A multi-signal fusion analysis method for coal-rock fracture instability

Abstract

The present invention discloses a multi-signal fusion analysis method for coal rock mass fracture instability, which belongs to coal rock mass analysis: performing time-frequency domain processing on acoustic emission signals to obtain cumulative ringing eigenvalues ​​and cumulative b eigenvalues; cumulative master frequency eigenvalues ​​and cumulative amplitude eigenvalues; spatial eigenvalues ​​of acoustic emission; fitting functions from eigenvalues ​​to eigenvalues ​​respectively; calculating the correlation coefficient matrix R between eigenvalues ​​to eigenvalues ​​based on the fitting functions; determining the weight coefficient of each eigenvalue based on the matrix R; obtaining the first-order and second-order derivative functions of the fitting functions, and determining the demarcation time point in the coal rock mass failure process based on the first-order and second-order derivative functions; and using the weight coefficients to correct the demarcation time point to obtain the corrected demarcation time point. In view of the fact that the existing single index for judging coal rock mass fracture instability and the low accuracy of the single index for coal rock mass fracture instability early warning are used, the present application improves the early warning accuracy through multi-signal fusion.

Description

Technical Field

[0001] The present application relates to coal rock mass analysis, and more specifically, to a multi-signal fusion analysis method for coal rock mass fracture instability. Background Art

[0002] Acoustic emission technology, a nondestructive dynamic monitoring technique, has been widely used in recent years in a variety of fields, including mineral resource development, slope protection, railway and highway construction, and tunnel stability. Based on the principle that strain energy is released and propagated outward as stress waves during the deformation and fracture of coal and rock masses in coal mine carbon storage spaces, this technology can invert fracture information and the dynamic evolution of internal microfractures during coal and rock mass failure, effectively understanding the fracture mechanisms and precursors of coal and rock mass failure, providing important technical support for early warning and analysis of coal and rock mass instability.

[0003] Acoustic emission waveforms generated during coal and rock mass failure carry rich information about their stress state, internal structure, fracture transformations, and physical and mechanical properties. Analyzing the acoustic emission spectrum and spatial characteristics provides a more comprehensive understanding of the deformation and failure mechanisms of coal and rock masses, as well as information on their precursors. This is of great significance for safety monitoring and early warning of underground rock projects.

[0004] However, existing technologies for early warning of coal and rock mass fracture instability have significant shortcomings. Traditional methods primarily rely on single time-domain or frequency-domain analysis, lacking cumulative effect analysis and arbitrarily selecting parameters. This method fails to fully reflect the specific characteristics of coal and rock mass fracture failure, resulting in inaccurate and untimely early warning information, making it difficult to meet actual engineering needs. Furthermore, existing technologies generally ignore the spatial distribution characteristics of acoustic emission signals, failing to effectively capture the spatially non-uniform failure characteristics caused by preferential fractures.

[0005] For example, patent document CN118688301A discloses a method for assessing fractured rock failure based on acoustic emission characteristic parameters. The method includes: constructing an acoustic emission system to acquire acoustic emission signals generated by rock samples during stress; inputting the acoustic emission signals into the system to obtain the acoustic emission amplitude, acoustic emission magnitude, and acoustic emission ring count rate; calculating the acoustic emission b-value based on the acoustic emission amplitude and magnitude; and calculating the correlation dimension value D2 based on the acoustic emission ring count rate. During the variation of the correlation dimension value D2, a process parameter sequence T is introduced to ensure that the calculation step length of the acoustic emission b-value and the correlation dimension value D2 are consistent. The failure point is found when the correlation dimension value D2 is maximum and the acoustic emission b-value is minimum. However, this method relies solely on the acoustic emission b-value and the correlation dimension D2 for judgment, resulting in a low signal characteristic dimension that cannot fully characterize the complex process of microcrack evolution within the coal rock mass. Furthermore, it completely ignores the interaction between fluids (such as gas and water) and the coal rock mass, which is a key factor in coal rock instability in real mine environments. Summary of the Invention

[0006] In view of the fact that the interaction mechanism between fluid and coal rock is often ignored in the existing coal rock instability warning technology and the single analysis index leads to low warning accuracy, this application provides a multi-signal fusion analysis method for coal rock fracture instability. The multi-signal fusion analysis method based on fluid-coal rock interaction improves the warning accuracy.

[0007] The present application provides a multi-signal fusion analysis method for coal rock fracture instability, comprising: collecting acoustic emission data during the deformation and destruction of coal rock; the acoustic emission data includes acoustic emission signals and spatial data corresponding to the acoustic emission signals; performing time-frequency domain processing on the acoustic emission signals to obtain the cumulative ringing characteristic value of the acoustic emission , cumulative b eigenvalue , the cumulative main control frequency characteristic value of acoustic emission and cumulative amplitude eigenvalues ; Extract the cumulative energy value of acoustic emission from the spatial data as the spatial characteristic value of acoustic emission ; Extract the cumulative energy value of acoustic emission from the spatial data as the spatial characteristic value of acoustic emission ; Fitting eigenvalues ​​respectively To eigenvalue The fitting function ; According to the fitting function , calculate the eigenvalue to The correlation coefficient matrix R between them; according to the correlation coefficient matrix R, determine the weight coefficient of each eigenvalue ; Get the fitting function The first derivative of and the second-order derivative function , and according to the first-order and the second-order derivative function , determine the demarcation time point in the coal rock destruction process ; Using weight coefficient Demarcation time point Make corrections to get the corrected demarcation time point , according to the revised demarcation time point The destruction process of coal rock mass is divided into stable period, development period, energy storage period and destruction period; the time range of coal rock mass entering the destruction period is obtained as a precursor warning signal of coal rock mass instability.

