J. Morgenroth, K. Kalenchuk, L. Moreau-Verlaan, M. Perras, U. T. Khan
{"title":"A novel long-short term memory network approach for stress model updating for excavations in high stress environments","authors":"J. Morgenroth, K. Kalenchuk, L. Moreau-Verlaan, M. Perras, U. T. Khan","doi":"10.1080/17499518.2023.2182889","DOIUrl":null,"url":null,"abstract":"ABSTRACT Digitalisation has increased access to large amounts of data for rock engineers. Machine learning presents an opportunity to aid data interpretation. The operators of Garson Mine use a microseismic database to calibrate a mine-scale finite difference model, used to assess seismic risk to inform mine operations. A Long-Short Term Memory (LSTM) network is proposed for stress model updating. The model is trained using microseismic data, geology, and geomechanical parameters from the FLAC3D model. Two LSTM networks are developed for Garson Mine: (1) predicting far field principal stresses in the FLAC3D model, and (2) predicting the far field six-component stress tensors in the model. Various LSTM network hyperparameters were analyzed to determine the architecture for the targets: input encoding and pre-processing, training solver, network layer architecture, and cost function. Architectures were chosen based on the corrected Akaike Information Criterion (AICc), coefficient of determination (R2), and percent capture (%C). When predicting principal stresses, AICc = −59.62, R2 = 0.996, and %C = 97%, and when predicting the six-component stress tensor AICc = −45.50, R2 = 0.997, and %C = 80%. This research represents progress towards continuous, automated updating of numerical models such that rapid, more accurate forecasts of changes in stress conditions will allow earlier reaction to challenging stress environments, increasing safety of excavations.","PeriodicalId":48524,"journal":{"name":"Georisk-Assessment and Management of Risk for Engineered Systems and Geohazards","volume":"17 1","pages":"196 - 216"},"PeriodicalIF":6.5000,"publicationDate":"2023-01-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"2","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Georisk-Assessment and Management of Risk for Engineered Systems and Geohazards","FirstCategoryId":"5","ListUrlMain":"https://doi.org/10.1080/17499518.2023.2182889","RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"ENGINEERING, GEOLOGICAL","Score":null,"Total":0}
引用次数: 2
Abstract
ABSTRACT Digitalisation has increased access to large amounts of data for rock engineers. Machine learning presents an opportunity to aid data interpretation. The operators of Garson Mine use a microseismic database to calibrate a mine-scale finite difference model, used to assess seismic risk to inform mine operations. A Long-Short Term Memory (LSTM) network is proposed for stress model updating. The model is trained using microseismic data, geology, and geomechanical parameters from the FLAC3D model. Two LSTM networks are developed for Garson Mine: (1) predicting far field principal stresses in the FLAC3D model, and (2) predicting the far field six-component stress tensors in the model. Various LSTM network hyperparameters were analyzed to determine the architecture for the targets: input encoding and pre-processing, training solver, network layer architecture, and cost function. Architectures were chosen based on the corrected Akaike Information Criterion (AICc), coefficient of determination (R2), and percent capture (%C). When predicting principal stresses, AICc = −59.62, R2 = 0.996, and %C = 97%, and when predicting the six-component stress tensor AICc = −45.50, R2 = 0.997, and %C = 80%. This research represents progress towards continuous, automated updating of numerical models such that rapid, more accurate forecasts of changes in stress conditions will allow earlier reaction to challenging stress environments, increasing safety of excavations.
期刊介绍:
Georisk covers many diversified but interlinked areas of active research and practice, such as geohazards (earthquakes, landslides, avalanches, rockfalls, tsunamis, etc.), safety of engineered systems (dams, buildings, offshore structures, lifelines, etc.), environmental risk, seismic risk, reliability-based design and code calibration, geostatistics, decision analyses, structural reliability, maintenance and life cycle performance, risk and vulnerability, hazard mapping, loss assessment (economic, social, environmental, etc.), GIS databases, remote sensing, and many other related disciplines. The underlying theme is that uncertainties associated with geomaterials (soils, rocks), geologic processes, and possible subsequent treatments, are usually large and complex and these uncertainties play an indispensable role in the risk assessment and management of engineered and natural systems. Significant theoretical and practical challenges remain on quantifying these uncertainties and developing defensible risk management methodologies that are acceptable to decision makers and stakeholders. Many opportunities to leverage on the rapid advancement in Bayesian analysis, machine learning, artificial intelligence, and other data-driven methods also exist, which can greatly enhance our decision-making abilities. The basic goal of this international peer-reviewed journal is to provide a multi-disciplinary scientific forum for cross fertilization of ideas between interested parties working on various aspects of georisk to advance the state-of-the-art and the state-of-the-practice.