ISRM 2023 - PUBLICATIONS

T05: Digitalization & AutomatisationMingliang ZHOU (1), Le ZHANG (2), Jiayao CHEN (3), Minglun TAN (2), Hongwei HUANG (1)1: Tongji University, Shanghai, People's Republic of China; 2: Qingdao Guoxin Jiaozhou Bay Second Submarine Tunnel Co., Ltd., Qingdao, China; 3: College of Civil Engineering, Beijing Jiaotong University, Beijing, ChinaQuantitative Risk Assessment of Rock Tunnel Faces Using Stacked Deep Learning and Multi-Source DataQuantitative tunnel face risk assessment is the characteristic challenge of rock tunnel excavation projects. This study establishes a multi-source database and proposes a stacked deep learning method for the quantitative tunnel face risk assessment. Both contact and non-contact methods are used to collect various data sources such as face images, site geological information, and rock mass properties. Thirteen multi-source variables describing the rock tunnel faces were considered as inputs, and Rock Mass Rating (RMR) system was adopted to generate the target output. The proposed model architecture combines a range of well-performing models to make accurate predictions of the tunnel face rock mass rating. Overall, this study establishes a comprehensive tunnel face risk assessment framework that leverages various data sources and stacked deep learning. The experimental results from a tunnel project in China demonstrate that our stacking deep learning model performs well in assessing rock tunnel face stability. | Rock tunnel, Face stability, Machine learning, Risk Assessment, Multi-source Data
T05: Digitalization & AutomatisationDavid HACKER (1), Martin ZEINDL (2)1: Dipl.-Ing. Bernd Gebauer Ingenieur GmbH, Germany; 2: Landesbaudirektion Bayern, GermanyProject management in connection with a digital construction site with BIM in tunnelingThe special features of tunneling create great potential for utilizing the advantages of the digital construction site and the BIM method during construction. The approach was applied to the execution of an approximately 600 m long road tunnel for mining excavation. Data sources, such as the digital tunneling documentation, were combined in a central live data management system and used efficiently, especially for project controlling and risk management. In addition to an As-planned model, an As-Built model was created during construction implementation on the basis of the digital construction site data. Each volume based calculation of time (4D) and costs (5D) conducted for the project was in accordance with the construction contract guidelines and specifications. These calculations formed the basis for checking monthly invoicing, performing precise target/actual comparisons and transparently presenting forecasts. | Digital Construction Site, BIM, Tunneling, Project Controlling, Monitoring Dashboard, Data Management
T05: Digitalization & AutomatisationYupeng CAO (1), Weiren LIN (1), Feng ZHANG (2)1: Graduate School of Engineering, Kyoto University, Japan; 2: Department of Civil Engineering, Nagoya Institute of Technology, JapanAn example of calibrating physical parameters in a constitutive model based on machine learning frameworkIn order to more accurately describe the behavior of geological materials, constitutive models with increasing complexity are being developed. The increase in the number of parameters and equations makes engineering applications difficult. On the other hand, data-driven machine learning approaches have shown great potential in addressing this issue. Therefore, from a data perspective, this paper recapitulates and discuss the task and framework of machine learning. For the parameter calibration task, 1-Dimensional Convolutional Neural Networks (1D-CNN), Recurrent Neural Network (RNN), Genetic Algorithm (GA), and Particle Swarm Optimization (PSO) were tested under the framework of representation learning and optimization. In the preset computing environment, the normalized RMSEs of representation learning are 0.21 and 0.17, and the calculation takes less than 0.1 second. Optimization frameworks perform better with RMSEs of 0.04 and 0.07, but cost over 10 hours. Finally, the advantages and disadvantages of each framework are discussed. | Machine Learning. Constitutive Model. Parameter Calibration. Representation
T05: Digitalization & AutomatisationJukka-Pekka UUSITALO, Tobias WENDEL-EICHHOLZ, Matleena MELASAARISandvik Mining & Rock Solutions, FinlandUtilizing sensor technology in drilling optimizationGlobal climate goals have increased the need to reduce CO2 emissions and material waste in the mining industry. Rock drilling, being the first phase of the mining process, affects the whole chain from bedrock to crushed stone. Modern sensor technology provides a new level of visibility to the drilling process, enabling a reduction of CO2 emissions and material waste. Sandvik Mining and Rock Solutions is pioneering in this field of technology with the RockPulse® product. The product provides direct online stress wave measurement enabling fact-based adaptation to varying rock conditions and optimization of drilling parameters. RockPulse enables finding the optimal percussion power and the reduction of harmful stress waves, which increases the penetration rate and the effectiveness of rock drilling while reducing tool consumption. The RockPulse system has been verified to enable annual reductions of several tons of CO2 emissions per equipment. | climate goals, sensor technology, drilling optimization, rock knowledge