[0008] Furthermore, the cumulative ringing characteristic value Indicates the cumulative value of ringing counts during the acoustic emission process, reflecting the number of microcracks inside the coal rock mass; the cumulative b eigenvalue It represents the cumulative value of the slope coefficient of the acoustic emission signal amplitude-frequency distribution, reflecting the change in the distribution of crack scale inside the coal rock mass during the interaction between the fluid and the coal rock mass; the cumulative main control frequency characteristic value It represents the time series cumulative value of the frequency corresponding to the maximum amplitude in the acoustic emission signal spectrum, reflecting the micro-crack growth rate inside the coal rock mass; the cumulative amplitude characteristic value It represents the time series cumulative value of the maximum amplitude of the acoustic emission signal, reflecting the intensity change of energy released by microcracks inside the coal and rock mass.

[0009] Furthermore, the cumulative b eigenvalue is obtained , including: extracting the absolute energy E of acoustic emission from the collected acoustic emission data, and the number of hits with absolute energy greater than E ; Establish the absolute energy E of acoustic emission and the number of hits The relationship between: , where a and b are constants; set the initial and step length , the acoustic emission data is segmented using a sliding window algorithm; the cumulative variance ratio (CVR) of the acoustic emission signal in the current window is calculated, where CVR represents the ratio of the signal energy variance to the average energy in the window; based on CVR, the optimal window for the current window is calculated. , , γ is a positive coefficient; according to the optimal window , calculate the acoustic emission b value of each window by the least square method; accumulate the acoustic emission b values ​​in each window in chronological order to obtain the acoustic emission cumulative b eigenvalue .

[0010] Furthermore, the cumulative master frequency characteristic value and cumulative amplitude eigenvalues , including: according to the acoustic emission data, the process of interaction between fluid and coal rock is divided into the initial adsorption stage, diffusion and penetration stage and crack expansion stage; the window lengths of the sliding windows in the initial adsorption stage, diffusion and penetration stage and crack expansion stage are set to 、 and ;in, ; According to the window 、 and , segment the acoustic emission wave data of each stage to obtain the acoustic emission time domain signals in multiple time windows Perform wavelet transform on the acoustic emission data in each window to obtain wavelet coefficients; calculate the wavelet energy distribution based on the wavelet coefficients and divide the wavelet energy distribution into low-frequency band, medium-frequency band and high-frequency band; perform peak analysis on each frequency band and extract the main control frequency of the low-frequency band in the initial adsorption stage and the corresponding maximum amplitude , mid-band main control frequency in the diffusion and penetration stage and the corresponding maximum amplitude , high-frequency band master frequency in the crack expansion stage and the corresponding maximum amplitude ; Among them, the main control frequency Indicates the frequency of the peak of the wavelet energy distribution in the corresponding frequency band; accumulates the master frequency of each time window , get the cumulative master control frequency characteristic value ; Accumulate the maximum amplitude of each time window , and get the cumulative amplitude eigenvalue .

[0011] In particular, coal-rock fracture instability often has complex characteristics such as nonlinear evolution, stage mutation and multi-field coupling. Existing technologies usually use Fourier transform or fast Fourier transform, but traditional FFT has defects in processing coal-rock fracture instability analysis: Quasi-FFT has the Heisenberg uncertainty principle limitation of time-frequency resolution, that is, the fixed number of sampling points N determines the frequency resolution. It also determines the length of the time window During the interaction between fluid and coal rock, the rock undergoes multi-stage damage, from microscopic adsorption to mesoscopic diffusion to macroscopic crack expansion, requiring different time-frequency resolutions for each stage. Rapid microcrack expansion in the coal rock requires higher time resolution, while monitoring changes in matrix adsorption requires higher frequency resolution. Fixed-parameter FFT cannot simultaneously meet both requirements.

[0012] In this application, based on the automatic discrimination algorithm of acoustic emission event rate and frequency band energy ratio, the interaction between fluid and coal rock is scientifically divided into the initial adsorption stage, diffusion and penetration stage and crack expansion stage; differentiated window lengths are used for each stage. The strategy solves the contradiction between the insufficient resolution of the traditional fixed window in the rapid change stage and the large noise interference in the stable stage; the characteristic parameters are extracted from the low-frequency band, medium-frequency band and high-frequency band for different stages, so that the system can accurately capture the microscopic signals of the whole process of crack expansion from micro to macro.

[0013] Furthermore, the process of interaction between fluid and coal rock is divided into the initial adsorption stage, diffusion and penetration stage and crack expansion stage, including: calculating the acoustic emission event rate and frequency band energy proportion based on acoustic emission data; setting the first threshold of the acoustic emission event rate and the second threshold ;in, ; Set the third threshold of the frequency band energy ratio of the low frequency band, the middle frequency band and the high frequency band respectively , the fourth threshold and the fifth threshold ; When the acoustic emission event rate is less than or equal to , and the low-frequency band energy ratio is greater than ε3, it is judged to be the initial adsorption stage; when the acoustic emission event rate is greater than and less than , and the frequency band energy of the mid-frequency band is greater than When the acoustic emission event rate is greater than , and the frequency band energy of the high frequency band accounts for more than When , it is judged to be the crack expansion stage.

[0014] In particular, the acoustic emission event rate actually reflects the number of microfractures generated per unit time. The physical mechanism of microfracture generation is different at different stages. Specifically, in the initial adsorption stage: the fluid molecules mainly adsorb on the surface of the coal rock mass. At this time, there are fewer microfractures, so the acoustic emission event rate is low (≤ Diffusion and penetration stage: Fluid molecules begin to diffuse and penetrate into the coal rock mass, causing changes in the internal structure, generating more micro-fractures, and increasing the event rate ( <Event Rate< ); Crack expansion stage: the formed microcracks rapidly expand and connect under the action of fluid, resulting in a large number of microfractures, and the event rate increases significantly (> ).

