{"id":5178,"date":"2024-02-08T11:56:11","date_gmt":"2024-02-08T10:56:11","guid":{"rendered":"https:\/\/ingenius.ecoledesponts.fr\/?p=5178"},"modified":"2024-11-15T10:28:01","modified_gmt":"2024-11-15T09:28:01","slug":"machine-learning-in-geotechnics","status":"publish","type":"post","link":"https:\/\/ingenius.ecoledesponts.fr\/en\/articles\/machine-learning-in-geotechnics\/","title":{"rendered":"Machine Learning in Geotechnics"},"content":{"rendered":"\n\n\n<figure class=\"wp-block-image alignwide size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"294\" src=\"https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2024\/01\/Image1-1024x294.png\" alt=\"\" class=\"wp-image-5162\" srcset=\"https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2024\/01\/Image1-1024x294.png 1024w, https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2024\/01\/Image1-300x86.png 300w, https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2024\/01\/Image1-768x221.png 768w, https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2024\/01\/Image1.png 1386w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><figcaption class=\"wp-element-caption\">Source : Lina-Mar\u00eda Guayac\u00e1n-Carrillo<\/figcaption><\/figure>\n\n\n\n<p>The design, construction and maintenance of geotechnical structures address socioeconomic and environmental concerns, especially in terms of safety, reliability, and optimization. Interdisciplinary approaches and the use of artificial intelligence tools can play an invaluable role in meeting the most pressing needs in this field of engineering.<\/p>\n\n\n\n<p>In recent years, this field has seen rapid growth in the use of artificial intelligence, and in particular the subfield of machine learning. These techniques have many advantages in terms of their advanced computational performance and applicability to problems involving many simultaneously interacting parameters or variables (high-dimensional, non-linear problems). It is therefore essential to take a closer look at existing techniques, analyze their advantages and disadvantages, and explore the opportunities they offer.<\/p>\n\n\n\n<h2 class=\"wp-block-heading has-red-color has-text-color\">Obtaining and producing geotechnical data<\/h2>\n\n\n\n<p>Datasets obtained during geotechnical projects still tend to be small compared with other fields (e.g., image recognition and robotics), even though the current technological transition is improving the situation. The first challenge in using machine learning is therefore the quality and sufficiency of the data.<\/p>\n\n\n\n<p>There are two types of data:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><em>raw data<\/em>, obtained from <em>in-situ<\/em> monitoring, and<\/li>\n<\/ul>\n\n\n\n<p><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><em>synthetic data<\/em>, generated from purely theoretical-numerical approaches and\/or based on fragmentary information from the field.<\/li>\n<\/ul>\n\n\n\n<p><em>Raw<\/em> or <em>in-situ data<\/em> can be noisy and incomplete, due in particular to the accuracy of sensors, their potential failure, lack of sufficient lighting, or obstructed access to the area monitored. In addition, this data is varied (in terms of ground behavior, excavation method parameters, structure-specific characteristics) and variable (in terms of both space and time).<\/p>\n\n\n\n<p>There are three key points to bear in mind to ensure this data is processed and used correctly:<\/p>\n\n\n\n<p>(1) Identifying the data available and required for project modeling;<\/p>\n\n\n\n<p>(2) Obtaining and cleansing the data (data organization, deleting duplicate data, etc.); and<\/p>\n\n\n\n<p>(3) Storing the data in a database for more efficient use.<\/p>\n\n\n\n<p>An example of how to set up a <a href=\"https:\/\/www.researchgate.net\/publication\/361819272_Constitution_d'une_base_de_donnees_des_mesures_obtenues_lors_du_creusement_de_deux_tunnels_du_Grand_Paris_Express\" target=\"_blank\" rel=\"noreferrer noopener\">database of measurements obtained during the excavation of two tunnels for the Grand Paris Express<\/a> is detailed in Richa et al.<\/p>\n\n\n\n<p>As far as <em>synthetic data<\/em> is concerned (i.e., data derived from analytical solutions or generated by numerical simulations and\/or artificial intelligence tools), there are two possible usage scenarios:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>When existing data is insufficient, artificial data can be generated using information already collected and initial simplified models. For example, <a href=\"https:\/\/www.presses-des-ponts.fr\/notre-librairie\/406-transitions--les-nouvelles-annales-des-ponts-et-chaussees-n3.html\" target=\"_blank\" rel=\"noreferrer noopener\">as part of a study on monitoring the stability of natural and man-made slopes by tracking surface displacements<\/a> (i.e., displacements tracked at the ground surface using optical targets), a data generation methodology covering various configurations (geometry and material properties) and various climatic scenarios was developed to generate 2,000 changes in movement speed over a 10-year period.[2]<\/li>\n<\/ul>\n\n\n\n<p><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Where access to field data is not possible, for example during a structure\u2019s design phase, data generation can be used to take account of a wide range of parameters (including geological conditions and structure characteristics) for preliminary design analyses. A more detailed example of the use of this type of data is presented in the following.<\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading has-red-color has-text-color\">Machine learning in geotechnics: efficient and reliable techniques?<\/h2>\n\n\n\n<p>Once the database is in place, work can begin on applying machine learning algorithms. In the case of small datasets, which are common in geotechnical projects, the learning model overfits the training data, resulting in poor predictions for new data . One practical solution to this problem is to combine several models to make an overall prediction. In most cases, this leads to better results.<\/p>\n\n\n\n<p>Various studies on tunnel excavation projects, using both<a href=\"https:\/\/eposter.at\/ISRM2023\/data\/PDF\/1163.pdf\" target=\"_blank\" rel=\"noreferrer noopener\"> synthetic dat<\/a>a and data from<a href=\"https:\/\/hal.science\/hal-04376951\"> <em>in-situ<\/em> monitoring<\/a>, have concluded that learning techniques based on these what are known as \u201censemble\u201d methods seem to perform best when using small datasets for model training, and deliver the most reliable predictions.<\/p>\n\n\n\n<h2 class=\"wp-block-heading has-red-color has-text-color\">How to optimize the use of machine learning in Geotechnics<\/h2>\n\n\n\n<p>Geotechnical data is highly variable in space and time. To accurately represent the phenomena observed <em>in situ<\/em>, the objective must be set in advance. For example, to represent the progress of the excavation of an underground structure and its impact on the surrounding area, the engineer aims to estimate what is happening in front of the cutting face <sup data-fn=\"14141f49-8f9e-4fed-9ee6-a12183629ef7\" class=\"fn\"><a href=\"#14141f49-8f9e-4fed-9ee6-a12183629ef7\" id=\"14141f49-8f9e-4fed-9ee6-a12183629ef7-link\">1<\/a><\/sup> They therefore need to collect information as excavation progresses, such as surface settlement measurements, convergence measurements <sup data-fn=\"849c2e54-b2ee-409f-b61d-8b23c4e0ab1c\" class=\"fn\"><a href=\"#849c2e54-b2ee-409f-b61d-8b23c4e0ab1c\" id=\"849c2e54-b2ee-409f-b61d-8b23c4e0ab1c-link\">2<\/a><\/sup> and information linked to excavation machine parameters. In other words, the engineer will try to use the data collected during excavation (everything behind the cutting face) for training, and everything in front of the cutting face (the area not yet excavated) for forecasting. <a href=\"https:\/\/ingenius.ecoledesponts.fr\/articles\/intelligence-artificielle-au-service-du-creusement-des-tunnels\/\" target=\"_blank\" rel=\"noreferrer noopener\">Machine learning has been used, for example, to predict the settlement caused by tunneling on two lines of the Grand Paris Express.<\/a><\/p>\n\n\n\n<p>Another example of estimating how convergence varies over time during deep tunneling is presented in <a href=\"https:\/\/www.sciencedirect.com\/science\/article\/pii\/S0266352X2400291X\" target=\"_blank\" rel=\"noreferrer noopener\">Guayac\u00e1n-Carrillo &amp; Sulem<\/a>.The methodology proposed involves using the information recorded during excavation for training purposes and finding a simple mathematical expression that captures the changes in convergence over time. This approach takes account of the intrinsic anisotropy of rock formation and the anisotropy of the initial stress state. This type of approach can provide additional and complementary information to support the observational method. It provides a dynamic approach whose predictions can be refined as new <em>in-situ<\/em> measurements are taken.