You can find a searchable list of my publications below. My Google Scholar profile contains an up-to-date overview of my citations. I also have a ResearchGate profile with most of my full-texts.
I am a strong proponent of Open Access, especially after having spent more than four years as a researcher at an institution with a very limited number of journal subscriptions. For each entry below where I am legally allowed to share a full-text, you can find it as the first link of the entry.
2023
Viggen, Erlend Magnus; Singstad, Bjørn-Jostein; Time, Eirik; Mishra, Siddharth; Berg, Eirik
Assisted cement log interpretation using machine learning Journal Article
In: SPE Drilling & Completion, vol. 38, iss. 02, pp. 220–234, 2023.
Abstract | BibTeX | Tags: machine learning, well logging | Links:
@article{viggen_assisted_2022,
title = {Assisted cement log interpretation using machine learning},
author = {Erlend Magnus Viggen and Bjørn-Jostein Singstad and Eirik Time and Siddharth Mishra and Eirik Berg},
url = {https://hdl.handle.net/11250/3062647, Post-print at NTNU Open},
doi = {10.2118/209529-PA},
year = {2023},
date = {2023-06-14},
urldate = {2023-06-14},
journal = {SPE Drilling & Completion},
volume = {38},
issue = {02},
pages = {220–234},
abstract = {The Assisted Cement Log Interpretation Project has used machine learning (ML) to create a tool that interprets cement logs by predicting a predefined set of annular condition codes used in the cement log interpretation process.
The development of a cement log interpretation tool speeds up the log interpretation process and enables expert knowledge to be efficiently shared when training new professionals. By using high-quality and consistent training data sets, the project has trained a model that will support unbiased and consistent interpretations over time.
The tool consists of a training and a prediction tool integrated with cased-hole logging interpretation software. By containerizing the code using an “API First” design principle (API: application programming interface), the applicability of this add-on tool is broad. The ML model is trained using selected and engineered features from cement logs, and the tool predicts an annular condition code according to the cement classification system for each depth segment in the log. The interpreters can easily fetch a complete cement log interpretation prediction for the log and use that as a template for their final interpretation. The ML model can easily be retrained with new data sets to improve accuracy even further.
To improve cement log interpretation consistency in the industry, the code will be made available as open source.},
keywords = {machine learning, well logging},
pubstate = {published},
tppubtype = {article}
}
The Assisted Cement Log Interpretation Project has used machine learning (ML) to create a tool that interprets cement logs by predicting a predefined set of annular condition codes used in the cement log interpretation process.
The development of a cement log interpretation tool speeds up the log interpretation process and enables expert knowledge to be efficiently shared when training new professionals. By using high-quality and consistent training data sets, the project has trained a model that will support unbiased and consistent interpretations over time.
The tool consists of a training and a prediction tool integrated with cased-hole logging interpretation software. By containerizing the code using an “API First” design principle (API: application programming interface), the applicability of this add-on tool is broad. The ML model is trained using selected and engineered features from cement logs, and the tool predicts an annular condition code according to the cement classification system for each depth segment in the log. The interpreters can easily fetch a complete cement log interpretation prediction for the log and use that as a template for their final interpretation. The ML model can easily be retrained with new data sets to improve accuracy even further.
To improve cement log interpretation consistency in the industry, the code will be made available as open source.
The development of a cement log interpretation tool speeds up the log interpretation process and enables expert knowledge to be efficiently shared when training new professionals. By using high-quality and consistent training data sets, the project has trained a model that will support unbiased and consistent interpretations over time.
The tool consists of a training and a prediction tool integrated with cased-hole logging interpretation software. By containerizing the code using an “API First” design principle (API: application programming interface), the applicability of this add-on tool is broad. The ML model is trained using selected and engineered features from cement logs, and the tool predicts an annular condition code according to the cement classification system for each depth segment in the log. The interpreters can easily fetch a complete cement log interpretation prediction for the log and use that as a template for their final interpretation. The ML model can easily be retrained with new data sets to improve accuracy even further.
To improve cement log interpretation consistency in the industry, the code will be made available as open source.
Johansen, Tonni Franke; Buschmann, Philip Erik; Røsberg, Knut Marius; Diez, Anja; Viggen, Erlend Magnus
Ultrasonic well integrity logging using phased array technology Proceedings Article
In: Proceedings of the ASME 2023 42nd International Conference on Ocean, Offshore and Arctic Engineering, pp. 9, American Society of Mechanical Engineers, 2023, ISBN: 978-0-7918-8691-5.
Abstract | BibTeX | Tags: guided waves, well logging | Links:
@inproceedings{johansen_ultrasonic_2023,
title = {Ultrasonic well integrity logging using phased array technology},
author = {Tonni Franke Johansen and Philip Erik Buschmann and Knut Marius Røsberg and Anja Diez and Erlend Magnus Viggen},
url = {https://hdl.handle.net/11250/3102191, Full-text on NTNU Open},
doi = {10.1115/OMAE2023-108101},
isbn = {978-0-7918-8691-5},
year = {2023},
date = {2023-06-11},
urldate = {2023-06-11},
booktitle = {Proceedings of the ASME 2023 42nd International Conference on Ocean, Offshore and Arctic Engineering},
volume = {9},
pages = {9},
publisher = {American Society of Mechanical Engineers},
abstract = {Ultrasonic well integrity logging is an important and common procedure for well completion and plug-and-abandonment operations. Typical logging tools employ single-element ultrasound transducers. In medical ultrasound imaging, however, more flexible phased arrays are the standard. This paper presents a first set of experimental results obtained with up to two linear 32-element phased arrays that are specifically designed for plug-and-abandonment operations in terms of their centre frequency. The experiments encompass pulse-echo and pitch-catch studies for different incidence angles and aperture sizes on plates and pipes with wall thickness as encountered in the offshore industry. The pulse-echo experiment is backed up by simulations, and shows that effect of the incidence angle on the pipe resonance’s frequency and strength is weak and more strong, respectively, and that the effect depends on the frequency response and directivity of the transducer. The pitch-catch experiment demonstrates the importance of carefully choosing the right angle of incidence to excite the intended wave modes.},
keywords = {guided waves, well logging},
pubstate = {published},
tppubtype = {inproceedings}
}
Ultrasonic well integrity logging is an important and common procedure for well completion and plug-and-abandonment operations. Typical logging tools employ single-element ultrasound transducers. In medical ultrasound imaging, however, more flexible phased arrays are the standard. This paper presents a first set of experimental results obtained with up to two linear 32-element phased arrays that are specifically designed for plug-and-abandonment operations in terms of their centre frequency. The experiments encompass pulse-echo and pitch-catch studies for different incidence angles and aperture sizes on plates and pipes with wall thickness as encountered in the offshore industry. The pulse-echo experiment is backed up by simulations, and shows that effect of the incidence angle on the pipe resonance’s frequency and strength is weak and more strong, respectively, and that the effect depends on the frequency response and directivity of the transducer. The pitch-catch experiment demonstrates the importance of carefully choosing the right angle of incidence to excite the intended wave modes.