[0015] In addition, different types of micro-fractures will produce acoustic emission signals with different characteristics. These signals are manifested as the distribution of energy in different frequency bands in the frequency domain. Low-frequency band energy (initial adsorption stage): mainly corresponds to the tiny deformation caused by surface adsorption, with low energy conversion efficiency and low frequency; medium-frequency band energy (diffusion and penetration stage): corresponds to the seepage and diffusion process of fluid inside the coal rock body, resulting in medium-scale internal structural changes; high-frequency band energy (crack expansion stage): corresponds to rapid crack expansion and connectivity, releasing high-frequency acoustic energy.

[0016] Therefore, this application uses both "acoustic emission event rate" and "frequency band energy proportion" as the basis for judgment. The acoustic emission event rate reflects the time density of the number of microfractures; the frequency band energy proportion reflects the spectral characteristics of the microfracture type.

[0017] Furthermore, the wavelet transform adopts the following formula: ,in, is the wavelet coefficient, s is the scale parameter, τ is the translation parameter, is the time domain signal; It is the conjugate function of the wavelet basis function; wavelet transform has better time-frequency localization characteristics than traditional Fourier transform, and can simultaneously obtain the time domain and frequency domain information of the signal; multi-band analysis realizes the classification and identification of coal rock cracks of different scales: low frequency mainly corresponds to large-scale structural changes, medium frequency corresponds to mesoscopic crack expansion, and high frequency corresponds to microcrack initialization; by analyzing the main control frequency of each band and maximum amplitude The changing laws of the coal-rock fracture and instability are accurately described.

[0018] Furthermore, the eigenvalues ​​are fitted separately To eigenvalue The fitting function ,include: ,in, They correspond to five eigenvalues ​​respectively, x is the time variable, which represents the duration from the beginning of the interaction between the fluid and the coal rock to the present. is the j-th coefficient of the fitting function of the i-th eigenvalue; is the constant term of the polynomial, which represents the initial state value of the eigenvalue; j represents the highest degree of polynomial fitting.

[0019] Furthermore, the weight coefficients of each eigenvalue are determined, including: setting the fitting function The common effective time interval of is taken as the minimum domain of definition; according to the minimum domain of definition, by fitting the function Calculate the eigenvalues ​​at the sampling time points , as sample data; according to the sample data, construct the acoustic emission eigenvalue sample matrix F: ,in, Indicates the i-th eigenvalue at time point The value of m is the number of sampling points; according to the acoustic emission eigenvalue sample matrix F, the eigenvalue is calculated to The correlation coefficient matrix R between them, the elements in the matrix R Represents eigenvalues and Correlation coefficient; Calculate the matrix eigenvalue of the correlation coefficient matrix R to , and the corresponding eigenvector ; Calculate the eigenvalues ​​of each matrix Contribution rate and cumulative contribution rate, select the matrix eigenvalue whose cumulative contribution rate is greater than the threshold ; According to the selected matrix eigenvalue and the corresponding eigenvector , construct the factor loading matrix ρ, where ; According to the factor loading matrix ρ, determine the eigenvalue to The weight coefficient , .

[0020] In particular, the principal component information is extracted by matrix eigenvalue decomposition, redundancy is removed, and the information most sensitive to the change of coal and rock mass state is retained; the factor load matrix construction realizes the mapping from the original feature to the principal component space and optimizes the feature combination; the weight coefficient The calculation method takes into account the feature contribution rate and realizes the reasonable quantification of feature importance.

[0021] Further, The expression is as follows: ;in, is the eigenvalue and The correlation coefficient of ;and , and They are and The average value of Represents the characteristic value at the kth time point The value of Represents the characteristic value at the kth time point The value of; k represents the sequence number of the time point, .

[0022] Furthermore, the demarcation time point in the coal-rock mass failure process is determined, including: calculating the first-order derivative function The time point is taken as the extreme point or stationary point of the fitting function; the second-order derivative function is calculated The time point is taken as the inflection point of the fitting function; according to the first-order derivative function and the second-order derivative function , as well as extreme points or stationary points, and inflection points, to determine the demarcation time points of different failure nodes of coal and rock mass .

[0023] In particular, first-order derivative analysis Accurately identify the extreme points or stationary points of characteristic value changes, which correspond to the turning points of energy accumulation and release in the process of coal and rock mass destruction; second-order derivative analysis Accurately capture the inflection point of the curve to reflect the acceleration or deceleration process of the change in the coal and rock mass destruction rate.

[0024] Furthermore, the destruction stages of coal and rock masses are divided into the following categories: ,in, is the jth revised demarcation time point, is the jth demarcation time point obtained by the derivative analysis of the fitting function of the i-th eigenvalue, is the weight coefficient of the ith eigenvalue; according to the modified demarcation time point , the destruction process of coal rock mass is divided into four stages: Stable period I: Corresponding to the coal rock crack closure stage; Development stage II: Corresponding to the linear elastic deformation stage of coal rock mass; energy storage period III: Corresponding to the crack expansion stage of coal rock mass; Destruction stage IV: Corresponding to the post-peak destruction stage of coal rock mass.

[0025] Compared with the existing technology, the advantages of this application are:

[0026] In improving the early warning analysis of coal rock instability, existing technologies use simple time domain characteristics (such as the number of events and duration) and basic frequency domain characteristics (such as peak frequency and frequency center). However, they lack cumulative effect analysis and cannot reflect the failure characteristics of coal rock. On the other hand, they often ignore the interaction between fluid and coal rock, resulting in the inability to identify the characteristic changes of fluid-induced coal rock failure in different stages, making it difficult to capture the evolution process of microcracks under the action of fluid, which ultimately affects the timeliness and accuracy of early warning.

