We propose a novel workflow for fast forward modeling of well logs in axially symmetric 2D models of the near-wellbore environment. The approach integrates the finite element method with deep residual neural networks to achieve exceptional computational efficiency and accuracy. The workflow is demonstrated through the modeling of wireline electromagnetic propagation resistivity logs, where the measured responses exhibit a highly nonlinear relationship with formation properties. The motivation for this research is the need for advanced modeling algorithms that are fast enough for use in modern quantitative interpretation tools, where thousands of simulations may be required in iterative inversion processes. The proposed algorithm achieves a remarkable enhancement in performance, being up to 3000 times faster than the finite element method alone when utilizing a GPU. While still ensuring high accuracy, this makes it well-suited for practical applications when reliable payzone assessment is needed in complex environmental scenarios. Furthermore, the algorithm's efficiency positions it as a promising tool for stochastic Bayesian inversion, facilitating reliable uncertainty quantification in subsurface property estimation.
{"title":"Fast 2D forward modeling of electromagnetic propagation well logs using finite element method and data-driven deep learning","authors":"A.M. Petrov, A.R. Leonenko, K.N. Danilovskiy, O.V. Nechaev","doi":"10.1016/j.aiig.2025.100112","DOIUrl":"10.1016/j.aiig.2025.100112","url":null,"abstract":"<div><div>We propose a novel workflow for fast forward modeling of well logs in axially symmetric 2D models of the near-wellbore environment. The approach integrates the finite element method with deep residual neural networks to achieve exceptional computational efficiency and accuracy. The workflow is demonstrated through the modeling of wireline electromagnetic propagation resistivity logs, where the measured responses exhibit a highly nonlinear relationship with formation properties. The motivation for this research is the need for advanced modeling algorithms that are fast enough for use in modern quantitative interpretation tools, where thousands of simulations may be required in iterative inversion processes. The proposed algorithm achieves a remarkable enhancement in performance, being up to 3000 times faster than the finite element method alone when utilizing a GPU. While still ensuring high accuracy, this makes it well-suited for practical applications when reliable payzone assessment is needed in complex environmental scenarios. Furthermore, the algorithm's efficiency positions it as a promising tool for stochastic Bayesian inversion, facilitating reliable uncertainty quantification in subsurface property estimation.</div></div>","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"6 1","pages":"Article 100112"},"PeriodicalIF":0.0,"publicationDate":"2025-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143776649","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-03-14DOI: 10.1016/j.aiig.2025.100111
Suraparb Keawsawasvong , Jim Shiau , Nhat Tan Duong , Thanachon Promwichai , Rungkhun Banyong , Van Qui Lai
This study examines the stability of three-dimensional rectangular tunnel headings in drained c-ϕ soils, incorporating surcharge effects using 3D Finite Element Limit Analysis (FELA). It focuses on the upper and lower bound solutions for three stability factors: cohesion, surcharge, and soil unit weight (Nc, Ns, and Nγ). Based on Terzaghi's principle of superposition, the analysis evaluates tunnel stability under varying parameters, such as cover-depth ratio (H/D), width-depth ratio (B/D), and friction angle (ϕ). The results align closely with previous studies, and practical design charts are provided for calculating minimum support pressures. Additionally, machine learning models (ANN and XGBoost) are used to develop accurate correlations between input parameters and stability results. A relative importance index analysis is conducted to assess the impact of these parameters. This research enhances understanding of tunnel stability and offers practical insights for tunnel design.
