Search Documents

By Hendri Silaen, Ishak Degei, Nicoll Soumilena, Arwin Edijanto, Avesiena Primadiansyah

PT Freeport Indonesia’s Deep Mill Level Zone (DMLZ) is one of three operating cave mines at the Grasberg mine complex. The DMLZ is situated in the East Ertsberg Skarn System (EESS). It is set 500 m below the operating DOZ mine and located approximately 1,500 m below the surface, making it one of the world’s deepest panel caving operations. Initial undercutting of the DMLZ commenced in the second half of 2015 and has expanded to more than 120,000 m2. High pre-mining stresses and the changes associated with caving, combined with a strong and brittle rock mass, led to significant seismicity and rock bursting. Stress and seismic management are the main challenges for the operation and have evolved significantly from initial undercutting to the present. Considerable amounts of empirical data associated with ground response have been collected, evaluated and combined with historical lessons learned from the Intermediate Ore Zone (IOZ) and Deep Ore Zone (DOZ) mines, which are located within the EESS. This has provided an improved understanding of how the various geological, geotechnical and mine design criteria relate to one another. Additionally, the information has allowed for the development and implementation of improved strategies for stress and seismic management in the DMLZ. This paper describes how these learnings have driven changes in mine geometry, layout, and support systems. Through comprehensive monitoring and detailed inspections, positive changes have been seen in the ground response, leading to an increase in footprint reliability and accelerated cave growth. The paper summarizes the key learnings at the DMLZ regarding stress and seismic management as well as the positive improvements towards sustained and safe caving operations in the DMLZ.

Apr 26, 2022

By Wenkai Wan, Charlie C. Li

Strain burst can occur in underground hard rock excavation when the in-situ stresses are high enough. The prerequisite for rock burst is that the total energy released from the rock mass is higher than the energy needed for rock fracture. The total released energy is contributed by both the burst rock and the surrounding rock. The part of the energy from the burst rock is constant per unit volume of the rock, but the part from the surrounding rock is associated with the burst volume. A greater amount of energy will be released from the surrounding rock in the case of a larger volume of burst rock. The postulation for the energy conversion in strain burst is verified through a FLAC3D model in the paper. It is assumed that the strain burst occurs in a circular tunnel that is located in the hydrostatically loaded rock mass in the model. Six pairs of notches with different depths are created in the opposite walls of the tunnel to simulate the bursts. The energies released from the notch rocks and from the surrounding rock are calculated during the burst. The simulation results show that the percentage of the energy released from the surrounding rock in the total released energy increases with the notch depth. The numerical modelling confirms the postulation that the surrounding rock contributes more energy than the burst rock itself in large-scale strain bursts

Apr 26, 2022

By Patrick Linzer, Lindsay Linzer

By its nature, seismology has been a data intensive field since the first digital computers were introduced in the 1950s. The first empirical models were developed in the same period and are still widely used today, for instance, the Gutenberg-Richter frequency-magnitude relationship (Gutenberg and Richter, 1956). Machine learning (ML) methods make it possible to redefine these empirical models using large datasets and examining considerably more parameters, where their significance to the problem being addressed is methodically and repeatably evaluated using robust statistical methods. In this paper we develop a simple logistic model for the short-term prediction of seismic events above a threshold magnitude from a mine’s event catalogue comprising more than a million events with 14 basic parameters from which a pool of over 830 derived parameters was developed, covering a series of 11 precursory time windows. The logistic model was fitted stepwise using SAS with α = 0.025 for both entry and acceptance. 25 variables from the pool met the acceptance criterion. The findings were surprising in that parameters a rock engineer or seismologist might have expected to be the most significant in the logistic model were not; instead, it was an increase in standard deviation of the immediately-prior seismic magnitudes that carried the most weight, followed by the average maximum source displacement and median apparent volume change. These findings make intuitive sense because the increasing scatter of the magnitude shows an increasing number of large events being recorded that could be indicative of instability in the system that could lead to large events. The model is rudimentary and enormous scope exists for more exploration in terms of data pre-processing, model types, and response variable definitions.