[0027] This application improves the accuracy and timeliness of coal-rock fracture instability warning by constructing a multivariate feature cumulative effect analysis system and integrating a three-stage division method of the fluid-coal-rock interaction mechanism.

[0028] Specifically, on the one hand, this application constructs a system including five cumulative characteristic values to A multi-characteristic system, each characteristic clearly corresponds to a different aspect of the microscopic damage of the coal and rock mass: the cumulative ringing eigenvalue Reflects the history of changes in the number of microcracks; cumulative b eigenvalue Record the evolution of crack size distribution; accumulate the main control frequency characteristic value Characterize the trend of microcrack growth rate; cumulative amplitude characteristic value Record the cumulative effect of energy release intensity; acoustic emission spatial characteristic value Characterize the cumulative changes in spatial dimensions; through feature accumulation, the system can "memorize" the historical path of coal and rock mass destruction, avoiding the defects of existing technologies that only focus on instantaneous parameters and ignore cumulative effects, and realize continuous and comprehensive monitoring of the progressive destruction process of coal and rock masses.

[0029] On the other hand, the fluid-coal rock interaction process is divided into three stages: initial adsorption, diffusion penetration, and crack expansion. Through the dual criteria of "acoustic emission event rate + frequency band energy ratio", the system can accurately identify the entire process of fluid from surface adsorption to internal penetration and then to induced crack expansion, thereby improving the early warning accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 This is a multi-signal fusion analysis method for coal rock mass fracture instability in this embodiment;

[0031] Figure 2 Schematic diagram of the arrangement of the acoustic emission sensor probes of this embodiment;

[0032] Figure 3 is an acoustic emission wavelet waveform diagram of a certain rupture event in this embodiment;

[0033] Figure 4 is a waveform diagram after wavelet transformation in this embodiment;

[0034] Figure 5 is the fitting function curve of this embodiment;

[0035] Figure 6 is the first-order derivative of the fitting function of this embodiment;

[0036] Figure 7 is the second-order derivative of the fitting function of this embodiment;

[0037] Figure 8 This is the rock failure stage division of this embodiment. DETAILED DESCRIPTION

[0038] The present application is described in detail below with reference to the accompanying drawings and specific embodiments.

[0039] like Figure 1As shown in the figure, acoustic emission data is collected during the deformation and destruction of coal and rock mass; the acoustic emission data includes acoustic emission signals and spatial data corresponding to the acoustic emission signals; the acoustic emission signals are processed in the time-frequency domain to obtain the cumulative ringing characteristic value of the acoustic emission , cumulative b eigenvalue , the cumulative main control frequency characteristic value of acoustic emission and cumulative amplitude eigenvalues ; Extract the cumulative energy value of acoustic emission from the spatial data as the spatial characteristic value of acoustic emission .

[0040] Fitting eigenvalues ​​separately To eigenvalue The fitting function ; According to the fitting function , calculate the eigenvalue to The correlation coefficient matrix R between them; according to the correlation coefficient matrix R, determine the weight coefficient of each eigenvalue ; Get the fitting function The first derivative of and the second-order derivative function , and according to the first-order and the second-order derivative function , determine the demarcation time point in the coal rock destruction process ; Using weight coefficient Demarcation time point Make corrections to get the corrected demarcation time point , according to the revised demarcation time point The destruction process of coal rock mass is divided into stable period, development period, energy storage period and destruction period; the time range of coal rock mass entering the destruction period is obtained as a precursor warning signal of coal rock mass instability.

[0041] Acoustic emission signal parameters for monitoring coal and rock mass deformation and failure processes were set, and a model for collecting and identifying signal waveforms and spatial parameters was established. Acoustic emission probes were used to detect coal and rock mass adhesion, and acoustic emission waveform and signal spatial parameter data were collected. First, this example was demonstrated using laboratory experiments. A 100mm×100mm×100mm CO2-water-coal sample was prepared to simulate an environment similar to that of a carbon-storage coal pillar in underground CO2 geological storage. A true triaxial loading fracturing test was then conducted on the coal, and monitoring was performed using acoustic emission methods.

[0042] The parameters for collecting the spatial positioning features of the acoustic emission signal are set as follows: In order to accurately locate the spatial position of the acoustic emission signal, two acoustic emission probes are set at the front, bottom, and right loading plate of the true triaxial warehouse. The layout of the acoustic emission probes is as follows: Figure 2 shown.

[0043] The parameters for collecting acoustic emission waveform signals were set as follows: the operating frequency range was 125kHz to 750kHz; the resonant frequency was 300kHz; the threshold and gain were both 40dB; the upper and lower limits of the analog filter were 1 and 400kHz, respectively, and the sampling frequency was 2MHz.

[0044] The parameters of the acoustic emission waveform signal recognition model are set as follows: time, channel, rise time, count, energy, duration, amplitude, average frequency, peak amplitude, back-calculated frequency, initial frequency, signal strength, and absolute energy.

[0045] Next, the acoustic emission signal data files and their spatial parameter data representing deformation and fracture events during the compression process of the coal and rock mass are extracted; the acoustic emission signal data is collected and extracted. Each acoustic emission waveform and spatial positioning data collected by the acoustic emission monitoring system is automatically stored as a *.csv format data file. Each data file uses an acoustic emission waveform to record a fracture event and its spatial positioning information. Each fracture event data file is processed and output as a *.TXT and *.XLS file for subsequent processing of the acoustic emission time-domain, frequency-domain, and spatial characteristic signals.

[0046] Then, the obtained acoustic emission data files are calculated and processed to obtain the acoustic emission time domain signals: acoustic emission ringing count eigenvalues, acoustic emission b-value eigenvalues, etc.; frequency domain signals: acoustic emission main control frequency eigenvalues ​​and corresponding acoustic emission amplitude eigenvalues, etc.; acoustic emission spatial eigenvalues.

