A mine surrounding rock stability evaluation method based on elastic wave response characteristics

Abstract

The present application belongs to the technical field of mine engineering, and specifically relates to a mine surrounding rock stability evaluation method based on elastic wave response characteristics, comprising: applying elastic wave excitation to obtain enhanced response signals, and extracting measured P-wave velocity and amplitude attenuation coefficient; performing wavelet packet decomposition on the enhanced response signals to extract the energy of each frequency band node, calculate the high-frequency energy proportion and information entropy; constructing a two-dimensional feature discrimination plane based on the measured P-wave velocity and high-frequency energy proportion, identifying the high confining pressure compaction type hidden damage state in combination with the information entropy to generate a damage mode marker, and calculating a damage index correction amount; inputting each index into a comprehensive scoring model, calculating a comprehensive score according to the adaptive switching weight configuration of the damage mode marker, and outputting the surrounding rock stability grade. The present application effectively distinguishes the crack closure compaction state and the real complete rock mass, and substantially improves the reliability of the stability grade determination.

Description

Technical Field

[0001] This invention belongs to the field of mining engineering technology, specifically relating to a method for evaluating the stability of surrounding rock in mines based on elastic wave response characteristics. Background Technology

[0002] As the mining depth of metal mines continues to increase, the in-situ stress level of the rock mass in deep engineering projects has risen significantly, and the surrounding rock structure of key areas such as the tunneling face, roof, sidewalls, and anchor-supported areas has become increasingly complex. Under high in-situ stress, the expansion and distribution of internal fractures in the rock mass often exhibit strong locality and concealment. Traditional passive monitoring methods such as microseismic monitoring, stress gauges, and displacement gauges can only passively acquire response signals after macroscopic fracturing or stress adjustment of the rock mass. They cannot actively trigger evaluation of the real-time status of specific monitoring locations, leaving on-site personnel facing a dilemma of information lag when they need to conduct point-to-point verification of a specific location.

[0003] The document with authorization announcement number CN113899811B discloses an acoustic wave testing system for cumulative damage to rock mass in coal mine roadways. The system includes three vertically drilled holes in the roadway roof, arranged in an equilateral triangle, with a test unit in each hole. Each test unit includes a measuring rod, a sealing assembly, a first acoustic transducer, and a second acoustic transducer. The sealing assembly and the second acoustic transducer are fixed sequentially from the bottom to the top of the measuring rod. The first acoustic transducer is fixed to the sealing assembly, and the second acoustic transducer is located at the top of the measuring rod. The measuring rod has a hollow structure, through which the data lines of the first and second acoustic transducers pass. This design utilizes a hole arrangement structure for acoustic wave testing of cumulative damage to rock mass in the roof of coal mine roadways, enabling both single-hole planar refraction wave testing and multi-hole penetrating direct wave testing. It can detect the differences in crack damage propagation in different directions of the roof rock mass, allowing for targeted design of support schemes.

[0004] To address the reactive nature of passive monitoring, active elastic wave detection methods have been introduced into the field of surrounding rock stability evaluation. These methods apply excitation signals to the rock surface and collect the propagation response, extracting rock mass state information from dimensions such as wave velocity, amplitude attenuation, and spectral characteristics, thus achieving proactive, triggered evaluation of the detected area. In existing engineering applications, the measured longitudinal wave velocity is typically used as the primary basis for integrity assessment, supplemented by the amplitude attenuation coefficient as a secondary reference for damage severity, achieving good application results in scenarios such as the detection of loosened zones in surrounding rock.