Diez, Anja; Johansen, Tonni Franke; Viggen, Erlend Magnus
From 3D to 1D: Effective numerical modelling of pulse-echo measurements in pipes Proceedings Article
In: Proceedings of the 46th Scandinavian Symposium on Physical Acoustics, pp. 23, Norwegian Physical Society, 2023, ISBN: 978-82-8123-023-1.
Abstract | BibTeX | Tags: acoustics, well logging | Links:
@inproceedings{diez_from_2023,
title = {From 3D to 1D: Effective numerical modelling of pulse-echo measurements in pipes},
author = {Anja Diez and Tonni Franke Johansen and Erlend Magnus Viggen},
url = {https://www.researchgate.net/publication/369926875_From_3D_to_1D_Effective_numerical_modelling_of_pulse-echo_measurements_in_pipes, Full-text on ResearchGate},
isbn = {978-82-8123-023-1},
year = {2023},
date = {2023-04-06},
urldate = {2023-04-06},
booktitle = {Proceedings of the 46th Scandinavian Symposium on Physical Acoustics},
pages = {23},
publisher = {Norwegian Physical Society},
abstract = {In the oil and gas industry, pulse echo measurements have for decades been used in cased holes to estimate the properties of the materials on the outside of the pipe. To investigate the methods used to analyse pulse echo measurements from pipes, we use numerical modelling in COMSOL Multiphysics. While 3D models correctly capture the real geometry, they are computationally heavy and, therefore, not appropriate for simulating a large range of geometric and material variations. Analytic 1D plane wave models are fast to calculate, but we observe large deviations between the 3D and 1D results, showing that corrections to 1D model results are necessary. Therefore, we investigate using 2D and axisymmetric 2.5D models instead and find good agreement between the 2.5D and 3D model results and larger deviations between the 2D and 3D model results. Further, we find that the time explicit formulation works reliably, with an effective absorbing layer, while using the time domain formulation requires more care and a larger domain due to the poorer performance of its perfectly matched layer. Nevertheless, the time domain formulation is preferable when introducing thin domains and additional domain boundaries to keep the computational time at an acceptable level.},
keywords = {acoustics, well logging},
pubstate = {published},
tppubtype = {inproceedings}
}
In the oil and gas industry, pulse echo measurements have for decades been used in cased holes to estimate the properties of the materials on the outside of the pipe. To investigate the methods used to analyse pulse echo measurements from pipes, we use numerical modelling in COMSOL Multiphysics. While 3D models correctly capture the real geometry, they are computationally heavy and, therefore, not appropriate for simulating a large range of geometric and material variations. Analytic 1D plane wave models are fast to calculate, but we observe large deviations between the 3D and 1D results, showing that corrections to 1D model results are necessary. Therefore, we investigate using 2D and axisymmetric 2.5D models instead and find good agreement between the 2.5D and 3D model results and larger deviations between the 2D and 3D model results. Further, we find that the time explicit formulation works reliably, with an effective absorbing layer, while using the time domain formulation requires more care and a larger domain due to the poorer performance of its perfectly matched layer. Nevertheless, the time domain formulation is preferable when introducing thin domains and additional domain boundaries to keep the computational time at an acceptable level.
2022
Arnestad, Håvard Kjellmo; Viggen, Erlend Magnus
A fast simulation method for Lamb wave propagation in coupled non-parallel plates Presentation
Poster presented at the IEEE International Ultrasonics Symposium 2022, 12.10.2022.
Abstract | BibTeX | Tags: acoustics, guided waves, well logging | Links:
@misc{nokey,
title = {A fast simulation method for Lamb wave propagation in coupled non-parallel plates},
author = {Håvard Kjellmo Arnestad and Erlend Magnus Viggen},
url = {https://www.researchgate.net/publication/364359505_A_fast_simulation_method_for_Lamb_wave_propagation_in_coupled_non-parallel_plates, Poster on ResearchGate},
year = {2022},
date = {2022-10-12},
urldate = {2022-10-12},
abstract = {Some systems in ultrasonic testing can be approximated as two non-parallel plates coupled by a fluid, where leaky Lamb waves propagate in each plate. This work develops a fast and accurate simulation method for such systems by extending methods based on the theory of layered media to non-parallel surfaces. The aim is to determine the presence of cement through two steel plates via inversion. The method runs roughly 10 000 times faster than equivalent simulations in COMSOL. Three-dimensional propagation is also shown, and a mechanism based on Lamb mode conversion between tilted plates is explained.},
howpublished = {Poster presented at the IEEE International Ultrasonics Symposium 2022},
keywords = {acoustics, guided waves, well logging},
pubstate = {published},
tppubtype = {presentation}
}
Some systems in ultrasonic testing can be approximated as two non-parallel plates coupled by a fluid, where leaky Lamb waves propagate in each plate. This work develops a fast and accurate simulation method for such systems by extending methods based on the theory of layered media to non-parallel surfaces. The aim is to determine the presence of cement through two steel plates via inversion. The method runs roughly 10 000 times faster than equivalent simulations in COMSOL. Three-dimensional propagation is also shown, and a mechanism based on Lamb mode conversion between tilted plates is explained.