<\/p>\n\n\n\n<h2 class=\"wp-block-heading has-red-color has-text-color\">Numerical modeling and machine learning: two complementary approaches<\/h2>\n\n\n\n<p>In a nutshell, surrogate models are based on artificial intelligence tools trained on synthetic datasets. In certain cases, using these models may be appropriate (especially when 3D numerical simulations are required). The main idea is to establish a complementary relationship between numerical modeling and machine learning to provide engineers with simple, reliable tools that avoid high computational efforts and costs.<\/p>\n\n\n\n<p>As part of their work on the sizing of deep tunnels, a team from the Navier laboratory used 3D numerical simulations to understand how rock reacts to tunnel construction. These simulations are especially important when the rock undergoes significant deformation, or when the supporting structures are very rigid and located close to the cutting face. For this study, <a href=\"https:\/\/eposter.at\/ISRM2023\/data\/PDF\/1163.pdf\" target=\"_blank\" rel=\"noreferrer noopener\">a synthetic dataset was used to take account of a wide range of rock and support structure characteristics, as well as tunnel sizes<\/a>.<\/p>\n\n\n\n<p>These models have three main advantages. They are trained on data managed by specialists, provide engineers with rapid estimates in a matter of seconds, and are suitable for statistical analyses requiring multiple simulations.<\/p>\n\n\n\n<h2 class=\"wp-block-heading has-red-color has-text-color\">What are the future prospects?<\/h2>\n\n\n\n<p>An interdisciplinary field of research is emerging, involving the development of simple behavioral machine learning models that can reproduce the overall behavior of the ground around the structure, both over the short and long term. The aim is to provide engineering with the new tools needed to design high-performance structures.<\/p>\n\n\n\n<p>While research at the interface between artificial intelligence and geotechnics has been ongoing for some years now, a number of questions remain unanswered. In particular, it is essential to take advantage of the wealth of data by making them a strategic part of the action plan. This involves thinking about how to share this data with the geotechnical community, possibly via Open Data platforms, and leveraging past experience to inform new projects. At the same time, it is crucial to develop frugal artificial intelligence, particularly when it comes to optimizing and harmonizing recovery\/preparation\/operation procedures and cycles. Finally, it is important to think about how to introduce these new machine learning approaches into the training of future engineers and into industry practice, without losing know-how and business expertise.<\/p>\n\n\n\n<p><\/p>\n\n\n<ol class=\"wp-block-footnotes\"><li id=\"14141f49-8f9e-4fed-9ee6-a12183629ef7\">Cutting face: the term used to refer to the working face of the tunnel excavation. <a href=\"#14141f49-8f9e-4fed-9ee6-a12183629ef7-link\" aria-label=\"Jump to footnote reference 1\">\u21a9\ufe0e<\/a><\/li><li id=\"849c2e54-b2ee-409f-b61d-8b23c4e0ab1c\">Convergence: the relative displacement of two opposite points on the wall in a tunnel cross-section. <a href=\"#849c2e54-b2ee-409f-b61d-8b23c4e0ab1c-link\" aria-label=\"Jump to footnote reference 2\">\u21a9\ufe0e<\/a><\/li><\/ol>","protected":false},"excerpt":{"rendered":"<p>The design, construction and maintenance of geotechnical structures address socioeconomic and environmental concerns, especially in terms of safety, reliability, and [&hellip;]<\/p>\n","protected":false},"author":6,"featured_media":5162,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_related_content_post":[],"_related_content_subject":[690,936],"_related_content_author":[5186,5188,5187],"_related_content_category":[1720],"_related_content_folder":[5213],"_excerpt":"Whether excavating a tunnel, building an embankment, or constructing a dam, the design of these types of structures requires excellent knowledge of soil properties. It also raises many scientific and technological challenges. Advances in artificial intelligence in recent years are opening up some exciting opportunities to meet these challenges.","_duration":5,"_manual_duration":false,"footnotes":"[{\"content\":\"Cutting face: the term used to refer to the working face of the tunnel excavation.\",\"id\":\"14141f49-8f9e-4fed-9ee6-a12183629ef7\"},{\"content\":\"Convergence: the relative displacement of two opposite points on the wall in a tunnel cross-section.\",\"id\":\"849c2e54-b2ee-409f-b61d-8b23c4e0ab1c\"}]"},"article-types":[27],"class_list":["post-5178","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","article-types-folder"],"has_blocks":true,"block_data":[{"blockName":"enpc\/excerpt","attrs":{"lock":[],"metadata":[],"className":"","style":""},"innerBlocks":[],"innerHTML":"","innerContent":[],"rendered":""},{"blockName":"core\/image","attrs":{"id":5162,"sizeSlug":"large","linkDestination":"none","align":"wide","blob":"","url":"https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2024\/01\/Image1-1024x294.png","alt":"","caption":"Source : Lina-Mar\u00eda Guayac\u00e1n-Carrillo","lightbox":[],"title":"","href":"","rel":"","linkClass":"","width":"","height":"","aspectRatio":"","scale":"","linkTarget":"","lock":[],"metadata":[],"className":"wp-block-image alignwide size-large","style":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<figure class=\"wp-block-image alignwide size-large\"><img src=\"https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2024\/01\/Image1-1024x294.png\" alt=\"\" class=\"wp-image-5162\"\/><figcaption class=\"wp-element-caption\">Source : Lina-Mar\u00eda Guayac\u00e1n-Carrillo<\/figcaption><\/figure>\n","innerContent":["\n<figure class=\"wp-block-image alignwide size-large\"><img src=\"https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2024\/01\/Image1-1024x294.png\" alt=\"\" class=\"wp-image-5162\"\/><figcaption class=\"wp-element-caption\">Source : Lina-Mar\u00eda Guayac\u00e1n-Carrillo<\/figcaption><\/figure>\n"],"rendered":"\n<figure class=\"wp-block-image alignwide size-large\"><img src=\"https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2024\/01\/Image1-1024x294.png\" alt=\"\" class=\"wp-image-5162\"\/><figcaption class=\"wp-element-caption\">Source : Lina-Mar\u00eda Guayac\u00e1n-Carrillo<\/figcaption><\/figure>\n"},{"blockName":"core\/paragraph","attrs":{"align":"","content":"The design, construction and maintenance of geotechnical structures address socioeconomic and environmental concerns, especially in terms of safety, reliability, and optimization. Interdisciplinary approaches and the use of artificial intelligence tools can play an invaluable role in meeting the most pressing needs in this field of engineering.","dropCap":false,"placeholder":"","direction":"","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<p>The design, construction and maintenance of geotechnical structures address socioeconomic and environmental concerns, especially in terms of safety, reliability, and optimization. Interdisciplinary approaches and the use of artificial intelligence tools can play an invaluable role in meeting the most pressing needs in this field of engineering.<\/p>\n","innerContent":["\n<p>The design, construction and maintenance of geotechnical structures address socioeconomic and environmental concerns, especially in terms of safety, reliability, and optimization. Interdisciplinary approaches and the use of artificial intelligence tools can play an invaluable role in meeting the most pressing needs in this field of engineering.<\/p>\n"],"rendered":"\n<p>The design, construction and maintenance of geotechnical structures address socioeconomic and environmental concerns, especially in terms of safety, reliability, and optimization. Interdisciplinary approaches and the use of artificial intelligence tools can play an invaluable role in meeting the most pressing needs in this field of engineering.<\/p>\n"},{"blockName":"core\/paragraph","attrs":{"align":"","content":"In recent years, this field has seen rapid growth in the use of artificial intelligence, and in particular the subfield of machine learning. These techniques have many advantages in terms of their advanced computational performance and applicability to problems involving many simultaneously interacting parameters or variables (high-dimensional, non-linear problems). It is therefore essential to take a closer look at existing techniques, analyze their advantages and disadvantages, and explore the opportunities they offer.","dropCap":false,"placeholder":"","direction":"","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<p>In recent years, this field has seen rapid growth in the use of artificial intelligence, and in particular the subfield of machine learning. These techniques have many advantages in terms of their advanced computational performance and applicability to problems involving many simultaneously interacting parameters or variables (high-dimensional, non-linear problems). It is therefore essential to take a closer look at existing techniques, analyze their advantages and disadvantages, and explore the opportunities they offer.