Research On Noise Suppression And Inversion Methods For Urban Towed Transient Electromagnetic Survey

With the rapid development of urbanization in China,the near-surface land resources are becoming increasingly scarce,and the development of urban underground space has become an inevitable trend.In the process of urban underground space development and construction,geological disasters such as road collapse caused by underground cavities seriously threaten people’s life and property security.Therefore,it is crucial to adopt an efficient and accurate surveying method to achieve "CT" for urban underground space and timely warn road collapse disasters.Transient electromagnetic instruments equipped with towed platform have become a research hotspot in recent years due to the advantages in efficient,non-destructive,and deep sounding survey of urban road movement.However,there are various types of electromagnetic noise in urban environments,with strong energy and extremely severe interference,which causes low signal-to-noise ratio of towed transient electromagnetic data and makes it difficult for the instrument to work reliably.In addition,as one of the main detection targets,potential cavities have high resistivity characteristics.How to accurately image the high resistivity targets is also an important challenge for towed transient electromagnetic methods.This study relies on the National Major Scientific Research Instrument Development Project of the National Natural Science Foundation of China to carry out relevant research.Combining the with the characteristics of towed transient electromagnetic mobility scanning survey,this study respectively has proposed a Power Pine Harmonics Adaptive Cancellation Method,a Random Noise Suppression Method Based On Minimum Noise Fraction and Deep Learning,and a Quasi-Occam Geo-GMP(Geophysics-Genetic Marine Predators)Inversion Algorithm for Urban Underground Space Targeted Imaging.On the basis of the high quality data after noise suppression,the geoelectric model of urban underground space is accurately inverted,and the high resistance anomalies are reliably identified.The main research contents and achievements of this paper are as follows:(1)The characteristics of electromagnetic noise in urban environment have been analyzed.According to multiple electromagnetic noise data samples collected by urban towed mode,the Intrinsic Mode Function between multiple noise samples is analyzed in time domain and power spectrum by data processing technology.It was determined that the power line harmonics has the characteristics of large amplitude,strong periodicity and fundamental frequency volatility,which is the main factor leading to the low signal-to-noise ratio of urban towed transient electromagnetic data.It is determined that random noise has the characteristics of complex composition and wide frequency band,and it contaminates the transient electromagnetic late time signal,which is difficult to be effectively suppressed by the conventional single processing method.In particular,the regularity of noise components is analyzed by using Pearson correlation coefficient,which reveals the law of strong correlation of each Intrinsic Mode Function of noise in the same area of a city.(2)Aiming at the problem that the power line harmonics contaminated the transient electromagnetic all-time signal in the process of urban towed measurement,resulting in low signal-to-noise ratio of the raw data and difficulty in reliable operation of the instrument,a Power Pine Harmonics Adaptive Cancellation Method suitable for mobile survey mode was proposed.This method uses Improved Nonlinear Adaptive Genetic Algorithm Based on Neighborhood Optimization to automatically search the most relevant components of the raw data and pure noise samples,and uses Block Subtraction method to suppress power line harmonics.This method effectively solves the problem of insufficient amount of data for stacking-averaging to denoise under towed measurement conditions.The performance and effectiveness of the proposed method are verified by power line harmonics single baseband noise suppression test,power line harmonics volatility baseband noise suppression test and stacking-averaging method comparison test.The signal-to-noise ratio of transient electromagnetic data is greatly improved by the proposed method.(3)Aiming at the problem that random noise contaminates transient electromagnetic late time signals in the process of urban towed measurement,resulting in late time data being submerged and low signal-to-noise ratio,a Random Noise Suppression Method Based on Minimum Noise Fraction and Deep Learning is proposed.This method firstly suppressed part of the random noise through Minimum Noise Fraction,and further designed the GRU-FCNN(Gated Recurrent Unit-Fully Connected Neural Network)deep neural networks to suppress the remaining random noise.The neural network model uses the methods of double loss function,cosine annealing learning rate and double regularization parameter to enhance the training effect,so as to accurately extract the transient electromagnetic secondary field signal.Compared with conventional noise suppression methods and conventional deep neural networks,the test results show that the Random Noise Suppression Method Based on Minimum Noise Fraction and Deep Learning can effectively improve the late time signal-to-noise ratio of towed transient electromagnetic data,and has better random noise suppression effect.(4)Aiming at the problem that conventional inversion algorithms were difficult to image potential cavities accurately and reliably,a Quasi-Occam Geo-GMP Inversion Algorithm for Urban Underground Space Targeted Imaging was proposed.The algorithm combined the advanced ideas from other meta-heuristic algorithms,while adding the relevant constraints of geophysics(Geo)electromagnetic law theory.The inversion algorithm is guided to approach the global optimal solution of the geoelectric model,which effectively improves the global search ability and improves the imaging effect of high-resistivity anomalies.The comprehensive performance of the proposed algorithm is verified through the comparative test of the performance and noise resistance of Geo-GMP algorithm.The inversion algorithm of urban underground space targeted imaging realizes the accurate identification of potential underground cavities targets,so as to achieve the purpose of early warning of geological disasters.In summary,the research content of this study improves the survey performance of towed transient electromagnetic from two aspects: electromagnetic data noise suppression processing and inversion imaging,solving the challenges faced by the application of towed transient electromagnetic in urban environment,and providing important technical support for efficient and accurate detection of potential cavities in urban strong electromagnetic noise environment.