Time, Eirik; Viggen, Erlend Magnus; Mishra, Siddharth; Berg, Eirik
Assisted cement log interpretation Proceedings Article
In: SPE Norway Subsurface Conference 2022, pp. 15, Society of Petroleum Engineers, 2022.
Abstract | BibTeX | Tags: machine learning, well logging | Links:
@inproceedings{time_assisted_2022,
title = {Assisted cement log interpretation},
author = {Eirik Time and Erlend Magnus Viggen and Siddharth Mishra and Eirik Berg},
doi = {10.2118/209529-MS},
year = {2022},
date = {2022-04-27},
urldate = {2022-04-27},
booktitle = {SPE Norway Subsurface Conference 2022},
pages = {15},
publisher = {Society of Petroleum Engineers},
abstract = {The Assisted Cement Log Interpretation project has used machine learning (ML) to create a tool that interprets cement logs by predicting a predefined set of annular condition codes used in the cement log interpretation process.
The development of a cement log interpretation tool speeds up the log interpretation process and enables expert knowledge to be efficiently shared when training new professionals in the cased hole logging unit. By using high quality and consistent training data sets, the project has trained a model that will support unbiased and consistent interpretations over time.
The tool consists of a training and a prediction tool integrated with the cased hole logging interpretation software. By containerizing the code using an "API First" design principle, the applicability of this add- on tool is broad. The ML model is trained using selected and engineered features from cement logs, and the tool predicts an annular condition code according to the cement classification system for each depth segment in the log. The interpreters can easily fetch a complete cement log interpretation prediction for the log and use that as a template for their final interpretation. The ML model can easily be retrained with new data sets to improve accuracy even further.
To improve cement log interpretation consistency in the industry, the results are made available as open source.},
keywords = {machine learning, well logging},
pubstate = {published},
tppubtype = {inproceedings}
}
The Assisted Cement Log Interpretation project has used machine learning (ML) to create a tool that interprets cement logs by predicting a predefined set of annular condition codes used in the cement log interpretation process.
The development of a cement log interpretation tool speeds up the log interpretation process and enables expert knowledge to be efficiently shared when training new professionals in the cased hole logging unit. By using high quality and consistent training data sets, the project has trained a model that will support unbiased and consistent interpretations over time.
The tool consists of a training and a prediction tool integrated with the cased hole logging interpretation software. By containerizing the code using an "API First" design principle, the applicability of this add- on tool is broad. The ML model is trained using selected and engineered features from cement logs, and the tool predicts an annular condition code according to the cement classification system for each depth segment in the log. The interpreters can easily fetch a complete cement log interpretation prediction for the log and use that as a template for their final interpretation. The ML model can easily be retrained with new data sets to improve accuracy even further.
To improve cement log interpretation consistency in the industry, the results are made available as open source.
The development of a cement log interpretation tool speeds up the log interpretation process and enables expert knowledge to be efficiently shared when training new professionals in the cased hole logging unit. By using high quality and consistent training data sets, the project has trained a model that will support unbiased and consistent interpretations over time.
The tool consists of a training and a prediction tool integrated with the cased hole logging interpretation software. By containerizing the code using an "API First" design principle, the applicability of this add- on tool is broad. The ML model is trained using selected and engineered features from cement logs, and the tool predicts an annular condition code according to the cement classification system for each depth segment in the log. The interpreters can easily fetch a complete cement log interpretation prediction for the log and use that as a template for their final interpretation. The ML model can easily be retrained with new data sets to improve accuracy even further.
To improve cement log interpretation consistency in the industry, the results are made available as open source.
2021
Viggen, Erlend Magnus; Løvstakken, Lasse; Måsøy, Svein-Erik; Merciu, Ioan Alexandru
Better automatic interpretation of cement evaluation logs through feature engineering Journal Article
In: SPE Journal, vol. 26, no. 05, pp. 2894–2913, 2021, ISSN: 1930-0220.
Abstract | BibTeX | Tags: machine learning, well logging | Links:
@article{viggen_better_2021b,
title = {Better automatic interpretation of cement evaluation logs through feature engineering},
author = {Erlend Magnus Viggen and Lasse Løvstakken and Svein-Erik Måsøy and Ioan Alexandru Merciu},
doi = {10.2118/204057-PA},
issn = {1930-0220},
year = {2021},
date = {2021-10-13},
urldate = {2021-10-13},
journal = {SPE Journal},
volume = {26},
number = {05},
pages = {2894–2913},
abstract = {We investigate systems to automatically interpret cement evaluation logs using supervised machine learning (ML). Such systems can provide instant rough interpretations that may then be used as a basis for human interpretation. Here, we compare the performance of two approaches, one previously published and one new. The previous approach is based on deep convolutional neural networks (CNNs) that autonomously learn to extract features from well log data, whereas the new approach uses feature engineering, in which we use our own domain knowledge to extract features.
We base this work on a data set of approximately 60 km of well log data. Specialist interpreters have classified these logs according to the bond quality (BQ; six ordinal classes) and hydraulic isolation (HI; two classes) of solids outside the casing. We train the ML systems to reproduce these reference interpretations in segments of 1 m in length. The CNNs directly receive log data as a collection of 2D images and 1D curves. In the feature-engineering approach, we combine the extracted features with various classifiers.
For BQ, the CNNs' interpretation exactly matches the reference 51.6% of the time. It does not miss by more than one class 88.5% of the time. For HI, the CNNs match the reference 86.7% of the time. The best-performingfeature-based classifier, which is an ensemble of individual classifiers, provides better results of 57.4, 89.5, and 88.9%, respectively.
Our results indicate two main reasons why feature-based classifiers may perform particularly well on this task. First, there is some subjectivity inherent in the well log interpretations that are used to train and test ML systems. Second, well logs comprise many different and complex pieces of data. For these reasons, this data set may be particularly liable to overfitting. This may favor approaches based on feature engineering, where we apply our domain knowledge to extract a few pieces of essential information from the data instead of leaving the job of understanding the data to an ML system that may misinterpret spurious patterns as generalizable. It may also favor simpler classifiers with less overfitting capacity.