T05: Digitalization & AutomatisationAdrián J. RIQUELME (1), José Luis PASTOR (1), Miguel CANO (1), Roberto TOMÁS (1), Antonio ABELLÁN (2)1: University of Alicante, Spain; 2: Centre de recherche sur l’environnement alpin (CREALP)Influence of SfM reconstruction techniques on the extraction of rock slope discontinuitiesRock slope discontinuities can be analyzed using 3D point clouds (3DPC), which are typically acquired through 3D laser scanning (3DLS) instruments or Structure from Motion (SfM) photogrammetry. While 3DLS is often regarded as a precise dataset, the use of SfM is gaining popularity due to the advent of remotely piloted aircraft systems and its relative affordability. Traditionally, extracting discontinuities from 3DPC relies on dense cloud reconstruction. However, the exploration of tiled models as an alternative approach remains limited. In order to assess the viability of using tiled models, we conducted an experiment on a pyramid-shaped sculpture reconstructed using both 3DLS and SfM techniques. The results demonstrate that the orientation of discontinuities obtained from the tiled model aligns well with those derived from both the TLS and dense cloud results. Interestingly, the extraction of discontinuities from the tiled model proves to be more efficient than from the SfM dense cloud. | Structure-from-Motion, 3D Point Clouds, discontinuities, rock slope, experiment, tiled-model
T05: Digitalization & AutomatisationAndreas-Nizar GRANITZER (1), Franz TSCHUCHNIGG (1), Wolfgang SUMMERER (2)1: Graz University of Technology, Graz, Austria; 2: ILF Consulting Engineers Austria GmbH, Rum b. Innsbruck, AustriaIntegration of Point Cloud Data for Numerical Simulations Using NURBS SurfacesThe numerical modelling of complex surface geometries represents a challenging task in computational geotechnics. Corresponding problem classes include non-horizontal topographic surfaces (TS) in mountainous terrain, rock mass discontinuities or steep slopes, presumably inducing significant stress redistributions in the project domain that need to be considered. This raises an obvious question at the interface with geographic information systems: How to integrate point-based surveying measurements defining the TS into the numerical modelling process? In this respect, we present a step-by-step methodology based on survey data from two infrastructure projects in Austria, at the core of which NURBS surfaces are used to generate TS models with high accuracy. | NURBS, GIS-to-FEM, point pattern analysis, optimal tunnel route, topographic surface
T05: Digitalization & AutomatisationMyung-Kyu SONG (1), Hyun-Koo LEE (1,2), Jae-Kyum LEE (1), Sean Seungwon LEE (1)1: Hanyang University, Republic of Korea (South Korea); 2: AICONT, Republich of Korea (South Korea)Identification of the optimal time series machine learning algorithm for the prediction of the ground subsidence with TBM machine dataControlling ground settlement during tunnel excavation in urban areas is a challenging task for contractors even with tight and comprehensive monitoring. In this study, utilizing the settlement monitoring and the sensor data collected during TBM drive, penetration and settlement prediction models are built. We postulated that TBM machine sensors may capture both actions of the machine and the reactions of the ground. Hence the prediction of settlement can be made if an appropriate algorithm is applied. There are a few sequential algorithms such as vanilla LSTM, LSTM with attention, Transformer, and Informer. This paper attempts to identify the optimal algorithm for training sensor data with a sub-workstation equipped in TBM. By comparing the performances of the algorithms, the DALSTM is identified as optimal algorithm for TBM machine data. Furthermore, subsequent analyses are carried out to develop a settlement prediction model, which demonstrates exceptional performance, marking a promising step towards deployment of the proposed method. | TBM, Subsidence Prediction, Machine Learning, Time Series, Sensor Data
T05: Digitalization & AutomatisationWolfgang DOLSAK (1), Mislav MIKULEC (2)1: DSI Underground Austria GmbH, Austria; 2: RHI Magnesita GmbH, AustriaDevelopment and implementation of a sensor-supported rock bolt system for underground monitoringAlongside with digitalization in underground mining and tunnel construction, user-friendly acquisition and processing of data is of paramount importance, considering the particularities of the working environment. Various conventional data measurement systems are currently in use for different applications; however, their scope is somehow isolated and not applied for a large-scale basis monitoring of default ground support systems such as rock bolts. For underground applications, system sourcing, and installation of special monitoring bolts is cost intensive and impracticable for a large-scale application. A low-cost intelligent rock bolt assembly concept was developed during the EU-funded illuMINEation project, part of the horizon 2020 research and innovation program. This low-cost intelligent rock bolt assembly allows an easy application with already installed rock bolts, or in combination with rock bolts featuring an integrated tendon sensor. The principal idea is to provide real-time recordings and visualization of geotechnical and environmental measurands on a large-scale collective basis. | ground support, rock bolt, monitoring, tendon sensor, illuMINEation
T05: Digitalization & AutomatisationAlexander ZÖHRER (1), Vincent WINTER (1), Anika TERBUCH (2), Paul O'LEARY (2), Negin KHALILI-MOTLAGH-KASMAEI (2)1: Keller Grundbau GesmbH, Austria; 2: University of Leoben, AustriaApplication of hybrid machine learning based quality control in daily site managementThis paper presents a system that combines KPI with autoencoders to implement a hybrid machine learning system. The goal here is to investigate workflows which permit the site manager to use the hybrid machine learning systems as a decision support tool. The workflows are explained by means of case studies, demonstrating the application of the hybrid system to detect both element as well as site related quality issues. In addition to that, the detection of anomalies regarding execution efficiency assist the project manager to optimize the sequence of work on site. | ground improvement, quality control, machine learning, key performance indicators
