Real-time warning method for disturbance catastrophe of high-stress rock mass based on comprehensive signal of fracture

Abstract

The application provides a high-stress rock mass disturbance disaster real-time warning method based on a broken comprehensive signal, belongs to the technical field of geotechnical engineering, and constructs a comprehensive acoustic parameter ASCP by fusing multiple parameters such as amplitude, main frequency, energy distribution, crack fractal characteristics and rock mass breaking event density in an acoustic emission signal; the calculation process combines the contribution weights of different parameters to the damage stage, and the weight values are jointly calibrated through data analysis. Based on a large amount of engineering site monitoring and indoor true triaxial test data, the evolution law of ASCP in the rock mass damage process is analyzed, the starting warning threshold of the ASCP evolution curve and the multiple warning thresholds of the ASCP disaster warning threshold are proposed, and the warning accuracy and timeliness are improved. The warning method of the application comprehensively considers various parameters of the acoustic signal, establishes a multiple threshold warning method, comprehensively considers the contribution degrees of various main sensitive parameters of the acoustic emission of the high-stress rock mass disturbance and damage, and improves the warning accuracy, timeliness and high efficiency and economy.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of geotechnical engineering, and particularly relates to a high-stress rock mass disturbance disaster real-time early warning method based on a comprehensive signal of rock failure, which is suitable for safety monitoring and disaster prevention and control of various deep geotechnical engineering. BACKGROUND

[0002] In deep geotechnical engineering (such as deep mine exploitation and long and large tunnel construction), the rock mass is in a high-stress state, and when affected by excavation, blasting and other dynamic disturbances, sudden failure and instability are prone to occur, which may cause rock burst, collapse and other disasters, and seriously threaten the safety of engineering and personnel life. At present, the industry mainly realizes early warning by monitoring acoustic emission (AE) signals generated when the rock mass fails, but the existing technology has obvious defects: most methods only rely on a single parameter (such as amplitude, energy) in the acoustic emission signal to judge the rock mass state, which cannot comprehensively reflect the complex process of rock failure (such as different stages of crack initiation, expansion and penetration), and is prone to false positives or false negatives due to one-sided parameters; the early warning threshold is mostly based on experience, and is not dynamically adjusted in combination with the rock mass characteristics (such as lithology, structure surface distribution) and laboratory test data in the engineering site, so it is difficult to adapt to the disturbance conditions of different engineering scenes; there is a lack of accurate identification of rock failure precursors, and early warning is only given when the rock mass is close to instability, so there is little time left for on-site emergency disposal, and the disaster risk cannot be effectively avoided.

[0003] Therefore, there is an urgent need for a high-stress rock mass disturbance disaster early warning method that can integrate multiple sensitive parameters, dynamically adjust thresholds and accurately identify precursors, to solve the problems of "low accuracy, poor adaptability and untimely" in the existing technology. SUMMARY

[0004] The purpose of the present application is to provide a high-stress rock mass disturbance disaster real-time early warning method based on a comprehensive signal of rock failure, which intuitively represents the precursor law of rock failure, provides an accurate early warning time, and improves the accuracy and timeliness of early warning.

[0005] In order to achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0006] A high-stress rock mass disturbance disaster real-time early warning method based on a comprehensive signal of rock failure, comprising the following steps:

[0007] S1, collecting acoustic emission signals of rock failure under dynamic disturbance in the engineering site by a sound signal sensor, extracting sound signal characteristic parameters and screening sensitive characteristic parameters, wherein the sound signal characteristic parameters include waveform amplitude, fractal dimension, LgN / b value, main frequency, activity coefficient and real volume;

[0008] S2, under true three-dimensional stress conditions, carry out rock mass dynamic disturbance failure tests under different test conditions, synchronously collect acoustic emission signals of the indoor tests and extract sensitive characteristic parameters consistent with step S1, and establish a database;

[0009] S3, based on the various sensitive characteristic parameters obtained in steps S1 and S2, an ASCP function of a comprehensive characteristic parameter is constructed through a weighted fusion method, and an expression of the ASCP function is , wherein is a weight coefficient of each sensitive characteristic parameter; is a proportional function of each sensitive characteristic parameter;

[0010] S4, the ASCP evolution curves of the rock mass failure at each stage are compared between the engineering site and the indoor tests, the change trend and evolution law of the ASCP are analyzed through a quantitative method, and a one-to-one correspondence between the ASCP and the stress state and the failure degree of the rock mass is established;

[0011] S5, according to the change trend and evolution law of the ASCP evolution curve in step S4, a starting early warning threshold is determined through a perpendicular bisector method, a disaster early warning threshold is determined through critical failure data of the indoor tests, and multiple early warning thresholds are formed;

[0012] S6, the ASCP function is inversed based on the rock mass dynamic disturbance failure results of the engineering site and the indoor tests in steps S1 and S2, and the weight coefficient and the threshold interval are dynamically adjusted;

[0013] S7, the data and early warning results of steps S1-S6 are stored in a sample library, an ASCP evolution-threshold criterion correlation model is constructed through MATLAB interpolation processing and neural network training;

[0014] S8, the trained ASCP evolution-threshold criterion correlation model is embedded into a monitoring system, acoustic emission signals of the engineering site are collected in real time and the ASCP is calculated, early warning is triggered when the ASCP reaches the starting threshold, and emergency warning is triggered when the ASCP reaches the disaster threshold.