[0005] However, in deep, high-stress mining environments, the aforementioned method suffers from a fundamental flaw that is systematically overlooked under extreme pressure conditions: when the surrounding rock is under high confining pressure, the microfracture network already developed within the rock mass will partially close under the pressure of normal stress. The blocked effect of the closed fracture surfaces on elastic wave propagation is significantly weakened, leading to an abnormally high measured longitudinal wave velocity in this compaction-damaged area. This value highly overlaps with the wave velocity response of a truly intact rock mass, making it difficult to distinguish between the two using wave velocity as a single indicator. Simultaneously, these closed fractures have a significant scattering and dissipation effect on the high-frequency components of elastic waves. The scattering loss of high-frequency components of elastic waves as they pass through the surface of microfractures does not disappear with fracture closure. The microscopic discontinuities in fracture surface roughness and interfacial wave impedance continue to consume high-frequency energy, resulting in a systematic reduction in the proportion of high-frequency band energy in the received signal. This characteristic is completely overlooked in existing evaluation systems that rely solely on wave velocity and overall amplitude attenuation. The ultimate consequence is that the evaluation system misclassifies the "high confining pressure compaction type hidden damage" rock mass as intact and stable surrounding rock, resulting in an inflated assessment of the stability level of this type of dangerous state, which poses a hidden danger to the safety of on-site mining operations. Summary of the Invention

[0006] This invention provides a method for evaluating the stability of surrounding rock in mines based on elastic wave response characteristics. This method addresses the technical problem that active detection of elastic waves in deep, high-stress mining environments leads to abnormally high wave velocities due to the compaction effect of high confining pressure fractures. Existing evaluation methods cannot identify latent compaction damage from the perspective of frequency band energy distribution characteristics, resulting in falsely high stability levels.

[0007] This invention provides a method for evaluating the stability of mine surrounding rock based on elastic wave response characteristics, comprising: S1, continuously apply elastic wave excitation to the surface of the surrounding rock to be tested, filter the effective response signals based on adaptive signal-to-noise ratio and perform aligned superposition and averaging to obtain the enhanced response signal, and extract the measured longitudinal wave velocity and amplitude attenuation coefficient. S2, perform wavelet packet decomposition on the enhanced response signal, extract the energy of each frequency band node based on the preset decomposition level, and calculate the high-frequency energy ratio and the information entropy of the wavelet packet frequency band node; S3. Based on the measured longitudinal wave velocity and the proportion of high-frequency energy, a two-dimensional feature discrimination plane is constructed. The information entropy is combined to identify the latent damage state of high confining pressure compaction to generate damage mode markers. When the damage is determined to be latent damage, the damage index correction amount is calculated. S4 inputs the measured longitudinal wave velocity deviation index, amplitude attenuation coefficient, normalized value of information entropy, normalized value of high frequency energy ratio, and damage index correction amount into the comprehensive scoring model. Based on the damage mode indicator, the weight configuration is adaptively switched to calculate the comprehensive score, and the surrounding rock stability level is output based on the comprehensive score.

[0008] Its effect is as follows: by introducing wavelet packet decomposition to refine the frequency band energy of the elastic wave response signal, a two-dimensional frequency domain feature discrimination system is constructed with the high frequency energy ratio and wavelet packet information entropy. In the high wave velocity region, it effectively distinguishes the crack closure compaction state from the real intact rock mass, making up for the blind spot of the existing method when it only relies on wave velocity and overall amplitude attenuation to identify hidden compaction damage.

[0009] Furthermore, based on adaptive signal-to-noise ratio, effective response signals are filtered and aligned, superimposed, and averaged to obtain enhanced response signals, including: The root mean square amplitude of the response signal acquired during the first excitation before the arrival of the first wave is calculated as the background noise baseline. The signal-to-noise ratio is obtained by extracting the peak value of the first arrival envelope of each subsequent excitation and calculating the ratio with the background noise baseline value. Excitation stops when the number of valid excitations that meet the signal-to-noise ratio threshold reaches a preset lower limit. The selected valid response signals are then time-domain aligned, superimposed, and averaged using the energy accumulation method to finally obtain the enhanced response signal.

[0010] Its effects are as follows: by using adaptive signal-to-noise ratio screening and energy accumulation method to align and superimpose, it effectively suppresses the interference of downhole random noise on the quality of the response signal, ensures the signal-to-noise ratio of the enhanced response signal on which subsequent feature extraction depends and the accuracy of the initial arrival time calibration, and avoids distortion of subsequent frequency band energy features due to occasional noise pollution from a single excitation.

