De-noising for NMR oil well logging signals based on empirical mode decomposition and independent component analysis

Inversions of T2-distribution can be severely disturbed by the noise in nuclear magnetic resonance (NMR) oil well logging. Methods to isolate and remove these disturbances are typically based on time-series editing. An alternative approach for noise removal is proposed based on a combination of empirical mode decomposition (EMD) and independent component analysis (ICA), called the EMD-ICA method. Firstly, the NMR oil well logging signals is decomposed into a series of IMFs (intrinsic mode function) with EMD. Then, the successive 3 orders IMF components are combined into a sequence sequentially, and ICA is applied for this sequence. Finally, the obtained results of ICA are used to reconstruct the de-noised signal. Principle and steps of method are presented, then, some simulated signal and measured logging data are processed. The de-noised results are compared with that from Wavelet method and EMD space-time filtering method. The results illustrate that free of noise data sections are preserved because logging data is analyzed through hierarchies, or scale levels, allowing separation of noise from signals with EMD-ICA method. After filtering stage, the two peak value points of T2 curve are highlighted and T2-distribution becomes more reliable comparing with before de-noising. The proposed method reduces the bias error of the estimated parameter and improves the quality of logging data significantly, as well as provides a good basis for further studies of the reservoir.