[0015] As a further description of the above technical solutions: the different test conditions in step S2 include different stress conditions, different stress levels, different structural planes and different surrounding rock unit types.

[0016] As a further description of the above technical solutions, the weight coefficient in step S3 is determined by an AHP hierarchical analysis method, and the specific process is as follows:

[0017] A hierarchical evaluation model is constructed according to the monitoring results, including a target layer, a criterion layer and a scheme layer;

[0018] A judgment matrix is constructed The two sensitive characteristic parameters are compared with each other, and the weight of each criterion layer to the target layer is determined, and a judgment matrix is obtained The middle element satisfies , and ;

[0019] The hierarchical single sorting is performed, all elements in the judgment matrix are compared with each other, and hierarchical sorting is carried out, and the important order is arranged, and the specific calculation is as follows:

[0020] , ;

[0021] The weight matrix is obtained, the maximum eigenvalue and C.I. value are calculated, and the random consistency index R.I. value is obtained by using Satty simulation multiple times, the C.R. value is calculated, and whether the consistency passes is judged:

[0022] , , ;

[0023] If C.R.<0.1, it indicates that the judgment matrix is within the error range, and the elements in it are used as weight coefficients ; if C.R.≥0.1, the judgment matrix needs to be modified.

[0024] As a further description of the above technical scheme: the proportional function in step S3 includes an amplitude proportional function , a fractal dimension proportional function , an LgN / b value proportional function and other sensitive characteristic parameter proportional functions;

[0025] Wherein x is the number of sensors, V i is the waveform voltage value of the i-th time node, B1 is the maximum waveform voltage value of the time node; D i is the fractal dimension value of the i-th time node, B2 is the maximum fractal dimension value of the time node; LgN i / b i is the LgN / b value of the i-th time node, and B3 is the maximum LgN / b value of the time node.

[0026] As a further description of the above technical scheme, the change trend and evolution law of ASCP are analyzed by the quantification method in step S4, which is specifically:

[0027] The ASCP time sequence is normalized;

[0028]

[0029] The sliding window method (window length >= 10 sampling periods) is adopted to smooth the noise.

[0030] The curve slope k of the ASCP evolution curve is calculated based on linear regression, and k >= 0.05 / min is defined as a significant upward trend (corresponding to a disaster precursor), 0.01 <= k < 0.05 / min is defined as a stable fluctuation (corresponding to a gradual destruction), and k < 0.01 / min is defined as a stable state.

[0031] The CUSUM algorithm is used to set the control limit L = 3 * (mean value of the last 10 data points) / 10. When the cumulative sum exceeds the limit and the synchronous acoustic emission single event energy increases by more than 200%, the critical point of disaster is determined; by comparing the field and experimental data, the evolution law of ASCP under different disturbance intensities, rock mass characteristics and failure modes is studied, thereby providing theoretical basis and data support for subsequent warning threshold setting.

[0032] As a further description of the above technical solution, the starting warning threshold determined by the perpendicular bisector method in step S5 is that the upper and lower straight line segments of the ASCP evolution curve are extended to intersect at a first point, the perpendicular bisectors of the upper and lower straight line segments are drawn to intersect at a second point, and the first point and the second point are connected to intersect the ASCP evolution curve at a third point, and the ASCP value corresponding to the third point is the starting warning threshold.

[0033] As a further description of the above technical solution, the ASCP evolution-threshold criterion correlation model constructed by MATLAB interpolation processing and neural network training in step S7 is that: through the MATLAB program, the interp1 function is used to interpolate the data, fill in the missing data or smooth the data, and the interpolation results are stored in the database using the MATLAB database toolbox; then the data is loaded from the database for preprocessing and converted into a format suitable for neural network training; then the ASCP evolution law-threshold criterion correlation model is constructed using the deep learning toolbox of MATLAB, and the model training is performed by specifying the training options.

[0034] As a further description of the above technical solution: after the ASCP evolution-threshold criterion correlation model is constructed, the performance of the model is verified by test data.

[0035] As a further description of the above technical solution: step S8 also continuously returns the monitoring data and warning results to the database for iterative optimization of the correlation model.