This paper shows how petroleum researchers and engineers can implement automatic interpretation systems for cement evaluation logs using ML methods that are easier to apply and deploy while also performing better than an approach based on autonomous feature extraction. This approach could also be adapted for automatic interpretation of other types of well log data.},
keywords = {machine learning, well logging},
pubstate = {published},
tppubtype = {article}
}
We investigate systems to automatically interpret cement evaluation logs using supervised machine learning (ML). Such systems can provide instant rough interpretations that may then be used as a basis for human interpretation. Here, we compare the performance of two approaches, one previously published and one new. The previous approach is based on deep convolutional neural networks (CNNs) that autonomously learn to extract features from well log data, whereas the new approach uses feature engineering, in which we use our own domain knowledge to extract features.
We base this work on a data set of approximately 60 km of well log data. Specialist interpreters have classified these logs according to the bond quality (BQ; six ordinal classes) and hydraulic isolation (HI; two classes) of solids outside the casing. We train the ML systems to reproduce these reference interpretations in segments of 1 m in length. The CNNs directly receive log data as a collection of 2D images and 1D curves. In the feature-engineering approach, we combine the extracted features with various classifiers.
For BQ, the CNNs' interpretation exactly matches the reference 51.6% of the time. It does not miss by more than one class 88.5% of the time. For HI, the CNNs match the reference 86.7% of the time. The best-performingfeature-based classifier, which is an ensemble of individual classifiers, provides better results of 57.4, 89.5, and 88.9%, respectively.
Our results indicate two main reasons why feature-based classifiers may perform particularly well on this task. First, there is some subjectivity inherent in the well log interpretations that are used to train and test ML systems. Second, well logs comprise many different and complex pieces of data. For these reasons, this data set may be particularly liable to overfitting. This may favor approaches based on feature engineering, where we apply our domain knowledge to extract a few pieces of essential information from the data instead of leaving the job of understanding the data to an ML system that may misinterpret spurious patterns as generalizable. It may also favor simpler classifiers with less overfitting capacity.
This paper shows how petroleum researchers and engineers can implement automatic interpretation systems for cement evaluation logs using ML methods that are easier to apply and deploy while also performing better than an approach based on autonomous feature extraction. This approach could also be adapted for automatic interpretation of other types of well log data.
We base this work on a data set of approximately 60 km of well log data. Specialist interpreters have classified these logs according to the bond quality (BQ; six ordinal classes) and hydraulic isolation (HI; two classes) of solids outside the casing. We train the ML systems to reproduce these reference interpretations in segments of 1 m in length. The CNNs directly receive log data as a collection of 2D images and 1D curves. In the feature-engineering approach, we combine the extracted features with various classifiers.
For BQ, the CNNs' interpretation exactly matches the reference 51.6% of the time. It does not miss by more than one class 88.5% of the time. For HI, the CNNs match the reference 86.7% of the time. The best-performingfeature-based classifier, which is an ensemble of individual classifiers, provides better results of 57.4, 89.5, and 88.9%, respectively.
Our results indicate two main reasons why feature-based classifiers may perform particularly well on this task. First, there is some subjectivity inherent in the well log interpretations that are used to train and test ML systems. Second, well logs comprise many different and complex pieces of data. For these reasons, this data set may be particularly liable to overfitting. This may favor approaches based on feature engineering, where we apply our domain knowledge to extract a few pieces of essential information from the data instead of leaving the job of understanding the data to an ML system that may misinterpret spurious patterns as generalizable. It may also favor simpler classifiers with less overfitting capacity.
This paper shows how petroleum researchers and engineers can implement automatic interpretation systems for cement evaluation logs using ML methods that are easier to apply and deploy while also performing better than an approach based on autonomous feature extraction. This approach could also be adapted for automatic interpretation of other types of well log data.
Estuariwinarno, Mikael Yuan; Viggen, Erlend Magnus
Determining inner geometry properties from eccentered pulse-echo measurements in a pipe Proceedings Article
In: Proceedings of the 44th Scandinavian Symposium on Physical Acoustics, pp. 23, Norwegian Physical Society, Online, 2021, ISBN: 978-82-8123-021-71.
Abstract | BibTeX | Tags: acoustics, well logging | Links:
@inproceedings{estuariwinarno_determining_2021,
title = {Determining inner geometry properties from eccentered pulse-echo measurements in a pipe},
author = {Mikael Yuan Estuariwinarno and Erlend Magnus Viggen},
url = {https://www.researchgate.net/publication/351064881_Determining_Inner_Geometry_Properties_From_Eccentered_Pulse-Echo_Measurements_in_a_Pipe, Full-text on ResearchGate},
isbn = {978-82-8123-021-71},
year = {2021},
date = {2021-04-01},
urldate = {2021-04-01},
booktitle = {Proceedings of the 44th Scandinavian Symposium on Physical Acoustics},
pages = {23},
publisher = {Norwegian Physical Society},
address = {Online},
abstract = {In the petroleum industry, well integrity evaluation is an essential part of maintaining the safety and sustainability of hydrocarbon production. Ultrasonic pulse-echo cased hole logging is a widely used type of measurement for well integrity evaluation. It gives insight on casing condition and cement quality through the use of an ultrasonic transducer that ideally rotates around the center of the casing. One of the outputs of this logging is a set of inner geometry properties that describe the position of the tool and the inner radius of the casing. However, inner geometry determination is not straightforward as it has to consider the influence of tool eccentering due to gravity and tool movement, which causes the tool to rotate around another axis than the casing center. Despite its importance and wide implementation, detailed information on inner geometry determination from eccentered measurements has not been published in the scientific literature. In this study, an inner geometry determination algorithm was developed and tested on ultrasonic well log data from from the Norwegian North Sea. This algorithm estimates the inner geometry properties, i.e. the tool eccentering properties and the casing inner radius. The results show that the algorithm produces results that give a good match with the results of a reference algorithm from a service company. Our algorithm is also able to handle poor travel time measurements in a more reliable way than the reference algorithm. Hence, this article attempts to enhance and spread the knowledge of ultrasonic cased hole logging, specifically in terms of the determination of casing inner geometry.},
keywords = {acoustics, well logging},
pubstate = {published},
tppubtype = {inproceedings}
}
In the petroleum industry, well integrity evaluation is an essential part of maintaining the safety and sustainability of hydrocarbon production. Ultrasonic pulse-echo cased hole logging is a widely used type of measurement for well integrity evaluation. It gives insight on casing condition and cement quality through the use of an ultrasonic transducer that ideally rotates around the center of the casing. One of the outputs of this logging is a set of inner geometry properties that describe the position of the tool and the inner radius of the casing. However, inner geometry determination is not straightforward as it has to consider the influence of tool eccentering due to gravity and tool movement, which causes the tool to rotate around another axis than the casing center. Despite its importance and wide implementation, detailed information on inner geometry determination from eccentered measurements has not been published in the scientific literature. In this study, an inner geometry determination algorithm was developed and tested on ultrasonic well log data from from the Norwegian North Sea. This algorithm estimates the inner geometry properties, i.e. the tool eccentering properties and the casing inner radius. The results show that the algorithm produces results that give a good match with the results of a reference algorithm from a service company. Our algorithm is also able to handle poor travel time measurements in a more reliable way than the reference algorithm. Hence, this article attempts to enhance and spread the knowledge of ultrasonic cased hole logging, specifically in terms of the determination of casing inner geometry.