[0011] Furthermore, the measured P-wave velocity and amplitude attenuation coefficient are extracted, including: Based on the initial arrival and take-off point of the enhanced response signal, the propagation time difference of the elastic wave at the known transmission and reception distance is calculated to obtain the measured longitudinal wave velocity. The peak value of the first arrival envelope of the enhanced response signal is extracted synchronously, and the ratio of the peak value of the first arrival envelope to the excitation reference amplitude is calculated. The degree of energy loss is calculated based on the ratio, and the amplitude attenuation coefficient is finally obtained.

[0012] Its effects are as follows: it clearly calculates the measured longitudinal wave velocity based on the first arrival jump time difference of the enhanced response, and derives the amplitude attenuation coefficient by the ratio of the first arrival envelope peak value to the excitation reference amplitude; it can accurately quantify the overall elastic propagation rate and total energy loss of the rock mass, obtain two types of basic time-domain indicators characterizing the macroscopic integrity of the rock mass, provide basic input parameters for the construction of a two-dimensional discrimination plane and a comprehensive scoring model, and intuitively reflect the overall rock mass's ability to impede and attenuate elastic waves.

[0013] Furthermore, based on a preset number of decomposition layers, the energy of each frequency band node is extracted to calculate the proportion of high-frequency energy and the information entropy of the wavelet packet frequency band nodes, including: The minimum bandwidth that meets the resolution requirements is determined based on the signal sampling frequency and the excitation center frequency, and the number of decomposition layers is calculated. The node energy is calculated for the reconstructed signals of all frequency band nodes under the decomposition layer, and the high frequency band and low frequency band are divided by the excitation center frequency. The high-frequency energy percentage is obtained by calculating the ratio of the sum of the high-frequency band energies to the total signal energy. Then, the Shannon entropy of the discrete distribution of the energy percentages of each node is calculated, and finally, the information entropy of the wavelet packet band nodes is obtained.

[0014] Its effect is that by decomposing the elastic wave response signal into frequency band nodes through wavelet packet decomposition, two complementary frequency domain features, namely the high-frequency energy ratio and the frequency band energy distribution entropy, can be obtained simultaneously. This allows the selective dissipation of high-frequency components to be captured even when the total energy of the signal does not change significantly, providing physically interpretable feature inputs for subsequent identification of high-compression compaction-type latent damage.

[0015] Furthermore, by combining information entropy to identify latent damage states caused by high confining pressure compaction to generate damage pattern markers, including: Determine whether the measured longitudinal wave velocity is higher than the preset ratio of the reference value of the wave velocity of intact rock mass; If it is higher than the preset ratio, then it is further determined whether the proportion of high-frequency energy is lower than the first quantile value and whether the information entropy is higher than the second quantile value. If the proportion of high-frequency energy and the information entropy simultaneously meet the above conditions, it is determined to be a latent damage state of high confining pressure compaction, and a damage mode marker marked as latent damage is generated.

[0016] Furthermore, when an injury is determined to be latent, a correction factor for the injury index is calculated, including: The downward deviation of the proportion of high-frequency energy relative to the median level of the intact rock mass was obtained. The upward deviation of the information entropy relative to the median level of the intact rock mass; Based on the pre-calculated magnitude balance coefficient, the downward deviation magnitude and the upward deviation magnitude are equivalently linearly superimposed to finally obtain the damage index correction amount.

[0017] Furthermore, a comprehensive score is calculated by adaptively switching weight configurations based on damage mode flags, including: The normalized deviation of the measured longitudinal wave velocity from the reference value of the intact rock mass is calculated to obtain the wave velocity deviation index, and the normalized values ​​of information entropy and high-frequency energy ratio are calculated respectively. Determine whether the damage pattern markers indicate a latent damage state; If the damage is latent, the weight of the wave velocity deviation index is reduced, and the weights of the normalized value of the high-frequency energy ratio and the damage index correction are increased. The weighted sum is then calculated based on the weighted configuration after the switch, and the final comprehensive score is obtained.