[0036] Compared with the prior art, the present application has the following advantages: ​

[0037] The application fuses multi-class sensitive parameters (amplitude, fractal dimension, LgN / b value, etc.) of acoustic emission signals, comprehensively reflects the whole process of rock mass damage through AHP weight and ASCP, avoids one-sidedness of single parameter, reduces false positive and false negative rates, establishes double early warning thresholds of starting early warning + disaster early warning, identifies rock mass damage precursors (such as inflection point of ASCP evolution curve) in advance, reserves sufficient time for on-site emergency disposal, dynamically adjusts parameter weight and threshold through model inversion and machine learning, adapts to different lithology and different disturbance conditions of engineering scene, does not need to additionally lay complex equipment, can be realized relying on existing acoustic emission monitoring system, reduces engineering cost, is based on indoor test database and engineering site data, has clear steps and quantifiable parameters, and simplifies data processing and model training through MATLAB tool, and is convenient for engineering personnel to apply. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1 The flow chart of the high stress rock mass disturbance disaster real-time early warning method based on rupture comprehensive signal of the application.

[0039] Figure 2 The demonstration schematic diagram for establishing the nonlinear starting point of ASCP accelerated change.

[0040] Figure 3 The ASCP evolution rule graph of fractured granite at the eve of multistage disturbance shear failure under true triaxial in the application example one.

[0041] Figure 4 The ASCP evolution rule graph of intact granite at the eve of disturbance rock burst under different disturbance frequencies under true triaxial in the application example two. DETAILED DESCRIPTION

[0042] The claims of the application will be further explained in detail in combination with specific implementation manners, but do not constitute any limitation on the application, and the limited number of modifications made by anyone within the protection scope of the claims of the application is still within the protection scope of the claims of the application.

[0043] The high stress rock mass disturbance disaster real-time early warning method based on rupture comprehensive signal, the flow thereof is as shown in Figure 1 The method comprises the following steps:

[0044] S1, laying acoustic signal sensors in the engineering site (such as the tunnel face, mine stope), collecting the acoustic emission signals of rock mass rupture in the dynamic disaster process of surrounding rock in the engineering site, obtaining acoustic signal characteristic parameters, the acoustic signal characteristic parameters including emission waveform amplitude V, fractal dimension D (reflecting crack complexity), LgN / b value (reflecting signal frequency and intensity relationship), main frequency, activity coefficient (reflecting signal frequency) and solid volume (reflecting rupture range) and the like; analyzing the response characteristics and evolution law of the acoustic signal characteristic parameters, and selecting the acoustic signal characteristic parameters most sensitive to rock mass damage as sensitive characteristic parameters.

[0045] S2, under true three-dimensional stress conditions, confirming the sample types of all indoor tests, carrying out rock mass dynamic disaster damage tests of different test conditions (different stress conditions, different stress levels, different structural planes and surrounding rock unit types), collecting a large amount of acoustic signal information data (including acoustic signal parameters, stress conditions, damage results and the like) by using indoor monitoring equipment, extracting the sensitive characteristic parameters consistent with step S1, storing them in the database, establishing a total database, and analyzing the response characteristics and evolution law of the rock mass rupture sensitivity characteristic parameters.

[0046] S3, combining the various sensitive characteristic parameters of rock mass dynamic disturbance damage obtained from the engineering site and indoor tests, using a weighted fusion method to construct a comprehensive acoustic parameter ASCP, and the function relationship between ASCP and rock damage is:

[0047]

[0048] Wherein ASCP is the function relationship between the weighted comprehensive characteristic value of acoustic signals and rock damage, n1, n2, n3, , n i are the weight coefficients corresponding to different sensitive characteristic parameters, l1, l2, l3, , l i represent the proportional functions corresponding to the waveform amplitude, fractal dimension, LgN / b, main frequency, activity coefficient, solid volume and the like of the sensitive characteristic parameters.

[0049] Wherein, the weight coefficients , , , , The values ​​were determined using the Analytic Hierarchy Process (AHP): First, a hierarchical evaluation model was constructed based on extensive monitoring results. The target layer represents the selection of optimal acoustic emission parameters for destruction; the criterion layer includes parameter variation patterns, numerical values, the severity of abrupt changes in acoustic emission parameters, the degree of rock mass failure response to changes in acoustic emission parameters, and the anti-interference level of acoustic emission parameters; the scheme layer includes emission waveform amplitude, fractal dimension, LgN / b value, dominant frequency, activity coefficient, and solid volume. Second, a judgment matrix A was constructed, and pairwise comparisons were performed between each element to determine the weight of each criterion layer on the target layer. The judgment matrix A is as follows:

[0050]

[0051] Determine if the elements in matrix A satisfy: , and ;

[0052] Then, a hierarchical single sort is performed. All elements in the above formula are compared pairwise and then hierarchically sorted to determine their order of importance. The specific calculation is as follows:

[0053]

[0054]

[0055] Derive the weight matrix Next, the maximum eigenvalue and CI value are calculated, and the random consistency index RI is obtained by using Satty simulation 2000 times. The CR value is then calculated to determine whether the consistency is passed.

