A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Estimation of Mechanical Properties of the Bakken Shales Through Convolutional Neural Networks
https://orcid.org/0000-0003-2220-9369
Abstract Effective mechanical properties of shale rocks can be determined by knowing the mechanical properties and distribution pattern of each comprising constituent. However, building the relationship between them is complicated and requires mathematical manipulations. In this study, by taking advantage of machine learning (ML) that is capable of delineating hidden patterns with the least sophistication, a new approach to estimate Young’s modulus of shales by integrating deep learning convolutional neural networks (CNNs) into 2D elemental intensity distribution maps is presented. The generated SEM–EDX maps contain spatial distribution and intensity information of nine major elements abundant in a shale, Al, Ca, C, Fe, K, Mg, Na, S, and Si. The ground truth data are Young’s modulus based on laboratory microindentation tests from ten samples. A total amount of 800 images were used for training and testing, and the trained CNNs were then used to predict Young’s modulus of shale samples by feeding the elemental images. The predicted Young’s modulus exhibited an acceptable relative error of 6.5% and in a much faster time and less effort compared to the laboratory tests. Ultimately, we believe that this novel method has great potential for field applications due to simplified requirements for sample preparation and laboratory apparatus.
Highlights · A deep learning Convolutional Neural Networks model was employed on 2D elemental intensity distribution maps to predict the mechanical properties of shales from the Bakken Formation. · Easy-obtained EDS maps were used as input and Young’s modulus obtained from laboratory microindentation tests was used as output to train the CNNs model. · The trained model exhibited good performance on predicting the Young’ s modulus of unseen samples, with an average relative error of 6.5%.
Estimation of Mechanical Properties of the Bakken Shales Through Convolutional Neural Networks
https://orcid.org/0000-0003-2220-9369
Abstract Effective mechanical properties of shale rocks can be determined by knowing the mechanical properties and distribution pattern of each comprising constituent. However, building the relationship between them is complicated and requires mathematical manipulations. In this study, by taking advantage of machine learning (ML) that is capable of delineating hidden patterns with the least sophistication, a new approach to estimate Young’s modulus of shales by integrating deep learning convolutional neural networks (CNNs) into 2D elemental intensity distribution maps is presented. The generated SEM–EDX maps contain spatial distribution and intensity information of nine major elements abundant in a shale, Al, Ca, C, Fe, K, Mg, Na, S, and Si. The ground truth data are Young’s modulus based on laboratory microindentation tests from ten samples. A total amount of 800 images were used for training and testing, and the trained CNNs were then used to predict Young’s modulus of shale samples by feeding the elemental images. The predicted Young’s modulus exhibited an acceptable relative error of 6.5% and in a much faster time and less effort compared to the laboratory tests. Ultimately, we believe that this novel method has great potential for field applications due to simplified requirements for sample preparation and laboratory apparatus.
Highlights · A deep learning Convolutional Neural Networks model was employed on 2D elemental intensity distribution maps to predict the mechanical properties of shales from the Bakken Formation. · Easy-obtained EDS maps were used as input and Young’s modulus obtained from laboratory microindentation tests was used as output to train the CNNs model. · The trained model exhibited good performance on predicting the Young’ s modulus of unseen samples, with an average relative error of 6.5%.
Estimation of Mechanical Properties of the Bakken Shales Through Convolutional Neural Networks
https://orcid.org/0000-0003-2220-9369
Abstract Effective mechanical properties of shale rocks can be determined by knowing the mechanical properties and distribution pattern of each comprising constituent. However, building the relationship between them is complicated and requires mathematical manipulations. In this study, by taking advantage of machine learning (ML) that is capable of delineating hidden patterns with the least sophistication, a new approach to estimate Young’s modulus of shales by integrating deep learning convolutional neural networks (CNNs) into 2D elemental intensity distribution maps is presented. The generated SEM–EDX maps contain spatial distribution and intensity information of nine major elements abundant in a shale, Al, Ca, C, Fe, K, Mg, Na, S, and Si. The ground truth data are Young’s modulus based on laboratory microindentation tests from ten samples. A total amount of 800 images were used for training and testing, and the trained CNNs were then used to predict Young’s modulus of shale samples by feeding the elemental images. The predicted Young’s modulus exhibited an acceptable relative error of 6.5% and in a much faster time and less effort compared to the laboratory tests. Ultimately, we believe that this novel method has great potential for field applications due to simplified requirements for sample preparation and laboratory apparatus.
Highlights · A deep learning Convolutional Neural Networks model was employed on 2D elemental intensity distribution maps to predict the mechanical properties of shales from the Bakken Formation. · Easy-obtained EDS maps were used as input and Young’s modulus obtained from laboratory microindentation tests was used as output to train the CNNs model. · The trained model exhibited good performance on predicting the Young’ s modulus of unseen samples, with an average relative error of 6.5%.
Estimation of Mechanical Properties of the Bakken Shales Through Convolutional Neural Networks
Li, Chunxiao (author) / Wang, Dongmei (author) / Kong, Lingyun (author) / Ostadhassan, Mehdi (author)
2022
Article (Journal)
Electronic Resource
English
BKL:
38.58
Geomechanik
/
56.20
Ingenieurgeologie, Bodenmechanik
/
38.58$jGeomechanik
/
56.20$jIngenieurgeologie$jBodenmechanik
RVK:
ELIB41
Ongeval - Hete broodjes bakken
Online Contents | 2014
Physical and mechanical properties of two shales
| British Library Conference Proceedings | 1999
Mechanical, acoustic, and failure properties of shales
| British Library Conference Proceedings | 2001
Mechanical properties of shales for estimation of damage zone dimensions around waste repositories
| British Library Conference Proceedings | 1997
Probing the Mechanical Properties of Shales by Nanoindentation
| British Library Conference Proceedings | 2017