T05: Digitalization & AutomatisationAli BONAKDARIndividual private geomechanical advisor, Tehran, Islamic Republic of IranIntroducing a new cloud computation paradigm for rock engineering problems based on XaaS Model and the proposed A4 FrameworkNowadays, High Performance Computing (HPC) is being used as one of the most interesting topics in both aspects of scientific and practical, relies on recent special and exquisite developments in software architectures and frameworks and also hardware improvements in distributed networks over the Internet. As a geo-engineering problem to be solved, there are some limitations against today single-node commercial applications, containing runtime, analysis costs, scale, and accessibility. In this research, a new paradigm has been proposed base on a new developed framework (A4) and XaaS model. Using the new paradigm, significant modifications will be applied in computation runtime, decrease in computation costs, high software accessibility, in addition to scalability and diversity of problem solving. An online web-based 3D visualization platform (containing pre-processing and post-processing of numerical modeling) has been implemented to remove the limitations of the available single-node conventional applications to enable them running on a simple Internet browser with an affordable cost. | Cloud Computation, High Performance Computing (HPC), Distributed Computation, Parallel Numerical Analysis, Computation Farm, Web 3D Visualization
T05: Digitalization & AutomatisationJinfan CHEN (1), Zhihong ZHAO (1), Xingguang ZHAO (2)1: Tsinghua University, Beijing, China; 2: Beijing Research Institute of Uranium Geology, Beijing, ChinaDeep learning-aided prediction of peak shear strength of rock fracturesA robust estimation of peak shear strength of rock fractures in engineering practice is significant, but the three-dimensional (3D) surface characteristics of fractures have not been comprehensively quantified in the existing models. In this study, a deep learning-aided prediction method is proposed to estimate peak shear strength of rock fractures by considering the 3D surface roughness characteristics. The dataset is generated by numerical simulations after experimentally calibrated, and contains four features (normal stress, rock mechanical properties, relative fracture elevation). The deep learning model (FracSNet) assisted by data augmentation and fine-tuning is developed to provide reliable peak shear strength prediction of artificially-split fractures. The prediction ability is validated utilizing experimental data, and the results demonstrate that FracSNet can provide reliable prediction. The deep learning model of rock fractures has great a potential in engineering application with limited access to experimental data. | Peak shear strength, Rock fractures, Particle-based discrete element method, Deep learning
T05: Digitalization & AutomatisationChuyen Ngoc PHAM (1,2), Li ZHUANG (1,2), Sun YEOM (1), Hyu-Soung SHIN (1,2)1: Dept. of Geo-Space Engineering, University of Science and Technology, Republic of Korea (South Korea); 2: Korea Institute of Civil Engineering and Building Technology, Republic of Korea (South Korea)Deep Learning based Automatic Fracture Identification in CT Images of RockComputed tomography (CT) scans have been widely used to study the internal structure of rock including fractures. However, accurately quantifying fractures based on CT data could be challenging due to blurry appearance of fractures in CT images, complexity of fracture patterns and heterogeneity of the host rock. In this study, deep learning-based method is presented for automatically and accurately segmenting fractures in CT images of rock. The method involves two steps: fracture detection using the Faster R-CNN algorithm and fracture segmentation using the U-Net algorithm. The detection step establishes spatial constraints for the segmentation process, significantly removing noise and artifacts in background and improving the final segmentation result. CT images of fractured granite, sandstone and shale were applied for verifying the proposed method. It was found that our method can achieve a Dice score of up to 0.93, outperforming other deep learning approaches including Mask R-CNN or U-Net alone. | Rock fracture, Computed tomography, Deep learning, Rotated bounding box
T05: Digitalization & AutomatisationJian LIU (1), Aohui OUYANG (1,3), Omar ALDAJANI (2), Zili LI (1,2), Herbert EINSTEIN (2)1: University College Cork, Ireland; 2: Massachusetts Institute of Technology; 3: The European Organization for Nuclear Research (CERN)Automatic Fracture Extraction in Laboratory Rock Sample using Deep Learning methodFractures play a crucial role in the hydromechanical behavior of rocks. To investigate the fundamental fracturing mechanism, the imagery of hydraulic fracture evolution is captured in laboratory testing of rock specimens. Conventionally, temporal-spatial characteristics of rock fractures must be identified and extracted manually or by image processing techniques (IPTs) for interpretation, requiring enormous time and labor with low accuracy. This paper develops a deep learning-based method that quickly and automatically identifies and extracts hydraulically induced fractures in rock specimens at the pixel level. The applicability of this method is validated through image datasets from hydraulic fracturing tests. This method shows better effectiveness and efficiency than previous IPTs. The accuracy of the deep learning method reaches 99 percent and the average processing speed is only 389 ms per image when adopting an NVIDIA Tesla T4 GPU, saving a large amount of time compared to human work. | deep learning, hydraulic fracture, fracture extraction, image processing
