Journal of Applied Geophysics最新文献

{"title":"Rotary detection using ground penetrating radar based on a B-Scan compressed sensing imaging algorithm","authors":"Xiaosong Tang ,&nbsp;Feng Yang ,&nbsp;Xu Qiao ,&nbsp;Haitao Zuo ,&nbsp;Chong Zhang ,&nbsp;Suping Peng","doi":"10.1016/j.jappgeo.2025.105783","DOIUrl":"10.1016/j.jappgeo.2025.105783","url":null,"abstract":"<div><div>In this work, we propose an imaging algorithm applicable to the Ground Penetrating Radar (GPR) rotary detection scenario. The constructed sparse matrix is interpretable, strictly following the actual physical model, and the convex optimization objective function is not the conventional L1 norm but the inverse Huber norm. A solver with a primal-dual interior point method is used to solve the convex optimization problem.Moreover, the study proposes a new metric for evaluating imaging performance. Firstly, we compared different algorithms on simulation data, demonstrating the superiority of the proposed algorithm in multi-target imaging;subsequently, a reasonable interpretation of the weight values involved in the proposed metric was provided;then, to test the robustness of the algorithm, it was proven on data with different Peak Signal-to-Noise Ratio (PSNR) and different sampling rates;next, sensitivity analysis was conducted on the three parameters involved in the inverse Huber norm and beam width, and empirical parameter values were proposed;finally,the study selected appropriate parameters within the optimal range of the proposed algorithm to perform imaging comparisons on experimental data.We look forward to this work promoting the development of the GPR detection field.</div></div>","PeriodicalId":54882,"journal":{"name":"Journal of Applied Geophysics","volume":"240 ","pages":"Article 105783"},"PeriodicalIF":2.2,"publicationDate":"2025-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144154547","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Estimation of anisotropic elastic moduli and fracture parameters by combined rock physics modeling and well-log measurements","authors":"Jikun Meng ,&nbsp;Qiaomu Qi ,&nbsp;Linxin Li ,&nbsp;Xiaobin Li","doi":"10.1016/j.jappgeo.2025.105779","DOIUrl":"10.1016/j.jappgeo.2025.105779","url":null,"abstract":"<div><div>Fractures act as main fluid flow conduits for oil and gas reservoirs and has a direct impact on storage capacity and permeability. The detection of fractures is important for evaluating reservoir production potential. Seismic modeling and geomechanical characterization of fractured reservoir require information, such as density and azimuth of fractures as well as fracture-induced anisotropy. Formations with vertical fractures exhibits horizontal transverse isotropic (HTI) anisotropy. This type of formation requires five independent elastic moduli to describe and it is challenging to obtain the full elastic tensors. We propose an effective scheme to determine the anisotropic tensors of fractured HTI medium based on well-log measurements. This method constructs four auxiliary parameters <span><math><mfenced><mi>A</mi><mi>B</mi><mi>C</mi><mi>D</mi></mfenced></math></span> using four well-log data: P-wave velocity <span><math><msub><mi>V</mi><mi>P</mi></msub></math></span>, fast and slow S-wave velocities <span><math><msub><mi>V</mi><msub><mi>s</mi><mn>1</mn></msub></msub></math></span>, <span><math><msub><mi>V</mi><msub><mi>s</mi><mn>2</mn></msub></msub></math></span>, and density <span><math><mi>ρ</mi></math></span>, from which we derive analytical expressions for anisotropy and fracture parameters. Although the elastic properties of the HTI medium are fundamentally controlled by the Lamé's constants (<span><math><mi>λ</mi></math></span>, <span><math><mi>μ</mi></math></span>) and the fracture compliance parameters <span><math><msub><mi>Z</mi><mi>N</mi></msub></math></span> and <span><math><msub><mi>Z</mi><mi>T</mi></msub></math></span> (which describe the normal and tangential mechanical responses of fractures), these auxiliary parameters enable us to simplify and unify the expressions, directly incorporate well-log data, and streamline the computation of anisotropy and fracture parameters for practical reservoir characterization. To obtain the scheme's input, we use slowness-time coherence (STC) and waveform matching inversion to extract high-resolution <span><math><mi>P</mi></math></span>- and <span><math><mi>S</mi></math></span>-wave velocities from monopole and dipole acoustic logging data, respectively. We demonstrate the effectiveness and applicability of the method by applying the workflow to a suite of well-log data obtained from a fractured reservoir. Interpretations with the obtained fracture properties show that the fracture density and ratio of fracture compliances are useful indicators of reservoir fracture development and gas-infill of fractures, respectively.