Viggen, Erlend Magnus; Løvstakken, Lasse; Merciu, Ioan Alexandru; Måsøy, Svein-Erik
Better automatic interpretation of cement evaluation logs through feature engineering Proceedings Article
In: SPE/IADC International Drilling Conference and Exhibition, pp. 28, 2021.
Abstract | BibTeX | Tags: machine learning, well logging | Links:
@inproceedings{viggen_better_2021a,
title = {Better automatic interpretation of cement evaluation logs through feature engineering},
author = {Erlend Magnus Viggen and Lasse Løvstakken and Ioan Alexandru Merciu and Svein-Erik Måsøy},
doi = {10.2118/204057-MS},
year = {2021},
date = {2021-03-09},
urldate = {2021-03-09},
booktitle = {SPE/IADC International Drilling Conference and Exhibition},
pages = {28},
abstract = {We build systems to automatically interpret cement evaluation logs using supervised machine learning (ML). Such systems can provide instant rough interpretations that may then be used as a basis for human interpretation. Here, we compare the performance of two approaches: A previously published approach based on deep convolutional neural networks (CNNs) that autonomously learn to extract features from well log data, and a feature-engineering approach where we use our own domain knowledge to extract features.
We base this work on a dataset of around 60 km of well log data. Specialist interpreters have classified these logs according to the bond quality (6 ordinal classes) and hydraulic isolation (2 classes) of solids outside the casing. We train the ML systems to reproduce these reference interpretations in segments of 1 m length. The CNNs directly receive log data as a collection of 2D images and 1D curves. In the feature-engineering approach, we combine the extracted features with various classifiers.
For bond quality, the CNNs’ interpretation exactly matches the reference 51.6% of the time. 88.5% of the time, it does not miss by more than one class. For hydraulic isolation, the CNNs match the reference 86.7% of the time. The best-performing feature-based classifier, which is an ensemble of individual classifiers, provides better results of 57.4%, 89.5%, and 88.9%, respectively.
Our results indicate two main reasons why feature-based classifiers may perform particularly well on this task. First, there is some subjectivity inherent in the well log interpretations that are used to train and test ML systems. Second, well logs comprise many different and complex pieces of data. For these reasons, this dataset may be particularly liable to overfitting. This may favour approaches based on feature engineering, where we apply our domain knowledge to extract a few pieces of essential information from the data instead of leaving the job of understanding the data to an ML system that may misinterpret spurious patterns as generalisable. It may also favour simpler classifiers with less overfitting capacity.
This article shows how petroleum researchers and engineers can implement automatic interpretation systems for cement evaluation logs using ML methods that are relatively easy to apply and deploy, with better results than an approach based on autonomous feature extraction. This approach could also be adapted for automatic interpretation of other types of well log data.},
keywords = {machine learning, well logging},
pubstate = {published},
tppubtype = {inproceedings}
}
We build systems to automatically interpret cement evaluation logs using supervised machine learning (ML). Such systems can provide instant rough interpretations that may then be used as a basis for human interpretation. Here, we compare the performance of two approaches: A previously published approach based on deep convolutional neural networks (CNNs) that autonomously learn to extract features from well log data, and a feature-engineering approach where we use our own domain knowledge to extract features.
We base this work on a dataset of around 60 km of well log data. Specialist interpreters have classified these logs according to the bond quality (6 ordinal classes) and hydraulic isolation (2 classes) of solids outside the casing. We train the ML systems to reproduce these reference interpretations in segments of 1 m length. The CNNs directly receive log data as a collection of 2D images and 1D curves. In the feature-engineering approach, we combine the extracted features with various classifiers.
For bond quality, the CNNs’ interpretation exactly matches the reference 51.6% of the time. 88.5% of the time, it does not miss by more than one class. For hydraulic isolation, the CNNs match the reference 86.7% of the time. The best-performing feature-based classifier, which is an ensemble of individual classifiers, provides better results of 57.4%, 89.5%, and 88.9%, respectively.
Our results indicate two main reasons why feature-based classifiers may perform particularly well on this task. First, there is some subjectivity inherent in the well log interpretations that are used to train and test ML systems. Second, well logs comprise many different and complex pieces of data. For these reasons, this dataset may be particularly liable to overfitting. This may favour approaches based on feature engineering, where we apply our domain knowledge to extract a few pieces of essential information from the data instead of leaving the job of understanding the data to an ML system that may misinterpret spurious patterns as generalisable. It may also favour simpler classifiers with less overfitting capacity.
This article shows how petroleum researchers and engineers can implement automatic interpretation systems for cement evaluation logs using ML methods that are relatively easy to apply and deploy, with better results than an approach based on autonomous feature extraction. This approach could also be adapted for automatic interpretation of other types of well log data.
We base this work on a dataset of around 60 km of well log data. Specialist interpreters have classified these logs according to the bond quality (6 ordinal classes) and hydraulic isolation (2 classes) of solids outside the casing. We train the ML systems to reproduce these reference interpretations in segments of 1 m length. The CNNs directly receive log data as a collection of 2D images and 1D curves. In the feature-engineering approach, we combine the extracted features with various classifiers.