Apr 26, 2022

By Ehsan Moosavi, Kamyar Tolouei, Mehran Gholinejad

The literature defines rockbursts in slightly different ways based on their generation mechanisms, field phenomena, and degree of damage, but in a way that is almost consistent. Generally, a rockburst refers to seismic events that mainly happen in underground excavations, for purposes such as underground mining, road and railway tunnels, nuclear power, etc., and that are followed by the collapse and outburst of rocks. Long-term rockburst prediction can be abstracted as a classification problem mathematically, in which we can introduce deep learning to solve it. Further, using deep learning in rockburst prediction can overcome drawbacks such as subjectivity and inconsistency brought from previous traditional approaches. A rockburst data-set, was employed to evaluate the current method for predicting rockburst grade, and the good results of overall success rate were obtained. The results indicated that the proposed algorithms can speed up parameter optimization search, the proposed method is robust model and might hold a high potential to become a useful tool in rockburst prediction research.

Apr 26, 2022

By M. Sears, W. H. Su, J. Wickline, M. A. Van Dyke, G. Esterhuizen, Z. Khademian

It is commonly recognized that seismicity can be induced by rock failures or slip along geologic structures around mining excavations. The unpredictable nature of these events and their proximity to the working area may raise safety concerns in underground mining operations. Modeling can be a useful tool for estimating seismic potential for a given mine layout, but the challenge is to detect a seismic event and estimate its intensity in the model. One well-documented approach is to monitor the variation in the energy content of the model and record potentially damaging seismic events when there is a significant release of energy. Researchers from the National Institute for Occupational Safety and Health (NIOSH) employed this approach using 3DEC modeling software to simulate the anomalous seismic events that occurred in a deep longwall mine in Virginia in 2005. A four-panel model of the mine was constructed utilizing a 2D cross section at the location of the mining face at the time of the event. Seismic potentials associated with the development of each panel are evaluated by recording the seismic energy released during the simulation. The results demonstrate the occurrence of two relatively significant events during the mining of the last panel, which are consistent with the historical observations at the mine.

Apr 26, 2022

By P K. Kaiser, D A. Malovichko

Highly stressed excavations in hard brittle rock may fail by self-initiated, triggered or dynamically loaded strainbursts (Moss and Kaiser 2021). The low-frequency part of seismic radiation associated with strainbursts is described by a crush-type seismic source mechanism with a significant negative (implosive) isotropic component and a ‘pancake-shape’ CLVD deviatoric component. The seismic source is co-located with the excavation damage site. For such strainbursts, the ejection velocity is primarily related to the depth of strainbursting, the rock mass bulking characteristics, the duration of the bulking process, and the effectiveness of the support system. For dynamically loaded strainbursts, the displacement and energy demand may be augmented by the impact of the stress wave from a distant seismic event. This article describes the failure process and factors contributing to rock deformation and energy demands for support design. Case examples are used to demonstrate the dependence of rock displacements, rock ejection and energy demand on the factors described above. A seismic source model is proposed to assess the depth of strainbursting and the duration of the related bulking process by using seismic monitoring data. The implications for support design are briefly discussed.

Apr 26, 2022

By Peter Zhang, Mark Van Dyke, Ted Klemetti, Morgan Sears

Room-and-pillar mining is a mining method used in the United States to extract coal from underground, and pillar recovery is widely practiced for maximizing coal recovery and mining economics. Under certain conditions, room-and-pillar mining can create high stresses in coal pillars resulting in a coal outburst. These outbursts represent one of the more severe safety hazards in underground coal mines, potentially causing injuries and fatalities. Over the past ten years, coal outbursts have resulted in three fatalities in U.S. room-and-pillar mines. Though coal outbursts generally occur under deep cover, they are more likely to occur during pillar recovery and under multiple-seam mining conditions. Coal outbursts are generally associated with strong roof and floor and heavy loading of coal pillars, but it is difficult to predict whether, where, and when they will occur. In this respect, risk management provides a proactive tool to prevent coal outbursts in room-and-pillar mining by identifying mining areas with high risk of outburst and mitigating the risk accordingly.