{"title":"Enhancing understanding of 3D rectangular tunnel heading stability in c-φ soils with surcharge loading: A comprehensive FELA analysis using three stability factors and machine learning","authors":"Suraparb Keawsawasvong , Jim Shiau , Nhat Tan Duong , Thanachon Promwichai , Rungkhun Banyong , Van Qui Lai","doi":"10.1016/j.aiig.2025.100111","DOIUrl":"10.1016/j.aiig.2025.100111","url":null,"abstract":"<div><div>This study examines the stability of three-dimensional rectangular tunnel headings in drained <em>c-ϕ</em> soils, incorporating surcharge effects using 3D Finite Element Limit Analysis (FELA). It focuses on the upper and lower bound solutions for three stability factors: cohesion, surcharge, and soil unit weight (Nc, Ns, and Nγ). Based on Terzaghi's principle of superposition, the analysis evaluates tunnel stability under varying parameters, such as cover-depth ratio (<em>H/D</em>), width-depth ratio (<em>B/D</em>), and friction angle (<em>ϕ</em>). The results align closely with previous studies, and practical design charts are provided for calculating minimum support pressures. Additionally, machine learning models (ANN and XGBoost) are used to develop accurate correlations between input parameters and stability results. A relative importance index analysis is conducted to assess the impact of these parameters. This research enhances understanding of tunnel stability and offers practical insights for tunnel design.</div></div>","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"6 1","pages":"Article 100111"},"PeriodicalIF":0.0,"publicationDate":"2025-03-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143637661","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-03-01DOI: 10.1016/j.aiig.2025.100107
Jiaqi Yan , Li Ma , Tianqi Jiang , Jing Zheng , Dewei Li , Xingzhi Teng
This paper proposed a moment tensor regression prediction technology based on ResNet for microseismic events. Taking the great advantages of deep networks in classification and regression tasks, it can realize the great potential of fast and accurate inversion of microseismic moment tensors after the network trained. This ResNet-based moment tensor prediction technology, whose input is raw recordings, does not require the extraction of data features in advance. First, we tested the network using synthetic data and performed a quantitative assessment of the errors. The results demonstrate that the network exhibits high accuracy and efficiency during the prediction phase. Next, we tested the network using real microseismic data and compared the results with those from traditional inversion methods. The error in the results was relatively small compared to traditional methods. However, the network operates more efficiently without requiring manual intervention, making it highly valuable for near-real-time monitoring applications.
{"title":"Microseismic moment tensor inversion based on ResNet model","authors":"Jiaqi Yan , Li Ma , Tianqi Jiang , Jing Zheng , Dewei Li , Xingzhi Teng","doi":"10.1016/j.aiig.2025.100107","DOIUrl":"10.1016/j.aiig.2025.100107","url":null,"abstract":"<div><div>This paper proposed a moment tensor regression prediction technology based on ResNet for microseismic events. Taking the great advantages of deep networks in classification and regression tasks, it can realize the great potential of fast and accurate inversion of microseismic moment tensors after the network trained. This ResNet-based moment tensor prediction technology, whose input is raw recordings, does not require the extraction of data features in advance. First, we tested the network using synthetic data and performed a quantitative assessment of the errors. The results demonstrate that the network exhibits high accuracy and efficiency during the prediction phase. Next, we tested the network using real microseismic data and compared the results with those from traditional inversion methods. The error in the results was relatively small compared to traditional methods. However, the network operates more efficiently without requiring manual intervention, making it highly valuable for near-real-time monitoring applications.</div></div>","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"6 1","pages":"Article 100107"},"PeriodicalIF":0.0,"publicationDate":"2025-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143551571","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-02-26DOI: 10.1016/j.aiig.2025.100110
A. Bolève, R. Eddies, M. Staring, Y. Benboudiaf, H. Pournaki, M. Nepveaux
Geotechnical parameters derived from an intrusive cone penetration test (CPT) are used to asses mechanical properties to inform the design phase of infrastructure projects. However, local, in situ 1D measurements can fail to capture 3D subsurface variations, which could mean less than optimal design decisions for foundation engineering. By coupling the localised measurements from CPTs with more global 3D measurements derived from geophysical methods, a higher fidelity 3D overview of the subsurface can be obtained. Machine Learning (ML) may offer an effective means to capture all types of geophysical information associated with CPT data at a site scale to build a 2D or 3D ground model. In this paper, we present an ML approach to build a 3D ground model of cone resistance and sleeve friction by combining several CPT measurements with Multichannel Analysis of Surface Waves (MASW) and Electrical Resistivity Tomography (ERT) data on a land site characterisation project in the United Arab Emirates (UAE). To avoid a potential overfitting problem inherent to the use of machine learning and a lack of data at certain locations, we explore the possibility of using a prior Geo-Statistical (GS) approach that attempts to constrain the overfitting process by “artificially” increasing the amount of input data. A sensitivity study is also performed on input features used to train the ML algorithm to better define the optimal combination of input features for the prediction. Our results showed that ERT data were not useful in capturing 3D variations of geotechnical properties compared to Vs due to the geographical location of the site (200 m east from the Oman Gulf) and the possible effect of saline water intrusion. Additionally, we demonstrate that the use of a prior GS phase could be a promising and interesting means to make the prediction of ground properties more robust, especially for this specific case study described in this paper. Looking ahead, better representation of the subsurface can lead to a number of benefits for stakeholders involved in developing assets. Better ground/geotechnical models mean better site calibration of design methods and fewer design assumptions for reliability-based design, creating an opportunity for value engineering in the form of lighter construction without compromising safety, shorter construction timelines, and reduced resource requirements.