[0047] Extraction of acoustic emission time domain signals: Extract acoustic emission ring counts from the raw data files obtained from the experiment. Based on the extracted acoustic emission ring count parameters, the cumulative acoustic emission ring counts are used as the acoustic emission ring count feature value.

[0048] According to the relationship between earthquake magnitude and frequency: , where M is the earthquake magnitude; N is the frequency of earthquakes with magnitudes between M + ∆M; and a and b are constants. The acoustic emission b-value is calculated by dividing the amplitude by ten instead of the magnitude M. The calculation relationship is as follows: , where E is the absolute energy of acoustic emission; N(E) is the number of acoustic emission hits with an absolute energy greater than E.

[0049] In particular, the dynamic changes in coal rock properties during CO2 injection lead to the problem of fixed window parameters not being adaptable. During the carbon storage process, the internal structure of the coal rock mass constantly changes with the CO2 adsorption, diffusion and infiltration processes, which makes the traditional sliding window method with fixed parameters have obvious defects: if the window is too large, it will mask key short-term changes, and if the window is too small, it will introduce too much noise. Therefore, this application establishes a correlation function between CVR and window size to achieve automatic adjustment of window parameters according to changes in coal rock structure and adapt to the signal characteristics of different injection stages.

[0050] Detailed, set the initial and step length , the acoustic emission data is segmented using a sliding window algorithm; the cumulative variance ratio (CVR) of the acoustic emission signal in the current window is calculated, where CVR represents the ratio of the signal energy variance to the average energy in the window; based on CVR, the optimal window for the current window is calculated. , , γ is a positive coefficient, is the initial window length; according to the optimal window , calculate the acoustic emission b value of each window by the least square method; accumulate the acoustic emission b values ​​in each window in chronological order to obtain the acoustic emission cumulative b eigenvalue .

[0051] ,in, is the variance of energy within the calculation window; is the average value of energy within the calculation window.

[0052] The adaptive window algorithm dynamically adjusts the window size based on the signal energy variance to mean energy ratio (CVR), adaptively matching the window length to the signal characteristics. When the signal energy fluctuates dramatically (large CVR values), the window length is automatically reduced, improving temporal resolution and capturing rapidly changing microcrack information. When the signal is stable (small CVR values), the window length is automatically expanded, enhancing statistical stability and suppressing noise interference. Compared to fixed-window algorithms, the adaptive window algorithm achieves optimal signal-to-noise ratio data processing at all stages of the fluid (e.g., CO2) injection process. A nonlinear mapping between window size and signal volatility is established through the e-exponential function, avoiding the problems of over-adjustment or under-response.

[0053] Extraction of acoustic emission frequency domain signals: First, calculate the acoustic emission event rate and frequency band energy proportion based on the acoustic emission data; set the first threshold of the acoustic emission event rate and the second threshold ;in, ; Set the third threshold of the frequency band energy ratio of the low frequency band, the middle frequency band and the high frequency band respectively , the fourth threshold and the fifth threshold ; When the acoustic emission event rate is less than or equal to , and the frequency band energy of the low frequency band accounts for more than When the acoustic emission event rate is greater than and less than , and the frequency band energy of the mid-frequency band is greater than When the acoustic emission event rate is greater than , and the frequency band energy of the high frequency band accounts for more than When , it is judged to be the crack expansion stage.

[0054] In this embodiment, Represents the critical point of transition from surface adsorption to internal diffusion, reflecting the physical critical state where fluid molecules begin to migrate from the surface to the interior of the coal rock mass, with a value range of 5-20 events / minute; Characterizes the transition point from steady-state diffusion to unsteady-state crack expansion, with a value range of 30-100 events / minute; Reflects the dominance of low-frequency energy in the surface adsorption process, with a value range of 40%-70%; Represents the energy ratio of internal pore deformation to initial microcrack expansion, ranging from 30% to 60%; Characterizes the proportion of high-frequency energy during the rapid expansion and connectivity of cracks, with a value range of 20%-50%.

[0055] Then, the window lengths of the sliding windows in the initial adsorption stage, diffusion and penetration stage, and crack expansion stage are set as 、 and ;in, ; among them, among them, The value range is 2000-5000 sampling points. The value range is 800-2000 sampling points. The value range is 200-800 sampling points.

[0056] According to the window 、 and , segment the acoustic emission wave data of each stage to obtain the acoustic emission time domain signals in multiple time windows ;

[0057] Perform wavelet transform on the acoustic emission data in each window to obtain the wavelet coefficients; the wavelet transform uses the following formula:

[0058] ,in, is the wavelet coefficient, s is the scale parameter, τ is the translation parameter, and f(t) is the time domain signal; is the conjugate function of the wavelet basis function. In particular, this embodiment uses the Daubechies wavelet family (dbN) as the wavelet basis function, where N is an even number between 4 and 10. This wavelet basis function has excellent time-frequency positioning capabilities and can effectively identify non-stationary signal characteristics and mutation points in coal and rock masses during the fluid (e.g., CO2) injection process, from adsorption and diffusion to crack expansion stages.

[0059] Wavelet transform has better time-frequency localization characteristics than traditional Fourier transform, and can simultaneously obtain the time domain and frequency domain information of the signal; multi-band analysis realizes the classification and identification of coal rock cracks of different scales: low frequency mainly corresponds to large-scale structural changes, medium frequency corresponds to mesoscopic crack expansion, and high frequency corresponds to microcrack initialization; by analyzing the main control frequency of each band and maximum amplitude The changing laws of the coal and rock mass are accurately described during the injection of fluids (such as CO2).