[0056]

[0057]

[0058]

[0059] If CR < 0.1, it indicates that the judgment matrix A is within the error range class, and its elements can be used as weight coefficients. To perform the calculation, if CR ≥ 0.1, then the judgment matrix A needs to be corrected, and the final weight coefficients will be... The value range is 0.15 to 0.30, and the weighting coefficient is... The value range is 0.10 to 0.25, and the weighting coefficient is... The value range is 0.20 to 0.35, and the remaining weight coefficients are adjusted according to the rock mass type. The sum of the weight coefficients is 1.

[0060] The proportional function expressions are as follows:

[0061]

[0062]

[0063] ;

[0064] where x is the number of sensors; V i is the waveform voltage value of the i th time node, B 1 is the maximum waveform voltage value of the node; D i is the fractal dimension value of the i th time node, B 2 is the maximum fractal dimension value of the node; LgN i / b i is the LgN / b value of the i th time node, B 3 is the maximum LgN / b value of the node; the proportional function of the remaining parameters refers to the above form, and is normalized to the interval [0, 1].

[0065] S4, compare the evolution curves of ASCP at each stage (stable period, crack initiation period, crack propagation period, and instability period) of the rock mass failure of the engineering site and the indoor test, and analyze the change trend and evolution law of ASCP by the following quantitative method:

[0066] The ASCP time series is normalized: ;

[0067] The sliding window method (window length ≥ 10 sampling periods) is used to smooth the noise;

[0068] The curve slope k of the ASCP evolution curve is calculated based on linear regression: define k≥0.05 / min as a significant upward trend (corresponding to a precursor of catastrophe), 0.01≤k<0.05 / min as a stable fluctuation (corresponding to gradual destruction), and k<0.01 / min as a stable state;

[0069] The CUSUM algorithm is used to set the control limit L=3× When the cumulative sum exceeds the limit and the acoustic emission single event energy suddenly increases by ≥200%, it is determined as a “catastrophic critical point”;

[0070] Finally, a one-to-one correspondence between ASCP and the stress state and damage degree of the rock mass is established.

[0071] S5, according to the evolution law of ASCP in step S4, set two levels of early warning thresholds:

[0072] The starting early warning threshold is determined by the perpendicular bisector method: the upper and lower straight line segments (corresponding to the crack propagation period and the stable period) of the ASCP evolution curve are extended to intersect at the first point A, as shown in Figure 2 The perpendicular bisectors of the upper and lower straight line segments are drawn, respectively Figure 2The dashed lines 1 and 2 intersect at the second point D. The dashed line 3 connects the first point A and the second point D and intersects the ASCP evolution curve at the third point E. The third point E is determined as the activation warning threshold. Based on the statistical data of the results in the database, the ASCP values ​​before the destruction under different conditions are obtained and used as the ASCP disaster warning threshold.

[0073] The steps for constructing the perpendicular bisectors of the upper and lower line segments include:

[0074] Determine the starting point for calculation, wherein the calculation formula for the starting point satisfies:

[0075]

[0076] In the formula, x1 is the point corresponding to the opening angle between the straight line segment and the fitted curve is 1 / 5000°, where 1 / 5000° makes the angle between the fitted curve and the straight line segment small enough to meet the accuracy requirements for the calculation of the starting point; x0 is the point of coincidence between the fitted line and the function; f1(x) is the ASCP evolution curve; and f2(x) is the slope function curve of the straight line segment.

[0077] Using the above formula, we determine that the starting point of the upward sloping line is point C and the starting point of the downward sloping line is point B. Starting from points B and C, we calculate and find the midpoints of the upward and downward line segments, and then draw the perpendicular bisectors.

[0078] The disaster warning threshold is determined by taking the combined value of ASCP values ​​of rock masses before critical failure from the statistical database and the test results of similar rock masses.

[0079] S6. Based on the results of rock mass dynamic disturbance and failure from engineering sites and laboratory tests, an ASCP calculation model is derived, and the weighting coefficient n is dynamically adjusted. i And threshold ranges to ensure that the model matches the actual working conditions.

[0080] S7. Store all data from steps S1 to S6 (including acoustic signal parameters, ASCP values, early warning results, and damage results) into the database as samples. Then, using MATLAB interpolation and neural network training, construct an ASCP evolution-threshold criterion association model, as detailed below:

[0081] Using the interp1 function in MATLAB, data is interpolated to fill in missing or smooth data, and the interpolation results are stored in a database using MATLAB's database toolbox.

[0082] Load data from the database, preprocess it, and convert it into a format suitable for neural network training;

[0083] The ASCP evolution rule-threshold criterion correlation model is built using the deep learning toolbox of MATLAB, and model training is carried out by specifying training options;

[0084] The trained correlation model is verified using test data, the predicted results and the actual values are compared, the model performance is verified, and the model accuracy is ensured.