Apr 26, 2022

By B. Ollila, L. Brown

As mining fronts converge in sill pillar extraction, the mine seismic hazard can be significantly altered and large seismic events have been known to occur. This study used common seismological tools to analyze the extraction of highly stressed sills from two deep Canadian mines with the objective of characterizing the rock mass failure process and relation to large seismic events. The approach treated a sill pillar as a mine-scale failure process, meaning the temporal and spatial analysis scales needed to exist on a similar scale to a significant part of the mine. The temporal analysis scale overlapped with the entire mining sequence, included several years of data to establish long term change in the characteristics of rock mass failure. The spatial analysis scale included the rock mass surrounding a large part of the convergent mining sequence. To accommodate the large spatial and temporal scales, the highest magnitude 2% of events in the dataset were analyzed. Similar results were shown for sill pillars in both mines. Changes in event location, time, magnitude, energy index and apparent volume were similar to expected behavior of rock failure. Indications of strain hardening and strain softening were observed in the sills. The hardening period was characterized by a high rate of small events, high energy index and low apparent volume, giving an indication of high stress and low rock yield. The softening phase was characterized by a decrease in the rate of small events concurrent with an increase in the rate of large events, low energy index and high apparent volume, indicative of release of stress and increase in deformation. Large magnitude seismicity was strongly related to the softening phase, with several Nuttli magnitude 3.0+ events occurring shortly after the transition from hardening to softening failure regimes. Sill failure was indicated by consistently low EI and low levels of seismicity and seismic hazard in the sill pillar.

Apr 26, 2022

By E Mutaz, M Bahaaddini, D J. Williams, M Serati

The contribution of the mining industry to the global economy has reached (and gone beyond) $1,800 billion in recent years that necessitates (more than before) the need to ensure workability, viability, health, and safety of mine sites and workers to meet the ever-increasing global demand for mineral resources. Amongst the many challenges facing the excavation industries, rockburst and spalling failures in deep mines in hard and brittle rock formations are yet happening that are still deemed as unsolved problems. During unconfined compressive testing of brittle rocks, microcracks initiate at approximately 0.3–0.5 of the peak strength (which is commonly known as the Crack Initiation stress level) mainly in the direction of the major principal stress, then grow unstably and coalesce to form macro fractures at approximately 0.7–0.8 of the peak load at a so-called Crack Damage stress threshold. From a stress-strain perspective, the crack damage stress threshold () is associated with the reversal point of the volumetric strain, which is also proved to be strongly affected by the intermediate principal stress. This paper extends the previous studies on the impact of true triaxial loading conditions on the CD stress level in high porosity rocks, and introduce a new Crack Mode-Changing Stress (CMCS) defined as the minimum principal stress required to change rock fracturing from splitting to a sliding failure mode. A large number of experiments on sandstone specimens were carried out at different true triaxial loading conditions and the effect of principal stresses on the CI and CD as well as CMCS were studied. The results show that not only the intermediate principal stress, but also the minor principal stress can significantly affect the CD and CMCS thresholds.

Apr 26, 2022

By Grażyna Dzik, Adam Mirek, Krzysztof Król

The Upper Silesian Coal Basin is located in the area of the Upper Silesian Syncline and belongs to one of the most complex tectonic units in Europe. It occupies 5,650 km2. In 2020, there were 20 hard coal mines operating in the area. Seismological research in this area consists of observation of precursors/symptoms of shock generation and understanding of their effects. Studies (Teisseyre R. 1972, Mutke G. & Stec K. 1997, Stec K. 2007, Dubiński J. & other 2010, Idziak A. 2010) of complex impacts of rock mass provided a background for determining the conditions for the formation of exploitation and regional seismic shocks. Issues related to tectonic and exploitation stresses, successive extraction of coal seams at higher depths, operation in the vicinity of fault zones, often complex geological and organic conditions result in the activation of a discontinuity surface due to stress caused by deformation of rock layers, which, as a result, causes the emergence of strong shocks above 107 J. The article was based on analyses of extraction value, high-energy shocks, both of mining and tectonic-mining (so-called regional shocks) origin, against the background of changing operating conditions, which, as a result, creates an image of periodically changing seismic activity and location of strong shocks. Observation results show that, with an increase in exploitation depth and declining extraction, the number of strong shocks caused by tectonic stress (107 - 109 J) and exploitation stress (≥105 J) has been steadily increasing over the last 25 years.

