A Method for Rockburst Prediction in the Deep Tunnels of Hydropower Stations Based on the Monitored Microseismicity and an Optimized Probabilistic Neural Network Model
Abstract:Hydropower is one of the most important renewable energy sources. However, the safe construction of hydropower stations is seriously affected by disasters like rockburst, which, in turn, restricts the sustainable development of hydropower energy. In this paper, a method for rockburst prediction in the deep tunnels of hydropower stations based on the use of real-time microseismic (MS) monitoring information and an optimized probabilistic neural network (PNN) model is proposed. The model consists of the mean imp… Show more
“…That is: the sensitivity, accuracy and specificity was 98%, 88% and 87%, respectively. This study not only increased the sensitivity of rockburst risk prediction from 87% [41] to 98%, but also achieved 88% accuracy and 87% specificity, which was not achieved in previous studies. Such improvements may attribute to more features extracted from multi-domains and utilizing MS energy as well as raw wave data.…”
Section: Resultscontrasting
confidence: 52%
“…MS waves before the rockburst were investigated, suggesting that characteristics of the wave's velocity, amplitude, and frequency as well as rock stress state can be used for the rockburst warning [28]- [36]. The relationship between MS waves and the rockburst was analysed, indicating the features extracted from MS waves together with rock stress can be utilized to predict rockburst [25], [37]- [41].…”
As a severe hazard in coal mining and rock excavation, the rockburst is usually induced by the high energy tremor. Microseismic (MS) monitoring is suggested to forecast the rockburst risk to reduce its damage. The paper aims to investigate contribution of multi-MS data, including MS raw wave data and MS energy data, to prediction of the high energy tremor, using support vector machine (SVM) together with genetic algorithm (GA). MS monitoring data recorded for more than 400 days at Wudong coal mine of Xinjiang, China, were used in the paper. 132 and 24 features are initially extracted from MS raw wave and energy data in the frequency domain, entropy and time-frequency domain, respectively. GA is not only used to select effective ones among initially extracted features, but also optimize hyperparameters for SVM to classify high energy tremors from general MS events. The performances of the proposed approach based on multi-MS data are evaluated by cross-validation. The results show that the classifier achieves 98% sensitivity, 88% accuracy and 87% specificity using both MS raw wave and energy data, which is better than solely utilizing MS raw wave (98% sensitivity, 84% accuracy and 83% specificity) or energy data (98% sensitivity, 86% accuracy and 85% specificity). These findings suggest that MS raw wave data makes important contribution to rockburst risk prediction as well as MS energy data, and the better performance can be achieved when utilizing two kinds of data simultaneously. INDEX TERMS Rockburst risk prediction, microseismic monitoring, microseismic raw wave data, support vector machine, genetic algorithm.
“…That is: the sensitivity, accuracy and specificity was 98%, 88% and 87%, respectively. This study not only increased the sensitivity of rockburst risk prediction from 87% [41] to 98%, but also achieved 88% accuracy and 87% specificity, which was not achieved in previous studies. Such improvements may attribute to more features extracted from multi-domains and utilizing MS energy as well as raw wave data.…”
Section: Resultscontrasting
confidence: 52%
“…MS waves before the rockburst were investigated, suggesting that characteristics of the wave's velocity, amplitude, and frequency as well as rock stress state can be used for the rockburst warning [28]- [36]. The relationship between MS waves and the rockburst was analysed, indicating the features extracted from MS waves together with rock stress can be utilized to predict rockburst [25], [37]- [41].…”
As a severe hazard in coal mining and rock excavation, the rockburst is usually induced by the high energy tremor. Microseismic (MS) monitoring is suggested to forecast the rockburst risk to reduce its damage. The paper aims to investigate contribution of multi-MS data, including MS raw wave data and MS energy data, to prediction of the high energy tremor, using support vector machine (SVM) together with genetic algorithm (GA). MS monitoring data recorded for more than 400 days at Wudong coal mine of Xinjiang, China, were used in the paper. 132 and 24 features are initially extracted from MS raw wave and energy data in the frequency domain, entropy and time-frequency domain, respectively. GA is not only used to select effective ones among initially extracted features, but also optimize hyperparameters for SVM to classify high energy tremors from general MS events. The performances of the proposed approach based on multi-MS data are evaluated by cross-validation. The results show that the classifier achieves 98% sensitivity, 88% accuracy and 87% specificity using both MS raw wave and energy data, which is better than solely utilizing MS raw wave (98% sensitivity, 84% accuracy and 83% specificity) or energy data (98% sensitivity, 86% accuracy and 85% specificity). These findings suggest that MS raw wave data makes important contribution to rockburst risk prediction as well as MS energy data, and the better performance can be achieved when utilizing two kinds of data simultaneously. INDEX TERMS Rockburst risk prediction, microseismic monitoring, microseismic raw wave data, support vector machine, genetic algorithm.
“…Regarding the use of ANN in rockburst and flying rock generated by blasting, the in-situ rockburst database is analyzed by ANNs, SVM, and other two different data mining techniques [14]. Based on the PNN model, Feng et al predict rockburst in the deep tunnels [71]. The flyrock distance generated by blasting is predicted by three hybrid ANN models, including ICA-ANN, GA-ANN, and PSO-ANN [41].…”
Due to the lack of living space and the increase in population, there has been a construction boom in the underground space to improve the quality of human life. Tunnel engineering plays a vital role in the development of underground space. In addition to traditional methods, some intelligent methods such as artificial neural networks (ANNs) have been applied to various problems in the tunnel domain in recent years. This paper systematically reviews the application of ANNs from different aspects of tunnel engineering. It reveals that the backpropagation algorithm (BPA) and Levenberg-Marquardt algorithm (LMA) are the most widely used. Due to the limitations of some original models, some scholars use optimization algorithms such as particle swarm optimization (PSO) and genetic algorithm (GA) to optimize the original ANNs to obtain better prediction results. A comparison between the ANN-based methods and methods like statistical methods is conducted. Finally, the following conclusions can be drawn: (1) The recommended ratio of the training set and test set is 3:1; (2) The advantage of optimized ANNs is not apparent when the optimization algorithm varies. Additionally, the performance of ANNs is always better than that of statistical methods.
“…Seismicity related to underground mining has been observed since the beginning of the 20th century [42]. Seismic activities are closely related to the safety of mining operations, and rock bursts are often the main cause of mining accidents [43,44]. Passive seismic/microseismic monitoring can effectively detect and evaluate the seismic activity surrounding underground mines and provide early warning of potential geological risks [3].…”
Seismic source location specifies the spatial and temporal coordinates of seismic sources and lays the foundation for advanced seismic monitoring at all scales. In this work, we firstly introduce the principles of diffraction stacking (DS) and cross-correlation stacking (CCS) for seismic location. The DS method utilizes the travel time from the source to receivers, while the CCS method considers the differential travel time from pairwise receivers to the source. Then, applications with three field datasets ranging from small-scale microseismicity to regional-scale induced seismicity are presented to investigate the feasibility, imaging resolution, and location reliability of the two stacking operators. Both of the two methods can focus the source energy by stacking the waveforms of the selected events. Multiscale examples demonstrate that the imaging resolution is not only determined by the inherent property of the stacking operator but also highly dependent on the acquisition geometry. By comparing to location results from other methods, we show that the location bias is consistent with the scale size, as well as the frequency contents of the seismograms and grid spacing values.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.