| In recent years,with the rapid development of science and technology,because Our Country's urbanization process is accelerating and the near surface resources are gradually decreasing,the development of urban underground space is bound to become the main trend of development in the future.Underground tunnel construction and pipeline laying are inseparable from the rational planning of underground space,but the unreasonable development and utilization of underground space may also cause serious geological disasters.Therefore,using geophysical technology to detect urban underground geological information is very important.Transient electromagnetic method is a geophysical exploration method based on the law of electromagnetic induction.However,due to the high-amplitude electromagnetic noise such as power frequency noise and random noise produced by dense power lines and complex traffic in the city,the quality of transient electromagnetic data is seriously reduced,which leads to the deterioration of exploration effect.Most of the existing transient electromagnetic noise suppression technologies are aimed at the field data outside the city,which can't meet the needs of strong noise suppression in urban environment.Therefore,an effective urban transient electromagnetic noise suppression method is of great significance.Relying on the national major scientific instrument development project "Array Focusing TFSAR EM Imaging System for Urban Underground Space Detection",Aiming at the data quality problem caused by urban transient electromagnetic noise,the research is based on Wavelet-Nyman,Gaiser and Saucier estimator(WaveletNGSE),Minimum Noise Fraction(MNF)and Convolutional Neural Networks-Gated Recurrent Unit(CNN-GRU)method for urban transient electromagnetic noise suppression.The main research contents and main research results of the paper are as follows:(1)The theory of transient electromagnetic method,urban transient electromagnetic data and urban electromagnetic noise are studied.It is concluded that urban transient electromagnetic data has low signal-to-noise ratio and serious distortion of late signal;By analyzing the data of urban electromagnetic noise,it is found that the noise mainly includes two main components: strong power harmonic noise and random noise;The suppression strategy of urban electromagnetic noise is determined: first deal with power harmonic noise,and then deal with random noise.Firstly,the WaveletNGSE method is used for frequency correction,and the measured real frequency is50.32 Hz.Then the bipolar cancellation algorithm is used for power frequency suppression.The signal-to-noise ratio before and after processing is increased from-37.46 d B to-2.07 d B,which preliminarily improves the data quality and suppresses the power harmonic components in urban noise;(2)The power harmonic suppression method based on Wavelet-NGSE frequency estimation is studied,which solves the processing error of the traditional bipolar cancellation method in the case of power harmonic fundamental frequency fluctuation.We use simulation experiments to verify the fundamental frequency correction accuracy of the algorithm,then a group of urban collected noise is used for power harmonic noise suppression test.(3)The random noise suppression method based on MNF is studied to alleviate the problem of residual random noise in the transient electromagnetic data after power harmonic suppression.Firstly,we test the suppression effect of four simulated random noises with different signal-to-noise ratios,and the signal-to-noise ratio is increased by about 10 ? 20 d B after data processing.Then,a group of urban random noise collected after power frequency preprocessing is used for testing and the signal-to-noise ratio before and after data processing is increased from-14.3 DB to 0.4 d B.In general,the random noise is preliminarily suppressed,but the suppression effect needs to be improved,especially when the noise amplitude is high.(4)The suppression method of transient electromagnetic random noise based on CNN-GRU method is studied.Based on the traditional MNF method,the deep learning method is added to suppress the random noise for the second time to further improve the suppression effect of urban random noise;Firstly,we test the simulated random noise with four different amplitude,and the processed signal-to-noise ratio is improved by about 10 ? 40 d B.In addition,we compare the performance with the Secondary Field Signal Denoising Stacked Autoencoders(SFSDSA)and the Long Short-Term Memory Autoencoder(LSTM-Autoencoder)respectively.Finally,a group of urban collected random noise after power harmonic preprocessing is used for testing,and the signal-to-noise ratio before and after processing is improved from-14.3 DB to 23.5 d B.In general,MNF method and CNN-GRU model jointly suppress random noise,improve the quality of transient electromagnetic data,and the processing effect is better than SFSDSA Method and LSTM-Autoencoder method.Finally,we test the whole line data of urban transient electromagnetic,and verify the effectiveness of the urban noise suppression strategy proposed in this paper through channel extraction analysis,which provides a basis for data inversion and interpretation in the future. |
PDF Full Text Request