Search

CN-122017973-A - TBM (Tunnel boring machine) -oriented pulse reconstruction method for passive seismic data

CN122017973ACN 122017973 ACN122017973 ACN 122017973ACN-122017973-A

Abstract

The invention discloses a pulse reconstruction method for TBM (Tunnel boring machine) driven seismic data. The method comprises the steps of segmenting original TBM (Tunnel boring machine) along with passive seismic data according to time windows, decomposing CEEMDAN based on energy self-adaptive noise injection to obtain a plurality of IMF components, screening reserved components by combining time domain, frequency domain and time domain indexes and multi-channel consistency constraint, constructing an effective signal gather, constructing a self-adaptive regularized multidimensional deconvolution inversion model by taking the effective signal gather as input, solving virtual seismic source impulse responses in a robust mode by combining phase whitening or spectral amplitude flattening, and finally carrying out cross-correlation alignment on impulse responses of different time periods in a target time window and superposing and outputting impulse reconstruction seismic data. The method can inhibit narrowband spectrum and non-uniform interference, reduce sensitivity to single reference channel quality, improve coherence of reflection events and stability of reconstruction results, and is suitable for TBM advanced detection and imaging.

Inventors

  • ZHANG LEI
  • ZHU GUOWEI
  • ZHANG YUE
  • WU YANHUI
  • LI JIANSHUANG

Assignees

  • 中国矿业大学(北京)

Dates

Publication Date
20260512
Application Date
20260128

Claims (11)

