摘要
矿山微震监测系统中所产生的微震信号数量巨大并且存在多种复杂背景噪声干扰,使得矿山微震事件的识别难度很高。现有的微震事件识别方法仍然存在降噪效率低、时延明显、精度差等问题。为了提高微震事件识别的准确度,提出一种基于改进小波分解和极限学习机(ELM)的矿山微震事件识别方法,该方法能更有效、更准确地识别矿山微震事件。针对微震信号具有不可预测、复杂、扩散等特性,提出一种改进阈值函数的小波降噪方法,其中在小波分解过程中,首先确定小波阈值和小波分解层数,再利用提出的改进的小波阈值函数对小波系数进行阈值量化处理,得到优化后的小波系数,最后对小波系数进行重构得到去噪的信号。该方法有效的改进了目前软、硬阈值函数所存在的伪吉布斯现象和不连续、误差大的缺陷。其次,提取去噪后微震信号特征并训练ELM隐藏层节点数量,并利用训练得到的ELM隐藏层节点数量构建改进的ELM,改进的ELM解决了普通ELM训练数据时无法有效选取隐藏层节点数量的问题,从而提升了微震事件识别精度。最后,通过改进后的ELM能够对矿山微震事件进行更加有效的识别。结果表明:本文基于改进小波分解和ELM的矿山微震事件识别方法的分类准确率达到91.1%,验证了本文方法的有效性和准确性,并且该方法可以通过新增微震信号数据进一步提高识别精度。
The widely used mine micro-seismic monitoring system generates a large number of micro-seismic signals and has a variety of complex background noise disturbances,making it difficult to identify mine micros-eismic events.However,the existing identification methods have problems such as low noise reduction efficiency,obvious delay,and poor accuracy.In the process of wavelet decomposition,the wavelet threshold and wavelet decomposition level are determined firstly,and then the wavelet coefficients are quantized by using the improved wavelet threshold function to get the optimized wavelet coefficients.Finally,the wavelet coefficients are reconstructed to get the de-noising signal.This method effectively improves the current soft and hard threshold function of the existence of pseudo Gibbs phenomenon and discontinuity,large error defects.Secondly,the characteristics of the de-noised micro-seismic signal are extracted and the number of nodes in the ELM hidden layer is trained,and the improved ELM solves the problem that the number of nodes in the hidden layer could not be correctly selected in the training data using traditional ELM,which improves the identification accuracy of micro-seismic events.Finally,the improved ultimate learning machine can identify mine micro-seismic events more effectively.The results show that the classification accuracy of the mine microseismic event recognition method based on improved wavelet decomposition and ELM is 91.1%,which verifies the effectiveness and accuracy of the proposed method,and the method can further improve the identification accuracy by adding micro-seismic signal data.
作者
陈泽
丁琳琳
罗浩
宋宝燕
张明
潘一山
CHEN Ze;DING Linlin;LUO Hao;SONG Baoyan;ZHANG Ming;PAN Yishan(Xinwen Mining Group Co.,Ltd.,Xintai 271200,China;School of Information,Liaoning University,Shenyang 110036,China;School of Resources and Civil Engineering,Northeastern University,Shenyang 110819,China;School of Environment,Liaoning University,Shenyang 110036,China)
出处
《煤炭学报》
EI
CAS
CSCD
北大核心
2020年第S02期637-648,共12页
Journal of China Coal Society
基金
国家自然科学基金资助项目(62072220)
中国博士后基金面上资助项目(2020M672134)
辽宁省教育厅科学研究资助项目(LJC201913)
