Underground rock mass structure damage microseismic signal feature identification method

Abstract

The invention belongs to the field of underground rock mass structure engineering safety monitoring, and discloses an underground rock mass structure damage micro-seismic signal feature identification method, which comprises the following steps: acquiring a structure micro-seismic signal during monitoring, and calculating four time-frequency domain parameters including a micro-seismic energy spectrum coefficient, a signal correlation coefficient, a signal correlation moment and a frequency response difference value; combining parameters to analyze fusion features, and then constructing fractal dimension indexes; and finally, carrying out damage state identification by utilizing a membership degree vector calculated by a fractal dimension index. The method is accurate in identification result, has good real-time performance and advancement, can early warn underground rock mass structure damage in advance, and guarantees rock mass structure functions and personnel life safety.

Description

Technical Field

[0001] The invention belongs to the field of underground rock mass structure engineering safety monitoring, and in particular relates to a method for identifying microseismic signal characteristics of underground rock mass structure damage. Background Art

[0002] Underground rock structures are affected by multi-directional stresses during service, and their macroscopic mechanical properties such as stiffness, compressive and shear strength will change, resulting in damage such as crack extension and surface peeling, which will increase the risk of structural instability and cracking accidents. Therefore, carrying out rock structure damage status monitoring and damage early warning work is of great scientific significance to ensure project progress and personnel safety.

[0003] Rock structure damage detection can be divided into two types: static assessment and dynamic assessment. Static assessment refers to the use of structural geometric scale, material etching effect, deformation aging rate and other indicators to assess the structural damage state. This method has low technical barriers and is easy to carry out. It is widely used in components with fixed scales and relatively simple physical composition. However, it has the characteristics of low accuracy and large interference in identifying the damage state of rock structures in complex underground environments.

[0004] The dynamic assessment method refers to the use of dynamic parameters to analyze the health status of the structure, which can be divided into local analysis and overall analysis based on the detection range. Local analysis refers to the use of infrared, ultrasound, CT, etc. to perceive the damage variables in the detection area, and overall analysis refers to the analysis of environmental impact effects on dynamic fingerprints such as natural frequency, vibration mode, and modal curvature. The local method has shortcomings in terms of sensitivity to full-cycle damage detection, equipment deployment costs in narrow underground environments, and long-term monitoring continuity. In addition, the underground environment has multi-field coupling characteristics, and the loads on underground rock structures come from multiple directions in the horizontal and vertical directions. Therefore, the overall dynamic analysis method needs to be further improved in terms of high-order modal parameter dependence and environmental anti-disturbance, and it is difficult to meet the needs of accurate real-time detection of the damage status of rock structures in complex underground environments. How to achieve low cost and high reliability while ensuring the sensitivity of identifying the damage status of rock structures is a difficult problem that needs to be solved urgently.

[0005] Rock microseismicity is a phenomenon in which a structure reaches a stable state by releasing transient ultrasonic waves under stress. Microseismic events occur during the sliding process of rock particles. Damage will change the combination of rock particles and the friction law between particles, thereby causing multi-parameter changes in microseismic signals. The use of microseismicity and structural damage state correlation to carry out damage monitoring has been widely recognized by academia and industry. In addition, microseismic technology is a passive detection method. In real-time monitoring, there is no need to emit excitation signals. It can achieve long-term and continuous damage monitoring with low power consumption, which can better control equipment energy consumption and cost issues; but there is currently no good method to use this feature to study the identification of rock structure damage state. Summary of the invention

[0006] To solve the above-mentioned problems, the present invention proposes a method for identifying the characteristics of microseismic signals of underground rock structure damage, extracts four characteristic parameters of rock microseismic signals and conducts combined analysis, calculates the fractal dimension index, and constructs a membership vector to identify the damage state of the rock structure.

[0007] The method for identifying microseismic signal characteristics of underground rock structure damage according to the present invention comprises the following steps:

[0008] Step 1: Obtain microseismic signals of underground rock structures in different damage states, and perform modal decomposition of the original signal through SVMD to obtain several modal components, calculate the correlation coefficient between each modal component and the original signal, and reconstruct the signal of each modal component greater than the correlation coefficient threshold to achieve denoising;

[0009] Step 2, extracting the energy spectrum coefficient, signal correlation coefficient, signal correlation moment, and frequency response difference parameter of the denoised microseismic signal;

[0010] Step 3, constructing an adaptive fractal dimension index using the four characteristic parameters extracted in step 2; the method for obtaining the adaptive fractal dimension index is: taking the microseismic parameters in the historical database as input, combining two parameters into a group, constructing a Euclidean space, calculating the distance of the sequence points in the Euclidean space, and calculating the fractal dimension index using the limit method;

[0011] Step 4: Use the fractal dimension index to calculate the membership vector and realize damage state identification.