[0085] S8, embed the trained ASCP evolution-threshold criterion correlation model into the monitoring system, extract the indoor test matched with the construction site from the database according to the acoustic emission response of the rock mass damage in the engineering site, set the AE equipment parameters, use the AE precursor characteristics of the indoor test for the rock damage monitoring in the construction site, and extract the start warning threshold and the catastrophe warning threshold of the ASCP evolution curve;

[0086] The area where the rock mass may be damaged is determined through AE positioning test, and the area is set as the AE monitoring space area;

[0087] After all the previous preparations are completed, the AE equipment is started, and the warning is started;

[0088] The rock mass in the construction process is monitored in real time according to the new acoustic emission parameter ASCP of rock damage precursor;

[0089] The change trend and evolution rule of the ASCP evolution curve in the monitoring process are analyzed, whether the ASCP reaches the start warning threshold, that is, whether the ASCP value reaches the ASCP value corresponding to the peak point of the nonlinear curvature of the new parameter ASCP evolution curve, is observed, if the ASCP value reaches the ASCP value corresponding to the peak point of the nonlinear curvature of the new parameter ASCP evolution curve, that is, the start warning threshold is reached, a start warning signal is sent out, and the next step is entered; if the curve slope k does not reach the ASCP parameter corresponding to the peak point of the nonlinear curvature of the ASCP evolution curve, the monitoring is continued;

[0090] After the start warning signal is sent out, whether the ASCP reaches the catastrophe warning threshold is observed, if the comprehensive value of the new parameter ASCP test result of the critical damage of the rock mass of this type of test is reached, that is, the catastrophe warning threshold is reached, it is indicated that the rock mass is about to be damaged, a catastrophe warning signal is immediately sent out, and the on-site personnel are reasonably arranged to evacuate; if the catastrophe warning threshold is not reached, the monitoring is continued;

[0091] At the same time, the monitoring data and the warning result are returned to the database for iterative optimization of the correlation model.

[0092] Application Example One:

[0093] Taking the acoustic emission warning process of fractured granite under true triaxial multi-stage disturbance shear failure as an example, the implementation steps of the present application are described in detail:

[0094] In step S2, the sample type of all indoor tests is confirmed, and in the present example, the sample type is granite with a size of 100 mm x 100 mm x 100 mm and a crack width d = 25 mm, the loading rate of heat release stress in three directions is 1 MPa / s, the loading is to σ1 = 5 MPa, σ2 = 30 MPa, σ3 = 25 MPa, the maintenance time of the static stage is 200 s; in the stress redistribution stage, σ2 rises to 35 MPa, σ3 drops to 11.8 MPa, σ1 rises to 15 MPa, and then the loading of each level of disturbance is started, and the disturbance load applied in each level is 50 cycles, wherein the loading and unloading rate in each level is 1 MPa / s, the AE monitoring equipment used is 5 nano30 type sensors, the multi-level disturbance true triaxial shear failure test of fractured rock is carried out, a large amount of acoustic signal information data is collected by the related indoor monitoring equipment, the sensitive feature data consistent with step S1 is extracted, and stored in the database to establish a total database.

[0095] In step S3, the sensitive feature parameters of the dynamic disturbance and failure of the rock mass in the engineering site and the indoor test are combined, the comprehensive feature parameters with strong response are selected, the calculation method of the corresponding comprehensive parameter ASCP is constructed, for example, the results of the test at 2280s are as follows:

[0096] wherein ASCP is a function relationship formula of the weighted comprehensive feature value of the acoustic signal and the rock failure, are the weight coefficients corresponding to different sensitive feature parameters, l1, l2 and l3 represent the proportional functions corresponding to the waveform amplitude, the fractal dimension and the LgN / b value of the sensitive feature parameters, The proportional function relationship formula after substitution is calculated as follows:

[0097]

[0098]

[0099]

[0100] wherein x is the number of sensors, V i is the waveform voltage value collected at the i th time node, D i is the acoustic emission fractal dimension value at the i th time node, N i and b i are the total number of acoustic emission signals and the b value at the i th time node, f i and A i ​​​​​Let be the dominant frequency and amplitude of the acoustic emission at the i-th time node;

[0101] Among them, the weighting coefficient , , The values ​​were determined using the Analytic Hierarchy Process (AHP). Taking a granite sample as an example, a judgment matrix A was constructed. The criterion layer contains five elements (C1: parameter variation law; C2: numerical magnitude; C3: mutation sensitivity; C4: destructive response; C5: anti-interference capability), and the scheme layer contains six acoustic emission parameters (P1: waveform amplitude; P2: fractal dimension; P3: lgN / b value; P4: dominant frequency; P5: activity coefficient; P6: actual volume). The criterion layer judgment matrix A was constructed as follows:

[0102]

[0103] By calculating the maximum eigenvalue λ max =5.23, consistency ratio CR=0.04<0.1, passes the test. Weight allocation: Criterion layer weight: W C =[0.32,0.12,0.35,0.15,0.06], the judgment matrix shows that P1 (amplitude) is better at representing the destructive trend than other parameters. The final comprehensive weight of the scheme layer is:

[0104] Parameter weights: n1=0.25, n2=0.18, n3=0.30, n4=0.12, n5=0.10, n6=0.05.