[0060] According to the wavelet coefficients, the wavelet energy distribution is calculated and divided into low-frequency band, medium-frequency band and high-frequency band; in this application, the low-frequency band ranges from 1 to 30 kHz, corresponding to the large-scale structural changes of coal and rock masses; the medium-frequency band ranges from 30 to 100 kHz, corresponding to the mesoscopic crack propagation process; the high-frequency band ranges from 100 to 300 kHz, corresponding to the microcrack initialization process; this frequency band division can effectively distinguish the microstructural changes of coal and rock masses at different scales during the injection of fluids (such as CO2).

[0061] Perform peak analysis on each frequency band to extract the main control frequency of the low-frequency band in the initial adsorption stage and the corresponding maximum amplitude , mid-band main control frequency in the diffusion and penetration stage and the corresponding maximum amplitude , high-frequency band master frequency in the crack expansion stage and the corresponding maximum amplitude ; Among them, the main control frequency Indicates the frequency of the peak of the wavelet energy distribution in the corresponding frequency band; accumulates the master frequency of each time window , get the cumulative master control frequency characteristic value ; Accumulate the maximum amplitude of each time window , and get the cumulative amplitude eigenvalue The acoustic emission wavelet waveform of a certain rupture event extracted in this embodiment is as follows: Figure 3 As shown in the figure, the waveform after wavelet transform is as follows Figure 4 shown.

[0062] Extraction of acoustic emission spatial signals: Extract the acoustic emission energy parameters through the spatial positioning information file in the original data file, and use the acoustic emission cumulative energy as the acoustic emission spatial characteristic value.

[0063] The calculated acoustic emission time-domain-frequency-space characteristic values ​​were normalized by using the deviation standardization method to eliminate the dimension difference, and the data were normalized. The acoustic emission time-domain-frequency-space characteristic fitting function was established for the normalized result data.

[0064] Since the acoustic emission time-frequency-space signal responses accompanying the coal destruction process have different dimensions, in order to eliminate the dimensional differences of this information, the normalization method of deviation standardization is used to process it and obtain a dimensionless scalar.

[0065] The expression of the deviation standardization method is: Where: is the normalized function value of the monitoring data, x is the monitoring data; k is the ordinal number of different multi-source information parameters; the value range of the monitoring data is [min, max].

[0066] Then, the normalized acoustic emission time-frequency-space eigenvalue data were fitted to construct Figure 5 The polynomial function of . The fitting function expression is: , where: In the function, i is 1, 2, ..., 5, which represent the cumulative ringing characteristic value of the acoustic emission. ;Acoustic emission cumulative b value characteristic value ;Acoustic emission cumulative main control frequency characteristic value ;Acoustic emission cumulative amplitude characteristic value ; Acoustic emission spatial characteristic value ; are the fitting function coefficients, is the constant term of the polynomial, which represents the initial state value of the eigenvalue; j represents the highest degree of polynomial fitting.

[0067] The value of the fitted function as follows:

[0068] In this embodiment, the fitting function parameters of the five acoustic emission characteristic values ​​are shown in the following table, where R2 represents the degree of fitting of the fitting function. The closer the value of R2 is to 1, the higher the degree of fitting.

[0069] As shown in the table below The fitting degree indices are 0.9869, 0.9954, 0.9415, 0.9749 and 0.9108 respectively, indicating that the fitting degree of the constructed polynomial function is relatively ideal, as shown in Table 1;

[0070] Table 1 Fitting function parameters

[0071]

[0072] The closely related quantity characteristics of the acoustic emission time-frequency-space eigenvalue fitting function are analyzed to obtain the influence weight of the acoustic emission time-frequency-space eigenvalue in the coal and rock mass destruction process;

[0073] First, by taking the minimum definition domain of the monitoring information in the above polynomial fitting function as the benchmark, the definition domain of the sample data matrix is ​​determined, and then the acoustic emission time domain-frequency domain-space eigenvalue sample matrix F is obtained:

[0074]

[0075] The correlation coefficient matrix R is calculated from the sample matrix F using the following formula:

[0076] The calculation formula is:

[0077] Where: The original variable and The correlation coefficient of and They are and The average value of , i, j = 1, 2, ..., n; and , the correlation coefficient matrix R is:

[0078] ;

[0079] The eigenvalues ​​of the correlation coefficient matrix R are calculated according to the following formula to and cumulative contribution rates, and then calculate the contribution rates and cumulative contribution rates of the main components. ; ; Where: E is the unit matrix; m is the number of eigenvalues ​​that determine the main component information, that is, the number of main components, see Table 2 for details.

[0080] Table 2 Matrix eigenvalues ​​and cumulative contribution rates

[0081]

[0082] To ensure that the main components reflect the original variable information to the greatest extent possible, the cumulative contribution rate is required to reach more than 80%. As shown in the table above, the cumulative contribution rate of the eigenvalue λ is 84.136%.

[0083] Eigenvalues ​​of the correlation coefficient matrix R The eigenvector corresponding to the eigenvalue The combined expression of forms the factor loading matrix ρ, that is, the factor loading matrix ρ is determined by the eigenvalue.

[0084] Therefore, the eigenvalue is 5.885 and the factor loading matrix ρ is [0.986, 0.839, 0.688, 0.991, 0.945, 0.958, 0.985].

[0085] According to the relationship between the eigenvalue and the factor loading matrix ρ, the function coefficient matrix is ​​obtained , where : ; By the coefficient matrix Get the principal component analysis function F(x):

[0086] ; F(x) can be regarded as The comprehensive variables mainly explain the comprehensive effects of the five acoustic emission characteristic parameters: acoustic emission ringing count, acoustic emission b value, acoustic emission main control frequency and amplitude, and acoustic emission spatial eigenvalue.

