Search

Blended acquisition offers significant cost and period reduction in seismic data acquisition. However, fired blended sources are usually deployed at off-the-grid (OffG) samples due to obstacle limitation and economic cost considerations. The irregular distribution of coordinates, along with the blending noise, has a detrimental effect on the performance of subsequent seismic processing and imaging. The interpolated multichannel singular spectrum analysis (I-MSSA) algorithm effectively provides on-the-grid deblended results by employing an interpolator, in conjunction with a projected gradient descent strategy. However, the deblending accuracy and computational efficiency of the I-MSSA are still a concern due to the limitations of the traditional singular value decomposition (SVD). To address these limitations, we propose an interpolated fast damped multichannel singular spectrum analysis (I-FDMSSA) rank-reduction algorithm. The proposed algorithm incorporates the damping operator, the randomized SVD (RSVD) and the fast Fourier transform (FFT) strategy. The damping operator can further attenuate the remaining noise in the estimated signal obtained from the truncated SVD, resulting in an improved deblending performance. The RSVD accelerates the rank-reduction process by shrinking the size of the Hankel matrix. To expedite the rank-reduction and anti-diagonal averaging stages without explicitly constructing large-scale block Hankel matrices, the FFT strategy is employed. By incorporating a 2D separable sinc interpolator, the I-FDMSSA enables an efficient and accurate deblending of 3D OffG blended data. The deblending performance and operational efficiency improvements of the proposed I-FDMSSA algorithm over the traditional I-MSSA algorithm are demonstrated through OffG synthetic and field blended data examples. [ABSTRACT FROM AUTHOR]

The potential of blended seismic acquisition to improve acquisition efficiency and cut acquisition costs is still open, particularly with efficient deblending algorithms to provide accurate deblended data for subsequent processing procedures. In recent years, deep learning algorithms, particularly supervised algorithms, have drawn much attention over conventional deblending algorithms due to their ability to nonlinearly characterize seismic data and achieve more accurate deblended results. Supervised algorithms require large amounts of labeled data for training, yet accurate labels are rarely accessible in field cases. We present a self-supervised multistep deblending framework that does not require clean labels and can characterize the decreasing blending noise level quantitatively in a flexible multistep manner. To achieve this, we leverage the coherence similarity of the common shot gathers (CSGs) and the common receiver gathers (CRGs) after pseudo-deblending. The CSGs are used to construct the training data adaptively, where the raw CSGs are regarded as the label with the corresponding artificially pseudo-deblended data as the initial training input. We employ different networks to quantitatively characterize decreasing blending noise levels in multiple steps for accurate deblending with the help of a blending noise estimation–subtraction strategy. The training of one network can be efficiently initialized by transfer learning from the optimized parameters of the previous network. The optimized parameters trained on CSGs are used to deblend all CRGs of the raw pseudo-deblended data in a multistep manner. Tests on synthetic and field data validate the proposed self-supervised multistep deblending algorithm, which outperforms the multilevel blending noise strategy. [ABSTRACT FROM AUTHOR]