Multi-horizon Simultaneous Tracking Based on Dynamic Seismic Waveform Matching

Summary Due to the rapid increase in the amount of seismic volumes, the traditional seismic interpretation mode based on manual structure interpretation and single-horizon automatic tracking has encountered many challenges. The seismic interpretation of large or super-large 3-D seismic surveys is facing serious accuracy and efficiency bottlenecks. Aiming to the goal of improving the accuracy and efficiency of seismic interpretation, we propose a dynamic seismic waveform matching technology based on the sparse dynamic time warping algorithm under the guidance of the relative geological time volume theory, and realize multi-horizon simultaneous tracking based on the technology. Has been verified by a model and a real seismic volume, it can realize simultaneous horizon automatic tracking, full spatial tracking and high-density tracking, and can significantly improve the accuracy and efficiency of structure interpretation.

Self-Supervised Delineation of Geological Structures using Orthogonal Latent Space Projection

We developed two machine learning frameworks that could assist in the automated litho-stratigraphic interpretation of seismic volumes without any manual hand labeling from an experienced seismic interpreter. The first framework is an unsupervised hierarchical clustering model to divide seismic images from a volume into certain number of clusters determined by the algorithm. The clustering framework uses a combination of density and hierarchical techniques to determine the size and homogeneity of the clusters. The second framework consists of a self-supervised deep learning framework to label regions of geological interest in seismic images. It projects the latent-space of an encoder-decoder architecture unto two orthogonal subspaces, from which it learns to delineate regions of interest in the seismic images. To demonstrate an application of both frameworks, a seismic volume was clustered into various contiguous clusters, from which four clusters were selected based on distinct seismic patterns: horizons, faults, salt domes and chaotic structures. Images from the selected clusters are used to train the encoder-decoder network. The output of the encoder-decoder network is a probability map of the possibility an amplitude reflection event belongs to an interesting geological structure. The structures are delineated using the probability map. The delineated images are further used to post-train a segmentation model to extend our results to full-vertical sections. The results on vertical sections show that we can factorize a seismic volume into its corresponding structural components. Lastly, we showed that our deep learning framework could be modeled as an attribute extractor and we compared our attribute result with various existing attributes in literature and demonstrate competitive performance with them.

3D seismic interpretation with deep learning: A brief introduction

Understanding the internal structure of our planet is a fundamental goal of the earth sciences. As direct observations are restricted to surface outcrops and borehole cores, we rely on geophysical data to study the earth's interior. In particular, seismic reflection data showing acoustic images of the subsurface provide us with critical insights into sedimentary, tectonic, and magmatic systems. However, interpretations of these large 2D grids or 3D seismic volumes are time-consuming, even for a well-trained person or team. Here, we demonstrate how to automate and accelerate the analysis of these increasingly large seismic data sets with machine learning. We are able to perform typical seismic interpretation tasks such as mapping tectonic faults, salt bodies, and sedimentary horizons at high accuracy using deep convolutional neural networks. We share our workflows and scripts, encouraging users to apply our methods to similar problems. Our methodology is generic and flexible, allowing an easy adaptation without major changes. Once trained, these models can analyze large volumes of data within seconds, opening a new pathway to study the processes shaping the internal structure of our planet.

Enhanced ant tracking: Using a multispectral seismic attribute workflow to improve 3D fault detection

Fault interpretation is a complex task that requires time and effort on behalf of the interpreter. Moreover, it plays a key role during subsurface structural characterization either for hydrocarbon exploration and development or well planning and placement. Seismic attributes are tools that help interpreters identify subsurface characteristics that cannot be observed clearly. Unfortunately, indiscriminate and random seismic attribute use affects the fault interpretation process. We have developed a multispectral seismic attribute workflow composed of dip-azimuth extraction, structural filtering, frequency filtering, detection of amplitude discontinuities, enhancement of amplitude discontinuities, and automatic fault extraction. The result is an enhanced ant-tracking volume in which faults are improved compared to common fault-enhanced workflows that incorporate the ant-tracking algorithm. To prove the effectiveness of the enhanced ant-tracking volume, we have applied this methodology in three seismic volumes with different random noise content and seismic characteristics. The detected and extracted faults are continuous, clean, and accurate. The proposed fault identification workflow reduces the effort and time spent in fault interpretation as a result of the integration and appropriate use of various types of seismic attributes, spectral decomposition, and swarm intelligence.