  1. 1. The pulse reconstruction method for TBM-oriented passive seismic data is characterized by comprising the following steps of: step 1, segmenting the TBM following passive seismic data of an original long-time sequence by adopting a fixed-length sliding window segment or an energy triggering mode to obtain a plurality of segments of seismic data; step 2, based on the energy level of the segmented seismic data, adaptively determining CEEMDAN the noise injection strength of decomposition, and executing CEEMDAN decomposition on the seismic data of each segment to obtain an IMF component set and a residual item of the seismic data of each channel in the time segment; Step 3, carrying out multi-domain joint screening on the IMF component set, determining a component reservation set by combining multiple channels of consistency constraints, and constructing an effective signal of a corresponding time period according to the component reservation set; Step 4, constructing an observation spectrum matrix and a reference spectrum matrix by using the effective signals, solving a virtual focus impulse response of a corresponding time period by adopting self-adaptive regularized multidimensional deconvolution inversion, and performing time-frequency conversion on the virtual focus impulse response to obtain a corresponding time domain impulse response; And 5, performing cross-correlation alignment and superposition on time domain impulse responses of different time periods in a target time window, and outputting the impulse reconstruction seismic data.
  2. 2. The method for pulse reconstruction of TBM-oriented passive seismic data according to claim 1, wherein in step 2, the noise injection intensity σ m is determined by the seismic data of the corresponding time period Root mean square energy adaptive determination of (2) satisfies: Wherein alpha is a proportionality coefficient, Is the mean of the root mean square energy of the seismic data over the period of time.
  3. 3. The method for pulse reconstruction of TBM-oriented mining passive seismic data according to claim 1, wherein the multi-domain joint screening in step 3 includes the following three types of criteria, and the IMF component is retained or removed according to the evaluation index corresponding to each criterion and a preset criterion: (1) The time domain criterion is used for representing the transient property and the energy concentration degree of the IMF component, eliminating the component with dominant continuous background oscillation or energy dispersion, and is expressed as follows: Where x is the IMF component to be evaluated, μ is its mean, When the Kur is not smaller than a preset threshold Kur 0 , judging that the IMF component meets a time domain criterion and is reserved, otherwise, eliminating; (2) Frequency domain criteria for characterizing spectral shape consistency and in-band concentration of IMF components, suppressing narrowband spectral dominant components and preserving components with valid dominant frequency band characteristics, expressed as: wherein X (f) is the spectrum of the IMF component, When R b is not smaller than a preset threshold R b0 , judging that the IMF component meets the frequency domain criterion and is reserved, otherwise, eliminating; (3) The time-frequency domain criterion is used for representing the time-frequency energy ridge line stability and the modal aliasing degree of the IMF component and avoiding abnormal drift or unstable frequency band of the invalid component in the time-frequency domain, and is expressed as: Wherein, the Resolving the instantaneous phase of the signal for IMF components, when said signal is in a predetermined time window The fluctuation amplitude of (2) does not exceed a preset threshold And whose principal distribution falls within the target primary frequency band When the IMF component is in the inner time, judging that the IMF component meets the time-frequency domain criterion and reserving the IMF component, otherwise, rejecting the IMF component; And 3, carrying out joint judgment on each IMF component according to the criteria, and reserving when the IMF components simultaneously meet at least two types of criteria, otherwise, rejecting.
  4. 4. The method for pulse reconstruction of passive seismic data for TBM mining according to claim 1, wherein the multi-channel consistency constraint in step 3 comprises a cross-channel correlation constraint and a cross-channel main frequency band consistency constraint, specifically: Respectively calculating a cross-channel correlation coefficient and a main frequency band statistic for IMF components corresponding to each channel in the same time period, removing cross-channel discontinuous components according to a preset criterion, and determining an IMF component retention set The cross-track consistency constraint includes: (1) Cross-channel correlation constraint, namely calculating a correlation coefficient rho of an IMF component to be evaluated in a multi-channel combination of adjacent channels or preset channel spacing, judging that the IMF component meets the cross-channel correlation consistency and is reserved when the correlation coefficient rho is not smaller than a preset threshold rho 0 , and eliminating if not; (2) Cross-channel main frequency band consistency constraint, namely extracting main frequency band statistic of IMF components to be evaluated, including main frequency f p and bandwidth B, and comparing deviation of the main frequency band statistic between different channels, when the main frequency deviation is the same as the main frequency deviation Not greater than a preset threshold And the bandwidth is deviated Not greater than a preset threshold When the IMF component meets the requirement of cross-channel main frequency band consistency, the IMF component is reserved, and otherwise, the IMF component is removed; wherein the same IMF component is incorporated into the IMF component retention set when the cross-channel correlation constraint and the cross-channel primary frequency band consistency constraint are simultaneously satisfied 。
  5. 5. The method for pulsated reconstruction of TBM-oriented passive seismic data according to claim 4, wherein a set is retained based on said IMF component Constructing an effective signal of the mth segment of seismic data :
  6. 6. The method for pulse reconstruction of TBM-oriented passive seismic data according to claim 1, wherein in step 4, a relation among an observation spectrum matrix, a reference spectrum matrix and a virtual seismic source impulse response is established in a frequency domain, and the expression is as follows: Wherein, the An observation spectrum matrix constructed for the mth time period seismic data, For the corresponding reference spectrum matrix, And (3) a frequency domain matrix of the virtual focus impulse response to be solved for the mth time period.
  7. 7. The method for pulse reconstruction of TBM-oriented passive seismic data as in claim 6, wherein said reference spectrum matrix in step 4 is used to construct said predetermined interval as a trace number interval by selecting reference traces at predetermined intervals in the effective signal And said And determining according to the total channel number N to avoid the adjacent channel redundancy which is highly relevant, thereby reducing the redundancy of the reference channel and improving the inversion pathological degree.
  8. 8. The method for pulse reconstruction of TBM-oriented passive seismic data according to claim 6, wherein in step 4, phase whitening or spectral amplitude flattening is performed on the observation spectrum matrix and the reference spectrum matrix to reduce the influence of TBM narrowband spectrum interference on inversion solution, thereby improving the robustness of inversion solution. The expression of phase whitening is: Spectral amplitude flattening is: Where ε is the stable term that avoids denominator zero.
  9. 9. The method for pulse reconstruction of TBM-oriented passive seismic data according to claim 6, wherein the virtual seismic source impulse response in the step 4 is solved by adopting self-adaptive regularized multidimensional deconvolution inversion, and the solving is realized by regularized optimization objective functions as follows: Wherein, the Is the Frobenius norm, Is an adaptive regularization parameter.
  10. 10. The method for pulse reconstruction of TBM-oriented passive seismic data as recited in claim 9, wherein said regularization parameters According to the frequency domain disease index self-adaptive determination of the inversion matrix, the expression is as follows: Wherein, the As a reference regularization coefficient, The condition number for inversion matrix is expressed as: Wherein, the Is a condition number operator. Adaptively increasing regularization parameters as the condition number statistics increase To suppress noise amplification and to improve stability of inversion solutions.
  11. 11. The method for pulse reconstruction of TBM-oriented passive seismic data according to claim 1, wherein in step 5, the reference impulse response is calculated within a target time window Ω Impulse response to be aligned And takes the time delay corresponding to the peak value of the cross-correlation function as the relative time delay : According to the relative time delay The time domain impulse responses of the M time periods are aligned and then superimposed: Wherein the target time window Determined according to the first arrival pickup of the direct wave or the effective signal energy concentration interval, the reference impulse response To be from the said And a reference impulse response selected from impulse responses of the time periods. By the aligned superposition, the in-phase energy of the impulse response sequence is enhanced and non-uniform components are suppressed, thereby outputting the pulsed reconstructed seismic data.

