Improved CEEMDAN–wavelet transform de-noising method and its application in well logging noise reduction. (5th March 2018)

• Record Type: Journal Article
• Authors: Zhang, Jingxia; Guo, Yinghai; Shen, Yulin; Zhao, Difei; Li, Mi
• Is Part Of: Journal of geophysics and engineering. Volume 15:Number 3(2018:Jun.)
• Page Start: 775
• Page End: 787
• Publication Date: 2018-03-05
• Subjects: well logging -- lithological identification -- noise reduction -- EMD -- wavelet transform -- improved CEEMDAN; Geophysics -- Periodicals; Prospecting -- Geophysical methods -- Periodicals; Engineering -- Periodicals; 622.1505
• DOI: 10.1088/1742-2140/aaa076
• Languages: English
• ISSNs: 1742-2132

Abstract

Abstract: The use of geophysical logging data to identify lithology is an important groundwork in logging interpretation. Inevitably, noise is mixed in during data collection due to the equipment and other external factors and this will affect the further lithological identification and other logging interpretation. Therefore, to get a more accurate lithological identification it is necessary to adopt de-noising methods. In this study, a new de-noising method, namely improved complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN)–wavelet transform, is proposed, which integrates the superiorities of improved CEEMDAN and wavelet transform. Improved CEEMDAN, an effective self-adaptive multi-scale analysis method, is used to decompose non-stationary signals as the logging data to obtain the intrinsic mode function (IMF) of N different scales and one residual. Moreover, one self-adaptive scale selection method is used to determine the reconstruction scale k . Simultaneously, given the possible frequency aliasing problem between adjacent IMFs, a wavelet transform threshold de-noising method is used to reduce the noise of the ( k –1)th IMF. Subsequently, the de-noised logging data are reconstructed by the de-noised ( k –1)th IMF and the remaining low-frequency IMFs and the residual. Finally, empirical mode decomposition, improved CEEMDAN, wavelet transform and the proposed method are applied for analysis of the simulation and the actual data. Results show diverse performance of these de-noising methods with regard to accuracy for lithological identification. Compared with the other methods, the proposed method has the best self-adaptability and accuracy in lithological identification.