[0012] Furthermore, in step 1, the collected underground rock structure damage microseismic signal x(t) is first transformed by Fourier transform to obtain a frequency domain signal Then, the acquired microseismic signals of underground rock mass structure are modally decomposed and reconstructed. Specifically, SVMD is used to obtain the modal components According to the energy distribution of each mode, update its center frequency ω k :

[0013]

[0014] Where n is the number of iterations, f represents the independent variable in the frequency domain, are the kth modal components at the nth and n+1th iterations in the frequency domain, respectively. They represent the center frequencies of the kth modal component at the nth and n+1th iterations in the frequency domain respectively; k represents the modal component number, λ is the Lagrange multiplier, and α is the smoothing parameter.

[0015] Furthermore, in step 1, the new underground rock structure damage microseismic signal is reconstructed by each modal component obtained to achieve denoising. Specifically, the correlation coefficient r between each modal component and the original microseismic signal is calculated.k , and then calculate the correlation coefficient limit r thr , select the modal components greater than the limit value for reconstruction:

[0016]

[0017] Among them, M(·) means finding the mean, E(·) means finding the expectation, are the modal component signals in the frequency domain, is the original microseismic signal in the frequency domain; is the reconstructed microseismic signal, and the reconstructed microseismic signal is inversely transformed by Fourier transform to obtain the reconstructed microseismic signal in the time domain.

[0018] Furthermore, in step 2, the energy spectrum coefficient, signal correlation coefficient, signal correlation moment, and frequency response difference parameters of the reconstructed microseismic signal are extracted and calculated. The energy spectrum coefficient parameter sequence under different damage levels is: The signal correlation coefficient parameter sequence is: The signal correlation moment parameter sequence is: The frequency response difference parameter sequence is: Where m is the damage level.

[0019] Furthermore, the order of parameter combination analysis in step 3 is: energy spectrum coefficient-signal correlation coefficient, signal correlation coefficient-signal correlation moment, signal correlation moment-frequency response difference, frequency response difference-energy spectrum coefficient.

[0020] Furthermore, the Euclidean space in step 3 is:

[0021]

[0022] Where τ is the time delay, v is the dimension of the Euclidean space, N is the number of microseismic signals, and N v is the dimension feature parameter, N v =N-(v-1)τ.

[0023] Furthermore, the distance d of the sequence points in the Euclidean space described in step 3 ij The calculation method is:

[0024]

[0025] where x i and x j are any two points in Euclidean space.

[0026] Furthermore, the fractal dimension index D(v) calculated by the limit method in step 3 is:

[0027]

[0028] Among them, r is a given scale parameter, H is the Heaviside function:

[0029]

[0030] Furthermore, step 4 is as follows: in the training phase, the fractal dimension index is used as input to train the BiLSTM to obtain the recognition model. In the application phase, the microseismic signal of unknown damage state is input into the trained model to obtain the membership vector; the order of the largest elements in the membership vector is used as the damage level state of the underground rock mass structure.

[0031] The beneficial effects of the present invention are as follows: the method of the present invention extracts four characteristic parameters including time domain and frequency domain, and analyzes the parameters in combination, which can avoid the shortcomings of insufficient consideration of the synergistic correlation characteristics between features in the traditional single-parameter analysis method, and has better scientificity and sensitivity; the BiLSTM model is used to fully mine historical microseismic data information. With the continuous enrichment of data in long-term monitoring projects, the constructed damage identification model also has better anti-interference ability. The hardware involved in the present invention is simple, the monitoring of the damage state of the rock structure is highly accurate and real-time, and it can provide early warning to ensure the progress of underground space rock engineering and the safety of construction personnel. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 is a flow chart of the method of the present invention;

[0033] Figure 2 It is a schematic diagram of microseismic signals in different damage states;

[0034] Figure 3 De-noising and reconstructing images for microseismic signals;

[0035] Figure 4 is a schematic diagram of microseismic signal parameter values;

[0036] Figure 5 It is a schematic diagram of the fractal dimension indicator with different parameter combinations. DETAILED DESCRIPTION

[0037] In order to make the contents of the present invention more clearly understood, the present invention is further described in detail below based on specific embodiments in conjunction with the accompanying drawings.