For bond quality, the CNNs’ interpretation exactly matches the reference 51.6% of the time. 88.5% of the time, it does not miss by more than one class. For hydraulic isolation, the CNNs match the reference 86.7% of the time. The best-performing feature-based classifier, which is an ensemble of individual classifiers, provides better results of 57.4%, 89.5%, and 88.9%, respectively.
Our results indicate two main reasons why feature-based classifiers may perform particularly well on this task. First, there is some subjectivity inherent in the well log interpretations that are used to train and test ML systems. Second, well logs comprise many different and complex pieces of data. For these reasons, this dataset may be particularly liable to overfitting. This may favour approaches based on feature engineering, where we apply our domain knowledge to extract a few pieces of essential information from the data instead of leaving the job of understanding the data to an ML system that may misinterpret spurious patterns as generalisable. It may also favour simpler classifiers with less overfitting capacity.
This article shows how petroleum researchers and engineers can implement automatic interpretation systems for cement evaluation logs using ML methods that are relatively easy to apply and deploy, with better results than an approach based on autonomous feature extraction. This approach could also be adapted for automatic interpretation of other types of well log data.
2020
Viggen, Erlend Magnus; Merciu, Ioan Alexandru; Løvstakken, Lasse; Måsøy, Svein-Erik
Automatic interpretation of cement evaluation logs from cased boreholes using supervised deep neural networks Journal Article
In: Journal of Petroleum Science and Engineering, vol. 195, pp. 17, 2020, ISSN: 0920-4105.
Abstract | BibTeX | Tags: machine learning, well logging | Links:
@article{viggen_automatic_2020,
title = {Automatic interpretation of cement evaluation logs from cased boreholes using supervised deep neural networks},
author = {Erlend Magnus Viggen and Ioan Alexandru Merciu and Lasse Løvstakken and Svein-Erik Måsøy},
url = {https://www.sciencedirect.com/science/article/pii/S0920410520306100},
doi = {10.1016/j.petrol.2020.107539},
issn = {0920-4105},
year = {2020},
date = {2020-12-01},
urldate = {2020-12-01},
journal = {Journal of Petroleum Science and Engineering},
volume = {195},
pages = {17},
abstract = {The integrity of cement in cased boreholes is typically evaluated using well logging. However, well logging results are complex and can be ambiguous, and decisions associated with significant risks may be taken based on their interpretation. Cement evaluation logs must therefore be interpreted by trained professionals. To aid these interpreters, we propose a system for automatically interpreting cement evaluation logs, which they can use as a basis for their own interpretation. This system is based on deep convolutional neural networks, which we train in a supervised manner using a dataset of around 60 km of interpreted well log data. Thus, the networks learn the connections between data and interpretations during training. More specifically, the task of the networks is to classify the bond quality (among 6 ordinal classes) and the hydraulic isolation (2 classes) in each 1 m depth segment of each well based on the surrounding 13 m of well log data. We quantify the networks' performance by comparing over all segments how well the networks' interpretations of unseen data match the reference interpretations. For bond quality, the networks’ interpretation exactly matches the reference 51.6% of the time and is off by no more than one class 88.5% of the time. For hydraulic isolation, the interpretations match the reference 86.7% of the time. For comparison, a random-guess baseline gives matches of 16.7%, 44.4%, and 50%, respectively. We also compare with how well human reinterpretations of the log data match the reference interpretations, finding that the networks match the reference somewhat better. This may be linked to the networks learning and sharing the biases of the team behind the reference interpretations. An analysis of the results indicates that the subjectivity inherent in the interpretation process (and thereby in the reference interpretations we used for training and testing) is the main reason why we were not able to achieve an even better match between the networks and the reference.},
keywords = {machine learning, well logging},
pubstate = {published},
tppubtype = {article}
}
The integrity of cement in cased boreholes is typically evaluated using well logging. However, well logging results are complex and can be ambiguous, and decisions associated with significant risks may be taken based on their interpretation. Cement evaluation logs must therefore be interpreted by trained professionals. To aid these interpreters, we propose a system for automatically interpreting cement evaluation logs, which they can use as a basis for their own interpretation. This system is based on deep convolutional neural networks, which we train in a supervised manner using a dataset of around 60 km of interpreted well log data. Thus, the networks learn the connections between data and interpretations during training. More specifically, the task of the networks is to classify the bond quality (among 6 ordinal classes) and the hydraulic isolation (2 classes) in each 1 m depth segment of each well based on the surrounding 13 m of well log data. We quantify the networks' performance by comparing over all segments how well the networks' interpretations of unseen data match the reference interpretations. For bond quality, the networks’ interpretation exactly matches the reference 51.6% of the time and is off by no more than one class 88.5% of the time. For hydraulic isolation, the interpretations match the reference 86.7% of the time. For comparison, a random-guess baseline gives matches of 16.7%, 44.4%, and 50%, respectively. We also compare with how well human reinterpretations of the log data match the reference interpretations, finding that the networks match the reference somewhat better. This may be linked to the networks learning and sharing the biases of the team behind the reference interpretations. An analysis of the results indicates that the subjectivity inherent in the interpretation process (and thereby in the reference interpretations we used for training and testing) is the main reason why we were not able to achieve an even better match between the networks and the reference.
Viggen, Erlend Magnus; Hårstad, Erlend; Kvalsvik, Jørgen
Getting started with acoustic well log data using the dlisio Python library on the Volve Data Village dataset Proceedings Article
In: Viggen, Erlend Magnus; Hoff, Lars (Ed.): Proceedings of the 43rd Scandinavian Symposium on Physical Acoustics, pp. 36, Norwegian Physical Society, Geilo, Norway, 2020, ISBN: 978-82-8123-020-0.
