Abstract

Geotechnical characterization of site materials is of paramount importance in the construction and mining industry. The analysis of large volumes of geotechnical information from multiple sources leads to data-driven decisions that help to minimize uncertainty. For this purpose, a unified digital information platform becomes handy to have a global perspective and improve the analysis of available ground information data. Access to historic ground investigation data from previous projects during the project planning stage might increase efficiency. However, accessing and processing legacy data from companies’ databases is time and resources consuming. In the recent years, software tools that are capable of extracting data in a digital format from images have become popular, but still require human-supervised interpretation. A novel tool combining Optical Character Recognition (OCR), digital data extraction technologies and AI-based data interpretation system is presented herein. The state-of-the-art OCR technology is capable of accurately recognizing and extracting text from various document types, such as scanned documents, images, and PDFs. It utilizes advanced machine learning algorithms to process text, even in challenging conditions, ensuring data is extracted accurately and reliably. Then, a data interpretation system has been trained to identify the type of site characterization data and its structure while retrieving all the content in a digital format. All components work seamlessly together to provide a comprehensive solution for automating the interpretation and extraction of site characterization data, streamlining data management and analysis processes. The capability of gathering data from multiple sources in a unique ground information system provides valuable information for planning and design stages while decreasing costs, time and uncertainties. In addition, all these data are then available within DAARWIN platform to feed the ground model workflow.

Abstract

Soil boundary delineation is an important task in geotechnical site characterization. It can be achieved by either extracting borehole samples, conducting laboratory tests, and classifying them according to a soil classification system such as the Unified Soil Classification System (USCS) or utilizing multiple cone penetration test (CPT) soundings, and identifying soil boundaries at the soundings from the Ic (soil behavior type index) profiles. However, most soil-layer delineation methods can only take a single type of test result as the input. For instance, the well-known Markov random field (MRF) method can only take soil-type data such as sand, silt, or clay at boreholes as the input. Recognizing that soil classifications and soil properties are correlated, this paper proposes a novel coupled MRF-Bayesian framework to infer the spatial variation of USCS classifications (e.g., sand, silt, and clay) as well as soil properties by integrating both CPT and borehole data. This integrated approach leverages both CPT and borehole data to address some main challenges e.g., uncertainties and multivariate soil data input in underground stratification problems by simultaneous sampling of soil properties and soil types. The new unified framework can accommodate multivariate data, hence the new framework is compatible with the geotechnical engineering practice. The uncertainties for the spatial variation of USCS classification at sounding locations are quantified through a “layer-specific” Bayesian updating i.e., updating posterior cross-correlation behaviors for different layers (such as sand, silt, and clay), independently. In this Bayesian updating, soil-type data can provide some information about the soil properties according to the unified soil classification system. Further, the soil boundaries can be identified across the entire domain by the realization of conditional random fields of soil properties once the spatial variation of USCS classification is inferred at sounding locations, followed by a 3-dimensional Markov random field process.

Abstract

In engineering geology and geotechnical engineering, subsurface soils and rocks are natural geomaterials and exhibit inherent variability in stratigraphy due to geological deposition process. Explicit knowledge of subsurface stratigraphy is a critical input for the analysis, design, and construction of geotechnical engineering systems. However, the accurate and reliable modelling of subsurface geological stratigraphy is challenging due to the limited number of available boreholes in practice and the complex nature of soil stratigraphy. This paper presents an innovative machine learning framework built upon the neighborhood aggregation technique for the prediction of digitized subsurface geological stratigraphy. To predict the stratigraphy at a given point of interest, neighborhood aggregation is first performed to intelligently consolidate the stratigraphy information from its neighboring boreholes, resulting in additional features associated with the target location. By combining the extra stratigraphy information with conventional location-specific features, the framework enhances the predictive capabilities of classical machine learning models at a finer scale. The proposed framework is implemented using common machine learning models and is validated using a simulated benchmark 3D example. The results of leave-one-out cross-validation demonstrate that the proposed framework can improve the performance of classical machine learning models, leading to more reasonable stratigraphy transition and associated uncertainty quantification.