{"title":"Innovative cone resistance and sleeve friction prediction from geophysics based on a coupled geo-statistical and machine learning process","authors":"A. Bolève, R. Eddies, M. Staring, Y. Benboudiaf, H. Pournaki, M. Nepveaux","doi":"10.1016/j.aiig.2025.100110","DOIUrl":"10.1016/j.aiig.2025.100110","url":null,"abstract":"<div><div>Geotechnical parameters derived from an intrusive cone penetration test (CPT) are used to asses mechanical properties to inform the design phase of infrastructure projects. However, local, in situ 1D measurements can fail to capture 3D subsurface variations, which could mean less than optimal design decisions for foundation engineering. By coupling the localised measurements from CPTs with more global 3D measurements derived from geophysical methods, a higher fidelity 3D overview of the subsurface can be obtained. Machine Learning (ML) may offer an effective means to capture all types of geophysical information associated with CPT data at a site scale to build a 2D or 3D ground model. In this paper, we present an ML approach to build a 3D ground model of cone resistance and sleeve friction by combining several CPT measurements with Multichannel Analysis of Surface Waves (MASW) and Electrical Resistivity Tomography (ERT) data on a land site characterisation project in the United Arab Emirates (UAE). To avoid a potential overfitting problem inherent to the use of machine learning and a lack of data at certain locations, we explore the possibility of using a prior Geo-Statistical (GS) approach that attempts to constrain the overfitting process by “artificially” increasing the amount of input data. A sensitivity study is also performed on input features used to train the ML algorithm to better define the optimal combination of input features for the prediction. Our results showed that ERT data were not useful in capturing 3D variations of geotechnical properties compared to V<sub>s</sub> due to the geographical location of the site (200 m east from the Oman Gulf) and the possible effect of saline water intrusion. Additionally, we demonstrate that the use of a prior GS phase could be a promising and interesting means to make the prediction of ground properties more robust, especially for this specific case study described in this paper. Looking ahead, better representation of the subsurface can lead to a number of benefits for stakeholders involved in developing assets. Better ground/geotechnical models mean better site calibration of design methods and fewer design assumptions for reliability-based design, creating an opportunity for value engineering in the form of lighter construction without compromising safety, shorter construction timelines, and reduced resource requirements.</div></div>","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"6 1","pages":"Article 100110"},"PeriodicalIF":0.0,"publicationDate":"2025-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143526930","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-02-25DOI: 10.1016/j.aiig.2025.100109
P. Zwartjes, J. Yoo
This work focuses on enhancing low frequency seismic data using a convolutional neural network trained on synthetic data. Traditional seismic data often lack both high and low frequencies, which are essential for detailed geological interpretation and various geophysical applications. Low frequency data is particularly valuable for reducing wavelet sidelobes and improving full waveform inversion (FWI). Conventional methods for bandwidth extension include seismic deconvolution and sparse inversion, which have limitations in recovering low frequencies. The study explores the potential of the U-net, which has been successful in other geophysical applications such as noise attenuation and seismic resolution enhancement. The novelty in our approach is that we do not rely on computationally expensive finite difference modelling to create training data. Instead, our synthetic training data is created from individual randomly perturbed events with variations in bandwidth, making it more adaptable to different data sets compared to previous deep learning methods. The method was tested on both synthetic and real seismic data, demonstrating effective low frequency reconstruction and sidelobe reduction. With a synthetic full waveform inversion to recover a velocity model and a seismic amplitude inversion to estimate acoustic impedance we demonstrate the validity and benefit of the proposed method. Overall, the study presents a robust approach to seismic bandwidth extension using deep learning, emphasizing the importance of diverse and well-designed but computationally inexpensive synthetic training data.