[0018] Its effect is that, through the adaptive switching of dual weights triggered by damage mode indicators, the comprehensive scoring model can actively avoid the misleading effect of artificially high wave velocity on the scoring results when facing high confining pressure compaction-type latent damage, and convert the quantitative deviation of frequency domain scattering characteristics into scoring penalties, thus ensuring that the stability level judgment results have reliable distinguishing ability in two completely different physical action modes: macroscopic cracking damage and compaction-type latent damage.

[0019] Furthermore, based on the comprehensive score, the surrounding rock stability level is output, including: Obtain multiple quantile values ​​of the comprehensive score distribution in the calibration dataset as boundary lines; The range of values ​​for the comprehensive score is divided into multiple stability level intervals using boundary lines; Determine the stability level range into which the comprehensive score falls, and output the corresponding surrounding rock stability level.

[0020] Furthermore, the adaptive switching of weight configuration based on damage mode flags also includes: During the system calibration phase, the damage level label of each known damage state sample in the calibration dataset is used as the target variable; Within subsets of different damage pattern markers, the absolute values ​​of Spearman rank correlation coefficients between each feature index and the damage level label are calculated. The weighting of feature indicators within each subset is determined by normalizing the absolute value of the Spearman rank correlation coefficient.

[0021] Furthermore, after outputting the corresponding surrounding rock stability level, it also includes: When the stability level of the surrounding rock falls into the abnormal warning or severe abnormal range, the audible and visual alarm module is triggered to issue a prompt. The system simultaneously displays the grade determination results, the values ​​of each indicator, and the status of compaction damage indicators, and writes the test data and evaluation records into the data storage module.

[0022] The beneficial effects of this invention are: by using damage mode indicators to drive the adaptive switching of weights in the comprehensive scoring model, the scoring results are no longer affected by the artificially inflated high wave velocity, and the reliability of stability level determination is substantially improved. All thresholds, weights, and level boundaries in the scheme are automatically derived from calibration data, eliminating subjective biases caused by manual experience-based assignment, and possessing engineering practicality for adaptive initialization and on-site operation without manual intervention in different mining environments. Attached Figure Description

[0023] Figure 1 This is a flowchart of a method for evaluating the stability of surrounding rock in mines based on elastic wave response characteristics.

[0024] Figure 2 This is a two-dimensional joint discrimination distribution diagram of longitudinal wave velocity and wavelet packet information entropy in this invention. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] An embodiment of the mine surrounding rock stability evaluation method based on elastic wave response characteristics provided by this invention: like Figure 1 As shown, the method for evaluating the stability of mine surrounding rock based on elastic wave response characteristics includes the following steps: S1, apply elastic wave excitation to obtain an enhanced response signal.

[0027] In an optional embodiment, a planar contact probe is pressed against the surface of the surrounding rock to be tested, and an excitation pulse is sent from the signal excitation module to the transducer unit to drive the probe to continuously apply excitation pulses to the same detection point. Secondary elastic wave excitation. The value is not predetermined manually, but is adaptively determined based on the background noise level at the site: after the signal acquisition module completes the initial excitation acquisition, the response signal is calculated before the arrival of the first arrival wave. The root mean square amplitude within the time window is recorded as the background noise baseline value. After each subsequent excitation, the peak value of the initial arrival wave envelope is extracted. Calculate the signal-to-noise ratio When satisfaction is obtained continuously The effective number of stimulations reached If the triggering stops and the subsequent processing steps are not completed, the triggering process continues until the maximum number of triggering attempts set by the system is reached. Among them, the signal-to-noise ratio threshold The lower quartile of the signal-to-noise ratio distribution of historical data collected in the calibration and testing of similar rock masses in the mine is used as the objective lower boundary of acceptable coupling quality under the current testing environment, rather than being given by human experience.