<\/p>\n","innerContent":["\n<p>In recent years, this field has seen rapid growth in the use of artificial intelligence, and in particular the subfield of machine learning. These techniques have many advantages in terms of their advanced computational performance and applicability to problems involving many simultaneously interacting parameters or variables (high-dimensional, non-linear problems). It is therefore essential to take a closer look at existing techniques, analyze their advantages and disadvantages, and explore the opportunities they offer.<\/p>\n"],"rendered":"\n<p>In recent years, this field has seen rapid growth in the use of artificial intelligence, and in particular the subfield of machine learning. These techniques have many advantages in terms of their advanced computational performance and applicability to problems involving many simultaneously interacting parameters or variables (high-dimensional, non-linear problems). It is therefore essential to take a closer look at existing techniques, analyze their advantages and disadvantages, and explore the opportunities they offer.<\/p>\n"},{"blockName":"core\/heading","attrs":{"textColor":"red","textAlign":"","content":"Obtaining and producing geotechnical data","level":2,"levelOptions":[],"placeholder":"","lock":[],"metadata":[],"align":"","className":"wp-block-heading has-red-color has-text-color","style":"","backgroundColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<h2 class=\"wp-block-heading has-red-color has-text-color\">Obtaining and producing geotechnical data<\/h2>\n","innerContent":["\n<h2 class=\"wp-block-heading has-red-color has-text-color\">Obtaining and producing geotechnical data<\/h2>\n"],"rendered":"\n<h2 class=\"wp-block-heading has-red-color has-text-color\">Obtaining and producing geotechnical data<\/h2>\n"},{"blockName":"core\/paragraph","attrs":{"align":"","content":"Datasets obtained during geotechnical projects still tend to be small compared with other fields (e.g., image recognition and robotics), even though the current technological transition is improving the situation. The first challenge in using machine learning is therefore the quality and sufficiency of the data.","dropCap":false,"placeholder":"","direction":"","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<p>Datasets obtained during geotechnical projects still tend to be small compared with other fields (e.g., image recognition and robotics), even though the current technological transition is improving the situation. The first challenge in using machine learning is therefore the quality and sufficiency of the data.<\/p>\n","innerContent":["\n<p>Datasets obtained during geotechnical projects still tend to be small compared with other fields (e.g., image recognition and robotics), even though the current technological transition is improving the situation. The first challenge in using machine learning is therefore the quality and sufficiency of the data.<\/p>\n"],"rendered":"\n<p>Datasets obtained during geotechnical projects still tend to be small compared with other fields (e.g., image recognition and robotics), even though the current technological transition is improving the situation. The first challenge in using machine learning is therefore the quality and sufficiency of the data.<\/p>\n"},{"blockName":"core\/paragraph","attrs":{"align":"","content":"There are two types of data:","dropCap":false,"placeholder":"","direction":"","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<p>There are two types of data:<\/p>\n","innerContent":["\n<p>There are two types of data:<\/p>\n"],"rendered":"\n<p>There are two types of data:<\/p>\n"},{"blockName":"core\/list","attrs":{"ordered":false,"values":"","type":"","start":0,"reversed":false,"placeholder":"","lock":[],"metadata":[],"className":"wp-block-list","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[{"blockName":"core\/list-item","attrs":{"placeholder":"","content":"raw data, obtained from in-situ monitoring, and","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<li><em>raw data<\/em>, obtained from <em>in-situ<\/em> monitoring, and<\/li>\n","innerContent":["\n<li><em>raw data<\/em>, obtained from <em>in-situ<\/em> monitoring, and<\/li>\n"],"rendered":"\n<li><em>raw data<\/em>, obtained from <em>in-situ<\/em> monitoring, and<\/li>\n"}],"innerHTML":"\n<ul class=\"wp-block-list\"><\/ul>\n","innerContent":["\n<ul class=\"wp-block-list\">",null,"<\/ul>\n"],"rendered":"\n<ul class=\"wp-block-list\">\n<li><em>raw data<\/em>, obtained from <em>in-situ<\/em> monitoring, and<\/li>\n<\/ul>\n"},{"blockName":"core\/paragraph","attrs":{"align":"","content":"","dropCap":false,"placeholder":"","direction":"","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<p><\/p>\n","innerContent":["\n<p><\/p>\n"],"rendered":"\n<p><\/p>\n"},{"blockName":"core\/list","attrs":{"ordered":false,"values":"","type":"","start":0,"reversed":false,"placeholder":"","lock":[],"metadata":[],"className":"wp-block-list","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[{"blockName":"core\/list-item","attrs":{"placeholder":"","content":"synthetic data, generated from purely theoretical-numerical approaches and\/or based on fragmentary information from the field.","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<li><em>synthetic data<\/em>, generated from purely theoretical-numerical approaches and\/or based on fragmentary information from the field.<\/li>\n","innerContent":["\n<li><em>synthetic data<\/em>, generated from purely theoretical-numerical approaches and\/or based on fragmentary information from the field.<\/li>\n"],"rendered":"\n<li><em>synthetic data<\/em>, generated from purely theoretical-numerical approaches and\/or based on fragmentary information from the field.<\/li>\n"}],"innerHTML":"\n<ul class=\"wp-block-list\"><\/ul>\n","innerContent":["\n<ul class=\"wp-block-list\">",null,"<\/ul>\n"],"rendered":"\n<ul class=\"wp-block-list\">\n<li><em>synthetic data<\/em>, generated from purely theoretical-numerical approaches and\/or based on fragmentary information from the field.<\/li>\n<\/ul>\n"},{"blockName":"core\/paragraph","attrs":{"align":"","content":"Raw or in-situ data can be noisy and incomplete, due in particular to the accuracy of sensors, their potential failure, lack of sufficient lighting, or obstructed access to the area monitored. In addition, this data is varied (in terms of ground behavior, excavation method parameters, structure-specific characteristics) and variable (in terms of both space and time).","dropCap":false,"placeholder":"","direction":"","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<p><em>Raw<\/em> or <em>in-situ data<\/em> can be noisy and incomplete, due in particular to the accuracy of sensors, their potential failure, lack of sufficient lighting, or obstructed access to the area monitored. In addition, this data is varied (in terms of ground behavior, excavation method parameters, structure-specific characteristics) and variable (in terms of both space and time).<\/p>\n","innerContent":["\n<p><em>Raw<\/em> or <em>in-situ data<\/em> can be noisy and incomplete, due in particular to the accuracy of sensors, their potential failure, lack of sufficient lighting, or obstructed access to the area monitored. In addition, this data is varied (in terms of ground behavior, excavation method parameters, structure-specific characteristics) and variable (in terms of both space and time).<\/p>\n"],"rendered":"\n<p><em>Raw<\/em> or <em>in-situ data<\/em> can be noisy and incomplete, due in particular to the accuracy of sensors, their potential failure, lack of sufficient lighting, or obstructed access to the area monitored. In addition, this data is varied (in terms of ground behavior, excavation method parameters, structure-specific characteristics) and variable (in terms of both space and time).