[0038] like Figure 1 As shown, the method for identifying microseismic signal characteristics of underground rock structure damage according to the present invention comprises the following steps:

[0039] Step 1: Obtain microseismic signals of underground rock structures in different damage states, and perform modal decomposition of the original signal through SVMD to obtain several modal components, calculate the correlation coefficient between each modal component and the original signal, and reconstruct the signal of each modal component greater than the correlation coefficient threshold to achieve denoising;

[0040] Step 2, extracting the energy spectrum coefficient, signal correlation coefficient, signal correlation moment, and frequency response difference parameter of the denoised microseismic signal;

[0041] Step 3, constructing an adaptive fractal dimension index using the four characteristic parameters extracted in step 2; the method for obtaining the adaptive fractal dimension index is: taking the microseismic parameters in the historical database as input, combining two parameters into a group, constructing a Euclidean space, calculating the distance of the sequence points in the Euclidean space, and calculating the fractal dimension index using the limit method;

[0042] Step 4: Use the fractal dimension index to calculate the membership vector and realize damage state identification.

[0043] In step 1, the collected underground rock structure damage microseismic signal x(t) is first transformed into a frequency domain signal by Fourier transform Then, the acquired microseismic signals of underground rock mass structure are modally decomposed and reconstructed. Specifically, SVMD is used to obtain the modal components According to the energy distribution of each mode, update its center frequency ω k :

[0044]

[0045] Where n is the number of iterations, f represents the independent variable in the frequency domain, are the kth modal components at the nth and n+1th iterations in the frequency domain, respectively. They represent the center frequency of the kth modal component at the nth and n+1th iterations in the frequency domain, respectively. k represents the modal component number, λ is the Lagrange multiplier, and α is the smoothing parameter.

[0046] In step 1, the new underground rock structure damage microseismic signal is reconstructed by the obtained modal components to achieve denoising. Specifically, the correlation coefficient r between each modal component and the original microseismic signal is calculated. k , and then calculate the correlation coefficient limit r thr , select the modal components greater than the limit value for reconstruction.

[0047]

[0048] Among them, M(·) means finding the mean, E(·) means finding the expectation, are the modal component signals in the frequency domain, is the original microseismic signal in the frequency domain. is the reconstructed microseismic signal, and the reconstructed microseismic signal is inversely transformed by Fourier transform to obtain the reconstructed microseismic signal in the time domain.

[0049] In step 2, the energy spectrum coefficient, signal correlation coefficient, signal correlation moment, and frequency response difference parameters of the reconstructed microseismic signal are extracted and calculated. The energy spectrum coefficient parameter sequence under different damage levels is: The signal correlation coefficient parameter sequence is: The signal correlation moment parameter sequence is: The frequency response difference parameter sequence is: Where m is the damage level.

[0050] The order of parameter combination analysis in step 3 is: energy spectrum coefficient-signal correlation coefficient, signal correlation coefficient-signal correlation moment, signal correlation moment-frequency response difference, frequency response difference-energy spectrum coefficient.

[0051] In step 3, the Euclidean space is:

[0052]

[0053] Where τ is the time delay, v is the dimension of the Euclidean space, N is the number of microseismic signals, and N v =N-(v-1)τ.

[0054] The distance calculation method of the sequence points in the Euclidean space is:

[0055]

[0056] The fractal dimension index D(v) calculated by the limit method in step 3 is:

[0057]

[0058] Where r is a given scale parameter, H is the Heaviside function:

[0059]

[0060] The identification method described in step 4 is: in the training stage, the fractal dimension index is used as input to train the BiLSTM to obtain the identification model. In the application stage, the microseismic signal of unknown damage state is input into the trained model to obtain the membership vector. The order of the largest elements in the membership vector is used as the damage level state of the underground rock mass structure.

[0061] Taking five types of damage states as an example, in order to simulate five damage levels, a hydraulic press is used to apply stress to the rock samples to stimulate cracks, with lengths of 2 cm, 4 cm, 6 cm, 8 cm, and 10 cm, corresponding to damage levels 1-5, respectively. Microseismic signals of samples with different damage states are collected. Figure 2 As shown, Figure 2 (a) is a schematic diagram of microseismic signals of damage level 5. Figure 2 (b) is a schematic diagram of microseismic signals of damage level 4. Figure 2 (c) is a schematic diagram of microseismic signals of damage level 3. Figure 2 (d) is a schematic diagram of microseismic signals of damage level 2. Figure 2 (e) is a schematic diagram of the microseismic signal of damage level 1. The signal comparison before and after denoising is as follows Figure 3 As shown, Figure 3 (a) is a signal comparison diagram of damage level 5. Figure 3 (b) is a signal comparison diagram of damage level 4. Figure 3 (c) is a signal comparison diagram of damage level 3. Figure 3 (d) is the signal comparison diagram of damage level 2. Figure 3 (e) is a signal comparison diagram of damage level 1. Four parameters of the microseismic signal after denoising are extracted, and the normalized parameter values ​​are as follows: Figure 4 The parameter combinations are analyzed and the fractal dimension index is calculated. The fractal dimension results of different parameter combinations are shown in Figure 5 shown.