[0028] Furthermore, for the selected The effective response signals are sample-by-sample aligned and averaged in the time domain to obtain the enhanced response signal after removing random interference. This serves as a unified input for subsequent feature extraction. The alignment benchmark for the superimposed average is the initial arrival time of each response, which is automatically picked up in the time domain using the energy accumulation method: accumulating the squared value of the signal sample by sample from the signal start point; when the accumulated energy exceeds the total signal energy... The moment is defined as the initial arrival at the take-off point. Since downhole background noise typically manifests as low-amplitude random disturbances in the time domain, the accumulated energy grows slowly within the time window before the arrival of the first arrival. However, the sharp increase in the squared value of the signal after the arrival of the first arrival causes a significant abrupt change in the slope of the accumulation curve. The energy threshold is precisely positioned before this slope abrupt change range, giving the method strong anti-interference capability against downhole background noise interference and avoiding the superposition misalignment problem caused by the drift of the start-up time in different probe coupling states of the fixed time window picking method.

[0029] Next, from Calculate elastic waves at a known transmit-receive distance Propagation time difference The measured longitudinal wave velocity at the detection point was obtained. Simultaneously extract the peak value of the first arrival wave envelope. With excitation reference amplitude The ratio of the two values ​​yields the amplitude attenuation coefficient. . The larger the value, the more severe the overall energy loss of the elastic wave along the propagation path, reflecting the stronger the overall attenuation of the elastic wave by the rock mass. The smaller the value, the lower the energy loss along the propagation path, and the weaker the rock mass's ability to block the wave. These two parameters will serve as the basic input variables for subsequent damage assessment and will be passed to the next step for further calculation.

[0030] By employing adaptive signal-to-noise ratio (SNR) filtering and energy accumulation method for alignment and superposition, the interference of downhole random noise on the response signal quality is effectively suppressed, ensuring the enhanced response signal upon which subsequent feature extraction depends. The signal-to-noise ratio and the accuracy of the initial arrival time calibration provide a reliable signal basis for refined frequency band analysis, avoiding distortion of subsequent frequency band energy characteristics caused by occasional noise pollution from a single excitation.

[0031] S2, wavelet packet decomposition extracts frequency band energy features.

[0032] In an optional embodiment, the enhanced response signal Wavelet packet decomposition is performed to decompose the signal into several frequency band nodes on the frequency axis, with each node corresponding to the vibration energy components within a specific frequency range. The wavelet basis functions must satisfy the compact support condition and have a sufficient vanishing moment order to ensure energy orthogonality between frequency band nodes and avoid interference from energy leakage between frequency bands in subsequent discrimination. It should be noted that there are various wavelet basis functions in the art that satisfy the compact support and orthogonality requirements; those skilled in the art can select a suitable wavelet basis based on the time-frequency characteristics of the actual excitation signal.

[0033] Furthermore, the wavelet packet... The thinnest bandwidth after layer decomposition is Number of decomposition layers Based on signal sampling frequency With the excitation center frequency It is determined that: to ensure a sufficient number of independent frequency band nodes within the high-frequency band to support subsequent... The statistical resolution requires that the narrowest bandwidth is no wider than [a certain value]. The preset ratio, namely:

[0034] For example, the preset ratio can be taken as follows: That is, the minimum bandwidth must meet the following requirements. Therefore, we can conclude that:

[0035] Take the smallest positive integer that satisfies the above inequality as the actual decomposition level. The data processing module automatically calculates and writes the values ​​based on the current excitation parameters before each detection, requiring no manual intervention. This is based on the excitation center frequency. Sampling frequency For illustrative purposes, substituting into the above formula yields... After rounding At this point, the narrowest bandwidth is approximately It can accommodate no less than 3 independent frequency band nodes in the high-frequency band to meet the resolution requirements.

