Automated recognition of stratigraphic marker shales from geophysical logs in iron ore deposits

Silversides, Katherine L.; Melkumyan, Arman; Wyman, Derek A.; Hatherly, Peter

doi:10.1016/j.cageo.2015.02.002

Cited by 26 publications

(4 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Machine learning has increasingly been utilized in various disciplines. In the field of geoscience, researchers are using machine learning to deal with geological problems, for example, parameter estimation (Iturrarán‐Viveros & Parra, 2014; Zerrouki et al, 2014), lithology characterization (Silva et al, 2015; G. Wang et al, 2014), and stratigraphic boundaries determination (Silversides et al, 2015; Singh, 2011). As for well log generation, cross‐plot and multiple regression have been applied to synthesize well logs (J. Wang et al, 2016).…”

Section: Introductionmentioning

confidence: 99%

Well Log Generation via Ensemble Long Short‐Term Memory (EnLSTM) Network

Chen

Zhang

2020

Geophysical Research Letters

View full text Add to dashboard Cite

In this study, we propose an ensemble long short-term memory (EnLSTM) network, which can be trained on a small data set and process sequential data. The EnLSTM is built by combining the ensemble neural network and the cascaded LSTM network to leverage their complementary strengths. Two perturbation methods are applied to resolve the issues of overconvergence and disturbance compensation. The EnLSTM is compared with commonly used models on a published data set and proven to be the state-of-the-art model in generating well logs. In the case study, 12 well logs that cannot be measured while drilling are generated based on the logs available in the drilling process. The EnLSTM is capable of reducing cost and saving time in practice. Plain Language Summary A novel neural network, called EnLSTM, is proposed by combining the ensemble neural network, which has good performance on small-data problems, and the cascaded long short-term memory network, which is effective at processing sequential data. The EnLSTM's capability of processing sequential data based on a small data set is especially suitable for generating synthetic well logs. In addition, two perturbation methods are used to ensure that the EnLSTM can be fully trained in practice. In the experiments, the EnLSTM achieved the current best results on a published well log data set, and its application value is verified in a case study.

show abstract

Section: Introductionmentioning

confidence: 99%

Well Log Generation via Ensemble Long Short‐Term Memory (EnLSTM) Network

Chen

Zhang

2020

Geophysical Research Letters

View full text Add to dashboard Cite

show abstract

“…However, machine learning has shown great potential in recent years, as it can describe these relationships with network parameters by using a large number of training datasets. Examples of relevant machine learning applications include geological parameter estimation (Ahmed Ali Zerrouki and Baddari, 2014;Iturrarán-Viveros and Parra, 2014), lithology discrimination (Wang et al, 2014;Silva et al, 2015), and stratigraphic boundary determination (Singh, 2011;Silversides et al, 2015). In recent years, there has been a surge of research on well log prediction, resulting in significant improvements in performance (Rolon et al, 2009;Alizadeh et al, 2012;Mo et al, 2015;Long et al, 2016;Salehi et al, 2017).…”

Section: Introductionmentioning

confidence: 99%

Well log prediction while drilling using seismic impedance with an improved LSTM artificial neural networks

Wang

Tang

et al. 2023

Front. Earth Sci.

View full text Add to dashboard Cite

Well log prediction while drilling estimates the rock properties ahead of drilling bits. A reliable well log prediction is able to assist reservoir engineers in updating the geological models and adjusting the drilling strategy if necessary. This is of great significance in reducing the drilling risk and saving costs. Conventional interactive integration of geophysical data and geological understanding is the primary approach to realize well log prediction while drilling. In this paper, we propose a new artificial intelligence approach to make the well log prediction while drilling by integrating seismic impedance with three neural networks: LSTM, Bidirectional LSTM (Bi-LSTM), and Double Chain LSTM (DC-LSTM). The DC-LSTM is a new LSTM network proposed in this study while the other two are existing ones. These three networks are thoroughly adapted, compared, and tested to fit the artificial intelligent prediction process. The prediction approach can integrate not only seismic information of the sedimentary formation around the drilling bit but also the rock property changing trend through the upper and lower formations. The Bi-LSTM and the DC-LSTM networks achieve higher prediction accuracy than the LSTM network. Additionally, the DC-LSTM approach significantly promotes prediction efficiency by reducing the number of training parameters and saving computational time without compromising prediction accuracy. The field data application of the three networks, LSTM, Bi-LSTM, and DC-LSTM, demonstrates that prediction accuracy based on the Bi-LSTM and DC-LSTM is higher than that of the LSTM, and DC-LSTM has the highest efficiency overall.

show abstract

“…For example, Kitzig et al (2017) demonstrated that the inclusion of petrophysical data with geochemistry (using unsupervised fuzzy c-means algorithm) can be useful for distinguishing rocks with similar chemistry but different textures and improve the overall classification rate. Silversides et al (2015) used Gaussian processes(supervised ML) to provide probabilistic values to classify characteristic shale bands in iron ore.…”

Section: Introductionmentioning

confidence: 99%

Improving Automated Geological Logging of Drill Holes by Incorporating Multiscale Spatial Methods

2020

View full text Add to dashboard Cite

Manually interpreting multivariate drill hole data is very time-consuming, and different geologists will produce different results due to the subjective nature of geological interpretation. Automated or semi-automated interpretation of numerical drill hole data is required to reduce time and subjectivity of this process. However, results from machine learning algorithms applied to drill holes, without reference to spatial information, typically result in numerous small-scale units. These small-scale units result not only from the presence of very small rock units, which may be below the scale of interest, but also from misclassification. A novel method is proposed that uses the continuous wavelet transform to identify geological boundaries and uses wavelet coefficients to indicate boundary strength. The wavelet coefficient is a useful measure of boundary strength because it reflects both wavelength and amplitude of features in the signal. This means that boundary strength is an indicator of the apparent thickness of geological units and the amount of change occurring at each geological boundary. For multivariate data, boundaries from multiple variables are combined and multiscale domains are calculated using the combined boundary strengths. The method is demonstrated using multi-element geochemical data from mineral exploration drill holes. The method is fast, reduces misclassification, provides a choice of scales of interpretation and results in hierarchical classification for large scales where domains may contain more than one rock type.

show abstract

Automated recognition of stratigraphic marker shales from geophysical logs in iron ore deposits

Cited by 26 publications

References 16 publications

Well Log Generation via Ensemble Long Short‐Term Memory (EnLSTM) Network

Well Log Generation via Ensemble Long Short‐Term Memory (EnLSTM) Network

Well log prediction while drilling using seismic impedance with an improved LSTM artificial neural networks

Improving Automated Geological Logging of Drill Holes by Incorporating Multiscale Spatial Methods

Contact Info

Product

Resources

About