基于IITD和PNN的广域电磁法数据信噪分离方法  

WFEM signal-noise separation method based on improved inherent time-scale decomposition and probabilistic neural network

作　　者：张贤 李帝铨[2] 胡艳芳[2] 朱云起 李富 ZHANG Xian;LI DiQuan;HU YanFang;ZHU YunQi;LI Fu(Hunan Provincial Key Laboratory of Finance and Economics Big Data Science and Technology,School of Information Technology and Management,Hunan University of Finance and Economics,Changsha 410205,China;School of Geosciences and Info-Physics,Central South University,Changsha 410083,China)

机构地区：[1]湖南财政经济学院,财经大数据科学与技术湖南省重点实验室,信息技术与管理学院,长沙410205 [2]中南大学地球科学与信息物理学院,长沙410083

出　　处：《地球物理学进展》2024年第1期241-252,共12页Progress in Geophysics

基　　金：国家自然科学基金项目“南方海相页岩气电磁法识别与预测方法研究”(41874081);有色金属成矿预测与地质环境监测教育部重点实验室(中南大学)开放基金项目(2023YSJS07);湖南省教育厅科学研究项目(22A0457)联合资助。

摘　　要：矿产资源开发需求加速了广域电磁法(Wide Field Electromagnetic Method,WFEM)的发展与应用,但强电磁干扰严重降低了原始数据质量与探测效果.为此,本文提出基于改进固有时间尺度分解(Improved Inherent Time-scale Decomposition,IITD)和概率神经网络(Probabilistic Neural Network,PNN)的广域电磁法数据信噪分离方法.首先,通过改进固有时间尺度分解算法,提高信号分解精度,消除原始数据中存在的趋势噪声;然后,构建WFEM数据样本库,提取最大值、峰值因子、脉冲因子、裕度因子等多个时域特征,结合概率神经网络进行信噪辨识;最后,将识别为有效信号的部分按原采样顺序进行整合与重构,实现WEFM数据去噪;通过数值模拟与实测分析,结果表明,趋势噪声和异常波形均能被有效辨识及剔除,处理后的电场曲线形态趋于平滑稳定,原始数据质量得到提升.The exploitation of mineral resources accelerates the development and application of the Wide Field ElectromagneticMethod(WFEM),butthestrong electromagnetic interference seriously degrades the quality of the raw data and the detection effect.In the paper,the Improved Inherent Time-scale Decomposition(ilTD)and Probabilistic Neural Network(PNN)is proposed for WFEM data signal-noise separation.First,the ITD is improved to improve the signal decomposition accuracy and eliminate the trend noise in the raw data.Then,the WFEM data sample library was constructed,multiple time-domainfeatures(maximum value,peak factor,pulse factor and margin factor)were extracted,and the signal-noise identification was applied by using PNN algorithm.Finally,the identified as effective signals are integrated and reconstructed according to the original sampling sequence to realize the WEFM data denoising,Through numerical simulation and measured analysis,the results show that the trend noise and abnormal waveform can be effectively identified and eliminated,the shape of the electric field curve after processing that tends to be smooth and stable,and the quality of raw data is improved.

关 键 词：广域电磁法 改进固有时间尺度分解 概率神经网络 信噪分离 

分 类 号：P631[天文地球—地质矿产勘探]