Transfer learning framework for multi-scale crack type classification with sparse microseismic networks

Rock fracture mechanisms can be inferred from moment tensors(MT)inverted from microseismic events.However,MT can only be inverted for events whose waveforms are acquired across a network of sensors.This is limiting for underground mines where the microseismic stations often lack azimuthal coverage.Thus,there is a need for a method to invert fracture mechanisms using waveforms acquired by a sparse microseismic network.Here,we present a novel,multi-scale framework to classify whether a rock crack contracts or dilates based on a single waveform.The framework consists of a deep learning model that is initially trained on 2400000+manually labelled field-scale seismic and microseismic waveforms acquired across 692 stations.Transfer learning is then applied to fine-tune the model on 300000+MT-labelled labscale acoustic emission waveforms from 39 individual experiments instrumented with different sensor layouts,loading,and rock types in training.The optimal model achieves over 86%F-score on unseen waveforms at both the lab-and field-scale.This model outperforms existing empirical methods in classification of rock fracture mechanisms monitored by a sparse microseismic network.This facilitates rapid assessment of,and early warning against,various rock engineering hazard such as induced earthquakes and rock bursts.

Microseismic source location using deep learning:A coal mine case study in China

Microseismic source location is crucial for the early warning of rockburst risks.However,the conventional methods face challenges in terms of the microseismic wave velocity and arrival time accuracy.Intelligent techniques,such as the full convolutional neural network(FCNN),can capture spatial information but struggle with complex microseismic sequence.Combining the FCNN with the long shortterm memory(LSTM)network enables better time-series signal classification by integrating multiscale information and is therefore suitable for waveform location.The LSTM-FCNN model does not require extensive data preprocessing and it simplifies the microseismic source location through feature extraction.In this study,we utilized the LSTM-FCNN as a regression learning model to locate the seismic focus.Initially,the method of short-time-average/long-time-average(STA/LTA)arrival time picking was employed to augment spatiotemporal information.Subsequently,oversampling the on-site data was performed to address the issue of data imbalance,and finally,the performance of LSTM-FCNN was tested.Meanwhile,we compared the LSTM-FCNN model with previous deep-learning models.Our results demonstrated remarkable location capabilities with a mean absolute error(MAE)of only 7.16 m.The model can realize swift training and high accuracy,thereby significantly improving risk warning of rockbursts.

Enhancing microseismic/acoustic emission source localization accuracy with an outlier-robust kernel density estimation approach

Monitoring sensors in complex engineering environments often record abnormal data,leading to significant positioning errors.To reduce the influence of abnormal arrival times,we introduce an innovative,outlier-robust localization method that integrates kernel density estimation(KDE)with damping linear correction to enhance the precision of microseismic/acoustic emission(MS/AE)source positioning.Our approach systematically addresses abnormal arrival times through a three-step process:initial location by 4-arrival combinations,elimination of outliers based on three-dimensional KDE,and refinement using a linear correction with an adaptive damping factor.We validate our method through lead-breaking experiments,demonstrating over a 23%improvement in positioning accuracy with a maximum error of 9.12 mm(relative error of 15.80%)—outperforming 4 existing methods.Simulations under various system errors,outlier scales,and ratios substantiate our method’s superior performance.Field blasting experiments also confirm the practical applicability,with an average positioning error of 11.71 m(relative error of 7.59%),compared to 23.56,66.09,16.95,and 28.52 m for other methods.This research is significant as it enhances the robustness of MS/AE source localization when confronted with data anomalies.It also provides a practical solution for real-world engineering and safety monitoring applications.

AMicroseismic Signal Denoising Algorithm Combining VMD and Wavelet Threshold Denoising Optimized by BWOA

The denoising of microseismic signals is a prerequisite for subsequent analysis and research.In this research,a new microseismic signal denoising algorithm called the Black Widow Optimization Algorithm(BWOA)optimized VariationalMode Decomposition(VMD)jointWavelet Threshold Denoising(WTD)algorithm(BVW)is proposed.The BVW algorithm integrates VMD and WTD,both of which are optimized by BWOA.Specifically,this algorithm utilizes VMD to decompose the microseismic signal to be denoised into several Band-Limited IntrinsicMode Functions(BLIMFs).Subsequently,these BLIMFs whose correlation coefficients with the microseismic signal to be denoised are higher than a threshold are selected as the effective mode functions,and the effective mode functions are denoised using WTD to filter out the residual low-and intermediate-frequency noise.Finally,the denoised microseismic signal is obtained through reconstruction.The ideal values of VMD parameters and WTD parameters are acquired by searching with BWOA to achieve the best VMD decomposition performance and solve the problem of relying on experience and requiring a large workload in the application of the WTD algorithm.The outcomes of simulated experiments indicate that this algorithm is capable of achieving good denoising performance under noise of different intensities,and the denoising performance is significantly better than the commonly used VMD and Empirical Mode Decomposition(EMD)algorithms.The BVW algorithm is more efficient in filtering noise,the waveform after denoising is smoother,the amplitude of the waveform is the closest to the original signal,and the signal-to-noise ratio(SNR)and the root mean square error after denoising are more satisfying.The case based on Fankou Lead-Zinc Mine shows that for microseismic signals with different intensities of noise monitored on-site,compared with VMD and EMD,the BVW algorithm ismore efficient in filtering noise,and the SNR after denoising is higher.