[0087] The coefficient of F(x) is the influence weight of each characteristic value parameter of acoustic emission in time domain, frequency domain and space. Therefore, the order of each variable reflecting the damage of the specimen from strong to weak is: ;

[0088] The derivative function of the acoustic emission time-frequency-space characteristic fitting function is solved, and the boundary time points of the coal and rock mass in different failure stages are determined according to the concavity and convexity and increase and decrease changes of the derivative function curve.

[0089] The derivative function of the acoustic emission time domain-frequency domain-space eigenvalue fitting function is solved to obtain the first-order derivative function of the fitting function as follows: Figure 6 and two The derivative function is as follows Figure 7 When the derivative function curve intersects the x-axis, the value of the derivative function is 0, and this intersection is the stationary point or inflection point. Based on the concave-convex and increasing-decreasing characteristics of the derivative function curve, the boundary time points of the coal body in different destruction stages can be determined;

[0090] The boundary time points of different damage stages are weightedly corrected through the obtained derivative function extreme points, inflection points and principal component analysis results: ,in, is the jth revised demarcation time point, is the jth demarcation time point obtained by the derivative analysis of the fitting function of the i-th eigenvalue, is the weight coefficient of the ith eigenvalue; according to the modified demarcation time point , the destruction process of coal rock mass is divided into four stages: Stability period (I): Corresponding to the coal rock crack closure stage; development stage (II): Corresponding to the linear elastic deformation stage of coal rock mass; energy storage period (III): Corresponding to the crack expansion stage of coal rock mass; destruction stage (IV): Corresponding to the post-peak destruction stage of coal and rock mass;

[0091] The obtained derivative function extreme points, inflection points and corresponding main component analysis results are used to perform weighted correction on the boundary time points of different failure stages by averaging the factor load matrix, as shown in the following table. Then, the precursor characteristics of the rock failure process are divided into the stable period I to the failure period IV, as shown in Table 3. The coal body failure stage diagram is obtained as shown in Figure 8 As shown;

[0092] Table 3 Classification of rock failure stages

[0093]

[0094] Based on the stage division results, the time range when the coal rock mass enters the destruction stage IV is used as the key discriminant information for the precursor warning of rock instability, as shown in Table 4 for details.

[0095] Table 4 Relationship between the precursor limit and bearing limit in the failure stage

[0096]

[0097] For example Figure 8 As shown, the classification of stages I to IV is combined with the stress-strain curve of the coal. Table 4 shows that during the coal failure process, stages II and III account for a significant portion of the time, as the coal transitions from the elastic stage to the crack expansion stage, representing the energy accumulation phase of the coal failure process. Stages I and IV, on the other hand, account for a smaller proportion of the time. Stage IV represents the process in which internal cracks develop and expand within the coal, gradually reaching their bearing capacity and leading to rapid failure. This corresponds to the crack expansion and post-peak failure phase of coal failure. Therefore, it is important to strengthen monitoring and early warning of stages III and IV.

[0098] The invention of the present application and its implementation methods are described schematically above. This description is not restrictive. Without departing from the spirit or basic features of the present application, the present application can be implemented in other specific forms. What is shown in the accompanying drawings is only one of the implementation methods of the invention of the present application, and the actual structure is not limited to this. Therefore, if a person of ordinary skill in the art is inspired by it, without departing from the purpose of the invention, a structural method and embodiment similar to the technical solution are designed without creativity, which should all fall within the scope of protection of the present application. In addition, the word "including" does not exclude other elements or steps, and the word "one" before an element does not exclude the inclusion of "multiple" elements. Words such as first and second are used to indicate names and do not indicate any specific order.

Claims

1. A multi-signal fusion analysis method for coal rock fracture instability, characterized by: include: Collecting acoustic emission data during the deformation and failure of coal and rock masses; the acoustic emission data includes acoustic emission signals and spatial data corresponding to the acoustic emission signals; Perform time-frequency domain processing on the acoustic emission signal to obtain the cumulative ringing characteristic value of the acoustic emission , cumulative b eigenvalue , the cumulative main control frequency characteristic value of acoustic emission and cumulative amplitude eigenvalues ; Extract the cumulative energy value of acoustic emission from the spatial data as the spatial characteristic value of acoustic emission ; Fitting eigenvalues ​​separately To eigenvalue The fitting function ,include: ； in, They correspond to five eigenvalues ​​respectively, x is the time variable, which represents the duration from the beginning of the interaction between the fluid and the coal rock to the present. is the j-th coefficient of the fitting function of the i-th eigenvalue; is the constant term of the polynomial, indicating the initial state value of the eigenvalue; j indicates the highest degree of polynomial fitting; According to the fitting function , calculate the eigenvalue to The correlation coefficient matrix R between them; according to the correlation coefficient matrix R, determine the weight coefficient of each eigenvalue , including: setting the fitting function The common effective time interval of is taken as the minimum domain of definition; according to the minimum domain of definition, by fitting the function Calculate the eigenvalues ​​at the sampling time points , as sample data; according to the sample data, construct the acoustic emission eigenvalue sample matrix F; according to the acoustic emission eigenvalue sample matrix F, calculate the eigenvalue to The correlation coefficient matrix R between them, the elements in the matrix R Represents eigenvalues and Correlation coefficient; Calculate the matrix eigenvalue of the correlation coefficient matrix R to , and the corresponding eigenvector ; Calculate the eigenvalues ​​of each matrix Contribution rate and cumulative contribution rate, select the matrix eigenvalue whose cumulative contribution rate is greater than the threshold ; According to the selected matrix eigenvalue and the corresponding eigenvector , construct the factor loading matrix ρ, where ; According to the factor loading matrix ρ, determine the eigenvalue to The weight coefficient , ; Get the fitting function The first derivative of and the second-order derivative function , and according to the first-order and the second-order derivative function , determine the demarcation time point in the coal rock destruction process ; Using weight coefficients Demarcation time point Make corrections to get the corrected demarcation time point , according to the revised demarcation time point , the destruction process of coal rock mass is divided into stable period, development period, energy storage period and destruction period; The time range when the coal rock mass enters the destruction period is obtained as a precursor warning signal for coal rock mass instability.