T05: Digitalization & AutomatisationTushar BHANDARI, Debasis DEB, Chamanth Sai Reddy VEMULAPATIIndian Institute of Technology, Kharagpur, IndiaGlobal Unstructured Digital Image Correlation for determining strains around circular openingDigital Image Correlation (DIC) is a non-contact displacement measurement method that uses image correlation algorithms to measure displacement. The algorithms used are often designed for structures having uniform geometry without cavities. However, in the rock engineering problems irregular geometry and cavities are common. Therefore, an unstructured finite element based DIC (FE-DIC) algorithm is developed, which can incorporate specimens with curved geometries, and cavities. The mathematical principles behind the algorithm are discussed briefly. The validation of the developed algorithm is conducted using a deformed image created using displacements based on the Kirsch solution. The algorithm is also employed to calculate displacement field and strain visualization around a circular cavity in a concrete specimen under compressive loading conditions. The results demonstrate the potential of the unstructured FE-DIC algorithm in providing insights into the material deformation and failure behavior in the specimens. | DIC, Unstructured Mesh, Circular Cavity, Strain Visualization, Kirsch Solution
T05: Digitalization & AutomatisationEngin USTA, Kamran ESMAEILIMine Modeling and Analytics Lab, Department of Civil and Mineral Engineering, University of Toronto, CanadaJoint surface characterization using manual and multi-sensor core logging systemsThe shear behavior of a discontinuity can be significantly influenced by its surface characteristics. This paper compares manual and sensor-based geotechnical core logging methods for joint surface characterization. More than 500 m of core samples were logged, in which 367 joints were both digitally and manually characterized. Manual logging was carried out by visually assessing joint surface roughness and alteration indices. Sensor-based logging includes creating 2D and 3D joint roughness profiles using a handheld 3D scanner and measuring joint wall hardness and alteration type using Equotip Leeb hardness and pXRF analyzer. The collected data were statistically analyzed. A comparison between manual and multi-sensor joint surface characterization was made to demonstrate the discrepancy between the manual and digital logging techniques regarding joint surface roughness, joint wall strength and joint alteration. | Joint surface characterization, Multi-sensor core logging, Leeb hardness test, Joint alteration, Joint roughness, 3D handheld scanner
T05: Digitalization & AutomatisationAlla SAPRONOVA (1), Kazuo SAKAI (2), Shuntaro MIYANAGA (2), Paul Johannes UNTERLASS (1), Abdallah Ahmed Fouad Elsayed SOLIMAN (1), Thomas MARCHER (1)1: Institute of Rock Mechanics and Tunnelling, University of Technology Graz, Austria; 2: Geotechnical Research Section, Infrastructure Technology Research Department, Taisei Advance Center of TechnologyMWD data analysis for optimization of tunnel excavationThe drill and blast tunneling method applies to various rock mass conditions and is widely used in underground construction. Optimization of drill and blast requires careful planning and currently depends on the engineers’ ability to execute the art of blasting. Intelligent analysis of measurement while drilling (MWD) data from blast holes can be used for process optimizations, responsible resource utilization, and risk minimization. For example, an Artificial Intelligence (AI) -based decision support system (DSS) can suggest the volume and content of explosive material. However, to develop a reliable and trustworthy DSS, one needs to understand the relation between MWD data logs and the underlying lithology conditions, like composition or type of rock mass. This work provides an overview of the most common methods for MWD data analysis. Selected methods are then utilized to develop predictive machine-learning (ML) models, which are further validated with available MWD data. | data analysis, MWD data, optimization, machine learning
T05: Digitalization & AutomatisationHaojun PANG, Fei JIA, Yingcai HOU, Feipeng HUANG, Yadong XUEDepartment of Geotechnical Engineering, College of Civil Engineering, Tongji University, Shanghai, ChinaIntelligent risk management for TBM hard rock tunnelling based on Knowledge GraphTunnel construction using TBM involves various factors that increase risks to the structure, including workers, machinery, operation, structure and surrounding environment. These factors interact in complex ways, making risk management rather complicated and challenging. To achieve a better risk management, state-of-the-art technologies such as knowledge graph (KG) can help manage construction risk by storing, managing and mining risk concepts and construction entities. In the paper, a risk management knowledge graph was created for the TBM hard rock tunnel constructed in West China using Neo4j graph database. Work breakdown tree (WBS) and risk breakdown tree (RBS) were created to subdivide the complex TBM tunnelling process and risk sources. WBS and RBS entities were then integrated into the knowledge graph, making the attributes and relations of various entities clear to engineers. The case study demonstrated that knowledge graph is effective, reliable and advanced in TBM hard rock tunnelling risk management. | TBM tunnelling, hard rock, risk management, knowledge graph, work breakdown structure, risk breakdown structure