<\/p>\n"},{"blockName":"core\/paragraph","attrs":{"align":"","content":"There are three key points to bear in mind to ensure this data is processed and used correctly:","dropCap":false,"placeholder":"","direction":"","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<p>There are three key points to bear in mind to ensure this data is processed and used correctly:<\/p>\n","innerContent":["\n<p>There are three key points to bear in mind to ensure this data is processed and used correctly:<\/p>\n"],"rendered":"\n<p>There are three key points to bear in mind to ensure this data is processed and used correctly:<\/p>\n"},{"blockName":"core\/paragraph","attrs":{"align":"","content":"(1) Identifying the data available and required for project modeling;","dropCap":false,"placeholder":"","direction":"","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<p>(1) Identifying the data available and required for project modeling;<\/p>\n","innerContent":["\n<p>(1) Identifying the data available and required for project modeling;<\/p>\n"],"rendered":"\n<p>(1) Identifying the data available and required for project modeling;<\/p>\n"},{"blockName":"core\/paragraph","attrs":{"align":"","content":"(2) Obtaining and cleansing the data (data organization, deleting duplicate data, etc.); and","dropCap":false,"placeholder":"","direction":"","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<p>(2) Obtaining and cleansing the data (data organization, deleting duplicate data, etc.); and<\/p>\n","innerContent":["\n<p>(2) Obtaining and cleansing the data (data organization, deleting duplicate data, etc.); and<\/p>\n"],"rendered":"\n<p>(2) Obtaining and cleansing the data (data organization, deleting duplicate data, etc.); and<\/p>\n"},{"blockName":"core\/paragraph","attrs":{"align":"","content":"(3) Storing the data in a database for more efficient use.","dropCap":false,"placeholder":"","direction":"","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<p>(3) Storing the data in a database for more efficient use.<\/p>\n","innerContent":["\n<p>(3) Storing the data in a database for more efficient use.<\/p>\n"],"rendered":"\n<p>(3) Storing the data in a database for more efficient use.<\/p>\n"},{"blockName":"core\/paragraph","attrs":{"align":"","content":"An example of how to set up a database of measurements obtained during the excavation of two tunnels for the Grand Paris Express is detailed in Richa et al.","dropCap":false,"placeholder":"","direction":"","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<p>An example of how to set up a <a href=\"https:\/\/www.researchgate.net\/publication\/361819272_Constitution_d'une_base_de_donnees_des_mesures_obtenues_lors_du_creusement_de_deux_tunnels_du_Grand_Paris_Express\" target=\"_blank\" rel=\"noreferrer noopener\">database of measurements obtained during the excavation of two tunnels for the Grand Paris Express<\/a> is detailed in Richa et al.<\/p>\n","innerContent":["\n<p>An example of how to set up a <a href=\"https:\/\/www.researchgate.net\/publication\/361819272_Constitution_d'une_base_de_donnees_des_mesures_obtenues_lors_du_creusement_de_deux_tunnels_du_Grand_Paris_Express\" target=\"_blank\" rel=\"noreferrer noopener\">database of measurements obtained during the excavation of two tunnels for the Grand Paris Express<\/a> is detailed in Richa et al.<\/p>\n"],"rendered":"\n<p>An example of how to set up a <a href=\"https:\/\/www.researchgate.net\/publication\/361819272_Constitution_d'une_base_de_donnees_des_mesures_obtenues_lors_du_creusement_de_deux_tunnels_du_Grand_Paris_Express\" target=\"_blank\" rel=\"noreferrer noopener\">database of measurements obtained during the excavation of two tunnels for the Grand Paris Express<\/a> is detailed in Richa et al.<\/p>\n"},{"blockName":"core\/paragraph","attrs":{"align":"","content":"As far as synthetic data is concerned (i.e., data derived from analytical solutions or generated by numerical simulations and\/or artificial intelligence tools), there are two possible usage scenarios:","dropCap":false,"placeholder":"","direction":"","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<p>As far as <em>synthetic data<\/em> is concerned (i.e., data derived from analytical solutions or generated by numerical simulations and\/or artificial intelligence tools), there are two possible usage scenarios:<\/p>\n","innerContent":["\n<p>As far as <em>synthetic data<\/em> is concerned (i.e., data derived from analytical solutions or generated by numerical simulations and\/or artificial intelligence tools), there are two possible usage scenarios:<\/p>\n"],"rendered":"\n<p>As far as <em>synthetic data<\/em> is concerned (i.e., data derived from analytical solutions or generated by numerical simulations and\/or artificial intelligence tools), there are two possible usage scenarios:<\/p>\n"},{"blockName":"core\/list","attrs":{"ordered":false,"values":"","type":"","start":0,"reversed":false,"placeholder":"","lock":[],"metadata":[],"className":"wp-block-list","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[{"blockName":"core\/list-item","attrs":{"placeholder":"","content":"When existing data is insufficient, artificial data can be generated using information already collected and initial simplified models. For example, as part of a study on monitoring the stability of natural and man-made slopes by tracking surface displacements (i.e., displacements tracked at the ground surface using optical targets), a data generation methodology covering various configurations (geometry and material properties) and various climatic scenarios was developed to generate 2,000 changes in movement speed over a 10-year period.[2]","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<li>When existing data is insufficient, artificial data can be generated using information already collected and initial simplified models. For example, <a href=\"https:\/\/www.presses-des-ponts.fr\/notre-librairie\/406-transitions--les-nouvelles-annales-des-ponts-et-chaussees-n3.html\" target=\"_blank\" rel=\"noreferrer noopener\">as part of a study on monitoring the stability of natural and man-made slopes by tracking surface displacements<\/a> (i.e., displacements tracked at the ground surface using optical targets), a data generation methodology covering various configurations (geometry and material properties) and various climatic scenarios was developed to generate 2,000 changes in movement speed over a 10-year period.[2]<\/li>\n","innerContent":["\n<li>When existing data is insufficient, artificial data can be generated using information already collected and initial simplified models. For example, <a href=\"https:\/\/www.presses-des-ponts.fr\/notre-librairie\/406-transitions--les-nouvelles-annales-des-ponts-et-chaussees-n3.html\" target=\"_blank\" rel=\"noreferrer noopener\">as part of a study on monitoring the stability of natural and man-made slopes by tracking surface displacements<\/a> (i.e., displacements tracked at the ground surface using optical targets), a data generation methodology covering various configurations (geometry and material properties) and various climatic scenarios was developed to generate 2,000 changes in movement speed over a 10-year period.[2]<\/li>\n"],"rendered":"\n<li>When existing data is insufficient, artificial data can be generated using information already collected and initial simplified models. For example, <a href=\"https:\/\/www.presses-des-ponts.fr\/notre-librairie\/406-transitions--les-nouvelles-annales-des-ponts-et-chaussees-n3.html\" target=\"_blank\" rel=\"noreferrer noopener\">as part of a study on monitoring the stability of natural and man-made slopes by tracking surface displacements<\/a> (i.e., displacements tracked at the ground surface using optical targets), a data generation methodology covering various configurations (geometry and material properties) and various climatic scenarios was developed to generate 2,000 changes in movement speed over a 10-year period.[2]<\/li>\n"}],"innerHTML":"\n<ul class=\"wp-block-list\"><\/ul>\n","innerContent":["\n<ul class=\"wp-block-list\">",null,"<\/ul>\n"],"rendered":"\n<ul class=\"wp-block-list\">\n<li>When existing data is insufficient, artificial data can be generated using information already collected and initial simplified models. For example, <a href=\"https:\/\/www.presses-des-ponts.fr\/notre-librairie\/406-transitions--les-nouvelles-annales-des-ponts-et-chaussees-n3.html\" target=\"_blank\" rel=\"noreferrer noopener\">as part of a study on monitoring the stability of natural and man-made slopes by tracking surface displacements<\/a> (i.e., displacements tracked at the ground surface using optical targets), a data generation methodology covering various configurations (geometry and material properties) and various climatic scenarios was developed to generate 2,000 changes in movement speed over a 10-year period.[2]<\/li>\n<\/ul>\n"},{"blockName":"core\/paragraph","attrs":{"align":"","content":"","dropCap":false,"placeholder":"","direction":"","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<p><\/p>\n","innerContent":["\n<p><\/p>\n"],"rendered":"\n<p><\/p>\n"},{"blockName":"core\/list","attrs":{"ordered":false,"values":"","type":"","start":0,"reversed":false,"placeholder":"","lock":[],"metadata":[],"className":"wp-block-list","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[{"blockName":"core\/list-item","attrs":{"placeholder":"","content":"Where access to field data is not possible, for example during a structure\u2019s design phase, data generation can be used to take account of a wide range of parameters (including geological conditions and structure characteristics) for preliminary design analyses. A more detailed example of the use of this type of data is presented in the following.","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<li>Where access to field data is not possible, for example during a structure\u2019s design phase, data generation can be used to take account of a wide range of parameters (including geological conditions and structure characteristics) for preliminary design analyses. A more detailed example of the use of this type of data is presented in the following.