2. The multi-signal fusion analysis method for coal rock fracture instability according to claim 1 is characterized by: Cumulative ringing characteristic value It indicates the cumulative value of ringing counts during the acoustic emission process, reflecting the number of microcracks inside the coal and rock mass; Cumulative b eigenvalue It represents the cumulative value of the slope coefficient of the acoustic emission signal amplitude-frequency distribution, reflecting the changes in the distribution of crack scale inside the coal rock mass during the interaction between fluid and coal rock mass; Cumulative master control frequency characteristic value It represents the time series cumulative value of the frequency corresponding to the maximum amplitude in the acoustic emission signal spectrum, reflecting the growth rate of microcracks inside the coal and rock mass; Cumulative amplitude eigenvalue It represents the time series cumulative value of the maximum amplitude of the acoustic emission signal, reflecting the intensity change of energy released by microcracks inside the coal and rock mass.

3. The multi-signal fusion analysis method for coal-rock mass fracture instability according to claim 2 is characterized by: Get the cumulative b eigenvalue ,include: Extract the absolute energy E of acoustic emission and the number of hits with absolute energy greater than E from the collected acoustic emission data ; Establish the absolute energy E of acoustic emission and the number of hits The relationship between: , where a and b are constants; Setting the initial and step length , the sliding window algorithm is used to segment the acoustic emission data; Calculate the cumulative variance ratio (CVR) of the acoustic emission signal in the current window. CVR represents the ratio of the signal energy variance to the average energy in the window. According to CVR, calculate the optimal window of the current window , , γ is a positive coefficient; According to the optimal window , calculate the acoustic emission b value of each window by the least square method; The acoustic emission b values ​​in each window are accumulated in chronological order to obtain the cumulative b characteristic value of the acoustic emission .

4. The multi-signal fusion analysis method for coal rock mass fracture instability according to claim 3 is characterized by: Cumulative master control frequency characteristic value and cumulative amplitude eigenvalues ,include: According to the acoustic emission data, the interaction process between fluid and coal rock is divided into the initial adsorption stage, diffusion and penetration stage and crack expansion stage. The window lengths of the sliding windows in the initial adsorption stage, diffusion and penetration stage, and crack expansion stage are set as 、 and ;in, ; According to the window 、 and , segment the acoustic emission wave data of each stage to obtain the acoustic emission time domain signals in multiple time windows ; Perform wavelet transform on the acoustic emission data in each window to obtain wavelet coefficients; According to the wavelet coefficients, the wavelet energy distribution is calculated and divided into low-frequency band, medium-frequency band and high-frequency band; Perform peak analysis on each frequency band to extract the main control frequency of the low-frequency band in the initial adsorption stage and the corresponding maximum amplitude , mid-band main control frequency in the diffusion and penetration stage and the corresponding maximum amplitude , high-frequency band master frequency in the crack expansion stage and the corresponding maximum amplitude ; Among them, the main control frequency Indicates the frequency at which the peak of wavelet energy distribution in the corresponding frequency band is located; Accumulate the master control frequency of each time window , get the cumulative master control frequency characteristic value ; Accumulate the maximum amplitude of each time window , and get the cumulative amplitude eigenvalue .

5. The multi-signal fusion analysis method for coal rock mass fracture instability according to claim 4 is characterized by: The process of fluid-coal rock interaction is divided into the initial adsorption stage, diffusion and penetration stage, and crack expansion stage, including: Calculate the acoustic emission event rate and frequency band energy ratio based on the acoustic emission data; Set the first threshold of acoustic emission event rate and the second threshold ;in, ; Set the third threshold of the frequency band energy proportion of the low frequency band, the middle frequency band and the high frequency band respectively , the fourth threshold and the fifth threshold ; When the acoustic emission event rate is less than or equal to , and when the frequency band energy proportion of the low frequency band is greater than ε3, it is judged to be the initial adsorption stage; When the acoustic emission event rate is greater than and less than , and the frequency band energy of the mid-frequency band is greater than When , it is judged to be the diffusion and penetration stage; When the acoustic emission event rate is greater than , and the frequency band energy of the high frequency band accounts for more than When , it is judged to be the crack expansion stage.

6. The multi-signal fusion analysis method for coal rock mass fracture instability according to claim 1 is characterized by: The expression is as follows: ； in, is the eigenvalue and The correlation coefficient of ;and , and They are and The average value of Represents the characteristic value at the kth time point The value of Represents the characteristic value at the kth time point The value of; k represents the sequence number of the time point, .

7. The multi-signal fusion analysis method for coal and rock mass fracture instability according to claim 6 is characterized by: Determine the demarcation time point in the coal and rock mass failure process, including: Calculate the first-order derivative function The time point of is taken as the extreme point or stationary point of the fitting function; Calculate the second-order derivative function The time point of is taken as the inflection point of the fitting function; According to the first-order derivative function and the second-order derivative function , as well as extreme points or stationary points, and inflection points, to determine the demarcation time points of different failure nodes of coal and rock mass .

8. The multi-signal fusion analysis method for coal and rock mass fracture instability according to claim 7 is characterized by: Use weight coefficients to modify the demarcation time points, including: ; in, is the jth revised demarcation time point, is the jth demarcation time point obtained by the derivative analysis of the fitting function of the i-th eigenvalue, is the weight coefficient of the i-th eigenvalue.

Images