Abstract | BibTeX | Tags: acoustics, well logging | Links:
@inproceedings{viggen_getting_2020,
title = {Getting started with acoustic well log data using the dlisio Python library on the Volve Data Village dataset},
author = {Erlend Magnus Viggen and Erlend Hårstad and Jørgen Kvalsvik},
editor = {Erlend Magnus Viggen and Lars Hoff},
url = {https://www.researchgate.net/publication/340645995_Getting_started_with_acoustic_well_log_data_using_the_dlisio_Python_library_on_the_Volve_Data_Village_dataset, Full-text on ResearchGate
https://github.com/equinor/dlisio-notebooks/blob/master/acoustic.ipynb, Companion Jupyter Notebook},
isbn = {978-82-8123-020-0},
year = {2020},
date = {2020-04-15},
urldate = {2020-04-15},
booktitle = {Proceedings of the 43rd Scandinavian Symposium on Physical Acoustics},
pages = {36},
publisher = {Norwegian Physical Society},
address = {Geilo, Norway},
abstract = {Three issues have long impeded academic research and teaching on well logging. First, real measured data has been hard to come by. This has now been alleviated by Equinor's 2018 release of the Volve Data Village dataset. Among its 5 TB of data, it contains 16.3 GB of various well log data, plots, and analyses. Second, no free and effective software tools to programmatically read DLIS files, one of the most common file formats for well log data today and by far the most common format in the Volve Data Village, have been available. This has now been remedied by the free and open-source Python library dlisio, first released by Equinor in 2018 and still under heavy development. Third, the data is often difficult to understand, as sufficient documentation is often not publicly available. As different tools measure, process, and store their data differently, different tools must be understood individually. This article aims to stimulate research into well logging, by showing how to use dlisio to investigate well log data from the Volve Data Village dataset. While the investigative methods used here can be adapted to other kinds of data, this article focuses on acoustic integrity logs. Specifically, we investigate data from a sonic tool (DSLT) and an ultrasonic tool (USIT), both extensively used in the dataset. In addition to identifying what the most fundamental pieces of data represent, we also show some simple examples of how this data can be reprocessed to find new results not provided in the well log file. We provide the code underlying this article in an accompanying Jupyter Notebook.},
keywords = {acoustics, well logging},
pubstate = {published},
tppubtype = {inproceedings}
}
Three issues have long impeded academic research and teaching on well logging. First, real measured data has been hard to come by. This has now been alleviated by Equinor's 2018 release of the Volve Data Village dataset. Among its 5 TB of data, it contains 16.3 GB of various well log data, plots, and analyses. Second, no free and effective software tools to programmatically read DLIS files, one of the most common file formats for well log data today and by far the most common format in the Volve Data Village, have been available. This has now been remedied by the free and open-source Python library dlisio, first released by Equinor in 2018 and still under heavy development. Third, the data is often difficult to understand, as sufficient documentation is often not publicly available. As different tools measure, process, and store their data differently, different tools must be understood individually. This article aims to stimulate research into well logging, by showing how to use dlisio to investigate well log data from the Volve Data Village dataset. While the investigative methods used here can be adapted to other kinds of data, this article focuses on acoustic integrity logs. Specifically, we investigate data from a sonic tool (DSLT) and an ultrasonic tool (USIT), both extensively used in the dataset. In addition to identifying what the most fundamental pieces of data represent, we also show some simple examples of how this data can be reprocessed to find new results not provided in the well log file. We provide the code underlying this article in an accompanying Jupyter Notebook.
2017
Viggen, Erlend Magnus; Johansen, Tonni Franke; Merciu, Ioan-Alexandru
Simulation and inversion of ultrasonic pitch-catch through-tubing well logging with an array of receivers Journal Article
In: NDT & E International, vol. 85, pp. 72–75, 2017, ISSN: 09638695.
Abstract | BibTeX | Tags: acoustics, guided waves, well logging | Links:
@article{viggen_simulation_2017,
title = {Simulation and inversion of ultrasonic pitch-catch through-tubing well logging with an array of receivers},
author = {Erlend Magnus Viggen and Tonni Franke Johansen and Ioan-Alexandru Merciu},
url = {https://erlend-viggen.no/wp-content/uploads/2018/04/viggen_simulation_2017_post-print.pdf, Full-text},
doi = {10.1016/j.ndteint.2016.10.008},
issn = {09638695},
year = {2017},
date = {2017-01-01},
urldate = {2017-01-01},
journal = {NDT & E International},
volume = {85},
pages = {72--75},
abstract = {Current methods for ultrasonic pitch-catch well logging use two receivers to log the bonded material outside a single casing. For two casings separated by a fluid, we find by simulation that increasing the number of receivers provides a better picture of the effect of the bonded material outside the second casing. Inverting simulated measurements with five receivers, using a simulated annealing algorithm and a simple forward model, we find for a subset of simulations that we can estimate the impedance of the material outside the outer casing.},
keywords = {acoustics, guided waves, well logging},
pubstate = {published},
tppubtype = {article}
}
Current methods for ultrasonic pitch-catch well logging use two receivers to log the bonded material outside a single casing. For two casings separated by a fluid, we find by simulation that increasing the number of receivers provides a better picture of the effect of the bonded material outside the second casing. Inverting simulated measurements with five receivers, using a simulated annealing algorithm and a simple forward model, we find for a subset of simulations that we can estimate the impedance of the material outside the outer casing.
2016
Viggen, Erlend Magnus; Johansen, Tonni Franke; Merciu, Ioan-Alexandru
Simulation and modeling of ultrasonic pitch-catch through-tubing logging Journal Article
In: Geophysics, vol. 81, no. 4, pp. D383-D393, 2016.