Description

TBM (Tunnel boring machine) -oriented pulse reconstruction method for passive seismic data Technical Field The invention belongs to the technical field of geophysical exploration and underground engineering detection, and particularly relates to a passive seismic signal processing method under the construction condition of a tunnel boring machine. Background Tunnel Boring Machines (TBMs) are widely used in traffic tunnels, mine roadways and various underground engineering constructions. Because geological conditions in front of the tunneling surface have uncertainty, bad geological bodies such as fault fracture zones, weak interlayers, water-rich zones and cavities can be encountered in the construction process, and if the distribution of the bad geological bodies cannot be recognized in advance, the construction safety and the tunneling efficiency can be seriously affected. Therefore, the development of efficient and reliable advanced detection in the TBM construction process is of great significance. The traditional active seismic detection method which depends on a manual seismic source is difficult to implement under TBM following conditions due to the limitation of construction space and safety regulations. The passive earthquake observation is carried out by utilizing the vibration signals generated by TBM rock breaking and equipment operation, continuous and non-interference type monitoring can be realized, but the data of the earthquake monitoring is usually in a strong and non-stable characteristic, and is accompanied by remarkable narrowband line spectrum vibration and background mechanical noise, and an effective wave field is easily mixed with direct waves and interference components, so that the signal to noise ratio is low and the signal structure is complex. The prior art still has the defects that the traditional filtering and time-frequency analysis method is difficult to stably inhibit the narrow-band line spectrum and retain an effective wave field under a strong non-stable background, the empirical mode decomposition method has the advantages of self-adaptive decomposition, but lacks a noise injection mechanism adaptive to signal energy fluctuation, is mostly limited to single-channel processing, lacks multi-channel space consistency constraint, and causes insufficient cross-channel continuity of a reconstruction result, and the method for constructing the virtual seismic source based on interference or deconvolution easily generates serious frequency domain disease state problems under the interference of the narrow-band line spectrum and is extremely sensitive to the selection of a reference channel, so that the impulse response stability and repeatability are poor. The problems are interwoven, so that stable extraction of the reflection signals which can be used for high-resolution imaging from complex mining data becomes extremely difficult, and the accuracy and reliability of advanced geological forecast are severely restricted. Therefore, a signal processing method suitable for TBM driven seismic data is needed, which can realize the stable extraction and the pulsed reconstruction of an effective wave field under the conditions of strong non-stationary and line spectrum interference, and improve the stability and the imaging reliability of virtual seismic source construction. Disclosure of Invention In view of the above, the invention provides a pulse reconstruction method for TBM-oriented passive seismic data, which aims to solve the problems that effective signals of the passive seismic data are difficult to stably extract under continuous excitation and strong non-stationary interference background, the virtual seismic source construction process is sensitive to reference channel selection, and the reflection event enhancement effect and reconstruction stability are insufficient, so that the reliability of TBM-oriented advanced detection and imaging results is improved. In order to achieve the above purpose, the present invention adopts the following technical scheme: a TBM-oriented pulse reconstruction method for passive seismic data mining comprises the following steps: step 1, segmenting TBM (Tunnel boring machine) following passive seismic data of an original long-time sequence by adopting a fixed-length sliding window segment or an energy triggering mode to obtain m-section seismic data ; Step2, calculating the m-th section of seismic dataAnd determining CEEMDAN the resolved noise injection intensity σ m from the energy adaptation: Wherein alpha is a proportionality coefficient, The mean value of root mean square energy of the seismic data; And adopts the corresponding noise injection intensity sigma m to perform seismic data Performing CEEMDAN decomposition to obtain k-order IMF component sets of the seismic data of each channel in the time periodResidual items; Step 3, for the IMF component set obtained in the step 2Performing multi-domain joint screening to remove noise