</div></div>","PeriodicalId":54882,"journal":{"name":"Journal of Applied Geophysics","volume":"240 ","pages":"Article 105779"},"PeriodicalIF":2.2,"publicationDate":"2025-05-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144166925","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Measurement of the effective dynamic viscosity of soil-rock mixture","authors":"Jiaxin Lan ,&nbsp;Fuguo Tong ,&nbsp;Biao Li ,&nbsp;Xiaowei Xu ,&nbsp;Chang Liu","doi":"10.1016/j.jappgeo.2025.105782","DOIUrl":"10.1016/j.jappgeo.2025.105782","url":null,"abstract":"<div><div>The effective dynamic viscosity of a soil-rock mixture (S-RM) serves as a essential parameter for simulating flow-like landslides in the context of fluid kinematics. Accurate measurement of this viscosity is significant for understanding the remote sustainability and rheological properties of landslide hazards. This study presents a method for determining dynamic viscosity, incorporating experimental measurements and numerical inversion. The experiment involves monitoring the movement of S-RMs with varying water content and rock block concentration, followed by the calculation of centroid displacements and velocities using digital image processing. The power-law model, combined with computational fluid dynamics, effectively captures the flow-like behavior of the S-RM. A grid search method is then employed to determine the optimal parameters by comparing the predicted centroid displacement with experimental results. A series of flume experiments were conducted, resulting in the observation of spatial mass distribution and centroid displacement variations over time during soil-rock movement. The dynamic viscosity model of the S-RM is derived from the experimental data. This dynamic viscosity model was then employed to simulate an additional flume experiment, with the results demonstrating excellent agreement between the simulated and experimental centroid displacements. Sensitivity analysis of the dynamic viscosity model indicates a dependence on shear rate and demonstrates a high sensitivity to water content and rock block concentration, following a parabolic trend within the measured range. This research contributes to the fields of geotechnical engineering and landslide risk assessment, offering a practical and effective method of measuring the dynamic viscosity of S-RM. Future research could explore additional factors influencing rheological behavior and extend the applicability of the proposed method to different geological environments.</div></div>","PeriodicalId":54882,"journal":{"name":"Journal of Applied Geophysics","volume":"240 ","pages":"Article 105782"},"PeriodicalIF":2.2,"publicationDate":"2025-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144166917","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Impact of hematite-silica composite nanofluids on interfacial tension and contact angle dynamics for electromagnetic-assisted enhanced oil recovery","authors":"Yarima Mudassir Hassan ,&nbsp;Beh Hoe Guan ,&nbsp;Lee Kean Chuan ,&nbsp;Nurul Hazlina Noordin ,&nbsp;Mohammed Falalu Hamza ,&nbsp;Surajudeen Sikiru","doi":"10.1016/j.jappgeo.2025.105788","DOIUrl":"10.1016/j.jappgeo.2025.105788","url":null,"abstract":"<div><div>The entrapment of a substantial volume of crude oil in subsurface reservoirs has caused extensive degradation of energy production globally. Various nanoparticles (NPs) were used in this regard; however, the performance of NPs was overwhelmed and consequently entrapped in the rock pores due to the unfavourable conditions of the reservoir. The electromagnetic (EM) driven approach was recently recognized as a suitable method for advancing nanofluid mobility in the geophysical porous media. Considering the combination of magnetic and dielectric attributes, forming smart composite nanofluids using hematite-silica (Fe₂O₃-SiO₂) under EM waves will be consequential. The Fe₂O₃-SiO₂ was synthesized using the sol-gel method. A visual test complemented by a zeta potential analyzer was used to determine the stability of the fluids. A goniometer was used for interfacial tension (IFT) and rock-oil wettability analysis, while a sandpack flooding method was used for the EOR experiment at 100<span><math><msup><mspace></mspace><mo>°</mo></msup><mi>C</mi></math></span>. The Fe₂O₃-SiO₂ nanofluid stability achieved the highest electrostatic repulsion at −39 mV. When EM waves were propagated to the Fe₂O₃-SiO₂ nanofluids, the IFT reduced from 17 ± 1.3 mN/m to 4.06 ± 0.6 mN/m. Similarly, the contact angle was reduced from 141 ± 3.5° to 71 ± 0.5°. The EM flooding experiment showed a total recovery of 56.68 to 79.94 % for Fe₂O₃-SiO₂ in advance of the same composite nanofluids without EM wave endorsement (61.13 to 69.53 %). The EM wave propagation has energized the combination of permeability and permittivity in the Fe₂O₃-SiO₂ nanofluids. Hence, an additional disturbance was generated at the oil-to-water interface, reducing IFT and contact angle, which enhanced fluids' transportation in a porous media and recovered additional oil. The environmental risks associated with Fe₂O₃-SiO₂ nanofluids discharge and mitigation strategies, such as magnetic recovery and biodegradable coating, have been evaluated. This work presents an innovative procedural development of EM-based techniques for subsurface fluid transport improvement to address environmental, engineering, and hydrological challenges in a geophysical porous media.