T05: Digitalization & AutomatisationAlla SAPRONOVA, Paul Johannes UNTERLASS, Vaibhav SHRINGI, Thomas MARCHERInstitute of Rock Mechanics and Tunnelling, Graz University of Technology, AustriaTowards the development of a harmonized inventory database for decision support: automatized information extractionDecisions made during tunnel construction are based on the opinion of safety-oriented engineers and utilize the knowledge of human experts. Because every tunnel is unique to some extent, it can be assumed that experts' decisions are often "reinvented" on-site. Given the number of completed, ongoing, and planned tunnel projects, it is possible to identify projects which could be used as an extra reference to assist the decision-making processes for new constructions. For human experts it is difficult and time-consuming to identify all similar reference projects. Aiming at developing a harmonized inventory database for decision support (DS) during the planning and construction phases, this work discusses a pathway for retrieving and processing data from archived projects. The major steps for information extraction are described, and the process of developing a harmonized database is discussed. This work specifically addresses the process of extracting tabular information from images using machine learning methods. | data analysis, information extraction, technical documentation, decision support
T05: Digitalization & AutomatisationYerkezhan MADENOVA (1,2), Fidelis SUORINENI (1), Shuai XU (3)1: Nazarbayev University, Astana, Kazakhstan; 2: Los Alamos National Laboratory, Los Alamos, NM, USA (Current); 3: Northeastern University, Shenyang, People's Republic of ChinaAutomated and digitalized web tool for open stope designThe Stability Graph is a widely used method for open stope design to control dilution in underground mining. The method considers several factors, which are determined from different empirical graphs and require additional computations. The manual application of the Stability Graph is time-consuming and has the potential to cause computational and human subjectivity errors in open stope stability prediction. The Stability Graph is developed empirically, and its reliability increases with increased data size and quality. However, its current application does not allow data sharing between practitioners that use the method in the mining industry. This work presents the digital analogous of the Stability Graph method through a web application, which is built to ease its use and eliminate limitations. The web-based tool for open stope design is named StopeSoft and is available at openstope.com. StopeSoft provides several benefits to users compared to the traditional Stability Graph including case history data sharing. | Open stope, Stability Graph, StopeSoft web tool, automation, digitalization
T05: Digitalization & AutomatisationZhansaya MAKSUT, Rakhat MEIRAMOV, Adnan YAZICI, Fidelis SUORINENINazarbayev University, KazakhstanA Machine Learning-based Microseismic Event Location and Wave Velocity PredictionHuman-induced seismicity in underground mining has significant impacts on productivity, safety, and operating costs. Accurate predictors for microseismic event sources are crucial to minimize disasters such as rockbursts. This study develops machine learning algorithms to predict real-time seismic wave velocities in deteriorating underground mines, using data from Nazarbayev University's School of Mining and Geosciences laboratory. Traditional constant velocity models are imprecise, so the study explores several machine learning models. The best-performing model was Gradient Boosted Decision Tree with an MAE of 7.146 m/s. These findings demonstrate that machine learning algorithms can accurately predict seismic wave velocities in underground mining environments. | neural network, decision tree, regression, rockbursts, wave velocity
T05: Digitalization & AutomatisationJunsu LEEM (1), Jineon KIM (1), Jiwon CHOI (1), Jae-Joon SONG (1,2)1: Seoul National Unviersity, Seoul, Korea; 2: Research Institute of Energy and ResourcesComparison of an underground rock face 3D modeling performance: SfM-MVS with optimum photographing settings and LiDAR technologyStructure-from-motion & multi-view-stereo (SfM-MVS) and light-detecting-and-ranging (LiDAR) are the representative methods to generate a 3D point cloud of a rock face. Both methods have pros and cons depending on the conditions including illumination, surveying time, resolution, accuracy, and cost. For application in underground space, SfM-MVS has been used less than LiDAR due to its lack of error pre-determination and ambiguity of photographing settings. Leem (2023) has developed a theoretical error prediction model for SfM-MVS and derived optimum photographing settings which minimize SfM-MVS error under light and time constraints. This work utilized the optimum photographing settings for the SfM-MVS and compared it with LiDAR when modeling a 70 m² rock face at an illumination of 25 lx within 5 minutes at a tunnel construction site (Yeoju-si, Korea). As a result, SfM-MVS could generate a point cloud with 20 times higher resolution and double accuracy at 10 times lower cost than LiDAR. | SfM (structure-from-motion), MVS (multi-view-stereo), camera settings, UAV flight method, underground digital survey, LiDAR (light-detecting-and-ranging)
T05: Digitalization & AutomatisationZizhuo XIANG, Zexin YU, Joung OH, Guangyao SI, Ismet CANBULATUniversity of New South Wales, AustraliaA machine learning model to estimate in-situ rock strength from borehole geophysical logsThis paper proposes an artificial neural network (ANN) model, which aims to improve the estimation accuracy of in-situ rock uniaxial compressive strength (UCS) for the Australian mining industry. The model utilises borehole geophysical logs (i.e., sonic, neutron, gamma and porosity logs) and rock density as inputs. A dataset of 274 samples from two mine sites in Australia is applied for the training, testing and validation of the model. Compared with the conventional sonic velocity model, the mean absolute percentage error of the predictions improves from 34.3% to 19.8% and the root mean squared error is reduced by over 4.6 MPa. In addition, it is also obtained that the accuracy of the model varies depending on the lithologies and mine locations. The proposed model is expected to provide more accurate rock strength estimations and be beneficial for further geotechnical analysis, such as estimating in-situ stresses based on borehole breakout. | Uniaxial compression strength, In-situ rock strength estimation, Borehole geophysical logs, Artificial neural network