<\/li>\n","innerContent":["\n<li>Where access to field data is not possible, for example during a structure\u2019s design phase, data generation can be used to take account of a wide range of parameters (including geological conditions and structure characteristics) for preliminary design analyses. A more detailed example of the use of this type of data is presented in the following.<\/li>\n"],"rendered":"\n<li>Where access to field data is not possible, for example during a structure\u2019s design phase, data generation can be used to take account of a wide range of parameters (including geological conditions and structure characteristics) for preliminary design analyses. A more detailed example of the use of this type of data is presented in the following.<\/li>\n"}],"innerHTML":"\n<ul class=\"wp-block-list\"><\/ul>\n","innerContent":["\n<ul class=\"wp-block-list\">",null,"<\/ul>\n"],"rendered":"\n<ul class=\"wp-block-list\">\n<li>Where access to field data is not possible, for example during a structure\u2019s design phase, data generation can be used to take account of a wide range of parameters (including geological conditions and structure characteristics) for preliminary design analyses. A more detailed example of the use of this type of data is presented in the following.<\/li>\n<\/ul>\n"},{"blockName":"core\/heading","attrs":{"textColor":"red","textAlign":"","content":"Machine learning in geotechnics: efficient and reliable techniques?","level":2,"levelOptions":[],"placeholder":"","lock":[],"metadata":[],"align":"","className":"wp-block-heading has-red-color has-text-color","style":"","backgroundColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<h2 class=\"wp-block-heading has-red-color has-text-color\">Machine learning in geotechnics: efficient and reliable techniques?<\/h2>\n","innerContent":["\n<h2 class=\"wp-block-heading has-red-color has-text-color\">Machine learning in geotechnics: efficient and reliable techniques?<\/h2>\n"],"rendered":"\n<h2 class=\"wp-block-heading has-red-color has-text-color\">Machine learning in geotechnics: efficient and reliable techniques?<\/h2>\n"},{"blockName":"core\/paragraph","attrs":{"align":"","content":"Once the database is in place, work can begin on applying machine learning algorithms. In the case of small datasets, which are common in geotechnical projects, the learning model overfits the training data, resulting in poor predictions for new data . One practical solution to this problem is to combine several models to make an overall prediction. In most cases, this leads to better results.","dropCap":false,"placeholder":"","direction":"","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<p>Once the database is in place, work can begin on applying machine learning algorithms. In the case of small datasets, which are common in geotechnical projects, the learning model overfits the training data, resulting in poor predictions for new data . One practical solution to this problem is to combine several models to make an overall prediction. In most cases, this leads to better results.<\/p>\n","innerContent":["\n<p>Once the database is in place, work can begin on applying machine learning algorithms. In the case of small datasets, which are common in geotechnical projects, the learning model overfits the training data, resulting in poor predictions for new data . One practical solution to this problem is to combine several models to make an overall prediction. In most cases, this leads to better results.<\/p>\n"],"rendered":"\n<p>Once the database is in place, work can begin on applying machine learning algorithms. In the case of small datasets, which are common in geotechnical projects, the learning model overfits the training data, resulting in poor predictions for new data . One practical solution to this problem is to combine several models to make an overall prediction. In most cases, this leads to better results.<\/p>\n"},{"blockName":"core\/paragraph","attrs":{"align":"","content":"Various studies on tunnel excavation projects, using both synthetic data and data from in-situ monitoring, have concluded that learning techniques based on these what are known as \u201censemble\u201d methods seem to perform best when using small datasets for model training, and deliver the most reliable predictions.","dropCap":false,"placeholder":"","direction":"","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<p>Various studies on tunnel excavation projects, using both<a href=\"https:\/\/eposter.at\/ISRM2023\/data\/PDF\/1163.pdf\" target=\"_blank\" rel=\"noreferrer noopener\"> synthetic dat<\/a>a and data from<a href=\"https:\/\/hal.science\/hal-04376951\"> <em>in-situ<\/em> monitoring<\/a>, have concluded that learning techniques based on these what are known as \u201censemble\u201d methods seem to perform best when using small datasets for model training, and deliver the most reliable predictions.<\/p>\n","innerContent":["\n<p>Various studies on tunnel excavation projects, using both<a href=\"https:\/\/eposter.at\/ISRM2023\/data\/PDF\/1163.pdf\" target=\"_blank\" rel=\"noreferrer noopener\"> synthetic dat<\/a>a and data from<a href=\"https:\/\/hal.science\/hal-04376951\"> <em>in-situ<\/em> monitoring<\/a>, have concluded that learning techniques based on these what are known as \u201censemble\u201d methods seem to perform best when using small datasets for model training, and deliver the most reliable predictions.<\/p>\n"],"rendered":"\n<p>Various studies on tunnel excavation projects, using both<a href=\"https:\/\/eposter.at\/ISRM2023\/data\/PDF\/1163.pdf\" target=\"_blank\" rel=\"noreferrer noopener\"> synthetic dat<\/a>a and data from<a href=\"https:\/\/hal.science\/hal-04376951\"> <em>in-situ<\/em> monitoring<\/a>, have concluded that learning techniques based on these what are known as \u201censemble\u201d methods seem to perform best when using small datasets for model training, and deliver the most reliable predictions.<\/p>\n"},{"blockName":"core\/heading","attrs":{"textColor":"red","textAlign":"","content":"How to optimize the use of machine learning in Geotechnics","level":2,"levelOptions":[],"placeholder":"","lock":[],"metadata":[],"align":"","className":"wp-block-heading has-red-color has-text-color","style":"","backgroundColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<h2 class=\"wp-block-heading has-red-color has-text-color\">How to optimize the use of machine learning in Geotechnics<\/h2>\n","innerContent":["\n<h2 class=\"wp-block-heading has-red-color has-text-color\">How to optimize the use of machine learning in Geotechnics<\/h2>\n"],"rendered":"\n<h2 class=\"wp-block-heading has-red-color has-text-color\">How to optimize the use of machine learning in Geotechnics<\/h2>\n"},{"blockName":"core\/paragraph","attrs":{"align":"","content":"Geotechnical data is highly variable in space and time. To accurately represent the phenomena observed in situ, the objective must be set in advance. For example, to represent the progress of the excavation of an underground structure and its impact on the surrounding area, the engineer aims to estimate what is happening in front of the cutting face 1 They therefore need to collect information as excavation progresses, such as surface settlement measurements, convergence measurements 2 and information linked to excavation machine parameters. In other words, the engineer will try to use the data collected during excavation (everything behind the cutting face) for training, and everything in front of the cutting face (the area not yet excavated) for forecasting. Machine learning has been used, for example, to predict the settlement caused by tunneling on two lines of the Grand Paris Express.","dropCap":false,"placeholder":"","direction":"","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<p>Geotechnical data is highly variable in space and time. To accurately represent the phenomena observed <em>in situ<\/em>, the objective must be set in advance. For example, to represent the progress of the excavation of an underground structure and its impact on the surrounding area, the engineer aims to estimate what is happening in front of the cutting face <sup data-fn=\"14141f49-8f9e-4fed-9ee6-a12183629ef7\" class=\"fn\"><a href=\"#14141f49-8f9e-4fed-9ee6-a12183629ef7\" id=\"14141f49-8f9e-4fed-9ee6-a12183629ef7-link\">1<\/a><\/sup> They therefore need to collect information as excavation progresses, such as surface settlement measurements, convergence measurements <sup data-fn=\"849c2e54-b2ee-409f-b61d-8b23c4e0ab1c\" class=\"fn\"><a href=\"#849c2e54-b2ee-409f-b61d-8b23c4e0ab1c\" id=\"849c2e54-b2ee-409f-b61d-8b23c4e0ab1c-link\">2<\/a><\/sup> and information linked to excavation machine parameters. In other words, the engineer will try to use the data collected during excavation (everything behind the cutting face) for training, and everything in front of the cutting face (the area not yet excavated) for forecasting. <a href=\"https:\/\/ingenius.ecoledesponts.fr\/articles\/intelligence-artificielle-au-service-du-creusement-des-tunnels\/\" target=\"_blank\" rel=\"noreferrer noopener\">Machine learning has been used, for example, to predict the settlement caused by tunneling on two lines of the Grand Paris Express.