[0105] Based on the acoustic emission response of rock mass failure at the engineering site, indoor tests matching the construction site were selected from the database, the AE equipment parameters were set, and the AE precursor characteristics of the indoor tests were used for rock failure monitoring at the construction site. The initiation warning threshold and ASCP disaster warning threshold of the ASCP evolution curve were extracted.

[0106] The area where the rock mass may be damaged was determined by AE positioning test, and this area was set as the AE monitoring space area.

[0107] After completing all the preliminary preparations, turn on the AE device and start the alert;

[0108] Based on the new acoustic emission parameter ASCP, which is a precursor to rock failure, the rock mass at the construction site is monitored in real time during the construction process.

[0109] Analysis of the ASCP evolution curve during the monitoring process ( Figure 3 The changing trend and evolution pattern of ASCP (as shown) are observed to determine whether it reaches the activation warning threshold. The numerical determination method is as described in steps S5 and... Figure 2As shown, the early warning threshold of the ASCP evolution curve in the present example is set to the curve slope k > 35°, and the emergency early warning threshold of the ASCP is set to ASCP > 2.0, that is, whether the ASCP value reaches the ASCP value corresponding to the peak point of the nonlinear curvature of the new parameter ASCP evolution curve, if the ASCP value reaches the ASCP value corresponding to the peak point of the nonlinear curvature of the new parameter ASCP evolution curve, that is, the starting early warning threshold is reached, an early warning signal is sent out, and the next step is entered; if the curve slope k does not reach the ASCP parameter corresponding to the peak point of the nonlinear curvature of the ASCP evolution curve, continue to monitor;

[0110] After sending out the starting early warning signal, it is observed whether the ASCP reaches the catastrophe early warning threshold, if the comprehensive value of the new parameter ASCP test result of the critical failure of the rock mass is reached, that is, the catastrophe early warning threshold is reached, it is indicated that the rock mass is about to be damaged, and a catastrophe early warning signal is immediately sent out, and the on-site personnel are reasonably arranged to evacuate; if the catastrophe early warning threshold is not reached, continue to monitor, and continuously return the monitoring data and the early warning result to the database for iterative optimization of the correlation model.

[0111] Application Example Two:

[0112] The following takes "true triaxial acoustic emission early warning process of complete granite at different disturbance frequencies before disturbance rock burst" as an example to illustrate the implementation steps of the present application in detail:

[0113] Based on the monitoring results of the rock mass fracture sound signals in the disaster process of the dynamic surrounding rock of the engineering site, the response characteristics and evolution law of the sound signal characteristic parameters are analyzed, the sound signal characteristic parameters include the emission waveform amplitude, the fractal dimension, the LgN / b value, the main frequency, the activity coefficient and the real volume, and the sensitive sound signal characteristic parameters are selected.

[0114] Confirm the sample type of all indoor tests, the sample type used in the present example is granite with a size of 100 mm x 100 mm x 200 mm, the stress loading rate in three directions is 0.5 MPa / s, the loading is to σ1 = 50 MPa, σ2 = 30 MPa, σ3 = 25 MPa, and the maintenance time of the static stage is 20 min; the load control mode is to load to 0.4σ c , and the disturbance induced rock burst stage is the host computer loading and unloading and applying disturbance load. When the early load is low, the load applied at each stage is 0.2σ c , with the increase of the load, the load at each stage is 0.1σ c , and when the later load is large, the load at each stage is 0.05σ cEach level of disturbance load is 50 cycles, with a loading rate of 1 MPa / s, the AE monitoring device is 3 nano30 sensors, the complete granite disturbance rock burst test under different disturbance frequencies is carried out, a large amount of acoustic signal information data is collected according to the relevant indoor monitoring equipment, stored in the database, and the database is established.