{"title":"Robust low frequency seismic bandwidth extension with a U-net and synthetic training data","authors":"P. Zwartjes, J. Yoo","doi":"10.1016/j.aiig.2025.100109","DOIUrl":"10.1016/j.aiig.2025.100109","url":null,"abstract":"<div><div>This work focuses on enhancing low frequency seismic data using a convolutional neural network trained on synthetic data. Traditional seismic data often lack both high and low frequencies, which are essential for detailed geological interpretation and various geophysical applications. Low frequency data is particularly valuable for reducing wavelet sidelobes and improving full waveform inversion (FWI). Conventional methods for bandwidth extension include seismic deconvolution and sparse inversion, which have limitations in recovering low frequencies. The study explores the potential of the U-net, which has been successful in other geophysical applications such as noise attenuation and seismic resolution enhancement. The novelty in our approach is that we do not rely on computationally expensive finite difference modelling to create training data. Instead, our synthetic training data is created from individual randomly perturbed events with variations in bandwidth, making it more adaptable to different data sets compared to previous deep learning methods. The method was tested on both synthetic and real seismic data, demonstrating effective low frequency reconstruction and sidelobe reduction. With a synthetic full waveform inversion to recover a velocity model and a seismic amplitude inversion to estimate acoustic impedance we demonstrate the validity and benefit of the proposed method. Overall, the study presents a robust approach to seismic bandwidth extension using deep learning, emphasizing the importance of diverse and well-designed but computationally inexpensive synthetic training data.</div></div>","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"6 1","pages":"Article 100109"},"PeriodicalIF":0.0,"publicationDate":"2025-02-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143512149","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-02-21DOI: 10.1016/j.aiig.2025.100108
Congcong Yuan , Jie Zhang
The availability of a tremendous amount of seismic data demands seismological researchers to analyze seismic phases efficiently. Recently, deep learning algorithms exhibit a powerful capability of detecting and picking on P- and S-wave phases. However, it remains a challenge to effeciently process enormous teleseismic phases, which are crucial to probe Earth's interior structures and their dynamics. In this study, we propose a scheme to detect and pick teleseismic phases, such as seismic phase that reflects off the core-mantle boundary (i.e., PcP) and that reflects off the inner-core boundary (i.e., PKiKP), from a seismic dataset in Japan. The scheme consists of three steps: 1) latent phase traces are truncated from the whole seismogram with theoretical arrival times; 2) latent phases are recognized and evaluated by convolutional neural network (CNN) models; 3) arrivals of good or fair phase are picked with another CNN models. The testing detection result on 7386 seismograms shows that the scheme recognizes 92.15% and 94.13% of PcP and PKiKP phases. The testing picking result has a mean absolute error of 0.0742 s and 0.0636 s for the PcP and PKiKP phases, respectively. These seismograms were processed in just 5 min for phase detection and picking, demonstrating the efficiency of the proposed scheme in automatic teleseismic phase analysis.