Predicting microseismic,acoustic emission and electromagnetic radiation data using neural networks

Microseism,acoustic emission and electromagnetic radiation(M-A-E)data are usually used for predicting rockburst hazards.However,it is a great challenge to realize the prediction of M-A-E data.In this study,with the aid of a deep learning algorithm,a new method for the prediction of M-A-E data is proposed.In this method,an M-A-E data prediction model is built based on a variety of neural networks after analyzing numerous M-A-E data,and then the M-A-E data can be predicted.The predicted results are highly correlated with the real data collected in the field.Through field verification,the deep learning-based prediction method of M-A-E data provides quantitative prediction data for rockburst monitoring.

Analysis of rockburst mechanism and warning based on microseismic moment tensors and dynamic Bayesian networks

One of the major factors inhibiting the construction of deep underground projects is the risk posed by rockbursts.A study was conducted on the access tunnel of the Shuangjiangkou hydropower station to determine the evolutionary mechanism of microfractures within the surrounding rock mass during rockburst development and develop a rockburst warning model.The study area was chosen through the combination of field studies with an analysis of the spatial and temporal distribution of microseismic(MS)events.The moment tensor inversion method was adopted to study rockburst mechanism,and a dynamic Bayesian network(DBN)was applied to investigating the sensitivity of MS source parameters for rockburst warnings.A MS multivariable rockburst warning model was proposed and validated using two case studies.The results indicate that fractures in the surrounding rock mass during the development of strain-structure rockbursts initially show shear failure and are then followed by tensile failure.The effectiveness of the DBN-based rockburst warning model was demonstrated using self-validation and K-fold cross-validation.Moment magnitude and source radius are the most sensitive factors based on an investigation of the influence on the parent and child nodes in the model,which can serve as important standards for rockburst warnings.The proposed rockburst warning model was found to be effective when applied to two actual projects.

Microseismic event waveform classification using CNN-based transfer learning models

The efficient processing of large amounts of data collected by the microseismic monitoring system(MMS),especially the rapid identification of microseismic events in explosions and noise,is essential for mine disaster prevention.Currently,this work is primarily performed by skilled technicians,which results in severe workloads and inefficiency.In this paper,CNN-based transfer learning combined with computer vision technology was used to achieve automatic recognition and classification of multichannel microseismic signal waveforms.First,data collected by MMS was generated into 6-channel original waveforms based on events.After that,sample data sets of microseismic events,blasts,drillings,and noises were established through manual identification.These datasets were split into training sets and test sets according to a certain proportion,and transfer learning was performed on AlexNet,GoogLeNet,and ResNet50 pre-training network models,respectively.After training and tuning,optimal models were retained and compared with support vector machine classification.Results show that transfer learning models perform well on different test sets.Overall,GoogLeNet performed best,with a recognition accuracy of 99.8%.Finally,the possible effects of the number of training sets and the imbalance of different types of sample data on the accuracy and effectiveness of classification models were discussed.

A 3D microseismic data-driven damage model for jointed rock mass under hydro-mechanical coupling conditions and its application

Rock mass is a fractured porous medium usually subjected to complex geostress and fluid pressure simultaneously.Moreover,the properties of rock mass change in time and space due to mining-induced fractures.Therefore,it is always challenging to accurately measure rock mass properties.In this study,a three-dimensional(3D)microseismic(MS)data-driven damage model for jointed rock mass under hydro-mechanical coupling conditions is proposed.It is a 3D finite element model that takes seepage,damage and stress field effects into account jointly.Multiple factors(i.e.joints,water and microseismicity)are used to optimize the rock mass mechanical parameters at different scales.The model is applied in Shirengou iron mine to study the damage evolution of rock mass and assess the crown pillar stability during the transition from open-pit to underground mining.It is found that the damage pattern is mostly controlled by the structure,water and rock mass parameters.The damage pattern is evidently different from the two-dimensional result and is more consistent with the field observations.This difference is caused by the MS-derived damage acting on the rock mass.MS data are responsible for gradually correcting the damage zone,changing the direction in which it expands,and promoting it to evolve close to reality.For the crown pillar,the proposed model yields a more trustworthy safety factor.In order to guarantee the stability of the pillar,it is suggested to take waterproof and reinforcement measures in areas with a high degree of damage.

