Contribution of remote sensing InSAR and optical imagery to the identification of seismic faults in northeastern Algeria

Accurate fault identification in coal mines is important to improve mine safety and economic benefits. We compare various intelligent algorithms for data preprocessing and optimization and analyze the construction methods of seismic attribute data sets and the performance of intelligent optimization algorithms using fault identification accuracy as the discrimination index to find a better combined model for seismic fault identification. First, the training data set is constructed by mining the fault and nonfault information revealed by the roadway. The distribution characteristics of the seismic attribute data indicate similarities among them, and they are nonlinearly separable. Directly using the attributes to construct the data set, the accuracy of fault identification using the support vector machine (SVM) model is 78.41%. Principal component analysis (PCA) and self-organizing mapping (SOM) neural networks are used to extract effective information and then combined with the SVM classification model, and the accuracy of fault identification is 83.82% and 87.47%, respectively. Compared with the original data and PCA dimensionality reduction data, the accuracy of fault detection is improved by 9.06% and 3.66%, respectively, indicating that SOM can effectively improve the accuracy of fault detection by eliminating similar attributes and reducing the weight of redundant information. Then, through a fixed attribute data set, genetic algorithm (GA), particle swarm optimization (PSO), and gray wolf optimizer (GWO) intelligent optimization algorithms are used to find the optimal kernel function parameter and penalty parameter of the SVM classifier. The accuracy rate of the SOM-GWO-SVM model reaches 91.12%, compared with the SOM-PSO-SVM and SOM-GA-SVM, and the model accuracy is increased by 5.2% and 5.61%, respectively. Compared with PSO and GA, the GWO algorithm has a better global search ability. The identification result of the SOM-GWO-SVM model is closest to the actual fault exposure, especially for the identification of “short” faults and associated faults, which has obvious advantages over the traditional manual interpretation in terms of efficiency and accuracy.

Fault identification has always been a challenging research task in the field of geophysical exploration, and accurate fault identification is the basis and key to the recovery of hydrocarbons. Manual fault marking is not only time-consuming and labor-intensive, but the application of deep learning is gradually being extended to the field of geophysical exploration. The U-Net model with encoder-decoder structure has better fault identification performance for seismic data, but the encoding process tends to lose the underlying features while expanding the sensory field. In this paper, we propose the Rotated-UNet model for fault identification, which is different from the U-Net model in that it incorporates a shallow network with upscaling and then downscaling before the encoder-decoder structure, with the aim of improving the completeness of the features. The results of the trained Rotated-UNet model on the test set and F3 seismic data show that the model is more effective in identifying faults in real seismic data than the U-Net network, and has stronger generalization capability.

The accurate identification of seismic faults, which serve as crucial fluid migration pathways in hydrocarbon reservoirs, is of paramount importance for reservoir characterization. Traditional interpretation is inefficient. It also struggles with complex geometries, failing to meet the current exploration demands. Deep learning boosts fault identification significantly but struggles with edge accuracy and noise robustness. To overcome these limitations, this research introduces SwiftSeis-AWNet, a novel lightweight and high-precision network. The network is based on an optimized MedNeXt architecture for better fault edge detection. To address the noise from simple feature fusion, a Semantics and Detail Infusion (SDI) module is integrated. Since the Hadamard product in SDI can cause information loss, we engineer an Attention-Weighted Semantics and Detail Infusion (AWSDI) module that uses dynamic multi-scale feature fusion to preserve details. Validation on field seismic datasets from the Netherlands F3 and New Zealand Kerry blocks shows that SwiftSeis-AWNet mitigates challenges like the loss of small-scale fault features and misidentification of fault intersection zones, enhancing the accuracy and geological reliability of automated fault identification.

Fault-Seg-Net: A method for seismic fault segmentation based on multi-scale feature fusion with imbalanced classification

Improving seismic fault mapping through data conditioning using a pre-trained deep convolutional neural network: A case study on Groningen field

A method for seismic fault identification based on self-training with high-stability pseudo-labels

Faults are geologic structures that can cause disasters and thereby affect the safety of coal mines. To achieve improved fault interpretation accuracy during 3D seismic exploration of coal mines, we develop a seismic fault identification method based on a combination of kernel principal component analysis (KPCA), genetic particle swarm optimization (GPSO), and a support vector machine (SVM). The Dongwupan area of the Sihe Coal Mine in Shanxi Province, which mainly contains small faults, is the research area, and we extract 20 types of seismic attributes. According to the median difference between fault and nonfault data, we select the 12 leading seismic attributes with differences greater than 0.1 in descending order. Considering the information redundancy and the nonlinear relationships among seismic attributes, we adopt the KPCA method to reduce and optimize the selected seismic attributes, thereby effectively capturing the main information contained in the data and eliminating noise. Moreover, we introduce the GPSO algorithm to effectively optimize the SVM model parameters, and we construct a KPCA-GPSO-SVM model to classify and predict faults. Through model testing, the average fault identification accuracy of the model is 98.89%. Relative to the KPCA-particle swarm optimization-SVM, principal component analysis-GPSO-SVM, and GPSO-SVM models, the accuracy is improved by 3.3%, 7.8%, and 16.7%, respectively. We apply the proposed model to predict the fault distribution in the research area, and we compare the predictions to the actual exposed faults. The results indicate that the KPCA-GPSO-SVM model can suitably realize fault distribution prediction in a mining area. Moreover, compared with manual fault interpretation, the developed method is faster, more intuitive, and can better identify small faults.