Apr 26, 2022

By Klein E., Kinscher J., Lizeur A., De Santis F., Lhoumaud E.

Seismic repeaters are rarely observed phenomena in underground mines. However, recent studies in Garpenberg mine (Sweden) have demonstrated that seismic repeaters and multiplets may represent significant portion of the total number of microseismic events (> 80 %) related to excavation operations. Detailed seismic analysis, involving template matching, multiplet clustering, double-difference relocation, source parameter and mechanism estimation, and interevent time analysis, suggest that seismic repeaters represent brittle frictional parts (asperity) of creeping, planar shaped, pre-exiting structures of several meters length. It is furthermore suggested that seismic repeater occur preferentially in weak rock-mass materials (e.g. talc) associated with strengthening friction behaviors. These results may imply that postblast related stress redistribution may be governed by “silent” aseismic movements in preferentially weak rock mass materials, what questions the reliability and performance of seismic monitoring to survey stress evolution during excavation operations. However, once the seismic and aseismic coupling processes and the geological nature linked to repeater occurrences are better understood, their detailed monitoring bears also opportunities to survey and quantify aseismic post-blast stress transfer. Here we present ongoing investigations on repeater occurrences aiming at better understating on their geological and geomechanical origin. Furthermore, we try to develop advanced monitoring tools helping to quantify aseismic slip and to provide criteria for seismic hazard as asperity density and the related potential to generate larger dynamic rupture zones. For this respect a local monitoring network has been set up close to certain repeater target families. Furthermore, geological well log drilling throughout repeater occurrences as well as geotechnical strain measurements (e.g. extensometer, optical fiber etc.) are currently conducted to provide crucial information on the geological nature of repeater and to quantify and characterize background creep induced from excavation.

Apr 26, 2022

By Jurgens PE Hamman, Faisal Putra, Hadnaltias Alpeki, Willem de Beer

The Deep Ore Zone mine is the third lift of caving at PT Freeport Indonesia’s East Ertsberg Skarn System. An increasing trend in seismicity along the northern boundary of the DOZ cave, as well as the northern fringes of the DOZ production and infrastructure levels, was taken as indicative of an interaction between the approaching DMLZ cave and the DOZ subsidence zone. Seismicity trends and locations indicated a reducing crown pillar and an associated stress increase along these DOZ excavations, as well as the regional area along the DOZ cave boundaries. Although rockbursts were not identified as an immediate hazard, seismic shakedown was a highly probable hazard in the case of remote large (≥3.0) seismic events. An evaluation process, as described in the Canadian Rockburst Support Handbook (1996), was followed to identify and quantify the various parameters that could contribute to a shakedown evaluation. A shakedown assessment, based on the Peak Particle Velocities scaling relation, was conducted to assess the current installed ground support systems along the northern fringes and the fixed facilities of the DOZ mine. The range in ground support systems were mostly due to different geotechnical environments. The load–deformation characteristics of these support systems were evaluated to determine the support capacity and compared to the potential support demand under varying shakedown conditions. Shakedown hazard maps were compiled to indicate the installed ground support system’s vulnerability to seismic shaking. Based on this information, areas were divided into zones indicating whether they were sufficiently supported, additional support was required, access was to be restricted, and / or they were to be abandoned. Mine management utilised these hazard maps to strategize actions for the various operational zones within the DOZ, including the fixed facilities. This paper briefly describes the derivation of the PPV scaling relationship. Seismic and geological data for the DMLZ – DOZ interaction areas were evaluated to determine the likely source of large events. Subsequent larger events and their effects are shown to assess the veracity of the magnitude and PPV estimates, as well as the effectiveness and appropriateness of the ground support systems.

