the modular function as 2, which avoids the redundant decomposition.