</div></div>","PeriodicalId":54882,"journal":{"name":"Journal of Applied Geophysics","volume":"240 ","pages":"Article 105788"},"PeriodicalIF":2.2,"publicationDate":"2025-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144177904","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Calibration of seismic parameters in the Zagros region, an application of generalized inversion technique","authors":"Reza Davoudian ,&nbsp;Hamid Zafarani ,&nbsp;Ahmad Sadidkhouy","doi":"10.1016/j.jappgeo.2025.105785","DOIUrl":"10.1016/j.jappgeo.2025.105785","url":null,"abstract":"<div><div>In this study, the Generalized Inversion Technique (GIT) was used to find seismic parameters (source and site effects) in the Zagros region, Iran. To this end, 91 events recorded at 205 three component accelerometers located in the Zagros region were analyzed. The dataset includes 745 acceleration waveforms in hypocentral distance range between 6 and 363 km and magnitude range between <span><math><msub><mi>M</mi><mi>w</mi></msub></math></span> 5.0 and <span><math><msub><mi>M</mi><mi>w</mi></msub></math></span> 7.3. All events included in the dataset were recorded between May 1997 and July 2022.</div><div>After applying a suitable processing procedure to prepare the data, near-surface high-frequency attenuation coefficient (Kappa) for both horizontal and vertical components was derived as 0.043 and 0.031, respectively. In the next step, the GIT was used to solve a linearized relationship and find spectra of site amplification and seismic source.</div><div>To find a parametric form for seismic source spectrum (<span><math><msub><mi>f</mi><mi>c</mi></msub><mspace></mspace><mi>and</mi><mspace></mspace><mi>γ</mi></math></span>, based on <span><math><mi>ω</mi></math></span> squared source model of <span><span>Brune, 1970</span></span>) and stress drop for each event, a Simple Grid Search routine was applied. The computed stress drop values were between 4 and 48.4 MPa for the west Zagros and 2 and 35 MPa for the east Zagros. The averaged values of stress drop for the west and east Zagros were 11.27 and 9.05 MPa, respectively.</div><div>Also, site amplifications derived from the H/V method were compared with the GIT results of site amplification at each station. This comparison showed that in the most stations the H/V method presents different amplification values than those of the GIT method. However, in the most stations resonance frequency reported by the H/V and the GIT methods were compatible.</div><div>The path effects including the quality factor and geometrical spreading model were adopted from <span><span>Zafarani and Hassani (2013)</span></span>. Plotting seismic moment versus corner frequency showed that the Zagros events do not follow scaling rule of <span><span>Aki (1967)</span></span> completely. The decreasing rate of seismic moment versus corner frequency equals to 1.98 in logarithmic scale.</div><div>Derived between-event and within-event residual in each frequency through studied frequency range indicated that there is a negligible trend in residual values against magnitude and distance.</div></div>","PeriodicalId":54882,"journal":{"name":"Journal of Applied Geophysics","volume":"240 ","pages":"Article 105785"},"PeriodicalIF":2.2,"publicationDate":"2025-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144154401","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Application of ground-airborne transient electromagnetic method with layered Green's function born approximation imaging in the Huang-ling coal mine goaf","authors":"Jie Ma ,&nbsp;Ziyuan Li ,&nbsp;Xiu Li ,&nbsp;Zhipeng Qi ,&nbsp;Qiding Wang","doi":"10.1016/j.jappgeo.2025.105784","DOIUrl":"10.1016/j.jappgeo.2025.105784","url":null,"abstract":"<div><div>The large-scale mining of coal resources has resulted in the formation of goaf areas, particularly the legacy issues caused by the disorderly mining of small coal mines in surrounding areas, which may lead to serious consequences such as ground subsidence, building damage, and casualties. Therefore, it is essential to conduct detailed exploration in these goaf areas. This study focuses on the goaf areas in the Huang-ling survey area, which have been preserved in the form of water filling, and the mountains are undulating, as well as the gullies are deep and wide, conventional ground geophysical methods are not suitable to carry out. In this paper, a Ground-Airborne Transient Electromagnetic (GATEM) method, which is highly sensitive to low-resistance anomalies, efficient in construction, and applicable to exploration in complex mountainous regions, is employed for detection. Aiming at the geophysical characteristics of coal mine goaf areas, a three-dimensional theoretical geoelectric model of low- and high-resistivity interlayers is established. The transient electromagnetic method (TEM) is used to define the apparent