A Convolutional Neural Network Approach for Predicting Tunnel Liner Yield at Cigar Lake Mine

Morgenroth, Josephine; Perras, Matthew A.; Khan, Usman T.

doi:10.1007/s00603-021-02563-3

Cited by 9 publications

(4 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Balancing may be built into the MLA itself by using ran- Error weighting is a technique used in classification to give particular targets preference during training by penalizing others [20]. This is illustrated with a case study where a convolutional neural network (CNN) was developed to classify tunnel liner yield at Cigar Lake Mine [21]. Four error weighting schemes were compared: uniform, linear, sigmoid, and inverse frequency (Figure 6).…”

Section: On Regression Algorithm Architecture Selection and Verificationmentioning

confidence: 99%

Practical recommendations for machine learning in underground rock engineering – On algorithm development, data balancing, and input variable selection

Morgenroth

Unterlaß

Sapronova

et al. 2022

Geomechanics and Tunnelling

Self Cite

View full text Add to dashboard Cite

Research has demonstrated that machine learning algorithms (MLAs) are a powerful addition to the rock engineering toolbox, and yet they remain a largely untapped resource in engineering practice. The reluctance to adopt MLAs as part of standard practice is often attributed to the ‘opaque’ nature of the algorithms, the complexity in developing them, and the difficulty in determining how the algorithms use the datasets. This article presents tools and processes for developing MLAs, input selection, and data balancing for practical underground rock engineering. MLAs for classification and regression – two main machine learning applications – are presented in terms of developing MLA to extract information from the dataset to obtain the desired output. Engineering verification metrics are selected based on their suitability for specific output. Methods for input selection and data balancing are discussed with a focus on selecting appropriate input data for the problem without introducing bias or excess complexity. Each tool and process for algorithm development, data preparation, and input selection is illustrated with a case study. This article demonstrates that geotechnical practitioners can extract additional value by applying MLAs to rock engineering problems. Once an understanding of the functions of MLAs is reached, the building blocks and open‐source code are available to be adapted to suit the rock mass behaviour of interest.

show abstract

Section: On Regression Algorithm Architecture Selection and Verificationmentioning

confidence: 99%

Practical recommendations for machine learning in underground rock engineering – On algorithm development, data balancing, and input variable selection

Morgenroth

Unterlaß

Sapronova

et al. 2022

Geomechanics and Tunnelling

Self Cite

View full text Add to dashboard Cite

show abstract

“…The Cigar Lake Mine was found to be an appropriate case study for a machine learning application because there are complex tunnel deformation mechanisms, namely the combination of a squeezing environment caused by poor geology and the varying imposed stress conditions caused by the ground freezing regime. The priority for the developed CNN approach was to accurately pre dict tunnel liner and support yield, which is critical for the safe operation of the tunnel and for forecasting rehabilitation works needed to maintain operations (Morgenroth et al, 2021).…”

Section: Cigar Lake Mine Convolutional Neural Networkmentioning

confidence: 99%

“…Each image is comprised of four channels, similar to the RGB (red, green, blue) channels in a typical image, however in this case the channels are GEO, SUPCL, FREEZE, and DISP. The predicted tunnel liner yield is a categorical output, also digitized from the GMPs, and characterized by failed rock bolts, spalled shotcrete, and compressed yield packs (Morgenroth et al, 2020): Class 0 is no yield, Class 1 is yield requiring minor repair, Class 2 is yield requiring major repair, and Class 3 is yield requiring total tunnel reprofiling. The Cigar Lake Mine CNN is trained on a subset of GMP images and then tested on the subsequent GMP image, to simulate prediction of tunnel liner yield forward in time, i.e.…”

Section: Cigar Lake Mine Convolutional Neural Networkmentioning

confidence: 99%

See 1 more Smart Citation

The Evolution of Geotech - 25 Years of Innovation

Hammah¹,

Yacoub²,

McQuillan³

et al. 2021

View full text Add to dashboard Cite

I have been teaching and researching rock engineering and practicing as a consultant since 1972. I have spent most of that time in underground mining. I have seen significant changes, both evolutionary and radical, in the sector. In this paper, I reflect on some of these changes in mining geomechanical practice. I will also comment on some of the positive and occasionally less positive impacts of these changes. One constant throughout my career is that geomechanical designs will always be fraught with uncertainty. As a result, we will discuss geomechanical design as being, in its essence, a risk management exercise. I will argue that, as impressive as modern numerical analysis techniques are, they do not on their own mitigate design risk; under worst-case scenarios, they make things worse. Empirical and simple models must be used as a check. The paper intends to provoke healthy skepticism in geomechanics practitioners, designers and managers regarding all the aspects of the geomechanical analysis and design process. A failure to recognize the risks in these processes can have catastrophic effects.

show abstract