T05: Digitalization & AutomatisationTae Young KO, Ju-Pyo HONGDepartment of Energy and Resources Engineering, Kangwon National University, Chuncheon, KoreaComparative Analysis of CAI Estimation using Symbolic Regression and Machine Learning ApproachesThe abrasiveness of rocks being excavated is a major challenge in TBM tunneling, as it affects the performance and durability of cutting tools. The Cerchar abrasivity Index (CAI) is a widely used method to assess rock abrasiveness and predict tool wear and cutter life in TBM tunneling. The CAI can be estimated from rock properties, such as compressive strength, tensile strength, and petrographic factors. A novel approach using symbolic regression was proposed to predict CAI. Symbolic regression can generate accurate and interpretable mathematical equations to capture the relationship between inputs and outputs. The proposed approach was compared to traditional machine-learning-based regression models using a dataset obtained from published articles and geotechnical data reports. Various machine-learning-based regression methods were also used to forecast the CAI, and their performances were compared. The proposed symbolic regression-based CAI prediction model has the potential to improve the performance of models for predicting rock abrasivity. | Cerchar Abrasivity Index, Wear, Symbolic Regression, Machine Learning
T05: Digitalization & AutomatisationMateusz JANISZEWSKI, Lauri UOTINEN, Masoud TORKAN, Mikael RINNEAalto University, FinlandVirtual learning environments for rock engineering education and training - a guideline for development, examples, and lessons learnedThis paper presents the research and educational development activity at Aalto University in creating virtual learning environments for rock engineering education. Virtual learning environments are increasingly recognized as tools to improve engineering education, but their creation requires specialized knowledge of 3D scanning, computer graphics, and game development. The paper discusses a method for creating 3D models of real environments using photogrammetry, along with hardware and software options. The models are then integrated into virtual learning systems built using game engines. Two case examples focusing on digitizing sites for virtual rock mass mapping are presented, and the outcomes and lessons learned are discussed. The paper concludes that accurate and photorealistic virtual learning environments can be developed to enhance rock engineering education and training. This has implications for the future development of virtual learning environments in engineering education and highlights the potential for using extended reality technology to communicate complex spatial data. | virtual reality, engineering education, photogrammetry, virtual learning environment
T05: Digitalization & AutomatisationHannah SALZGEBER, Larissa SCHNEIDERBAUER, Kathrin GLAB, Matthias FLORAUniversity of Innsbruck, iBT, AustriaStudy on comparing current software for parametric modelling in Tunnel Information ModellingThe Tunnel Information Modelling (TIM) method has presented the tunnelling industry with certain modelling challenges. Tunnel structures are characterised by the arrangement of recurring components along an alignment and the therefore resulting lengthy and repetitive modelling task requires automation through parametric design. This paper presents an evaluation of currently used software solutions for TIM, which are able to implement parametric modelling via extensions or scripting. The comparison includes aspects of geometrical modelling, integration of alphanumerical information into model components and drawing derivation. The resulting table gives a consolidated overview of the findings and caters to anybody who is currently looking at the implementation of a viable software options for the application of TIM. | BIM, TIM, parametric modelling, digitalisation, tunnelling
T05: Digitalization & AutomatisationInes MASSIMO-KAISER (1), Hans EXENBERGER (1), Hannah SALZGEBER (1), Hannah WERKGARNER (1), Richard LOIDL (2), Matthias FLORA (1)1: Universität Innsbruck, Austria; 2: ASFINAG Bau Management GmbHFrom prognosis Ground Model to Tender Model and Tunnel Construction Framework Plan with Tunnel Information ModellingThis article presents a concept for a digital ground model usable throughout all project phases and visualizes causalities between geological conditions and the tunnel structure. Specifically, a dynamic modeling approach is introduced to represent the geological conditions from design to construction. Three process phases are dealt with: (i) a preliminary phase defining the model area, (ii) the geological prognoses, including parameterization and schematization along the alignment as a base for further planning steps, and (iii) the creation of a cumulative tunnel construction model. A schematic, parameterized, small-scale tunnel excavation element model is the basis for a dynamic adaptation of geology. In addition, detailed predictions modeled along the alignment are introduced to represent the predicted geology in higher detail, adjustable in the event of changes, and thus represent the current geological information in the construction area. These models form the basis for a tender model and a digital tunnel construction framework plan. | TIM, framework plan, ground model, design process