Uncertainty and Interpretability analysis of encoder-decoder architecture for channel detection

We have adopted a method to understand uncertainty and interpretability of a Bayesian convolutional neural network for detecting 3D channel geobodies in seismic volumes. We measure heteroscedastic aleatoric uncertainty and epistemic uncertainty. Epistemic uncertainty captures the uncertainty of the network parameters, whereas heteroscedastic aleatoric uncertainty accounts for noise in the seismic volumes. We train a network modified from U-Net architecture, on 3D synthetic seismic volumes, and then we apply it to field data. Tests on 3D field data sets from the Browse Basin, offshore Australia, and from Parihaka in New Zealand, prove that uncertainty volumes are related to geologic uncertainty, model mispicks, and input noise. We analyze model interpretability on these data sets by creating saliency volumes with gradient-weighted class activation mapping. We find that the model takes a global to local approach to localize channel geobodies as well as the importance of different model components in overall strategy. Using channel probability, uncertainty, and saliency volumes, interpreters can accurately identify channel geobodies in 3D seismic volumes and also understand the model predictions

Joint Learning for Spatial Context-based Seismic Inversion of Multiple Datasets for Improved Generalizability and Robustness

Seismic inversion plays a very useful role in detailed stratigraphic interpretation of migrated seismic volumes by enabling the estimation of reservoir properties over the complete volume. Traditional and machine learning-based seismic inversion workflows are limited to inverting each seismic trace independently of other traces to estimate impedance profiles, leading to lateral discontinuities in the presence of noise and large geological variations in the seismic data. In addition, machine learning-based approaches suffer the problem of overfitting if there is only a small number of wells on which the model is trained. We propose a two-pronged strategy to overcome these problems. We present a temporal convolutional network that models seismic traces temporally. We further inject spatial context for each trace into its estimations of the impedance profile. To counter the problem of limited labeled data, we also present a joint learning scheme whereby multiple datasets are simultaneously used for training, sharing beneficial information among each other. This results in the improvement in generalization performance on all datasets. We present a case study of acoustic impedance inversion using the open-source SEAM and Marmousi 2 datasets. Our evaluations show that our proposed approach is able to estimate impedance in the presence of noisy seismic data and a limited number of well logs with greater robustness and spatial consistency. We compare and contrast our approach to other learning-based seismic inversion methodologies in the literature. On SEAM, we are able to obtain an average MSE of 0.0476, the lowest among all other methodologies.

ChannelSeg3D: channel simulation and deep learning for channel interpretation in 3D seismic images

Seismic channel interpretation involves detecting channel structures which often appear as meandering shapes in 3D seismic images. Many conventional methods are proposed for delineating channel structures using different seismic attributes. However, these methods are often sensitive to seismic discontinuities (e.g., noise and faults) that are not related to channels. We propose a convolutional neural network (CNN) method to improve the automatic channel interpretation. The key problem in applying the CNNs method into channel interpretation is the absence of the labeled field seismic images for training the CNNs. To solve this problem, we propose a workflow to automatically generate numerous synthetic training datasets with realistic channel structures. In this workflow, we first randomly simulate various meandering channel models based on geological numerical simulation. We further simulate structural deformation in the form of stratigraphic folding referred to as “folding structures” and combine them with the previously generated channel models to create reflectivity models and the corresponding channel labels. Convolved with a wavelet, the reflectivity models can be transformed into learnable synthetic seismic volumes. By training the designed CNN with synthetic seismic data, we obtain a CNN which learns the characterization of channel structures. Although trained on only synthetic seismic volumes, this CNN shows an outstanding performance on field seismic volumes. This indicates that the synthetic seismic images created in this workflow are realistic enough to train the CNN for channel interpretation in field seismic images.

See the Unseen through Target-Oriented Dip-Steered Seismic Resolution

Seismic resolution plays an important role not only in interpretation and reservoir characterization but also in seismic inversion and seismic attributes analysis. The resolution depends on several factors, including seismic frequency bandwidth, dominant frequency, and layer velocity. This paper presents a spectral resolution enhancement approach that is based on Non-stationary Differential Resolution (NSDR) that honors the local structural dip, better preserves amplitude and improves target-oriented seismic interpretation. The proposed technology is applied to both 2D and 3D seismic volumes and findings are compared with the oil industry common spectral enhancement algorithms. We demonstrate the target-oriented dip steering spectral enhancement method on two 3D field datasets and compare the resulting outcome with those obtained by conventional techniques. It is found that thinly layered subsurface geological features with steeply dipping beds are better defined, with artifacts from the conflicting dips removed.