One of the major challenges in Abu Dhabi onshore oil fields is the substantial increased water production with time. This is mainly attributed to unforeseen the presence of sub seismic faults and fractures along the drain hole. To address this, smart completions are being introduced to control water production by moderating the flow profile across the completed interval, whereby water and gas breakthrough are delayed in producing wells and injection rates are optimized in water and gas injectors across the full wellbore face. The smart completion design is supported by dynamic and static reservoir information. In this study, production logging tool (PLT) data was available to benchmark and confirm the results from dynamic simulation modeling. However, geologic features that cause the uncontrolled water inflow were not confirmed due to lack of lack of image data. A novel methodology has been implemented allowing resistivity image acquisition in workover wells by utilizing a compact memory tool that addresses the challenge of accessing old wells in Abu Dhabi onshore oil fields. The poor condition of the well is a result of prolonged production, post-stimulation and subsequent borehole degradation. The technique consists of acquiring data using drill pipe with significant savings in rig time and a higher chance of accessibility when compared to alternative techniques. This technique also protects the tool from becoming stuck and incurring damage during entry, thus improving data acquisition quality and minimizing the possibility of re-runs. When compared with common pipe-conveyed wireline alternatives (TLC) in previous wells, up to 33% savings in logging time was observed with a much safer operation. The procedure was achieved with success in terms of image data quality and execution time. In the absence of core data, the microresistivity imaging technique gives the best possible picture of the rock. It has been effective in characterizing rock fabric required for structural analysis (including fracture identification and fault identification and analysis), geomechanics, sedimentology (including paleocurrents analysis and facies determination) and image petrophysics applications such as identifying vuggy secondary porosity. A quick- turnaround in data processing allows the introduction of detailed fractures and fault interpretation, which in turn allows optimization of the completion design.

Fault identification is vital for geological structure analysis and the optimization of oil–gas extraction. Deep neural networks, especially U-Net and its variants, are widely used for seismic fault interpretation. However, when applied to 3D seismic data volume, these models typically require substantial computation resources and memory consumption. For one reason, they do not take into consideration the obvious differences in characteristics of seismic data in space and time dimensions; therefore, they require a huge number of parameters to capture inherent information for seismic fault detection. This paper presents a lightweight 3D seismic fault interpretation network based on a spatial–temporal asymmetric convolution set (STA-Fault3D) to mitigate the aforementioned issue. STA-Fault3D uses the spatial–temporal asymmetric convolution set to construct a lightweight network and take into consideration seismic data dimension discrepancies. Multi-scale feature fusion operation and an enhanced-training workflow are adopted to improve the performance of the network on field data. Compared with the classic model, FaultSeg3D, it demonstrates improved performance on fault detection continuity with only 12.33% of the parameters and 18.57% of the computational quantity. Compared with the state-of-the-art (SOTA) lightweight network, Fault3DNnet, it reduces parameters by 10% and computational quantity by 4.2% for marginally improved detection results.

Automated seismic fault detection using artificial intelligence (AI) represents a transformative advance in subsurface interpretation, offering unprecedented precision and efficiency in identifying fault networks. Traditional manual interpretation of seismic volumes is time-consuming and subject to interpreter bias, often leading to inconsistent fault mapping. This paper reviews state-of-the-art AI methodologies—such as convolutional neural networks, deep learning architectures, and unsupervised feature extraction—for automated fault identification. We evaluate the performance of these models on diverse geological settings, highlighting their ability to detect subtle discontinuities, leverage transfer learning across basins, and integrate multi-attribute seismic data. Case studies demonstrate significant improvements in fault continuity, reduced false-positive rates, and accelerated interpretation workflows. Challenges—including training data scarcity, network generalization across varying seismic quality, and the need for explainable AI—are critically discussed. Finally, we outline best practices for integrating AI-driven fault detection into existing geoscience workflows, propose strategies for model validation and uncertainty quantification, and identify future research directions aimed at real-time monitoring and adaptive interpretation. The review underscores AI’s potential to revolutionize seismic fault mapping, improve reservoir characterization, and enhance decision-making in exploration and production.