GPU-acceleration 3D rotated-staggered-grid solutions to microseismic anisotropic wave equation with moment tensor implementation

To improve the accuracy of microseismic inversion,seismic anisotropy and moment tensor source should be carefully considered in the forward modelling stage.In this study,3D microseismic anisotropy wave forward modelling with a moment tensor source was proposed.The modelling was carried out based on a rotated-staggered-grid(RSG)scheme.In contrast to staggered-grids,the RSG scheme defines the velocity components and densities at the same grid,as do the stress components and elastic parameters.Therefore,the elastic moduli do not need to be interpolated.In addition,the detailed formulation and implementation of moment-tensor source loaded on the RSG was presented by equating the source to the stress increments.Meanwhile,the RSG-based 3D wave equation forward modelling was performed in parallel using compute unified device architecture(CUDA)programming on a graphics processing unit(GPU)to improve its efficiency.Numerical simulations including homogeneous and anisotropic models were carried out using the method proposed in this paper,and compared with other methods to prove the reliability of this method.Furthermore,the high efficiency of the proposed approach was evaluated.The results show that the computational efficiency of proposed method can be improved by about two orders of magnitude compared with traditional central processing unit(CPU)computing methods.It could not only help the analysis of microseismic full wavefield records,but also provide support for passive source inversion,including location and focal mechanism inversion,and velocities inversion.

Classification of mine blasts and microseismic events using starting-up features in seismograms

To find discriminating features in seismograms for the classification of mine seismic events,signal databases of blasts and microseismic events were established based on manual identification.Criteria including the repetition of waveforms,tail decreasing,dominant frequency and occurrence time of day were considered in the establishment of the databases.Signals from databases of different types were drawn into a unified coordinate system.It is noticed that the starting-up angles of the two types tend to be concentrated into two different intervals.However,it is difficult to calculate the starting-up angle directly due to the inaccuracy of the P-wave arrival＇s picking.The slope value of the starting-up trend line,which was obtained by linear regression,was proposed to substitute the angle.Two slope values associated with the coordinates of the first peak and the maximum peak were extracted as the characteristic parameters.A statistical model with correct discrimination rate of greater than 97.1% was established by applying the Fisher discriminant analysis.

Integrated application of 3D seismic and microseismic data in the development of tight gas reservoirs

The development of unconventional resources, such as shale gas and tight sane gas, requires the integration of multi-disciplinary knowledge to resolve many engineering problems in order to achieve economic production levels. The reservoir heterogeneit3 revealed by different data sets, such as 3D seismic and microseismic data, can more full3 reflect the reservoir properties and is helpful to optimize the drilling and completioT programs. First, we predict the local stress direction and open or close status of the natura fractures in tight sand reservoirs based on seismic curvature, an attribute that reveals reservoi heterogeneity and geomechanical properties. Meanwhile, the reservoir fracture network is predicted using an ant-tracking cube and the potential fracture barriers which can affec hydraulic fracture propagation are predicted by integrating the seismic curvature attribute anc ant-tracking cube. Second, we use this information, derived from 3D seismic data, to assis in designing the fracture program and adjusting stimulation parameters. Finally, we interpre the reason why sand plugs will occur during the stimulation process by the integration of 3E seismic interpretation and microseismic imaging results, which further explain the hydraulic fracure propagation controlling factors and open or closed state of natural fractures in tigh sand reservoirs.

Tomographic inversion for microseismic source parameters in mining

We propose a new method for inverting source function of microseismic event induced in mining. The observed data from microseismic monitoring during mining are represented by a wave equation in a spherical coordinate system and then the data are transformed from the time-space domain to the time-slowness domain based on tomographic principle, from whichwe can obtain the signals related to the source in the time-slowness domain. Through analyzing the relationship between the signal located at the maximum energy and the source function, we derive the tomographic equations to compute the source function from the signals and to calculate the effective radiated energy based on the source function. Moreover, we fit the real amplitude spectrum of the source function computed from the observed data into the co-2 model based on the least squares principle and determine the zero-frequency level spectrum and the corner frequency, finally, the source rupture radius of the event is calculated and The synthetic and field examples demonstrate that the proposed tomographic inversion methods are reliable and efficient