[0115] Combined with the engineering site and indoor test rock mass dynamic disturbance damage of multiple sensitive characteristic parameters, the response strong comprehensive characteristic parameters are selected, and the corresponding calculation method with comprehensive parameters ASCP is constructed, for example, the results of the 7400 s experiment are as follows:

[0116]

[0117] Wherein ASCP is the function relationship of the weighted comprehensive characteristic value of the acoustic signal and the rock damage 、 、 The weight coefficient corresponding to different acoustic signal parameters, l1, l2, l3 represent the proportional function corresponding to the acoustic emission parameter waveform amplitude, fractal dimension, LgN / b in the acoustic signal, 、 、 Wherein the proportional function relationship of ASCP after being brought in is as follows:

[0118]

[0119]

[0120]

[0121] Wherein x is the number of sensors, V i is the waveform voltage value collected at the ith time node, D i is the acoustic emission fractal dimension value at the ith time node, N i and b i are the total number of acoustic emission signals and b value at the ith time node, f i and A i are the main frequency and amplitude of the acoustic emission at the ith time node;

[0122] The weight coefficient 、 、 The values are determined by AHP analysis method. In the AHP analysis method, a certain granite sample is taken as an example to construct a judgment matrix A. The criterion layer includes five elements (C1: parameter variation law; C2: numerical magnitude; C3: mutation sensitivity; C4: damage response degree; C5: anti-interference), and the scheme layer includes six acoustic emission parameters (P1: waveform amplitude; P2: fractal dimension; P3: LgN / b value; P4: main frequency; P5: activity coefficient; P6: solid volume), and a criterion layer judgment matrix is constructed:

[0123]

[0124] By calculating the maximum eigenvalue λ max =5.23, the consistency ratio C.R.=0.051<0.1, and the test is passed. The weight distribution is: criterion layer weight: W C =[0.25,0.18,0.30,0.12,0.10,0.05], the judgment matrix shows that P1 (waveform amplitude) is better than other parameters in representing the damage trend, and the final scheme layer comprehensive weight is:

[0125] Parameter weight: n1=0.22, n2=0.16, n3=0.30, n4=0.14, n5=0.10, n6=0.08.

[0126] According to the acoustic emission response of the rock mass during the damage of the engineering site, the indoor test matched with the construction site is selected from the database, the AE equipment parameters are set, the AE precursor characteristics of the indoor test are used for the rock damage monitoring of the construction site, and the starting warning threshold and the ASCP disaster warning threshold of the ASCP evolution curve are extracted;

[0127] The area where the rock mass may be damaged is determined by AE positioning test, and the area is set as the AE monitoring space area;

[0128] After all the previous preparations are completed, the AE equipment is started, and the warning is started;

[0129] According to the new parameter ASCP of rock damage precursor acoustic emission, the rock mass in the construction process is monitored in real time;

[0130] The change trend and evolution law of the ASCP evolution curve (as shown in Figure 4 ) are analyzed, and whether the ASCP reaches the starting warning threshold is observed, and the numerical value determination method is as shown in Figure 2As shown, the early warning threshold of the ASCP evolution curve in this example is set to the curve slope k>35°, and the emergency warning threshold of the ASCP is set to ASCP>2.0, that is, whether the ASCP value reaches the ASCP value corresponding to the peak point of the nonlinear curvature of the new parameter ASCP evolution curve, if the ASCP value reaches the ASCP value corresponding to the peak point of the nonlinear curvature of the new parameter ASCP evolution curve, that is, the start warning threshold is reached, a start warning signal is sent out, and the next step is entered; if the curve slope k does not reach the ASCP parameter corresponding to the peak point of the nonlinear curvature of the ASCP evolution curve, monitoring continues;

[0131] After the start warning signal is sent out, it is observed whether the ASCP reaches the catastrophe warning threshold, if the comprehensive value of the new parameter ASCP test result of the critical failure of the rock in this type of test is reached, that is, the catastrophe warning threshold is reached, it is indicated that the rock mass is about to be damaged, a catastrophe warning signal is immediately sent out, and the on-site personnel are reasonably arranged to evacuate; if the catastrophe warning threshold is not reached, monitoring continues, and the monitoring data and the warning result are continuously returned to the database, and the correlation model is iteratively optimized.

[0132] The above examples are only used to illustrate the technical solutions of the present application, but not to limit them; those skilled in the art should understand that the technical solutions recorded in the foregoing examples can still be modified, or some technical features can be replaced by equivalents; and these modifications or replacements will not make the essence of the corresponding technical solutions deviate from the protection scope of the present application.