<\/a><\/p>\n","innerContent":["\n<p>Geotechnical data is highly variable in space and time. To accurately represent the phenomena observed <em>in situ<\/em>, the objective must be set in advance. For example, to represent the progress of the excavation of an underground structure and its impact on the surrounding area, the engineer aims to estimate what is happening in front of the cutting face <sup data-fn=\"14141f49-8f9e-4fed-9ee6-a12183629ef7\" class=\"fn\"><a href=\"#14141f49-8f9e-4fed-9ee6-a12183629ef7\" id=\"14141f49-8f9e-4fed-9ee6-a12183629ef7-link\">1<\/a><\/sup> They therefore need to collect information as excavation progresses, such as surface settlement measurements, convergence measurements <sup data-fn=\"849c2e54-b2ee-409f-b61d-8b23c4e0ab1c\" class=\"fn\"><a href=\"#849c2e54-b2ee-409f-b61d-8b23c4e0ab1c\" id=\"849c2e54-b2ee-409f-b61d-8b23c4e0ab1c-link\">2<\/a><\/sup> and information linked to excavation machine parameters. In other words, the engineer will try to use the data collected during excavation (everything behind the cutting face) for training, and everything in front of the cutting face (the area not yet excavated) for forecasting. <a href=\"https:\/\/ingenius.ecoledesponts.fr\/articles\/intelligence-artificielle-au-service-du-creusement-des-tunnels\/\" target=\"_blank\" rel=\"noreferrer noopener\">Machine learning has been used, for example, to predict the settlement caused by tunneling on two lines of the Grand Paris Express.<\/a><\/p>\n"],"rendered":"\n<p>Geotechnical data is highly variable in space and time. To accurately represent the phenomena observed <em>in situ<\/em>, the objective must be set in advance. For example, to represent the progress of the excavation of an underground structure and its impact on the surrounding area, the engineer aims to estimate what is happening in front of the cutting face <sup data-fn=\"14141f49-8f9e-4fed-9ee6-a12183629ef7\" class=\"fn\"><a href=\"#14141f49-8f9e-4fed-9ee6-a12183629ef7\" id=\"14141f49-8f9e-4fed-9ee6-a12183629ef7-link\">1<\/a><\/sup> They therefore need to collect information as excavation progresses, such as surface settlement measurements, convergence measurements <sup data-fn=\"849c2e54-b2ee-409f-b61d-8b23c4e0ab1c\" class=\"fn\"><a href=\"#849c2e54-b2ee-409f-b61d-8b23c4e0ab1c\" id=\"849c2e54-b2ee-409f-b61d-8b23c4e0ab1c-link\">2<\/a><\/sup> and information linked to excavation machine parameters. In other words, the engineer will try to use the data collected during excavation (everything behind the cutting face) for training, and everything in front of the cutting face (the area not yet excavated) for forecasting. <a href=\"https:\/\/ingenius.ecoledesponts.fr\/articles\/intelligence-artificielle-au-service-du-creusement-des-tunnels\/\" target=\"_blank\" rel=\"noreferrer noopener\">Machine learning has been used, for example, to predict the settlement caused by tunneling on two lines of the Grand Paris Express.<\/a><\/p>\n"},{"blockName":"core\/paragraph","attrs":{"align":"","content":"Another example of estimating how convergence varies over time during deep tunneling is presented in Guayac\u00e1n-Carrillo & Sulem.The methodology proposed involves using the information recorded during excavation for training purposes and finding a simple mathematical expression that captures the changes in convergence over time. This approach takes account of the intrinsic anisotropy of rock formation and the anisotropy of the initial stress state. This type of approach can provide additional and complementary information to support the observational method. It provides a dynamic approach whose predictions can be refined as new in-situ measurements are taken.","dropCap":false,"placeholder":"","direction":"","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<p>Another example of estimating how convergence varies over time during deep tunneling is presented in <a href=\"https:\/\/www.sciencedirect.com\/science\/article\/pii\/S0266352X2400291X\" target=\"_blank\" rel=\"noreferrer noopener\">Guayac\u00e1n-Carrillo &amp; Sulem<\/a>.The methodology proposed involves using the information recorded during excavation for training purposes and finding a simple mathematical expression that captures the changes in convergence over time. This approach takes account of the intrinsic anisotropy of rock formation and the anisotropy of the initial stress state. This type of approach can provide additional and complementary information to support the observational method. It provides a dynamic approach whose predictions can be refined as new <em>in-situ<\/em> measurements are taken.<\/p>\n","innerContent":["\n<p>Another example of estimating how convergence varies over time during deep tunneling is presented in <a href=\"https:\/\/www.sciencedirect.com\/science\/article\/pii\/S0266352X2400291X\" target=\"_blank\" rel=\"noreferrer noopener\">Guayac\u00e1n-Carrillo &amp; Sulem<\/a>.The methodology proposed involves using the information recorded during excavation for training purposes and finding a simple mathematical expression that captures the changes in convergence over time. This approach takes account of the intrinsic anisotropy of rock formation and the anisotropy of the initial stress state. This type of approach can provide additional and complementary information to support the observational method. It provides a dynamic approach whose predictions can be refined as new <em>in-situ<\/em> measurements are taken.<\/p>\n"],"rendered":"\n<p>Another example of estimating how convergence varies over time during deep tunneling is presented in <a href=\"https:\/\/www.sciencedirect.com\/science\/article\/pii\/S0266352X2400291X\" target=\"_blank\" rel=\"noreferrer noopener\">Guayac\u00e1n-Carrillo &amp; Sulem<\/a>.The methodology proposed involves using the information recorded during excavation for training purposes and finding a simple mathematical expression that captures the changes in convergence over time. This approach takes account of the intrinsic anisotropy of rock formation and the anisotropy of the initial stress state. This type of approach can provide additional and complementary information to support the observational method. It provides a dynamic approach whose predictions can be refined as new <em>in-situ<\/em> measurements are taken.<\/p>\n"},{"blockName":"core\/heading","attrs":{"textColor":"red","textAlign":"","content":"Numerical modeling and machine learning: two complementary approaches","level":2,"levelOptions":[],"placeholder":"","lock":[],"metadata":[],"align":"","className":"wp-block-heading has-red-color has-text-color","style":"","backgroundColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<h2 class=\"wp-block-heading has-red-color has-text-color\">Numerical modeling and machine learning: two complementary approaches<\/h2>\n","innerContent":["\n<h2 class=\"wp-block-heading has-red-color has-text-color\">Numerical modeling and machine learning: two complementary approaches<\/h2>\n"],"rendered":"\n<h2 class=\"wp-block-heading has-red-color has-text-color\">Numerical modeling and machine learning: two complementary approaches<\/h2>\n"},{"blockName":"core\/paragraph","attrs":{"align":"","content":"In a nutshell, surrogate models are based on artificial intelligence tools trained on synthetic datasets. In certain cases, using these models may be appropriate (especially when 3D numerical simulations are required). The main idea is to establish a complementary relationship between numerical modeling and machine learning to provide engineers with simple, reliable tools that avoid high computational efforts and costs.","dropCap":false,"placeholder":"","direction":"","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<p>In a nutshell, surrogate models are based on artificial intelligence tools trained on synthetic datasets. In certain cases, using these models may be appropriate (especially when 3D numerical simulations are required). The main idea is to establish a complementary relationship between numerical modeling and machine learning to provide engineers with simple, reliable tools that avoid high computational efforts and costs.<\/p>\n","innerContent":["\n<p>In a nutshell, surrogate models are based on artificial intelligence tools trained on synthetic datasets. In certain cases, using these models may be appropriate (especially when 3D numerical simulations are required). The main idea is to establish a complementary relationship between numerical modeling and machine learning to provide engineers with simple, reliable tools that avoid high computational efforts and costs.