resistivity, and a comprehensive interpretation is performed through virtual wavefield imaging, proving the high effectiveness of this method. Finally, taking the exploration of the coal bed goaf zone in a mining area in Huang-ling, Shaanxi Province, as an example, the apparent resistivity is defined using TEM to delineate low-resistance, water-rich areas. The interface morphology of water-bearing structures is described by combining the results with virtual wavefield imaging. The results show that the combination of virtual wavefield imaging can suppress the volume effect and obtain richer geological information. This study offers a theoretical basis and practical reference for the application of the transient electromagnetic method in multi-layer goaf exploration.</div></div>","PeriodicalId":54882,"journal":{"name":"Journal of Applied Geophysics","volume":"240 ","pages":"Article 105784"},"PeriodicalIF":2.2,"publicationDate":"2025-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144154548","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Simultaneous estimation of basement depth and density contrast by gravity anomaly via multi-task deep learning","authors":"Lin Wang ,&nbsp;Giovanni Florio ,&nbsp;Maurizio Fedi ,&nbsp;Shengqing Xiong ,&nbsp;Wanyin Wang","doi":"10.1016/j.jappgeo.2025.105781","DOIUrl":"10.1016/j.jappgeo.2025.105781","url":null,"abstract":"<div><div>We propose a multi-task deep learning (DL) method to simultaneously estimate the basement depth and the density contrast from gravity field anomalies. The method is based on a specially designed hybrid architecture, which comprises a convolutional neural network branch and a Multilayer Perceptron branch. This hybrid architecture fully leverages the benefits of multi-task DL, enabling simultaneous estimation of basement depth and density contrast, where the input is a gravity map. In the training phase, useful statistical prior information is incorporated from a global basin dataset. Our idea is that the learning based on such dataset helps to restrict the solution to a limited domain, so leading to a reasonable estimation of the basement depth and the density contrast.</div><div>We utilize a Deep Convolutional Generative Adversarial Network (DCGAN) to generate high-quality maps of basement depths based on a global catalog of basins. The preliminary real basement maps originate from the re-interpolations and nonstandard coordinate transformations of the sediment data inside the global basins, and more additional basement samples are generated by the trained DCGAN architecture, thereby forming our dataset.</div><div>We apply the method to synthetic dataset and to two real cases, thus demonstrating the feasibility and effectiveness of our DL method. The results show good performance of our DL architecture not only for the estimated basement models, but also for the density contrast. The method candidates as a valid tool for practical applications, especially when there is a lack of constraint information in complex real cases.</div></div>","PeriodicalId":54882,"journal":{"name":"Journal of Applied Geophysics","volume":"240 ","pages":"Article 105781"},"PeriodicalIF":2.2,"publicationDate":"2025-05-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144154546","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Decoupled elastic wave reverse time migration imaging for tunnel seismic data","authors":"Xuhua Liu ,&nbsp;Jun Wang ,&nbsp;Xinglin Lu ,&nbsp;Wei Wang ,&nbsp;Yufeng Liu ,&nbsp;Zhihong Fu","doi":"10.1016/j.jappgeo.2025.105786","DOIUrl":"10.1016/j.jappgeo.2025.105786","url":null,"abstract":"<div><div>Tunnel seismic advanced prediction is the main technique for identifying unusual geological bodies along the tunneling path. The investigation accuracy of hidden karst caves is still not ideal due to the limited tunnel observation space and the slight physical differences between fissure structures and surrounding rock. Tunnel seismic data consist of full-wave data, which includes both P-waves and S-waves. In contrast to the P-wave, the S-wave moves at a slower velocity, has a smaller wavelength, and provides better resolution. The current methods for processing tunnel data include the <em>F-K</em> method, τ-p transformation, and polarization filtering for separating P- and S-waves, followed by migration imaging. It is challenging to suggest only P wave and S wave for actual data. Reducing the accuracy of migration imaging will occur due to errors in separating the P- and S-waves. Migration imaging plays a crucial role in seismic data processing for tunnels. The decoupled elastic wave reverse time migration (DE-RTM) method utilizes both P- and S-waves effectively for high-resolution imaging, making it the most precise migration technique available. DE-RTM is employed in this paper for tunnel seismic data migration imaging, suggesting an interpretation approach for multi-wave multi-component tunnel seismic data and establishing a thorough process for processing tunnel seismic data. The tunnel seismic data is used to examine in depth the wavefield characteristics of the P- and S-wave. The results from the data simulation and actual data demonstrate that DE-RTM has the ability to prevent wavefield separation errors, enhance signal accuracy, and utilize the S-wave effectively for improved imaging resolution. The dot-sum migration data in multi-wave migration data can offer low-frequency contour details, whereas the dot-product migration data can offer high-frequency detail information. By merging the benefits of each, one can achieve more dependable interpretation results.