Research on Random Noise Attenuation Algorithm and Dual-Redundancy Receiving System for Magnetic Resonance Sounding

Author:Make public

Supervisor:Lin Tingting


Degree Year:2018





Facing the huge demand of groundwater investigation and the urgency of water inrush in underground engineering,such as mine and tunnel,an effective method for groundwater exploration is significantly needed to develope.Magnetic resonance sounding(MRS)has been a modern geophysical method used to detect water due to its nondestructive in groundwater detecting.This method has advantages of abundant information and has the ability to determine water content and rock properties,so it has been widely applied in hydrogeological investigation and disaster water exploration.However,the environmental electromagnetic noise cannot be neglected as the signal is notably weak and the Earth’s magnetic field cannot be shielded.The traditional instruments and de-noising methods cannot meet the requirements.Besides,they have problems such as low detecting efficiency,poor reliability and difficult to obtain signal.Hence,in order to improve the measurement efficiency and reliability,develping the methods and system is of great significance and practical values.The noise disturbing MRS signal mainly includes spiky noise,power-line harmonic noise and random noise.The recent studies suggest that the characteristic noise like power-line harmonic noise and spiky noise can be methodologically cancelled by model-based noise reduction approaches etc.However,the random noise contributed by a variety of unpredictable factors do not have a specific mathematical expression and difficult to be separated from the desired MRS signal.At present,the widely used method is stacking,that is averaging the multiple recordings.But this method can only suppress partial random noise which depends on the number of stacks,it is time-consuming.Besides,there has not appears more effective method for random noise attenuation of MRS signal.To solve the above problems,this paper is supported by the China NSFC under Grant 41722405 and the National Key Foundation for Exploring Scientific Instrument under Grant 2011YQ030133,and focuses on the research on random noise attenuation algorithm and dual-redundancy receiving system.Joint detection of MRS envelope signal and MRS oscillating signal is adopted and according to the characteristics of the signals acquired by different paths.This paper has been promoted including time-frequency analysis,error of instantaneous frequency(IF)estimation,developemt of dual-redundancy receiving system,numerical simulations and field cases,etc.The main research contents are as follows:1.In terms of the problem that random noise disturbs seriously,low efficiency and limited denoising effect of traditional method in random noise attenuation,time-frequency peak filtering(TFPF)method is proposed according to the analysis in time-frequency domain and the principle of IF estimation,which provides new ideas for random nosie attenuation of MRS signals.Meanwhile,the desired MRS signal can be recovered from a single recording by TFPF method.The random noise is effectively suppressed and the measurement efficiency is significantly improved.2.In terms of the error of filtering results caused by TFPF method,the formula of IF estimated error is derived according to the property of Wigner-Ville distribution.The error sources are analysed and the main factor is found.The unbiased IF estimation can be obtained when singal is linear.Hence,for the processing of MRS envelope singal with high linearity,pseudo Wigner-Ville distribution with large window length is adopted and the relationship of window length is derived.Taking the advantage of energy concentrated along the peak of time-frequency distribution,the IF is estimated.Additionally,numerical simulations and field cases also show that signal parameters such as water content and pore size can be effectively achieved in noise level down to a signal-to-noise ratio(SNR)of-17 dB.3.In terms of processing the MRS oscillating signal with high nonlinearity and large data samples,that computer memory always overflow when doing time-frequency analysis.The joint application of pseudo Wigner-Ville distribution with small window length and segmented processing method is adopted.An appropriate window length should be selected to make the signal as close to linear as possible across the window length.Thus,the bias of the filtering result caused by signal nonlinearity is cancelled.In addition,the signal is segmented to realize the processing of MRS oscillating signal.Model simulations show that the desired signal can be recovered at the SNR down to-5 dB by segmented time-frequency peak filtering(STFPF)method.4.In terms of the problem that the high level random noise corrupted MRS oscillating signal is difficult to be suppressed,combing with empirical mode decomposition(EMD)method and STFPF method.The MRS oscillating signal is decomposed into several different components by EMD,and the components dominated by noise are discarded to remove a part of random noise.Then,the residual random noise embedded in the components dominated by signal is suppressed by STFPF method.The suppression of random noise in MRS oscillating signal provides an effective solution for the actual signal processing.5.In terms of the problem that poor reliability of receiving system and low measurement efficiency,taking the advantages of high sampling speed of envelope signal for on-site measurement and abundant imformation of oscillating signal for laboratory analysis,compensating for the envelope acquisition cannot be matched with the subsequent denoising method and low efficiency of oscillating signal acquisition,a combined work mode is adopted by combining envelope signal measurment and oscillating signal measurment.The system noise is measured in shielded room and channel consistency is tested by the modeled experiments.Finally,the effectiveness of the system and methods is verified by the field experiment conducted in Shaoguo town,Changchun city.The main innovation lies as follows:1.Random noise attenuation method according to the time-frequency characteristic of MRS signal has been first proposed.It is an innovated method for MRS random noise attenuation,more than the time domin or frequency domain filtering methods.By using TFPF method,it achieved better result with single signal than stacking,which improve the measurement efficiency.2.Random noise attenuation methods of envelope signal and oscillating signal based on TFPF and EMD-STFPF are proposed,respectively.The bias of the filtering result caused by signal nonlinearity is canceled,and the problems that noise suppression and signal preserving difficult to weigh and computer memory always overflow are solved.The TFPF method is practically applied in the field signal processing,which effectively improves the SNR and the accuracy of signal parameters.It lays the foundation for obtaining better inversion result.3.The measurement mode combining MRS envelope signal recording and MRS oscillating signal recording is proposed.It takes the advantage of fast acquisition speed of envelope signal which can be used for improving the measurement efficiency on-site measurement,and also takes the advantage of the abundant information of oscillating signal for laboratory analysis.So,dual-redundancy receiving system is developed,when there is something wrong with one path,the other path can be supplemented aming at improving the measurement efficiency and reliability,as well as providing a suit of reliable equipment for the exploration of groundwater and detection of disaster water.