Abstract

Soil boundary delineation is an important task in geotechnical site characterization. It can be achieved by either extracting borehole samples, conducting laboratory tests, and classifying them according to a soil classification system such as the Unified Soil Classification System (USCS) or utilizing multiple cone penetration test (CPT) soundings, and identifying soil boundaries at the soundings from the Ic (soil behavior type index) profiles. However, most soil-layer delineation methods can only take a single type of test result as the input. For instance, the well-known Markov random field (MRF) method can only take soil-type data such as sand, silt, or clay at boreholes as the input. Recognizing that soil classifications and soil properties are correlated, this paper proposes a novel coupled MRF-Bayesian framework to infer the spatial variation of USCS classifications (e.g., sand, silt, and clay) as well as soil properties by integrating both CPT and borehole data. This integrated approach leverages both CPT and borehole data to address some main challenges e.g., uncertainties and multivariate soil data input in underground stratification problems by simultaneous sampling of soil properties and soil types. The new unified framework can accommodate multivariate data, hence the new framework is compatible with the geotechnical engineering practice. The uncertainties for the spatial variation of USCS classification at sounding locations are quantified through a “layer-specific” Bayesian updating i.e., updating posterior cross-correlation behaviors for different layers (such as sand, silt, and clay), independently. In this Bayesian updating, soil-type data can provide some information about the soil properties according to the unified soil classification system. Further, the soil boundaries can be identified across the entire domain by the realization of conditional random fields of soil properties once the spatial variation of USCS classification is inferred at sounding locations, followed by a 3-dimensional Markov random field process.

Abstract

In engineering geology and geotechnical engineering, subsurface soils and rocks are natural geomaterials and exhibit inherent variability in stratigraphy due to geological deposition process. Explicit knowledge of subsurface stratigraphy is a critical input for the analysis, design, and construction of geotechnical engineering systems. However, the accurate and reliable modelling of subsurface geological stratigraphy is challenging due to the limited number of available boreholes in practice and the complex nature of soil stratigraphy. This paper presents an innovative machine learning framework built upon the neighborhood aggregation technique for the prediction of digitized subsurface geological stratigraphy. To predict the stratigraphy at a given point of interest, neighborhood aggregation is first performed to intelligently consolidate the stratigraphy information from its neighboring boreholes, resulting in additional features associated with the target location. By combining the extra stratigraphy information with conventional location-specific features, the framework enhances the predictive capabilities of classical machine learning models at a finer scale. The proposed framework is implemented using common machine learning models and is validated using a simulated benchmark 3D example. The results of leave-one-out cross-validation demonstrate that the proposed framework can improve the performance of classical machine learning models, leading to more reasonable stratigraphy transition and associated uncertainty quantification.

Abstract

Quantifying uncertainties in subsurface properties and stratigraphy can lead to better understanding of the ground conditions and enhance the design and assessment of geotechnical structures. Several studies have utilized Cone Penetration Test (CPT) data and employed Bayesian and Machine Learning methods to quantify the geological uncertainty, based on the Robertson’s soil classification charts and the Soil Behaviour Type Index (Ic). The incorporation of borehole data can reduce the stratigraphic uncertainty. Significant challenges can arise, however, mainly due to the intrinsic differences between field and laboratory-based soil classification systems, which can potentially lead to inconsistent soil classification. To this end, this study proposes a multivariate Gaussian Process model that utilizes site-specific data and: i) jointly models multiple categorical (USCS labels) and continuous (Ic) variables, ii) learns a (shared) spatial correlation structure and the betweenoutputs covariance, and iii) produces two types of dependent classification outputs. The results indicate that the integration of geotechnical and geological information into a unified model can provide more reliable predictions of the subsurface stratification, by allowing simultaneous interpretation of USCS and Ic profiles. Importantly, the model demonstrates the potential to integrate multiple variables of different types, aiming to contribute to the development of a methodology for joint modeling of geotechnical, geological and geophysical data.