SZSibo Zhang

Sibo Zhang

*This is an algorithmically generated author record based on publicly available information.

Sibo Zhang has a total of 16 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Top Cited Papers

    Seismic foundation model: A next generation deep-learning model in geophysics

    by Hanlin Sheng, Xinming Wu, Xu Si, Jintao Li, Sibo Zhang, Xudong Duan

    Although computer science has seen remarkable advancements in foundation models, they remain underexplored in geoscience. Addressing this gap, we introduce a workflow to develop geophysical foundation models, such as data preparation, model pretraining, and adaption to downstream tasks. From 192 globally collected 3D seismic volumes, we create a carefully curated data set of 2,286,422 2D seismic images. To fully use these unlabeled images, we use self-supervised learning to pretrain a transformer-based seismic foundation model for producing all-purpose seismic features that work across various tasks and surveys. Through experiments on seismic facies classification, geobody identification, interpolation, denoising, and inversion, our pretrained model demonstrates versatility, generalization, scalability, and superior performance compared with baseline models. In conclusion, we provide a foundation model and vast data set to advance artificial intelligence (AI) in geophysics, addressing the challenges (poor generalization, a lack of labels, and repetitive training for task-specific models) of applying AI to geophysics and paving the way for future innovations in geoscience.


  • Building subsurface models with horizon-guided interpolation and deep learning: Application to the Volve field data

    by Shangsheng Yan, Xiaoming Sun, Xinming Wu, Sibo Zhang, Hongjie Si

    Subsurface modeling plays an important role in hydrocarbon exploration but remains a challenging task that typically requires a full and reasonable integration of geophysical observations and geologic constraints. We have developed a workflow to fully use seismic amplitudes, well-log properties, and interpreted seismic structures to build geologically reasonable models. We take the Volve field data as an example and apply our workflow step by step as follows. First, we perform some preprocessing on the provided Volve seismic data, horizons, and well logs to remove anomalous values and adjust seismic-well ties in the depth domain. Second, we use a dynamic-programming-based method to infill gaps and refine the vertical positions of the provided horizons and efficiently pick more horizons. We further use the horizon surfaces to interpolate a relative geologic time (RGT) volume which can be considered as an implicit structural model representing seismic structural and stratigraphic features. Third, we integrate the provided well logs and the computed RGT volume to interpolate a subsurface model that conforms to well-log properties and seismic structural and stratigraphic features. Finally, we propose a multiscale neural network to predict a final subsurface model by using a combination of the seismic data and the interpolated model as inputs and using the well-log data as training labels. Inputting the interpolated model to the network helps to provide a low-frequency control for obtaining a more stable prediction. The results indicate that our workflow is able to produce geologically reasonable subsurface models with high lateral continuity and vertical resolution.


  • Geologic-time-based interpolation of borehole data for building high-resolution models: Methods and applications

    by Zhengfa Bi, Xinming Wu, Yaxing Li, Shangsheng Yan, Sibo Zhang, Hongjie Si

    Integrating borehole data into a uniformly sampled grid is an essential but challenging task for seismic data processing and reservoir characterization. We propose an efficient geologic-time-based interpolation method to build subsurface models where the structures are consistent with the seismic structures and the vertical resolution of model properties is as high as well-log records. Based on a seismic image, we first compute a relative geologic time (RGT) volume that provides an implicit map of all the geologic structures in the seismic image. We then construct an interpolated model from borehole data by following constant RGT values (each one corresponds to a same geologic layer), and thus obtain a high-resolution model honoring both seismic structures and well-log values. Such a model could provide a low-frequency control for a deep learning or conventional inversion method to estimate reservoir properties, or it can be used as a reliable initial background model to improve the performance of full-waveform inversion. We use both synthetic and field data examples to demonstrate the effectiveness of our method even when reservoir properties are only observed at sparsely scattered locations. In comparison to the existing approaches, our method can produce a more geologically consistent subsurface model which can be used as a better initial model for a deep learning method to estimate a refined rock-property model from seismic data.


  • Deep learning for multidimensional seismic impedance inversion

    by Xinming Wu, Shangsheng Yan, Zhengfa Bi, Sibo Zhang, Hongjie Si

    Deep-learning (DL) methods have shown promising performance in predicting acoustic impedance from seismic data that is typically considered as an ill-posed problem for traditional inversion schemes. Most of DL methods are based on a 1D neural network that is straightforward to implement, but they often yield unreasonable lateral discontinuities while predicting a multidimensional impedance model trace by trace. We have developed an improvement over the 1D network by replacing it with a 2D convolutional neural network (CNN) and incorporating the constraints of an initial impedance model. The initial model is fed to the network to provide low-frequency trend control, which is helpful for 1D and 2D CNNs to yield stable impedance predictions. Our 2D CNN architecture is quite simple; however, due to the lack of 2D impedance labels, training it is not straightforward. To prepare a 2D training data set, we first define a random path that passes through multiple well logs. We then follow the path to extract a 2D seismic profile and an initial impedance profile that together form an input to the 2D CNN. The set of well logs (traversed by the path) serves as a partially labeled target. We train the 2D CNN with weak supervision by using an adaptive loss in which the output 2D impedance model is adaptively evaluated at the well logs only in the partially labeled target. Because the training data sets are randomly extracted in all directions in a 3D survey, the trained 2D CNN can predict a consistent 3D impedance model section by section in either the inline or crossline directions. Synthetic and field examples indicate that our 2D CNN is more robust to noise, recovers thin layers better, and yields a laterally more consistent impedance model than a 1D CNN with the same network architecture and the same training logs.


  • Global Low-Rank Image Restoration With Gaussian Mixture Model

    by Sibo Zhang, Licheng Jiao, Fang Liu, Shuang Wang

    Low-rank restoration has recently attracted a lot of attention in the research of computer vision. Empirical studies show that exploring the low-rank property of the patch groups can lead to superior restoration performance, however, there is limited achievement on the global low-rank restoration because the rank minimization at image level is too strong for the natural images which seldom match the low-rank condition. In this paper, we describe a flexible global low-rank restoration model which introduces the local statistical properties into the rank minimization. The proposed model can effectively recover the latent global low-rank structure via nuclear norm, as well as the fine details via Gaussian mixture model. An alternating scheme is developed to estimate the Gaussian parameters and the restored image, and it shows excellent convergence and stability. Besides, experiments on image and video sequence datasets show the effectiveness of the proposed method in image inpainting problems.