Apr 26, 2022

By KOKOWSKI J., RUDZIŃSKI Ł.

The LUMINEOS network was established in 2013 to monitor mining-induced seismicity in Legnica-Głogów Copper District (LGCD), located in south-western Poland. Since the starting time, the network was improved and today consists of 17 seismometers and 10 accelerometers. Up to now, the strongest recorded earthquake in the region had Mw4.0. Both raw seismic data and earthquakes catalog for events recorded by the LUMINEOS network are fully open and available on the EPISODES platform (https://tcs.ah-epos.eu/#episode:LGCD). During the last 7 years, the network improvements gave us a possibility to increase location accuracy for shallow mining events. This accuracy is influenced by factors such as the configuration of seismic stations, noise level at seismic stations, and the influence of the velocity model. Here we encountered two main issues which are worth to discuss. The first one is the low detection threshold of two seismic stations and their significant impact on the location errors. The second one is the large uncertainty in the S-wave picking. Even though S-waves are not so clearly visible on seismograms, they are very helpful in the estimation of source location and strength of seismic events in the region. In our work, we present the history of LUMINEOS network and corresponding improvement on location accuracy as well as the influence of noisy stations on the error distribution and how it is connected with location results. We also investigate the effect of the S-wave phases and how they are related to the accuracy of the earthquakes locations in the network.

Apr 26, 2022

By Kathy Kalenchuk

To develop strategic and tactical strategies to manage seismic hazard in a safe and economic manner, it is necessary to understand the spatial and temporal distribution of mine wide seismogenic behavior. Global mine-scale numerical stress modelling provides a means to forward predict seismic susceptibility thereby facilitating informed decision making for the development and implementation of strategic and tactical risk mitigation measures. To make forward predictions of hazard susceptibility using global stress models with reasonable confidence, it is necessary to rigorously calibrate numerical stress models to the observed history of ground reaction. This paper demonstrates quantifiable methods for calibrating numerical models and developing probabilistic prediction criterion using seismic data. The appropriate applications of model inputs (including boundary conditions, constitutive models and material parameterization) specific to global mine-scale modelling are also discussed. Lastly, this paper demonstrates how the calibration process adds value not only to informed decision making, but also to the improved understanding of site geomechanical conditions; including enhanced identification of geotechnical domains and mechanical indicators required for seismic mechanism domaining.

Apr 26, 2022

By Musa D. Manzi, Raymond J. Durrheim, Richard T. Masethe

Large mining-induced seismic events and their aftershocks cause hazardous working conditions, damage infrastructure, delay production and pose a risk to miners in the deep-level gold mines of South Africa. Stress changes affecting pre-existing zones of weakness are the main drivers of large mining-induced seismic events, which are often followed by aftershocks triggered by quasi-static stress transfer within the rockmass. A better understanding of seismic source mechanisms is important if the seismic data is to be used more effectively for mine planning and to reduce the rockburst risk. Four large seismic events (ML≥2.0) associated with the mining of the Ventersdorp Contact Reef (VCR) at Kloof Gold Mine occurred on 13 June 2017. The events and their fore- and aftershocks were studied in detail to establish the circumstances and the cause of the incidents. The local magnitudes ranged from ML 2.3 to ML 3.0, the seismic moment from 0.69 to 2.25 x 1013 Nm and the source radius from 92 – 150 m. The main shocks were preceded by foreshock sequences, which were especially vigorous in the four hours before the main shock. Seismicity was recorded by in-mine arrays of three-component 4.5Hz geophones (detection threshold approx. ML -2.9). Following all four main seismic events, ML 2.3, ML 2.3, ML 2.9 and ML 3.0, the network detected a total number of 300 aftershocks ranging from ML -2.9 to ML 1.6 spread within a 1300 m radius from the centre of the seismic network. The statistical properties of the aftershocks were analysed using three scaling relationships: (i) Gutenberg–Richter frequency magnitude relationship; (ii) modified Omori’s law; and (iii) Båth’s law for the magnitude of the largest aftershock. The geological or mining feature that controlled the failure (e.g., fault, dyke, pillar, abutment) was determined by calculating the orientation of the nodal planes of the large seismic events using moment tensor analysis and studying the aftershock distributions.

