土舱压力数据序列的混合降噪方法

Noise Reduction of Chamber Earth Pressure Data Sequence Using a Hybrid Approach

针对具有复杂趋势和季节成分的盾构机土舱压力数据序列,提出了一种基于经验模态分解(empirical mode decomposition,EMD)和奇异谱分析(singular spectrum analysis,SSA)的数据序列信噪分离新方法。首先采用经验模态分解对原始数据序列进行分解,以分离出包含噪声成分的固有模态函数(intrinsic mode function,IMF)。对这些固有模态函数进行序列重构,并利用奇异谱分析对重构后的序列进行奇异值分解(singular value decomposition,SVD)和重构,提取出其中的主要趋势成分和季节成分,同时得到包含噪声成分的残差序列,再次利用经验模态分解对残差序列进行处理,重复该过程直至得到满足纯随机性假设的噪声序列。采用该方法对土压平衡盾构机土舱压力监测数据进行降噪处理,分析结果表明该方法能够在保持原始数据序列结构的前提下准确提取出土舱压力数据中的噪声成分,验证了所提方法的有效性。

A new hybrid denoising approach based on singular spectrum analysis( SSA) and empirical mode decomposition( EMD) is proposed to reduce the chamber earth pressure data sequence with complex trends and periodicities. Firstly,the original data sequence is decomposed into intrinsic mode functions( IMF) using empirical mode decomposition method; and then the IMFs which contain noise components are reconstructed into a new data sequence for singular spectrum analysis. By decomposition and reconstruction of singular spectrum analysis,the characteristic components are extracted out from the new data sequence,meanwhile the remaining residuals are decomposed into IMFs again. This process iterates until the data sequence reconstructed from noise IMF's conforms to the pure random hypothesis. A chamber earth data sequence is applied to evaluate the performance. The results imply that the proposed approach can separate noise with useful signals without damaging the inner structure of the original sequence,which illustrates the validity of this new hybrid approach.