</div></div>","PeriodicalId":54882,"journal":{"name":"Journal of Applied Geophysics","volume":"240 ","pages":"Article 105786"},"PeriodicalIF":2.2,"publicationDate":"2025-05-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144125236","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Using transfer learning to enhance void detection and shear wave velocity model inversion from near-surface seismic shot gathers","authors":"Yanwei Zhang ,&nbsp;Dmitry Borisov ,&nbsp;Salman Abbasi ,&nbsp;Richard D. Miller ,&nbsp;Steven D. Sloan","doi":"10.1016/j.jappgeo.2025.105780","DOIUrl":"10.1016/j.jappgeo.2025.105780","url":null,"abstract":"<div><div>A Convolutional Neural Network (CNN) has been designed to delineate the shear-wave velocity (<em>Vs</em>) models and detect subsurface void locations. Addressing the processing and interpretation challenges posed on real seismic data, our strategy emphasizes that leveraging the ground truth, which is the void location in this study, enables the CNN to catch the identical features in real waveforms. Initially, a synthetic dataset is employed, imparting foundational knowledge to the CNN regarding the <em>Vs</em> model and void locations. Drawing inspiration from transfer learning, this pre-trained CNN serves as an initial model and is refined using a real dataset focused on void locations. After refining, the CNN shows enhanced reliability to detect the void and extract the <em>Vs</em> model, as evidenced by the improved alignment between forward modeling and real waveforms. Our findings underscore how leveraging the ground truth can actualize the potential of CNN on velocity model extraction.</div></div>","PeriodicalId":54882,"journal":{"name":"Journal of Applied Geophysics","volume":"240 ","pages":"Article 105780"},"PeriodicalIF":2.2,"publicationDate":"2025-05-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144125237","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Automated hydraulic flow unit determination using FZI-Z-score probability, K-means clustering with autoencoders, and machine learning classification: A case study of the Subei Basin, China","authors":"Stephen Adjei ,&nbsp;Deon Yeboah Takyi-Amponsem ,&nbsp;Jonathan Atuquaye Quaye ,&nbsp;Yen Adams Sokama-Neuyam ,&nbsp;John Ojuu Oleka ,&nbsp;George Kobla Asigbi ,&nbsp;Turkson Kwesi Duodu","doi":"10.1016/j.jappgeo.2025.105778","DOIUrl":"10.1016/j.jappgeo.2025.105778","url":null,"abstract":"<div><div>Subsurface characterization is fundamental to understanding key reservoir properties such as fluid flow through rock pores. The concept of Hydraulic Flow Units (HFUs) is a subsurface characterization technique that helps identify zones with similar flow characteristics. However, the conventional technique for HFU identification using probability by ranking is not statistically robust and is unsuitable for large datasets. This study utilizes RCAL data gathered from five selected wells in the Subei Basin oilfields, in China. The data consists of the sample's diameter, length, dry weight, particle volume, bulk volume, pore volume, particle density, porosity, volumetric density, location, and permeability. A <em>Z</em>-score probability approach is proposed as a statistically superior alternative to probability computation based on ranking. Additionally, the present study adopts an autoencoder-based K-means clustering approach to delineate HFUs. Subsequently, the identified HFUs are predicted using three supervised machine learning classification models: Neural Networks Multilayer Perceptron (MLP), Logistic Regression (LG), and Random Forest Classifier (RFC). The autoencoder-based K-means clustering method identified eight (8) HFUs. Feature selection was based on the strength of the relationship between features and targets and inter-feature correlations, as defined by the correlation coefficient. Using the random subsampling approach with an 80 %–20 % train-test split, the MLP emerged as the most effective and robust model for predicting HFUs, achieving a test accuracy of 94 %. By integrating the <em>Z</em>-score probability and autoencoder-based k-means approach into HFU prediction using machine learning approaches, subsurface characterization is greatly improved.</div></div>","PeriodicalId":54882,"journal":{"name":"Journal of Applied Geophysics","volume":"240 ","pages":"Article 105778"},"PeriodicalIF":2.2,"publicationDate":"2025-05-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144125235","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}