Application and development trend of artificial intelligence in petroleum exploration and development

In recent years, there has been a rapid development of the computer-aided interpretation of seismic data to reduce the otherwise intensive manual labor. A variety of seed detection algorithms for horizon and fault identification are integrated into popular seismic software packages. Recently, there has been an increasing focus on using neural networks for fully automatic fault detection without manually seeding each fault. These networks are usually trained with synthetic fault data sets. These data sets can be used across multiple seismic data sets; however, they are not as accurate as real seismic data, particularly in structurally complex regions associated with several generations of faults. The approach taken here is to combine the accuracy of manual fault identification in certain parts of the data set with a convolutional neural network that can then sweep through the entire data set to identify faults. We have implemented our method using 3D seismic data acquired from the Arabian Basin in Saudi Arabia covering an area of 1051 km2. The network is trained, validated, and tested with samples that included a seismic cube and fault images that are labeled manually corresponding to the seismic cube. The model successfully identifies faults with an accuracy of 96% and an error rate of 0.12 on the training data set. To achieve a robust model, we further enhance the prediction results using postprocessing by linking discontinued segments of the same fault line, thus reducing the number of detected faults. The postprocessing improves the prediction results from the test data set by 77.5%. In addition, we introduce an efficient framework to correlate the predictions and the ground truth by measuring their average distance value. Furthermore, tests using this approach also have been conducted on the F3 Netherlands survey with complex fault geometries and find promising results. As a result, fault detection and diagnosis are achieved efficiently with structures similar to the trained data set.

ABSTRACTThe existence of active faults hidden below Quaternary alluvium is a common geological scenario for intermontane basins, such as the areas struck by the recent earthquakes in Central Italy, and is of great importance for seismic hazard evaluation. Finding hidden faults is a challenging task from the geophysicist's point of view since the goal is twofold: to identify the seismic bedrock at a certain depth; and to detect lateral variations or dislocations that may indicate the presence of a fault. We propose a mixed approach encompassing at first single‐station seismic noise measurements, to detect sudden lateral variations in the bedrock surface in a fast and cost‐effective way, which might serve as a proxy for the potential identification of fault zones. Then, more accurate electrical resistivity tomography investigations are carried out only at selected sites as indicated by the preliminary noise analysis, as electrical methods cannot effectively be employed at a large scale for time and economic limitations. Surface‐wave dispersion analysis is jointly interpreted together with ambient noise data to improve the seismic characterization of the alluvium, giving further insight on the assessment of the depth to bedrock. The proposed approach can be an effective way to manage and investigate a large portion of the territory within a sensible budget, as commonly needed in seismic hazard assessment and microzonation studies. We present a real‐world application to the San Vittorino Plain (Central Italy) close to the epicentre of the 24 August 2016 Amatrice earthquake, where the geological faulted bedrock is covered by alluvial sediments of the Velino River up to a maximum estimated thickness of 150–200 m. Although engineered for the post‐earthquake reconstruction emergency, the approach employed in our study can be adopted in other areas of similar geology, to ease the application of seismic microzonation in time of seismic silence as a tool for long‐term land planning and management.

SUMMARY In the French western Alps, east of Grenoble, we identify the Belledonne Border Fault as an active seismic fault. This identification is based on the seismic monitoring of the Grenoble area by the Sismalp seismic network over the past 12 yr (1989–2000). We located a set of earthquakes with magnitudes ranging from 0 to 3.5 along a ∼50 km long alignment which runs in a N30°E direction on the western flank of the Belledonne crystalline massif. Available focal solutions for these events are consistent with this direction (N36°E strike-slip fault with right-lateral displacement). These events along the Belledonne Border Fault have a mean focal depth of ∼7 km (in the crystalline basement), with a probably very low slip rate. The Belledonne Border Fault has never been mapped at the surface, where the otherwise heavily folded and faulted Mesozoic cover makes this identification difficult. Historical seismicity also shows that, over the past two and a half centuries, a few events located mainly along the southern part of the Belledonne Border Fault caused damage, with the magnitude 4.9 1963 Monteynard earthquake reaching intensity VII. The most recent damaging event in the study area is the magnitude 3.5 1999 Laffrey earthquake (intensity V–VI). Although its epicentre lies at the southern tip of the Belledonne Border Fault, there is clear evidence that aftershocks were activated by the left-lateral slip of a N122°E-striking fault. The length of the Belledonne Border Fault, which could easily accommodate a magnitude 6 event, as well as the proximity to the Isere valley with its unlithified Quaternary deposits up to 500 m thick known to generate marked site effects, make the identification of the Belledonne Border Fault an important step in the evaluation of seismic risk in the Grenoble area. Besides, the activity observed on the fault will now have to be taken into account in future geodynamic models of the western Alps.

Fault-Seg-LNet: A method for seismic fault identification based on lightweight and dynamic scalable network