Claims

1. A high-stress rock mass disturbance disaster real-time warning method based on a rupture comprehensive signal, characterized in that, It comprises the following steps: S1, collecting the acoustic emission signals of rock mass rupture under the dynamic disturbance of the construction site by an acoustic signal sensor, extracting acoustic signal characteristic parameters and screening sensitive characteristic parameters, wherein the acoustic signal characteristic parameters include waveform amplitude, fractal dimension, LgN / b value, dominant frequency, activity coefficient and real volume; S2, under true three-dimensional stress conditions, carrying out rock mass dynamic disturbance damage tests under different test conditions, synchronously collecting the acoustic emission signals of the indoor tests and extracting the sensitive characteristic parameters consistent with step S1, and establishing a database; wherein the different test conditions include different stress conditions, different stress levels, different structural planes and different surrounding rock unit types; S3, based on the plurality of sensitive characteristic parameters obtained in steps S1 and S2, an ASCP function of a comprehensive characteristic parameter is constructed through a weighted fusion method, and an expression of the ASCP function is wherein is a weight coefficient of each sensitive characteristic parameter; is a proportional function of each sensitive characteristic parameter; S4, comparing the ASCP evolution curves of the rock mass damage at each stage of the construction site and the indoor tests, analyzing the change trend and evolution law of ASCP by a quantitative method, and establishing a one-to-one correspondence between ASCP and the stress state and damage degree of rock mass; S5, determining the starting early warning threshold value by the perpendicular bisector method according to the change trend and evolution law of the ASCP evolution curve in step S4, determining the disaster early warning threshold value by the critical damage data of the indoor tests, and forming multiple early warning threshold values; S6, based on the damage results of the rock mass dynamic disturbance of the construction site and the indoor tests in steps S1 and S2, inversing the ASCP function, and dynamically adjusting the weight coefficient and threshold interval; S7, storing the data of steps S1-S6 into the database as samples, processing by MATLAB interpolation and training by neural network, and constructing an ASCP evolution-threshold criterion correlation model; S8, embedding the trained ASCP evolution-threshold criterion correlation model into the monitoring system, real-time collecting the acoustic emission signals of the construction site and calculating ASCP, triggering early warning when the ASCP reaches the starting threshold value, and triggering emergency warning when the ASCP reaches the disaster threshold value.

2. The method according to claim 1, wherein, The weight coefficient in step S3 is determined by AHP hierarchical analysis method, and the specific process is as follows: A hierarchical evaluation model is constructed according to the monitoring results, including target layer, criterion layer and scheme layer; Constructing judgment matrix , the weight of each criterion layer to the target layer is determined by comparing each sensitive characteristic parameter with each other, and the judgment matrix satisfies , and ; The hierarchical single sorting is performed, all elements in the judgment matrix are compared with each other, and the hierarchical sorting is carried out, and the important order is arranged, and the specific calculation is as follows: ， ； The weight matrix is obtained Then, the maximum eigenvalue is calculated C.I. value, and the C.R. value is calculated by using Satty simulation multiple random consistency index R.I. value, to determine whether the consistency passes. ， ， ； If C.R. < 0.1, it indicates that the judgment matrix Within the error range, use the elements therein as weight coefficients ; If C.R. ≥ 0.1, the judgment matrix needs to be revised. If C.R. ≥ 0.1, the judgment matrix needs to be revised.

3. The method according to claim 1, wherein the method is characterized in that: The scaling function described in step S3 includes an amplitude scaling function a fractal dimension scaling function an LgN / b value scaling function and other sensitive characteristic parameter scaling functions; wherein x is the number of sensors, respectively the waveform amplitude, the fractal dimension and the LgN / b value of the i-th time node, respectively the maximum value of the sensitive characteristic parameter corresponding to the i-th time node.

4. The method according to claim 1, wherein, The change trend and evolution law of ASCP are analyzed by a quantitative method in step S4, and the specific process is as follows: The ASCP time series is normalized: ; The sliding window method is used to smooth the noise; The curve slope k of the ASCP evolution curve is calculated based on linear regression: defining k≥0.05 / min as a significant upward trend, 0.01≤k<0.05 / min as a stable fluctuation, and k<0.01 / min as a stable state; The control limit L = 3 x ASCP is set using the CUSUM algorithm norm When the cumulative sum exceeds the limit and the acoustic emission single event energy increases by ≥ 200%, it is determined that the critical point of catastrophe has been reached.

5. The method according to claim 1, wherein, The starting early warning threshold value is determined by the perpendicular bisector method in step S5, and the specific process is as follows: the upper and lower straight line segments of the ASCP evolution curve are extended to intersect at a first point, the perpendicular bisectors of the upper and lower straight line segments are intersected at a second point, the first point and the second point are connected to intersect the ASCP evolution curve at a third point, and the ASCP value corresponding to the third point is the starting early warning threshold value.

6. The method according to claim 1, wherein, The ASCP evolution-threshold criterion correlation model constructed by the MATLAB interpolation processing and neural network training in step S7 is specifically as follows: through the MATLAB program, the interp1 function is used to interpolate the data, fill in the missing data or smooth the data, and the interpolation results are stored into the database by using the database toolbox of MATLAB; subsequently, the data are loaded from the database for pretreatment and converted into a format suitable for neural network training; then, the deep learning toolbox of MATLAB is used to construct the ASCP evolution rule-threshold criterion correlation model, and the model training is performed by specifying the training options.

7. The method according to claim 6, characterized in that: After the ASCP evolution-threshold criterion correlation model is constructed, the performance of the model is verified by using test data.

8. The method according to claim 1, characterized in that: Step S8 also continuously returns the monitoring data and the early warning results to the database, and iteratively optimizes the correlation model.

Images