Abstract | BibTeX | Tags: acoustics, guided waves, well logging | Links:
@article{viggen_simulation_2016,
title = {Simulation and modeling of ultrasonic pitch-catch through-tubing logging},
author = {Erlend Magnus Viggen and Tonni Franke Johansen and Ioan-Alexandru Merciu},
url = {https://erlend-viggen.no/wp-content/uploads/2018/04/viggen_simulation_2016.pdf, Full-text},
doi = {10.1190/geo2015-0251.1},
year = {2016},
date = {2016-07-01},
urldate = {2016-07-01},
journal = {Geophysics},
volume = {81},
number = {4},
pages = {D383-D393},
abstract = {Cased petroleum wells must be logged to determine the bonding and hydraulic isolation properties of the sealing material and to determine the structural integrity status. Although ultrasonic pitch-catch logging in single-casing geometries has been widely studied and is commercially available, this is not the case for logging in double-casing geometries despite its increasing importance in plug and abandonment operations. It is therefore important to investigate whether existing logging tools can be used in such geometries. Using a finite-element model of a double-casing geometry with a two-receiver pitch-catch setup, we have simulated through-tubing logging, with fluid between the two casings. We found that there appears a cascade of leaky Lamb wave packets on both casings, linked by leaked wavefronts. By varying the geometry and materials in the model, we have examined the effect on the pulse received from the second wave packet on the inner casing, sometimes known as the third interface echo. The amplitude of this pulse was found to contain information on the bonded material in the outer annulus. Much stronger amplitude variations were found with two equally thick casings than with a significant thickness difference; relative thickness differences of up to one-third were simulated. Finally, we have developed a simple mathematical model of the wave packets’ time evolution to encapsulate and validate our understanding of the wave packet cascade. This model shows a more complex time evolution in the later wave packets than the exponentially attenuated primary packet, which is currently used for single-casing logging. This indicates that tools with more than two receivers, which could measure wave packets’ amplitude at more than two points along their time evolution, would be able to draw more information from these later packets. The model was validated against simulations, finding good agreement when the underlying assumptions of the model were satisfied.},
keywords = {acoustics, guided waves, well logging},
pubstate = {published},
tppubtype = {article}
}
Cased petroleum wells must be logged to determine the bonding and hydraulic isolation properties of the sealing material and to determine the structural integrity status. Although ultrasonic pitch-catch logging in single-casing geometries has been widely studied and is commercially available, this is not the case for logging in double-casing geometries despite its increasing importance in plug and abandonment operations. It is therefore important to investigate whether existing logging tools can be used in such geometries. Using a finite-element model of a double-casing geometry with a two-receiver pitch-catch setup, we have simulated through-tubing logging, with fluid between the two casings. We found that there appears a cascade of leaky Lamb wave packets on both casings, linked by leaked wavefronts. By varying the geometry and materials in the model, we have examined the effect on the pulse received from the second wave packet on the inner casing, sometimes known as the third interface echo. The amplitude of this pulse was found to contain information on the bonded material in the outer annulus. Much stronger amplitude variations were found with two equally thick casings than with a significant thickness difference; relative thickness differences of up to one-third were simulated. Finally, we have developed a simple mathematical model of the wave packets’ time evolution to encapsulate and validate our understanding of the wave packet cascade. This model shows a more complex time evolution in the later wave packets than the exponentially attenuated primary packet, which is currently used for single-casing logging. This indicates that tools with more than two receivers, which could measure wave packets’ amplitude at more than two points along their time evolution, would be able to draw more information from these later packets. The model was validated against simulations, finding good agreement when the underlying assumptions of the model were satisfied.
Viggen, Erlend Magnus; Johansen, Tonni Franke; Merciu, Ioan-Alexandru
Analysis of outer-casing echoes in simulations of ultrasonic pulse-echo through-tubing logging Journal Article
In: Geophysics, vol. 81, no. 6, pp. D679–D685, 2016, ISSN: 0016-8033, 1942-2156.
Abstract | BibTeX | Tags: acoustics, well logging | Links:
@article{viggen_analysis_2016,
title = {Analysis of outer-casing echoes in simulations of ultrasonic pulse-echo through-tubing logging},
author = {Erlend Magnus Viggen and Tonni Franke Johansen and Ioan-Alexandru Merciu},
url = {https://erlend-viggen.no/wp-content/uploads/2018/04/viggen_analysis_2016.pdf, Full-text},
doi = {10.1190/geo2015-0376.1},
issn = {0016-8033, 1942-2156},
year = {2016},
date = {2016-01-01},
urldate = {2017-10-11},
journal = {Geophysics},
volume = {81},
number = {6},
pages = {D679--D685},
abstract = {Cased petroleum wells must be logged to determine the bonding and hydraulic isolation properties of the cement. Ultrasonic logging of single casings has been widely studied and is commercially available. However, ultrasonic logging in multiple-casing geometries is an unexplored topic despite its importance in plug and abandonment operations. Therefore, current logging technologies should be studied to evaluate whether they indicate the potential for multiple-casing logging. In this study, we used two finite-element models of pulse-echo logging. The first model represents logging in the transverse cross section of a double-casing well. The second model is a copy of the first, but with the outer casing and formation removed so that the pulse-echo transducer receives only a resonant first interface echo from the inner casing. By subtracting the received signals of the second model from those of the first, we can recover the third interface echo (TIE) signal representing the resonant reflection from the outer casing. This signal is used to study what information can, in principle, be drawn from TIEs in double-casing geometries, with the caveat that TIEs can only approximately be recovered in practical cases. We simulated variations of the material in the annulus beyond the outer casing, of the thickness of the outer casing, and of the eccentering of the outer casing. We have determined that the first two of these variations have only weak effects on the TIE, but that the eccentering of the outer casing can, in principle, be found using the TIE arrival time.},
keywords = {acoustics, well logging},
pubstate = {published},
tppubtype = {article}
}
Cased petroleum wells must be logged to determine the bonding and hydraulic isolation properties of the cement. Ultrasonic logging of single casings has been widely studied and is commercially available. However, ultrasonic logging in multiple-casing geometries is an unexplored topic despite its importance in plug and abandonment operations. Therefore, current logging technologies should be studied to evaluate whether they indicate the potential for multiple-casing logging. In this study, we used two finite-element models of pulse-echo logging. The first model represents logging in the transverse cross section of a double-casing well. The second model is a copy of the first, but with the outer casing and formation removed so that the pulse-echo transducer receives only a resonant first interface echo from the inner casing. By subtracting the received signals of the second model from those of the first, we can recover the third interface echo (TIE) signal representing the resonant reflection from the outer casing. This signal is used to study what information can, in principle, be drawn from TIEs in double-casing geometries, with the caveat that TIEs can only approximately be recovered in practical cases. We simulated variations of the material in the annulus beyond the outer casing, of the thickness of the outer casing, and of the eccentering of the outer casing. We have determined that the first two of these variations have only weak effects on the TIE, but that the eccentering of the outer casing can, in principle, be found using the TIE arrival time.
This page is powered by the excellent TeachPress plugin for WordPress.