{"title":"Applying deep learning to teleseismic phase detection and picking: PcP and PKiKP cases","authors":"Congcong Yuan , Jie Zhang","doi":"10.1016/j.aiig.2025.100108","DOIUrl":"10.1016/j.aiig.2025.100108","url":null,"abstract":"<div><div>The availability of a tremendous amount of seismic data demands seismological researchers to analyze seismic phases efficiently. Recently, deep learning algorithms exhibit a powerful capability of detecting and picking on P- and S-wave phases. However, it remains a challenge to effeciently process enormous teleseismic phases, which are crucial to probe Earth's interior structures and their dynamics. In this study, we propose a scheme to detect and pick teleseismic phases, such as seismic phase that reflects off the core-mantle boundary (i.e., PcP) and that reflects off the inner-core boundary (i.e., PKiKP), from a seismic dataset in Japan. The scheme consists of three steps: 1) latent phase traces are truncated from the whole seismogram with theoretical arrival times; 2) latent phases are recognized and evaluated by convolutional neural network (CNN) models; 3) arrivals of good or fair phase are picked with another CNN models. The testing detection result on 7386 seismograms shows that the scheme recognizes 92.15% and 94.13% of PcP and PKiKP phases. The testing picking result has a mean absolute error of 0.0742 s and 0.0636 s for the PcP and PKiKP phases, respectively. These seismograms were processed in just 5 min for phase detection and picking, demonstrating the efficiency of the proposed scheme in automatic teleseismic phase analysis.</div></div>","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"6 1","pages":"Article 100108"},"PeriodicalIF":0.0,"publicationDate":"2025-02-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143488648","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-02-13DOI: 10.1016/j.aiig.2025.100105
Mohanad Diab , Polychronis Kolokoussis , Maria Antonia Brovelli
The use of AI technologies in remote sensing (RS) tasks has been the focus of many individuals in both the professional and academic domains. Having more accessible interfaces and tools that allow people of little or no experience to intuitively interact with RS data of multiple formats is a potential provided by this integration. However, the use of AI and AI agents to help automate RS-related tasks is still in its infancy stage, with some frameworks and interfaces built on top of well-known vision language models (VLM) such as GPT-4, segment anything model (SAM), and grounding DINO. These tools do promise and draw guidelines on the potentials and limitations of existing solutions concerning the use of said models. In this work, the state of the art AI foundation models (FM) are reviewed and used in a multi-modal manner to ingest RS imagery input and perform zero-shot object detection using natural language. The natural language input is then used to define the classes or labels the model should look for, then, both inputs are fed to the pipeline. The pipeline presented in this work makes up for the shortcomings of the general knowledge FMs by stacking pre-processing and post-processing applications on top of the FMs; these applications include tiling to produce uniform patches of the original image for faster detection, outlier rejection of redundant bounding boxes using statistical and machine learning methods. The pipeline was tested with UAV, aerial and satellite images taken over multiple areas. The accuracy for the semantic segmentation showed improvement from the original 64% to approximately 80%–99% by utilizing the pipeline and techniques proposed in this work. GitHub Repository:MohanadDiab/LangRS.