<\/p>\n"],"rendered":"\n<p>In a nutshell, surrogate models are based on artificial intelligence tools trained on synthetic datasets. In certain cases, using these models may be appropriate (especially when 3D numerical simulations are required). The main idea is to establish a complementary relationship between numerical modeling and machine learning to provide engineers with simple, reliable tools that avoid high computational efforts and costs.<\/p>\n"},{"blockName":"core\/paragraph","attrs":{"align":"","content":"As part of their work on the sizing of deep tunnels, a team from the Navier laboratory used 3D numerical simulations to understand how rock reacts to tunnel construction. These simulations are especially important when the rock undergoes significant deformation, or when the supporting structures are very rigid and located close to the cutting face. For this study, a synthetic dataset was used to take account of a wide range of rock and support structure characteristics, as well as tunnel sizes.","dropCap":false,"placeholder":"","direction":"","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<p>As part of their work on the sizing of deep tunnels, a team from the Navier laboratory used 3D numerical simulations to understand how rock reacts to tunnel construction. These simulations are especially important when the rock undergoes significant deformation, or when the supporting structures are very rigid and located close to the cutting face. For this study, <a href=\"https:\/\/eposter.at\/ISRM2023\/data\/PDF\/1163.pdf\" target=\"_blank\" rel=\"noreferrer noopener\">a synthetic dataset was used to take account of a wide range of rock and support structure characteristics, as well as tunnel sizes<\/a>.<\/p>\n","innerContent":["\n<p>As part of their work on the sizing of deep tunnels, a team from the Navier laboratory used 3D numerical simulations to understand how rock reacts to tunnel construction. These simulations are especially important when the rock undergoes significant deformation, or when the supporting structures are very rigid and located close to the cutting face. For this study, <a href=\"https:\/\/eposter.at\/ISRM2023\/data\/PDF\/1163.pdf\" target=\"_blank\" rel=\"noreferrer noopener\">a synthetic dataset was used to take account of a wide range of rock and support structure characteristics, as well as tunnel sizes<\/a>.<\/p>\n"],"rendered":"\n<p>As part of their work on the sizing of deep tunnels, a team from the Navier laboratory used 3D numerical simulations to understand how rock reacts to tunnel construction. These simulations are especially important when the rock undergoes significant deformation, or when the supporting structures are very rigid and located close to the cutting face. For this study, <a href=\"https:\/\/eposter.at\/ISRM2023\/data\/PDF\/1163.pdf\" target=\"_blank\" rel=\"noreferrer noopener\">a synthetic dataset was used to take account of a wide range of rock and support structure characteristics, as well as tunnel sizes<\/a>.<\/p>\n"},{"blockName":"core\/paragraph","attrs":{"align":"","content":"These models have three main advantages. They are trained on data managed by specialists, provide engineers with rapid estimates in a matter of seconds, and are suitable for statistical analyses requiring multiple simulations.","dropCap":false,"placeholder":"","direction":"","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<p>These models have three main advantages. They are trained on data managed by specialists, provide engineers with rapid estimates in a matter of seconds, and are suitable for statistical analyses requiring multiple simulations.<\/p>\n","innerContent":["\n<p>These models have three main advantages. They are trained on data managed by specialists, provide engineers with rapid estimates in a matter of seconds, and are suitable for statistical analyses requiring multiple simulations.<\/p>\n"],"rendered":"\n<p>These models have three main advantages. They are trained on data managed by specialists, provide engineers with rapid estimates in a matter of seconds, and are suitable for statistical analyses requiring multiple simulations.<\/p>\n"},{"blockName":"core\/heading","attrs":{"textColor":"red","textAlign":"","content":"What are the future prospects?","level":2,"levelOptions":[],"placeholder":"","lock":[],"metadata":[],"align":"","className":"wp-block-heading has-red-color has-text-color","style":"","backgroundColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<h2 class=\"wp-block-heading has-red-color has-text-color\">What are the future prospects?<\/h2>\n","innerContent":["\n<h2 class=\"wp-block-heading has-red-color has-text-color\">What are the future prospects?<\/h2>\n"],"rendered":"\n<h2 class=\"wp-block-heading has-red-color has-text-color\">What are the future prospects?<\/h2>\n"},{"blockName":"core\/paragraph","attrs":{"align":"","content":"An interdisciplinary field of research is emerging, involving the development of simple behavioral machine learning models that can reproduce the overall behavior of the ground around the structure, both over the short and long term. The aim is to provide engineering with the new tools needed to design high-performance structures.","dropCap":false,"placeholder":"","direction":"","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<p>An interdisciplinary field of research is emerging, involving the development of simple behavioral machine learning models that can reproduce the overall behavior of the ground around the structure, both over the short and long term. The aim is to provide engineering with the new tools needed to design high-performance structures.<\/p>\n","innerContent":["\n<p>An interdisciplinary field of research is emerging, involving the development of simple behavioral machine learning models that can reproduce the overall behavior of the ground around the structure, both over the short and long term. The aim is to provide engineering with the new tools needed to design high-performance structures.<\/p>\n"],"rendered":"\n<p>An interdisciplinary field of research is emerging, involving the development of simple behavioral machine learning models that can reproduce the overall behavior of the ground around the structure, both over the short and long term. The aim is to provide engineering with the new tools needed to design high-performance structures.<\/p>\n"},{"blockName":"core\/paragraph","attrs":{"align":"","content":"While research at the interface between artificial intelligence and geotechnics has been ongoing for some years now, a number of questions remain unanswered. In particular, it is essential to take advantage of the wealth of data by making them a strategic part of the action plan. This involves thinking about how to share this data with the geotechnical community, possibly via Open Data platforms, and leveraging past experience to inform new projects. At the same time, it is crucial to develop frugal artificial intelligence, particularly when it comes to optimizing and harmonizing recovery\/preparation\/operation procedures and cycles. Finally, it is important to think about how to introduce these new machine learning approaches into the training of future engineers and into industry practice, without losing know-how and business expertise.","dropCap":false,"placeholder":"","direction":"","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<p>While research at the interface between artificial intelligence and geotechnics has been ongoing for some years now, a number of questions remain unanswered. In particular, it is essential to take advantage of the wealth of data by making them a strategic part of the action plan. This involves thinking about how to share this data with the geotechnical community, possibly via Open Data platforms, and leveraging past experience to inform new projects. At the same time, it is crucial to develop frugal artificial intelligence, particularly when it comes to optimizing and harmonizing recovery\/preparation\/operation procedures and cycles. Finally, it is important to think about how to introduce these new machine learning approaches into the training of future engineers and into industry practice, without losing know-how and business expertise.<\/p>\n","innerContent":["\n<p>While research at the interface between artificial intelligence and geotechnics has been ongoing for some years now, a number of questions remain unanswered. In particular, it is essential to take advantage of the wealth of data by making them a strategic part of the action plan. This involves thinking about how to share this data with the geotechnical community, possibly via Open Data platforms, and leveraging past experience to inform new projects. At the same time, it is crucial to develop frugal artificial intelligence, particularly when it comes to optimizing and harmonizing recovery\/preparation\/operation procedures and cycles. Finally, it is important to think about how to introduce these new machine learning approaches into the training of future engineers and into industry practice, without losing know-how and business expertise.