T05: Digitalization & AutomatisationSarp SAYDAM (2,1), Chengpei XU (1), Binghao LI (1), Birgul TOPAL (1), Serkan SAYDAM (1)1: MERE, Faculty of Engineering, University of New South Wales, Australia; 2: DYWIDAG-Systems International Pty Limited, AustraliaFeature Sampling and Balancing for Detecting Rock Bolts from the LiDAR Point CloudsRock bolts play a crucial role in enhancing the stability of tunnel structures. However, designing bolt detection methods from LiDAR point clouds often faces the challenge of data imbalance. Despite this, the state-of-the-art deep learning-based detection methods adopt uniform sampling strategy for both bolt and background points without considering their unequal distribution, leading to a substantial loss of bolt features and overabundance of noise points. We propose a novel bolt feature sampling and grouping strategy by integrating principal component and surface curvature analysis to balance the noise background points and the bolt points. The balanced point features are fed into a newly designed deep neural network with a weighted loss function to accurately detect the position of rock bolts. The proposed method achieves state-of-the-art results on the Civil Tunnel dataset and Mining Tunnel dataset, outperforming the state-of-the-art 3D deep learning-based detection methods with uniform sampling strategy. | Rock bolt detection, point cloud, LiDAR, neural network
T05: Digitalization & AutomatisationAmanda HUANG (1), Frank LI (1), Tong Joo SIA (1), Qianbing ZHANG (2)1: SMEC, Australia; 2: Monash University, AustraliaDigitalization and creep modelling for trinocular-cavern based metro stationMetro stations are complex underground infrastructure encompassing wide-span caverns. Their interactions with tunnels, adits, and utilities in vicinity can have impacts on ground stability. This paper aims to undertake a novel approach leveraging digitalization of design and 3D numerical modelling to better visualize and quantify the ground-structure interaction based on a metro station featuring a trinocular cavern platform. Knowledge-based prediction is important for critical areas of underground construction. By engaging numerical modelling and using appropriate constitutive relationships for both short- and long-term deformation criteria, predictions into the ground movement and responses of ground support can be made to offer insights into the adequacy of the support system and propose monitoring scheme correspondingly, which forms an essential part of structural health management for the underground infrastructure when coupled with data-driven analytics. | Digitalization, Long-term deformation, Numerical modelling, Trinocular-cavern, Monitoring
T05: Digitalization & AutomatisationDaisuke ASAHINA (1), Takato TAKEMURA (2), Takahisa ENDO (2), Yu TSUKURIMICHI (2), Yu AOJI (2)1: Geological Survey of JAPAN, AIST, Japan; 2: Geomechanics Lab., Department of Earth and Environmental Sciences, Nihon UniversityUsing webcam to visually observe the mechanical behaviors of Shirahama sandstone under triaxial compressionWe present a newly developed monitoring system to directly visually observe the shear behavior of Shirahama sandstones under triaxial compression testing. The proposed visual monitoring system contains a commercially available webcam and a pressure-resistant steel housing device. The webcam records the video clip of the rock specimen surface during the triaxial compression test. The recorded video frames and their associated digital image correlation (DIC) results allow a better understanding of the evolution of the strain fields and fracture development. The implementation of the proposed monitoring system is simple, safe, and inexpensive, but allow new data to be measure and provides new insights into the progressive shear behavior under triaxial compression loads. | webcam, visual monitoring system, triaxial compression test, DIC, sandstone
T05: Digitalization & AutomatisationJulia C. GODLEWSKA, Marc S. OGAN, Julia GALLAS, Sarah SCHUMSKI, Mandy DUDA, Tobias BACKERSRuhr-University, Bochum, GermanyComparison between manual and automated determination of discontinuity orientations in different rock mass typesIn this study, the use of automated discontinuity orientation analysis is evaluated as a potential addition to manual data collection. Point clouds of outcrops of three different rock masses in Germany with distinct structural characteristics are analysed using Open-Source Software Cloud Compare and the plug-in Facets. The accuracy of the automated analysis data is compared to manually collected data. The findings of this study confirm that the discontinuity orientation from automated analysis corresponds to the manually generated data regardless of the geological setting but shows differences caused by exposed area of the discontinuity surfaces and undetected discontinuities due to minimal apertures. Further there remain differences between the results of the methods caused by complex morphology, especially in the context of human bias. The automated approach allows for the investigation of areas that are inaccessible by manual methods, and can also reduce human bias through careful interpretation of the results. | Automated discontinuity analysis, UAV, DOM, photogrammetry
T05: Digitalization & AutomatisationHans EXENBERGER (1), Ines MASSIMO-KAISER (1), Hannah SALZGEBER (1), Peter KOMPOLSCHEK (2), Frédéric HEIL (3), Matthias FLORA (1)1: AB für Baumanagement, Baubetrieb und Tunnelbau (iBT), Universität Innsbruck; 2: DI Arch. Peter Kompolschek; 3: ÖBB-Infrastruktur AGConcept for Tunnel Information Modelling based work-preview and documentation during construction at Tunnel AngathFor the realisation of the ÖBB project Tunnel Angath, the idea was developed to use BIM for various controlling purposes. This publication takes a closer look at the implementation for the areas of work preview and documentation. Derived from established methods of conventional tunnel construction, the necessary models and their data requirements are defined. These include a target model based on the tender, several evolutions of work-preview models as well as singular as-built models, representing each excavation round. All models combined allow for a full documentation of the construction processes and its complete history. The accompanying data structure offers automated dependencies and future-proof expansion capabilities. Finally, this paper presents a concept, which still needs to be proven feasible within the pilot project. | Tunnel Information Modelling, TIM, Building Information Modelling, BIM, Tunnelling, Digitalisation