{"title":"Optimizing zero-shot text-based segmentation of remote sensing imagery using SAM and Grounding DINO","authors":"Mohanad Diab , Polychronis Kolokoussis , Maria Antonia Brovelli","doi":"10.1016/j.aiig.2025.100105","DOIUrl":"10.1016/j.aiig.2025.100105","url":null,"abstract":"<div><div>The use of AI technologies in remote sensing (RS) tasks has been the focus of many individuals in both the professional and academic domains. Having more accessible interfaces and tools that allow people of little or no experience to intuitively interact with RS data of multiple formats is a potential provided by this integration. However, the use of AI and AI agents to help automate RS-related tasks is still in its infancy stage, with some frameworks and interfaces built on top of well-known vision language models (VLM) such as GPT-4, segment anything model (SAM), and grounding DINO. These tools do promise and draw guidelines on the potentials and limitations of existing solutions concerning the use of said models. In this work, the state of the art AI foundation models (FM) are reviewed and used in a multi-modal manner to ingest RS imagery input and perform zero-shot object detection using natural language. The natural language input is then used to define the classes or labels the model should look for, then, both inputs are fed to the pipeline. The pipeline presented in this work makes up for the shortcomings of the general knowledge FMs by stacking pre-processing and post-processing applications on top of the FMs; these applications include tiling to produce uniform patches of the original image for faster detection, outlier rejection of redundant bounding boxes using statistical and machine learning methods. The pipeline was tested with UAV, aerial and satellite images taken over multiple areas. The accuracy for the semantic segmentation showed improvement from the original 64% to approximately 80%–99% by utilizing the pipeline and techniques proposed in this work. <strong>GitHub Repository:</strong> <span><span>MohanadDiab/LangRS</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"6 1","pages":"Article 100105"},"PeriodicalIF":0.0,"publicationDate":"2025-02-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143429918","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-27DOI: 10.1016/j.aiig.2025.100106
Kaoutar Clero , Said Ed-Diny , Mohammed Achalhi , Mouhamed Cherkaoui , Imad El Harraki , Sanaa El Fkihi , Intissar Benzakour , Tarik Soror , Said Rziki , Hamd Ait Abdelali , Hicham Tagemouati , François Bourzeix
Rockfalls are among the frequent hazards in underground mines worldwide, requiring effective methods for detecting unstable rock blocks to ensure miners' and equipment's safety. This study proposes a novel approach for identifying potential rockfall zones using infrared thermal imaging and image segmentation techniques. Infrared images of rock blocks were captured at the Draa Sfar deep underground mine in Morocco using the FLUKE TI401 PRO thermal camera. Two segmentation methods were applied to locate the potential unstable areas: the classical thresholding and the K-means clustering model. The results show that while thresholding allows a binary distinction between stable and unstable areas, K-means clustering is more accurate, especially when using multiple clusters to show different risk levels. The close match between the clustering masks of unstable blocks and their corresponding visible light images further validated this. The findings confirm that thermal image segmentation can serve as an alternative method for predicting rockfalls and monitoring geotechnical issues in underground mines. Underground operators worldwide can apply this approach to monitor rock mass stability. However, further research is recommended to enhance these results, particularly through deep learning-based segmentation and object detection models.
{"title":"Loosening rocks detection at Draa Sfar deep underground mine in Morocco using infrared thermal imaging and image segmentation models","authors":"Kaoutar Clero , Said Ed-Diny , Mohammed Achalhi , Mouhamed Cherkaoui , Imad El Harraki , Sanaa El Fkihi , Intissar Benzakour , Tarik Soror , Said Rziki , Hamd Ait Abdelali , Hicham Tagemouati , François Bourzeix","doi":"10.1016/j.aiig.2025.100106","DOIUrl":"10.1016/j.aiig.2025.100106","url":null,"abstract":"<div><div>Rockfalls are among the frequent hazards in underground mines worldwide, requiring effective methods for detecting unstable rock blocks to ensure miners' and equipment's safety. This study proposes a novel approach for identifying potential rockfall zones using infrared thermal imaging and image segmentation techniques. Infrared images of rock blocks were captured at the Draa Sfar deep underground mine in Morocco using the FLUKE TI401 PRO thermal camera. Two segmentation methods were applied to locate the potential unstable areas: the classical thresholding and the K-means clustering model. The results show that while thresholding allows a binary distinction between stable and unstable areas, K-means clustering is more accurate, especially when using multiple clusters to show different risk levels. The close match between the clustering masks of unstable blocks and their corresponding visible light images further validated this. The findings confirm that thermal image segmentation can serve as an alternative method for predicting rockfalls and monitoring geotechnical issues in underground mines. Underground operators worldwide can apply this approach to monitor rock mass stability. However, further research is recommended to enhance these results, particularly through deep learning-based segmentation and object detection models.</div></div>","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"6 1","pages":"Article 100106"},"PeriodicalIF":0.0,"publicationDate":"2025-01-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143143632","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号