| Electromagnetic method is one of the main technical means of geophysical prospecting.Due to the frequent human activities,the electromagnetic interference is increasingly serious,which leads to the low signal-to-noise ratio of the data collected in the field,and the detection effect is less and less ideal.The traditional electromagnetic signal processing usually involves time domain denoising,frequency domain filtering,anomaly screening,etc.,which is difficult to meet the high fidelity denoising effect.With the development and application of feature extraction,intelligent optimization and machine learning,the qualitative analysis of electromagnetic method data characteristics and the use of signal-noise identification are important technical means to improve the accuracy of signal-noise separation.It provides a new idea for further studying the data processing effect of electromagnetic method.Based on feature clustering and optimization learning method,this paper carries out adaptive and high-precision signal-noise identification and separation for electromagnetic method data.The main research contents are as follows:Firstly,the paper analyzes the electromagnetic basic theory and the characteristics of signal-noise.Secondly,the signal and noise types in natural source and controllable source electromagnetic data are studied.Thirdly,the influence range and effect of electromagnetic noise are described in detail.Finally,the multivariate performance evaluation parameters and the measured data evaluation index are used to measure the effectiveness of the method.Secondly,aiming at the problem that traditional Magnetotelluric(MT)denoising methods in full time sequence.Combined with the characteristic parameters of MT data and Fuzzy C-Means clustering(FCM),a high-precision signal-noise separation algorithm is used for targeted denoising.The experimental results show that the remote reference method and the full time sequence denoising method are compared,the proposed method can achieve more accurate MT noise suppression,and its denoising effect,efficiency and accuracy are effectively improved.Thirdly,In order to improve the problem that variational mode decomposition(VMD)can not get key parameters adaptively in MT data analysis,this study uses the envelope entropy function as the fitness function of grey wolf optimizer(GWO)algorithm.By iteratively searching for the optimal VMD parameters,the MT data can be adaptively decomposed with high precision.In addition,the scale index in detrended fuctuation analysis(DFA)algorithm is used to determine the corresponding intrinsic mode function(IMF)component for superposition and reconstruct MT useful data.Experimental results show that the method can suppress the charge-discharge triangle noise and irregular noise adaptively.Fourthly,The wide field electromagnetic method(WFEM)usually uses time-frequency filtering,screening or eliminating abnormal frequency points in signal processing,which is difficult to meet the needs of targeted denoising.Multi-domain features,clustering and optimization learning methods are used to identify noise and eliminate abnormal waveform.Through digital coherent technique to reconstruct WFEM data extraction of available frequency spectrum amplitude,calculating electric field curve and the apparent resistivity curve.The experimental results show that not only the multi-domain and multi-feature parameters can comprehensively and precisely characterize WFEM signal and noise,but also the machine learning method and its optimization algorithm can effectively improve the signal-noise identification effect and efficiency of WFEM data.Finally,Summarize the main research results of this paper,discuss the application of feature clustering and optimization learning in electromagnetic signal-noise separation and the existing shortcomings,and look forward to the next research plan.Figure 66,table 7,reference 201... |
PDF Full Text Request