Research On Inversion Imaging Of Microseismic Signals Propagation Velocity Models And Source Localization

Rock fracture in deep underground construction tends to induce severe disasters such as rock bursts,collapses and impact ground pressure,causing massive casualties and economic losses.Real-time monitoring of critical information about the rock rupture process is an effective way to avoid disasters.Microseismic monitoring has become the mainstream method for global monitoring of rock fracture evolution,which captures microseismic elastic waves generated by the rock fracture process in real-time through sensors and then analyzes the microseismic localization source.The first arrival moment of the microseismic signal at each sensor and the distribution of microseismic signal propagation velocity are two critical factors for the accurate localization of microseismic sources.However,the complex environment of deep underground construction leads to a low signal-to-noise ratio of microseismic signals and inaccurate acquisition of the first arrival moment.Furthermore,the microseismic signal’s propagation velocity varies substantially within the inhomogeneous deep rock mass,severely limiting the accuracy of the classic time-difference localization method in microseismic source localization.In view of the above problems,this thesis mainly carries out the following research work:(1)To deal with the low signal-to-noise ratio of microseismic signals,a noise reduction method with improved wavelet thresholding is studied.The multisynchrosqueezing transform with a strong time-frequency aggregation capability is introduced to accurately extract each frequency component of the microseismic signal,which solves the problem of poor signal focus in the frequency domain.The method improves the signal-to-noise ratio of the microseismic reconstruction signal and,even better,reduces the computation time.(2)To address the inaccurate acquisition of the first arrival moment of microseismic signals,this thesis develops a new method particularly from the perspectives of both computational accuracy and efficiency.Firstly,a higher-order finite-difference operator is introduced into ray tracing to obtain the travel time distribution,forming an improved higher-order fast travel method.This method obtains precise first arrival moment acquisition and improves the accuracy of the first arrival moment acquisition for the entire model by compensating for its significant error in the microseismic source’s main diagonal direction.In addition,automatic features selection and clustering analysis(with ReliefF and K-means methods)is investigated based on the waveform features of the microseismic signal.This method utilizes both the saliency features and the microseismic time-domain signal to reduce the redundancy of microseismic signal features and improve the accuracy of the clustering analysis,which further in turn improves the efficiency of the first arrival moment calculation and enables fast,accurate and automatic acquisition of the first arrival moments.(3)Conventional inversion methods suffer from various technical issues,including the significantly different speeds of microseismic signal propagation in inhomogeneous media,ambiguously non-unique solutions,long computation time,and coarse inversion interfaces.An intelligent inversion method applicable to microseismic inhomogeneous propagation velocity models is proposed in this thesis to address these issues.This method incorporates the physical laws of microseismic signal propagation into a data-driven model and is focused on constructing interface-sensitive fully variational regularization constraints(TVR).The proposed method effectively solves the problems of unclear interface boundaries and large fluctuations of propagation values in the inverse propagation velocity model,which achieves accurate inversion of complex microseismic propagation velocity distributions such as multi-layer interfaces,cavities and faults.(4)In the conventional microseismic source location method,a considerable difference often exists between the model setting and the actual formation propagation velocity.This mismatch can trap the iterative algorithm into local minima and hinder the location of the valid source point.A hybrid PSO-SM microseismic source location method with complementary search characteristics(including strengthened global search and more efficient local search)is developed based on the accurate acquisition of the microseismic signal’s first arrival moment and the microseismic signal propagation velocity distribution.This method demonstrates an improved global search capability and a faster computational speed for source localization.Single-source and multi-source microseismic source localization tests are then conducted systematically,which involves disturbances in the arrival time and propagation velocity.Results that the developed methods can achieve accurate localization of microseismic sources.