Fault detection based on microseismic events

In unconventional reservoirs, small faults allow the flow ofoil and gas as well as act as obstacles to exploration; for, （1） fracturing facilitates fluid migration, （2） reservoir flooding, and （3） triggering of small earthquakes. These small faults are not generally detected because of the low seismic resolution. However, such small faults are very active and release sufficient energy to initiate a large number of microseismic events （MEs） during hydraulic fracturing. In this study, we identified microfractures （MF） from hydraulic fracturing and natural small faults based on microseismicity characteristics, such as the time-space distribution, source mechanism, magnitude, amplitude, and frequency. First, I identified the mechanism of small faults and MF by reservoir stress analysis and calibrated the ME based on the microseismic magnitude. The dynamic characteristics （frequency and amplitude） of MEs triggered by natural faults and MF were analyzed; moreover, the geometry and activity types of natural fault and MF were grouped according to the source mechanism. Finally, the differences among time-space distribution, magnitude, source mechanism, amplitude, and frequency were used to differentiate natural faults and manmade fractures.

Discrimination of mining microseismic events and blasts using convolutional neural networks and original waveform

Microseismic monitoring system is one of the effective methods for deep mining geo-stress monitoring.The principle of microseismic monitoring system is to analyze the mechanical parameters contained in microseismic events for providing accurate information of rockmass.The accurate identification of microseismic events and blasts determines the timeliness and accuracy of early warning of microseismic monitoring technology.An image identification model based on Convolutional Neural Network(CNN)is established in this paper for the seismic waveforms of microseismic events and blasts.Firstly,the training set,test set,and validation set are collected,which are composed of 5250,1500,and 750 seismic waveforms of microseismic events and blasts,respectively.The classified data sets are preprocessed and input into the constructed CNN in CPU mode for training.Results show that the accuracies of microseismic events and blasts are 99.46%and 99.33%in the test set,respectively.The accuracies of microseismic events and blasts are 100%and 98.13%in the validation set,respectively.The proposed method gives superior performance when compared with existed methods.The accuracies of models using logistic regression and artificial neural network(ANN)based on the same data set are 54.43%and 67.9%in the test set,respectively.Then,the ROC curves of the three models are obtained and compared,which show that the CNN gives an absolute advantage in this classification model when the original seismic waveform are used in training the model.It not only decreases the influence of individual differences in experience,but also removes the errors induced by source and waveform parameters.It is proved that the established discriminant method improves the efficiency and accuracy of microseismic data processing for monitoring rock instability and seismicity.

Weighted-elastic-wave interferometric imaging of microseismic source location

Knowledge of the locations of seismic sources is critical for microseismic monitoring. Time-window-based elastic wave interferometric imaging and weighted- elastic-wave （WEW） interferometric imaging are proposed and used to locate modeled microseismic sources. The proposed method improves the precision and eliminates artifacts in location profiles. Numerical experiments based on a horizontally layered isotropic medium have shown that the method offers the following advantages： It can deal with Iow-SNR microseismic data with velocity perturbations as well as relatively sparse receivers and still maintain relatively high precision despite the errors in the velocity model. Furthermore, it is more efficient than conventional traveltime inversion methods because interferometric imaging does not require traveltime picking. Numerical results using a 2D fault model have also suggested that the weighted-elastic-wave interferometric imaging can locate multiple sources with higher location precision than the time-reverse imaging method.

Automatically positioning microseismic sources in mining by the stereo tomographic method using full wavefields

For microseisimic monitoring it is difficult to determine wave modes and their propagation velocity. In this paper, we propose a new method for automatically inverting in real time the source characteristics of microseismic events in mine engineering without wave mode identification and velocities. Based on the wave equation in a spherical coordinate system, we derive a tomographic imaging equation and formulate a scanning parameter selection criterion by which the microseisimic event maximum energy and corresponding parameters can be determined. By determining the maximum energy positions inside a given risk district, we can indentify microseismic events inside or outside the risk districts. The synthetic and field examples demonstrate that the proposed tomographic imaging method can automatically position microseismic events by only knowing the risk district dimensions and range of velocities without identifying the wavefield modes and accurate velocities. Therefore, the new method utilizes the full wavefields to automatically monitor microseismic events.