Apr 26, 2022

By Rene Padilla, Siyanda B. Mngadi, Gerrie van Aswegen, Peter Stonestreet

While many mine tunnels are developed through ground that must be riddled with the sources of small (and larger) mining induced seismic events, recorded by the mine seismic monitoring systems, we do not see planar discontinuities classified as seismic sources in the geotechnical databases. Here we list criteria that may be useful to identify discontinuities in drill core which are dynamic slip surfaces, parts of mining induced seismic sources. Such shear rupture surfaces in core would generally be powdery, cohesionless, and rough on cm scale. Finely broken rock and rock flour may be found in the core boxes at the discontinuities. Slickensides would be soft (easily destroyed) and kinematically consistent with the local stress field. In some cases, a cross fabric would be present, orthogonal to the slickensides, being the intersections of en-echelon extension fractures, within the dynamic shear zones, with the shear zone walls.

Apr 26, 2022

By David Beck

In Australia, Safe Work legislation requires mines to have a documented ‘Safety Case’ explaining how a mine will safely manage all major hazards. A similar requirement is legislated or inferred by regulation in all jurisdictions. For small scale, instabilities leading to day-to-day seismic hazards, a ‘Stability Safety Case’ built around observational methods and measurement triggered responses can be effective because adverse situations are easily identified and the costs of lower control measures are not onerous. Also, if the influence of a volume of yield on the mine load system is small, the response of the system will be more predictable. However, for episodes of large scale mine instability, the hazard derives from the widespread, multi component influence of the event on the mine load system. As the system response depends on interactions between many parts, the resolution of predictions will be lower and situations can arise without local precedent. Further, as episodes of large scale instability change the stiffness and nature of the connectedness between parts of the mine load system, the behavior of the system before the event will be a poor empirical basis for forecasts of event related hazards. Most importantly, Triggered Action Response Plans (TARPs) and ‘traffic light’ warning systems cannot make design choices or identify new hazard phenomena. They are ‘lower’ control measures targeting explicit human-made instability hypotheses. As such, the Safety Cases related to management of hazard from potential large scale seismic events and mine instability cannot rely solely on TARPs. Mines that successfully manage meaningful dynamic instability always acknowledge the ultimate reliance on expert human intuition, knowledge and imagination to appreciate the range of possible hazard scenarios, to make stability hypotheses and propose solutions. Complex analysis, TARPs and measurements are critical steps in the human decision cycle, but only support the making and testing of ‘Stability Hypotheses’ made by expert engineers. Relying on expert skill and judgement as essential components of a Safety Case, however essential, introduces human vulnerabilities. Another challenge is that the expert decision making process is hard to document. Beck et al 2019 first described an organizational and procedural framework for maintenance of an ongoing large-scale ‘Stability Safety Case' where a degree of expert judgement and imagination is needed to manage a seismically active mine. The structured data appreciation process outlined sufficient measurements, analysis and appreciation to confirm the stability state of the mine and was applicable to all underground mines. After several years of implementation in various forms in stoping, caving and room and pillar mines, this update addresses lessons learned and describes some considerations for numerical forecasting to establish a base case, and some challenges for empirical hazard identification.