T05: Digitalization & AutomatisationAna Raquel Sena LEITE, Hatem MRADUniversité du Québec en Abitibi-Témiscamingue, CanadaThe creative process for the development of an autonomous bolting arm for underground minesGround support installation is essential in the underground mining industry. However, it is one of the riskiest activities due to the operator's exposure to unsupported faces recently excavated. From this reality, this work was performed to develop autonomous drilling and bolting arm for underground support purposes. The primary considerations for machinery are the rock mass interaction, artificial intelligence for recognition and mechanical requirements. First, some of the latest technologies that can be related to this research of the autonomous underground mine will be discussed. Secondly, the insight from the miners obtained from interviews will be reported. Finally, the challenges, premises and steps taken will be discussed. This research cannot be compared to one standard automatization due to the extreme variability encountered in rock masses during excavations. | Autonomation, machine learning, rock mechanics, ground support
T05: Digitalization & AutomatisationJosephine MORGENROTH (1), Jim HAZZARD (2), Shelby YEE (1), Davide ELMO (3)1: RockMass Technologies, Toronto, Canada; 2: Itasca Consulting Group, Inc., Minneapolis, USA; 3: The University of British Columbia, Vancouver, CanadaGeneration of a Discrete Fracture Network from digital discontinuity data captured using the 3D Axis Mapping methodAccurate mapping of geological structures and their orientations is a crucial step for analysis of underground excavation stability. A case is presented of a novel digital mapping workflow for generating a discrete fracture network (DFN) model. The RockMass Eon applies the 3-Dimensial Axis Mapping (3DAM) method to capture orientations and locations of discontinuities, producing a statistical distribution of their natural variability. The Eon uses infrared LiDAR and an altitude heading reference system to capture a point cloud, and a clustering algorithm is applied to extract statistically significant orientations representing the discontinuities. In this case study, 83 measurements are generated for 12 structures to provide a statistical dataset of the discontinuities. The mapped data from Glencore Kidd Operations in Canada are imported into ITASCA software to generate a DFN. This paper demonstrates that the use of digital data collection can streamline analyses and increase model confidence in underground rock engineering. | digital mapping, Discrete Fracture Network, numerical modelling, digitalization
T05: Digitalization & AutomatisationNaval SINGH (1), Oscar LUNDHEDE (2)1: ORCX AB, FORCIT Explosives, Gothenburg, Sweden; 2: L-Consulting, B&R Sverige, Orebro, SwedenThe autonomous system cycle approach in off-road industrial applications – a holistic viewThe fundamental need in ‘Off-road industrial applications’, e.g., mining, tunnelling, construction sites etc. are – safety, reliability, operational / economic viability, and sustainability. Available ‘automation and autonomous technologies’ has numerous benefits in off-road industrial applications. In the last 15 to 20 years ‘the automation and autonomous technologies’ has gained popularity in off-road industrial applications and addresses many of its above-mentioned fundamental needs. Although with all its operational and long-term economic benefits, still spread of automation and autonomous technologies are limited to few geographies and certain applications only. In this paper, we are presenting ‘the autonomous system cycle approach in off-road industrial applications’, and its potential benefits. We will also briefly review the available autonomous technologies, its status, and applications e.g., in mineral exploration, rock breaking, material haulage, rock processing etc. Through ‘the autonomous system cycle approach’, goal of this paper is to present a step-by-step guideline for integration and implementation of the new technologies in off-road industries. | Automation, Mining, Construction, Civil, Digitalization, Tunneling
T05: Digitalization & AutomatisationMasanari NAKATA (1), Karnallisa Desmy HALIM (1), Yeboon YUN (1), Harushige KUSUMI (1), Akinobu NISHIO (2)1: Kansai University, Osaka, Japan; 2: Kinki Regional Development Bureau, Osaka, JapanRock evaluation of NATM tunnel face using deep learningBecause of the complexity of the geological features, when the NATM method is used in Japan, the rock mass is evaluated in nine categories (A. condition of tunnel face, B. condition of excavation face, C. compressive rock strength, D. weathering and alteration, E. spacing of discontinuities, F. condition of discontinuities, G. direction of discontinuities, H. presence of water inflow, I. deterioration due to water) The evaluation is graded on four levels. The objective of this study is to use deep learning to quantitatively evaluate the frequency, condition, and morphology of fractures, as well as weathering and alteration of the tunnel faces; CNN was used to grade the three criteria regarding fractures. Furthermore, ratio of weathering area was detected by HSV color space for categories regarding weathering and alteration. We also applied Grad-CAM to verify whether the CNN model could actually evaluate rock fractures as a decision criterion. | New Austrian Tunneling Method, Rock mass rating, Convolutional neural network, Gradient-weighted class activation map