<\/p>\n"],"rendered":"\n<p>While research at the interface between artificial intelligence and geotechnics has been ongoing for some years now, a number of questions remain unanswered. In particular, it is essential to take advantage of the wealth of data by making them a strategic part of the action plan. This involves thinking about how to share this data with the geotechnical community, possibly via Open Data platforms, and leveraging past experience to inform new projects. At the same time, it is crucial to develop frugal artificial intelligence, particularly when it comes to optimizing and harmonizing recovery\/preparation\/operation procedures and cycles. Finally, it is important to think about how to introduce these new machine learning approaches into the training of future engineers and into industry practice, without losing know-how and business expertise.<\/p>\n"},{"blockName":"core\/paragraph","attrs":{"align":"","content":"","dropCap":false,"placeholder":"","direction":"","lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","gradient":"","fontSize":"","fontFamily":"","borderColor":"","anchor":""},"innerBlocks":[],"innerHTML":"\n<p><\/p>\n","innerContent":["\n<p><\/p>\n"],"rendered":"\n<p><\/p>\n"},{"blockName":"core\/footnotes","attrs":{"lock":[],"metadata":[],"className":"","style":"","backgroundColor":"","textColor":"","fontSize":"","fontFamily":"","borderColor":""},"innerBlocks":[],"innerHTML":"","innerContent":[],"rendered":""}],"seo":{"title":"Machine Learning in Geotechnics"},"media":{"img":"<img width=\"1386\" height=\"398\" src=\"https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2024\/01\/Image1.png\" class=\"attachment-full size-full\" alt=\"\" decoding=\"async\" loading=\"lazy\" srcset=\"https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2024\/01\/Image1.png 1386w, https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2024\/01\/Image1-300x86.png 300w, https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2024\/01\/Image1-1024x294.png 1024w, https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2024\/01\/Image1-768x221.png 768w\" sizes=\"auto, (max-width: 1386px) 100vw, 1386px\" \/>","src":"https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2024\/01\/Image1.png"},"url":"\/en\/articles\/machine-learning-in-geotechnics\/","related":{"post":[],"author":[{"title":"Lina Mar\u00eda Guayacan","url":"\/en\/authors\/lina-maria-guayacan\/","id":"5186","media":"<img width=\"60\" height=\"60\" src=\"https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2024\/01\/Lina-Guayacan-60x60.png\" class=\"attachment-author-thumb size-author-thumb wp-post-image\" alt=\"\" decoding=\"async\" loading=\"lazy\" srcset=\"https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2024\/01\/Lina-Guayacan-60x60.png 60w, https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2024\/01\/Lina-Guayacan-150x150.png 150w\" sizes=\"auto, (max-width: 60px) 100vw, 60px\" \/>","slug":"lina-maria-guayacan"},{"title":"Jean-Michel Pereira","url":"\/en\/authors\/jean-michel-pereira\/","id":"5188","media":"<img width=\"60\" height=\"60\" src=\"https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2023\/01\/Jean-Michel_Pereira-60x60.png\" class=\"attachment-author-thumb size-author-thumb wp-post-image\" alt=\"\" decoding=\"async\" loading=\"lazy\" srcset=\"https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2023\/01\/Jean-Michel_Pereira-60x60.png 60w, https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2023\/01\/Jean-Michel_Pereira-150x150.png 150w\" sizes=\"auto, (max-width: 60px) 100vw, 60px\" \/>","slug":"jean-michel-pereira"},{"title":"Jean Sulem","url":"\/en\/authors\/jean-sulem\/","id":"5187","media":"<img width=\"60\" height=\"60\" src=\"https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2024\/01\/Jean-Sulem-60x60.png\" class=\"attachment-author-thumb size-author-thumb wp-post-image\" alt=\"\" decoding=\"async\" loading=\"lazy\" srcset=\"https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2024\/01\/Jean-Sulem-60x60.png 60w, https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2024\/01\/Jean-Sulem-150x150.png 150w\" sizes=\"auto, (max-width: 60px) 100vw, 60px\" \/>","slug":"jean-sulem"}],"subject":[{"title":"Digital Technology, Modeling &#038; Artificial Intelligence","url":"\/en\/subjects\/digital-technology-modeling-artificial-intelligence\/","id":"690","media":"<img width=\"1920\" height=\"1080\" src=\"https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2022\/11\/Ecole-des-ponts-webmagazine-numerique.jpg\" class=\"attachment- size- wp-post-image\" alt=\"\" decoding=\"async\" loading=\"lazy\" srcset=\"https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2022\/11\/Ecole-des-ponts-webmagazine-numerique.jpg 1920w, https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2022\/11\/Ecole-des-ponts-webmagazine-numerique-300x169.jpg 300w, https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2022\/11\/Ecole-des-ponts-webmagazine-numerique-1024x576.jpg 1024w, https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2022\/11\/Ecole-des-ponts-webmagazine-numerique-768x432.jpg 768w\" sizes=\"auto, (max-width: 1920px) 100vw, 1920px\" \/>","slug":"digital-technology-modeling-artificial-intelligence"},{"title":"Cities, Urban planning &#038; Construction","url":"\/en\/subjects\/cities-urban-planning-construction\/","id":"936","media":"<img width=\"1920\" height=\"1080\" src=\"https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2022\/11\/Ecole-des-ponts-webmagazine-ville.jpg\" class=\"attachment- size- wp-post-image\" alt=\"\" decoding=\"async\" loading=\"lazy\" srcset=\"https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2022\/11\/Ecole-des-ponts-webmagazine-ville.jpg 1920w, https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2022\/11\/Ecole-des-ponts-webmagazine-ville-300x169.jpg 300w, https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2022\/11\/Ecole-des-ponts-webmagazine-ville-1024x576.jpg 1024w, https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2022\/11\/Ecole-des-ponts-webmagazine-ville-768x432.jpg 768w\" sizes=\"auto, (max-width: 1920px) 100vw, 1920px\" \/>","slug":"cities-urban-planning-construction"}],"category":[{"title":"Article collection","url":"\/en\/articles\/category\/dossier\/","id":"1720","media":"","slug":"dossier","_related_post_type":"folder"}],"folder":[{"title":"Artificial intelligence for engineering","url":"\/en\/folders\/artificial-intelligence-for-engineering\/","id":"5213","media":"<img width=\"945\" height=\"392\" src=\"https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2024\/02\/Image2.png\" class=\"attachment- size- wp-post-image\" alt=\"\" decoding=\"async\" loading=\"lazy\" srcset=\"https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2024\/02\/Image2.png 945w, https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2024\/02\/Image2-300x124.png 300w, https:\/\/ingenius.ecoledesponts.fr\/wp-content\/uploads\/2024\/02\/Image2-768x319.png 768w\" sizes=\"auto, (max-width: 945px) 100vw, 945px\" \/>","slug":"artificial-intelligence-for-engineering"}]},"translated":"https:\/\/ingenius.ecoledesponts.fr\/articles\/quand-lapprentissage-automatique-sinvite-en-geotechnique\/","icon":"icon-folder","duration":"5","custom_excerpt":"Whether excavating a tunnel, building an embankment, or constructing a dam, the design of these types of structures requires excellent knowledge of soil properties. It also raises many scientific and technological challenges. Advances in artificial intelligence in recent years are opening up some exciting opportunities to meet these challenges.","duration_type":"","_links":{"self":[{"href":"https:\/\/ingenius.ecoledesponts.fr\/en\/wp-json\/wp\/v2\/posts\/5178","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/ingenius.ecoledesponts.fr\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/ingenius.ecoledesponts.fr\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/ingenius.ecoledesponts.fr\/en\/wp-json\/wp\/v2\/users\/6"}],"replies":[{"embeddable":true,"href":"https:\/\/ingenius.ecoledesponts.fr\/en\/wp-json\/wp\/v2\/comments?post=5178"}],"version-history":[{"count":5,"href":"https:\/\/ingenius.ecoledesponts.fr\/en\/wp-json\/wp\/v2\/posts\/5178\/revisions"}],"predecessor-version":[{"id":7267,"href":"https:\/\/ingenius.ecoledesponts.fr\/en\/wp-json\/wp\/v2\/posts\/5178\/revisions\/7267"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/ingenius.ecoledesponts.fr\/en\/wp-json\/wp\/v2\/media\/5162"}],"wp:attachment":[{"href":"https:\/\/ingenius.ecoledesponts.fr\/en\/wp-json\/wp\/v2\/media?parent=5178"}],"wp:term":[{"taxonomy":"article-types","embeddable":true,"href":"https:\/\/ingenius.ecoledesponts.fr\/en\/wp-json\/wp\/v2\/article-types?post=5178"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}