Apr 26, 2022

By Stanisław Lasocki

We present an algorithm to integrate the uncertainties of mean activity rate and magnitude cumulative distribution function in the interval estimation of the primary seismic hazard parameters: the exceedance probability and the mean return period. The algorithm is applicable for both the parametric and nonparametric modeling of magnitude distribution. When the Gutenberg-Richter relation-based parametric model of magnitude distribution is used, the interval estimation of its parameters is based on the asymptotic normality of the maximum likelihood estimators. When the nonparametric kernel estimation of magnitude distribution is used, we apply the iterated bias-corrected and accelerated method for the interval estimation, based on the smoothed bootstrap and second-order bootstrap samples. Previous studies indicated that the uncertainty of mean activity rate significantly affects the interval estimates of hazard functions only when the activity rate is low and the inference period is not long (Orlecka-Sikora & Lasocki 2017). With increasing mean activity rate, the uncertainty of magnitude distribution dominates over the uncertainty of mean activity rate in the aggregated uncertainty of the hazard functions. The provided algorithm is generic and can be applied to capture the uncertainty of any multiparameter function from the uncertainties of its parameters. Here we present a further development of the algorithm developed by Orlecka-Sikora & Lasocki (2017) by presenting its implementation in the Traffic Light System approach. As an example of use, we present the confidence intervals estimation of hazard parameters for the datasets related to induced seismicity accompanying underground exploitation of copper ore in the Legnica-Głogów Copper District (LGCD) in Poland. The algorithm is under implementation to the EPISODES platform (https://tcs.ah-epos.eu/) - an IT platform that provides access to integrated research infrastructures of the Thematic Core Service of Anthropogenic Hazards (TCS AH) of EPOS (European Plate Observing System). The algorithm is also a component of the pilot developed within the EPOS SP project (No 871121), which aims to strengthen the interaction with the private sector and the added value for society. The pilot demonstrates the potential of the IS-EPOS platform software applications for the technology-related probabilistic assessment of time-dependent seismic hazard. The pilot supports a Traffic Light System approach. The current seismic hazard is compared to threshold values based on the exceedance probability for a given magnitude.

Apr 26, 2022

By Evan Dressel, Mark S. Diederichs, Kathy Kalenchuk

As hard rock mining develops deeper over time, there is an increasing need to predict the spatial and temporal distribution of adverse seismogenic rockmass reaction to mining. Real-time monitoring and feedback of mine data are used to tactically manage risk associated with induced-seismicity; however, strategic risk management requires reliable tools for forward predictions of seismic hazard. The forward prediction of seismic hazard is strengthened by understanding of potential seismic intensity, which often relies on well calibrated mine-wide numerical models. This paper uses a case study to explore how future seismogenic potential and the nature of future seismicity can be domained according to numerically predicted stress-strain paths at past seismic event locations. Delineation of high confinement seismogenic volumes vs. lower-confinement, high-strain seismogenic volumes at the mine-scale can be achieved, based on a loading and unloading stress path analogue, to refine understanding of seismic intensity. Practical qualitative and quantitative methods to achieve this delineation will be demonstrated. Once seismogenic mode (high-stress vs. high-strain) is identified, detailed analyses of seismic source parameters demonstrates change in seismic hazard between the domained volumes.

Apr 26, 2022

By Kazmierczak J. B., Klein E., Kinscher J., Lizeur A., De Santis F.

In this work, we propose to implement machine learning methods for the automatic classification of mining-related seismic data. The approach retained for the classification is based on a Convolutional Neural Network (CNN) which has been tested and applied to two different mining datasets: the active ore mine of Garpenberg (Sweden), currently under exploitation, and the abandoned coal mine of Gardanne (France), where mining stopped in 2003. The first step of the approach consists in creating a well classified (manually) database of seismic signals separated in different classes as a function of the seismic sources. The selected database is then divided into two groups: train, containing the 70% of the signals, and test, which contains the remaining 30% of the signals within the selected database. The learning phase of the algorithm consists in creating a classification model based on the train group, which is further validated on the test group. The application of CNN for the automatic classification of Garpenberg and Gardanne seismic data shows, respectively, that about 88% and 98% of the seismic events of the two sites can be properly classified. These results demonstrate that machine learning could be a powerful tool for microseismic monitoring in mining and post-mining environments, which could be further extended to the whole automatic processing of seismic data, including signal’s classification as well as microseismic event’s localization.

Apr 26, 2022