{"title":"Predicting microseismic sensitive feature data using variational mode decomposition and transformer","authors":"Xingli Zhang, Duanduan Hou, Qian Mao, Zhihui Wang","doi":"10.1007/s10950-024-10193-9","DOIUrl":null,"url":null,"abstract":"<div><p>Rock burst is one of the major disasters that endanger coal safety production. If a rock burst occurs, it will cause terrible casualties and significant property losses. Therefore, this article proposes to predict the sensitive characteristics of microseisms, which can achieve the prediction and early warning of rock burst disasters to a certain extent. To effectively improve the prediction accuracy and robustness of microseismic sensitive feature data, a hybrid model called VMD-Transformer is suggested in this study for predicting time series of microseismic sensitive features. This model is based on the variational mode decomposition (VMD) and transformer model and aims to predict future eigenvalue from the historical data of sensitive features. To a certain extent, the transformer model is used to predict the future eigenvalue, while the VMD is used to extract the features of the time series data at various frequency domain scales, which solves the problem of non-stationary time series data being difficult to predict accurately due to high fluctuations. This study extracts sensitive features from microseismic events that the same source registered by a certain geophone after locating, decomposes the time series of the sensitive features using VMD, predicts each component of the decomposition separately using the transformer model, and then combines the component prediction results to produce the final prediction results. Experimental results indicate that our method has the features of a simple algorithm, strong adaptivity, and high prediction accuracy and can effectively predict time series of sensitive features extracted from microseismic signals.</p></div>","PeriodicalId":16994,"journal":{"name":"Journal of Seismology","volume":"28 1","pages":"229 - 250"},"PeriodicalIF":1.6000,"publicationDate":"2024-01-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Seismology","FirstCategoryId":"89","ListUrlMain":"https://link.springer.com/article/10.1007/s10950-024-10193-9","RegionNum":4,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"GEOCHEMISTRY & GEOPHYSICS","Score":null,"Total":0}
引用次数: 0
Abstract
Rock burst is one of the major disasters that endanger coal safety production. If a rock burst occurs, it will cause terrible casualties and significant property losses. Therefore, this article proposes to predict the sensitive characteristics of microseisms, which can achieve the prediction and early warning of rock burst disasters to a certain extent. To effectively improve the prediction accuracy and robustness of microseismic sensitive feature data, a hybrid model called VMD-Transformer is suggested in this study for predicting time series of microseismic sensitive features. This model is based on the variational mode decomposition (VMD) and transformer model and aims to predict future eigenvalue from the historical data of sensitive features. To a certain extent, the transformer model is used to predict the future eigenvalue, while the VMD is used to extract the features of the time series data at various frequency domain scales, which solves the problem of non-stationary time series data being difficult to predict accurately due to high fluctuations. This study extracts sensitive features from microseismic events that the same source registered by a certain geophone after locating, decomposes the time series of the sensitive features using VMD, predicts each component of the decomposition separately using the transformer model, and then combines the component prediction results to produce the final prediction results. Experimental results indicate that our method has the features of a simple algorithm, strong adaptivity, and high prediction accuracy and can effectively predict time series of sensitive features extracted from microseismic signals.
期刊介绍:
Journal of Seismology is an international journal specialising in all observational and theoretical aspects related to earthquake occurrence.
Research topics may cover: seismotectonics, seismicity, historical seismicity, seismic source physics, strong ground motion studies, seismic hazard or risk, engineering seismology, physics of fault systems, triggered and induced seismicity, mining seismology, volcano seismology, earthquake prediction, structural investigations ranging from local to regional and global studies with a particular focus on passive experiments.