Preliminary engineering application of microseismic monitoring technique to rockburst prediction in tunneling of Jinping Ⅱ project

Monitoring and prediction of rockburst remain to be worldwide challenges in geotechnical engineering.In hydropower,transportation and other engineering fields in China,more deep,long and large tunnels have been under construction in recent years and underground caverns are more evidently featured by ＂long,large,deep and in group＂,which bring in many problems associated with rock mechanics problems at great depth,especially rockburst.Rockbursts lead to damages to not only underground structures and equipments but also personnel safety.It has been a major technical bottleneck in future deep underground engineering in China.In this paper,compared with earthquake prediction,the feasibility in principle of monitoring and prediction of rockbursts is discussed,considering the source zones,development cycle and scale.The authors think the feasibility of rockburst prediction can be understood in three aspects：（1） the heterogeneity of rock is the main reason for the existence of rockburst precursors;（2） deformation localization is the intrinsic cause of rockburst;and（3） the interaction between target rock mass and its surrounding rock mass is the external cause of rockburst.As an engineering practice,the application of microseismic monitoring techniques during tunnel construction of Jinping II Hydropower Station was reported.It is found that precursory microcracking exists prior to most rockbursts,which could be captured by the microseismic monitoring system.The stress concentration is evident near structural discontinuities（such as faults or joints）,which shall be the focus of rockburst monitoring.It is concluded that,by integrating the microseismic monitoring and the rock failure process simulation,the feasibility of rockburst prediction is expected to be enhanced.

Developments and prospects of microseismic monitoring technology in underground metal mines in China

Microseismic monitoring technology has become an important technique to assess stability of rock mass in metal mines.Due to the special characteristics of underground metal mines in China,including the high tectonic stress,irregular shape and existence of ore body,and complex mining methods,the application of microseismic technology is more diverse in China compared to other countries,and is more challenging than in other underground structures such as tunnels,hydropower stations and coal mines.Apart from assessing rock mass stability and ground pressure hazards induced by mining process,blasting,water inrush and large scale goaf,microseismic technology is also used to monitor illegal mining,and track personnel location during rescue work.Moreover,microseismic data have been used to optimize mining parameters in some metal mines.The technology is increasingly used to investigate cracking mechanism in the design of rock mass supports.In this paper,the application,research development and related achievements of microseismic technology in underground metal mines in China are summarized.By considering underground mines from the perspective of informatization,automation and intelligentization,future studies should focus on intelligent microseismic data processing method,e.g.,signal identification of microseismic and precise location algorithm,and on the research and development of microseismic equipment.In addition,integrated monitoring and collaborative analysis for rock mass response caused by mining disturbance will have good prospects for future development.

Comprehensive early warning of rock burst utilizing microseismic multi-parameter indices

Rock bursts have become one of the most severe risks in underground coal mining and its early warning is an important component in the safety management. Microseismic(MS) monitoring is considered potentially as a powerful tool for the early warning of rock burst. In this study, an MS multi-parameter index system was established and the critical values of each index were estimated based on the normalized multi-information warning model of coal-rock dynamic failure. This index system includes bursting strain energy(BSE) index, time-space-magnitude independent information(TSMII) indices and timespace-magnitude compound information(TSMCI) indices. On the basis of this multi-parameter index system, a comprehensive analysis was conducted via introducing the R-value scoring method to calculate the weights of each index. To calibrate the multi-parameter index system and the associated comprehensive analysis, the weights of each index were first confirmed using historical MS data occurred in LW402102 of Hujiahe Coal Mine(China) over a period of four months. This calibrated comprehensive analysis of MS multi-parameter index system was then applied to pre-warn the occurrence of a subsequent rock burst incident in LW 402103. The results demonstrate that this multi-parameter index system combined with the comprehensive analysis are capable of quantitatively pre-warning rock burst risk.

Characteristics of microseismic b-value associated with rock mass large deformation in underground powerhouse caverns at different stress levels

Rock mass large deformation in underground powerhouse caverns has been a severe hazard in hydropower engineering in Southwest China.During the development of rock mass large deformation,a sequence of fractures was generated that can be monitored using microseismic(MS)monitoring techniques.Two MS monitoring systems were established in two typical underground powerhouse caverns featuring distinct geostress levels.The MS b-values associated with rock mass large deformation and their temporal variation are analysed.The results showed that the MS bvalue in course of rock mass deformation was less than 1.0 in the underground powerhouse caverns at a high stress level while larger than 1.5 at a low stress level.Prior to the rock mass deformation,the MS b-values derived from both the high-stress and low-stress underground powerhouse caverns show an incremental decrease over 10%within 10 d.The results contribute to understanding the fracturing characteristics of MS sources associated with rock mass large deformation and provide a reference for early warning of rock mass large deformation in underground powerhouse caverns.