31 August 2023 in Geophysics
In field seismic data, abnormal environmental noise from the acquisition environment often is unavoidable, and its characteristics are complex. Abnormal environmental noise has a strong amplitude (tens of thousands times the average seismic amplitudes) and masks the effective seismic signal. The traditional method for removing such abnormal environmental noise relies on energy scanning to identify the noise location. However, energy scanning cannot always accurately identify such noise due to the sudden energy change at the locations of first-breaks and surface waves. A deep learning method for removing abnormal environmental noise has been developed. The presence of environmental noise with large amplitude also results in an uneven energy distribution. The existing denoising networks rarely consider this situation. Based on the characteristics of environmental noise received by the nodal seismic acquisition system, a workflow for automatically generating the training data sets has been established. The network is built based on the Unet with the incorporation of a residual block. The Batchnorm layer is specifically omitted to adapt to the uneven energy distribution. Synthetic and field data testing have confirmed the effectiveness and applicability of the proposed method. The new method indicates better denoising performance and greater flexibility than the traditional method. Comparison between different networks also indicates that the new Residual Block Connected Unet is much more effective in denoising seismic data that contain abnormal environmental noise.
25 July 2023 in Geophysics
The quality of wavefield separation on vertical seismic profiling (VSP) data directly affects subsequent imaging and inversion processes. However, the traditional methods have various defects in separating upgoing and downgoing waves. The application of deep learning brings another train of thought to solve related problems. The traditional methods, such as f-k filtering and Radon transform, produce results with spatial aliasing and inaccurate amplitudes. The median filtering method relies on accurate first-break picking and waveform consistency. Moreover, the results of traditional methods are subject to manual intervention. To overcome these problems, a deep-learning method is developed for the automatic wavefield separation of VSP data. First, the traditional Radon transform method is used to produce training data sets. Then, to ensure amplitude preservation and suppress spatial aliasing, new input data is formed by recombining the upgoing and downgoing waves extracted by Radon transform. The developed deep-learning network is highly efficient, and its accuracy exceeds that of the traditional methods only after one epoch training. A practical workflow of wavefield separation based on deep learning is established, and it is applied to synthetic data and field distributed acoustic sensing VSP data. The results indicate that our method is superior in terms of amplitude preservation and spatial aliasing suppression. The time consumption of our method is very acceptable and can be further minimized by training the network using downsampling data.
7 January 2025 in Geophysics
Distributed acoustic sensing (DAS) has emerged as an important seismic observation method due to the advantages of low cost, easy deployment, and high-density spatial sampling. The integration of DAS recording with vertical seismic profiling (VSP) has become increasingly prevalent across various applications of borehole geophysics. In seismic imaging and inversion of VSP data, it is essential to decompose the up-/downgoing and P-/S-wave modes as necessary preconditioning steps. However, conventional wavefield decomposition methods based on physical principles face significant challenges when handling 1C DAS-VSP data. We use a deep-learning method to address this mode decomposition problem. In the framework of supervised learning, the labeled data are constructed through elastic forward modeling aided by the Hilbert transform and the Helmholtz decomposition. Then, multitask convolutional neural network models are trained for one- or two-pass operation of up/down and P/S decomposition. To process field data, we rely on transfer learning to improve the generalization ability of the trained network models. Examples with synthetic and field walk-away DAS-VSP data demonstrate the effectiveness of our method in wavefield decomposition for the borehole axial strain data.
5 January 2025 in Geophysics
Distributed acoustic sensing (DAS) generally records the seismic signals by detecting the axial strain or strain rate that is stimulated by the impinging elastic wavefields along the optical fibers. It has become an important seismic observation technology, especially in vertical seismic profiling (VSP) applications, due to its low cost, easy deployment, and high-density spatial sampling. Although current DAS-VSP acquisitions typically offer only a single-component observation, they still provide valuable elastic information about the subsurface. Separating P and S waves from the DAS-VSP data and leveraging this wavefield information is very important for the inversion of elastic parameters and seismic imaging. Therefore, a polarization projection method is introduced to deal with the P/S separation of walkaway DAS-VSP data. First, the polarization directions of P and S waves are estimated using the dispersion relation derived from elastodynamic wave equations. Then, the P/S-wave separation is achieved through a nonstationary polarization projection that can take into account the effects of spatially varying wave velocities. Finally, a two-step workflow is developed to successively separate the P and S waves in common-shot and common-receiver gathers. The results of synthetic and real walkaway DAS-VSP data demonstrate that this method can effectively separate P- and S-wave signals from DAS-VSP data, and provide effective data preconditioning for subsequent velocity model building and seismic imaging.