[0062] Mining historical microseismic data, training the BiLSTM model to identify damage states. Taking the 8cm crack length (damage level 4) as an example, the maximum membership vector output by the model appears at the 4th position, indicating that this sample has the highest confidence in being at the 4th damage level, which is consistent with the actual sample damage state.

[0063] The above description is only a preferred embodiment of the present invention and is not intended to be a further limitation of the present invention. All equivalent changes made using the contents of the present specification and drawings are within the protection scope of the present invention.

Claims

1. A method for identifying microseismic signal characteristics of underground rock structure damage, characterized in that: The method comprises: Step 1: Obtain microseismic signals of underground rock structures in different damage states, and perform modal decomposition of the original signal through SVMD to obtain several modal components, calculate the correlation coefficient between each modal component and the original signal, and reconstruct the signal of each modal component greater than the correlation coefficient threshold to achieve denoising; Step 2, extracting the energy spectrum coefficient, signal correlation coefficient, signal correlation moment, and frequency response difference parameter of the denoised microseismic signal; Step 3, constructing an adaptive fractal dimension index using the four characteristic parameters extracted in step 2; the method for obtaining the adaptive fractal dimension index is: taking the microseismic parameters in the historical database as input, combining two parameters into a group, constructing a Euclidean space, calculating the distance of the sequence points in the Euclidean space, and calculating the fractal dimension index using the limit method; Step 4: Use the fractal dimension index to calculate the membership vector and realize damage state identification.

2. The method for identifying microseismic signal characteristics of underground rock structure damage according to claim 1, characterized in that: In step 1, the collected microseismic signal x(t) is first transformed by Fourier transform to obtain the frequency domain signal Secondly, the original signal is modally decomposed; specifically: update each modal component According to the energy distribution of each mode, update its center frequency ω k : Where n is the number of iterations, f represents the independent variable in the frequency domain, are the kth modal components at the nth and n+1th iterations in the frequency domain, respectively. They represent the center frequencies of the kth modal component at the nth and n+1th iterations in the frequency domain respectively; k represents the modal component number, λ is the Lagrange multiplier, and α is the smoothing parameter.

3. The method for identifying microseismic signal characteristics of underground rock structure damage according to claim 2, characterized in that: In step 1, the new underground rock structure damage microseismic signal is reconstructed by the obtained modal components to achieve denoising. First, the correlation coefficient r between each modal component and the original signal is calculated. k , and then calculate the correlation coefficient limit r thr , select the modal components greater than the limit value for reconstruction; Among them, M(·) means finding the mean, E(·) means finding the expectation, are the modal component signals in the frequency domain, is the original microseismic signal in the frequency domain; is the reconstructed microseismic signal, and the reconstructed microseismic signal is inversely transformed by Fourier transform to obtain the reconstructed microseismic signal in the time domain.

4. The method for identifying microseismic signal characteristics of underground rock structure damage according to claim 1, characterized in that: In step 2, the energy spectrum coefficient, signal correlation coefficient, signal correlation moment, and frequency response difference parameters of the reconstructed microseismic signal are extracted and calculated. The energy spectrum coefficient parameter sequence under different damage levels is: The signal correlation coefficient parameter sequence is: The signal correlation moment parameter sequence is: The frequency response difference parameter sequence is: Where m is the damage level.

5. The method for identifying microseismic signal characteristics of underground rock structure damage according to claim 1, characterized in that: In step 3, the order of parameter combination analysis is: energy spectrum coefficient-signal correlation coefficient, signal correlation coefficient-signal correlation moment, signal correlation moment-frequency response difference, frequency response difference-energy spectrum coefficient.

6. The method for identifying microseismic signal characteristics of underground rock structure damage according to claim 4, characterized in that: In step 3, the Euclidean space is: Where τ is the time delay, v is the dimension of the Euclidean space, N is the number of microseismic signals, and N v is the dimension feature parameter, N v =N-(v-1)τ.

7. The method for identifying microseismic signal characteristics of underground rock structure damage according to claim 6, characterized in that: The distance d of the sequence points in the Euclidean space described in step 3 ij The calculation method is: where x i and x j are any two points in Euclidean space.

8. The method for identifying microseismic signal characteristics of underground rock structure damage according to claim 7, characterized in that: The fractal dimension index D(v) calculated by the limit method in step 3 is: Where r is a given scale parameter, H is the Heaviside function:

9. The method for identifying microseismic signal characteristics of underground rock structure damage according to claim 1, characterized in that: Step 4 is as follows: in the training stage, the fractal dimension index is used as input to train the BiLSTM to obtain the recognition model; in the application stage, the microseismic signal of unknown damage state is input into the trained model to obtain the membership vector; the order of the largest elements in the membership vector is used as the damage level state of the underground rock structure.

Images