[0036] Then, after the decomposition is completed, the first... All layers The reconstructed signals of each frequency band node are used to calculate the node energy. Energy of each node The sum of the squares of the amplitudes of each sampling point of the reconstructed signal at this node, and the total signal energy. High-frequency and low-frequency bands are used to excite the center frequency. As a dividing line: the frequency range is higher than The nodes are classified into the high-frequency band, and the sum of their energy is denoted as... The frequency range is not higher than The nodes are classified as low-frequency bands, and their sum of energy is denoted as... This division is based on the physical center frequency of the excitation signal itself and is independent of the statistical results of the calibration data, making it fully reproducible to those skilled in the art. High-frequency energy proportion. for:

[0037] The larger the value, the higher the retention of the high-frequency components of the elastic wave along the propagation path, and the weaker the scattering and dissipation of the high-frequency components by the rock mass. The smaller the value, the more severe the loss of high-frequency components during propagation, and the higher the density of scattering sources along the corresponding path.

[0038] Next, in the extraction Simultaneously, the information entropy of the wavelet packet frequency band nodes is calculated. It is defined as the Shannon entropy, which is a discrete distribution of the energy proportions of each node. The elastic wave response energy of a complete rock mass is highly concentrated near the excitation center frequency. Lower; because energy diffuses to multiple non-central frequency bands during crack scattering. Significantly increased. and Together they form a frequency domain feature pair, which, along with the extracted features... and They will then proceed to the next step of the judgment process.

[0039] The elastic wave response signal is finely decomposed to the frequency band node level by wavelet packet decomposition, and the high-frequency energy ratio is obtained simultaneously. With frequency band energy distribution entropy Two complementary frequency domain features, compared to existing methods that rely solely on overall amplitude attenuation, can capture the selective dissipation of high-frequency components without significant changes in the total signal energy, providing physically interpretable feature inputs for subsequent identification of high-compression compaction-type latent damage.

[0040] S3 identifies latent damage states and calculates correction amounts.

[0041] In an optional embodiment, with and A two-dimensional feature discrimination plane is constructed to classify and identify the damage patterns at each detection point. Three typical distribution regions exist within this plane: the first type... High and High, corresponding to a real, intact rock mass, where high confining pressure did not induce a fracture network, and high-frequency components maintained intact propagation; the second type, low and Low or moderate damage corresponds to macroscopically cracked rock mass, where crack opening leads to reduced wave velocity, and scattering at the crack interface causes high-frequency dissipation; the third type... High but Abnormally low at the same time The high confining pressure compaction-type latent damage state, which is the focus of this invention, is characterized by the recovery of wave velocity after crack closure, but the rough interface of the closed crack surface continues to scatter high-frequency energy, causing... The systemic density is low, and scattering causes band energy diffusion. Elevation. It is precisely the third type of state that depends solely on... The existing evaluation system makes it difficult to distinguish from the first type of intact rock mass, which constitutes the root cause of hidden misjudgments.

[0042] Furthermore, the specific execution logic for damage mode determination is as follows: First, determine... Is it higher than the reference value for wave velocity in intact rock mass corresponding to the lithology of this detection point? If the preset ratio is not met, the damage is directly classified as macroscopic cracking, and subsequent compaction damage identification is skipped. For example, this preset ratio can be taken as... If so, then further judgment is needed. Is it lower than the calibration data of intact rock mass? The lower quarter quantile is denoted as ,at the same time Is it higher than the calibration data of intact rock mass? The upper quarter quantile value is denoted as If both conditions are met simultaneously, the detection point is marked as a compaction-type latent damage state, and the damage mode indicator is recorded. Otherwise, mark it as complete. The two quantile values ​​above are both from batch data collected during the calibration phase of the mine to confirm the location of the complete rock mass. They were automatically statistically entered by the data processing module during the initialization process and do not rely on manual experience parameters.

[0043] Next, regarding The detection point introduces a damage index correction. The correction amount is from Deviation The degree and Exceeding The degree of determination is jointly determined by:

[0044] in intact rock mass The median of intact rock mass The medians of both terms are derived from calibration data. The first term reflects the downward deviation of the proportion of high-frequency energy relative to the median level of intact rock mass. A larger value indicates more severe high-frequency scattering dissipation and more significant disturbance of high-frequency components by compaction fractures. The second term reflects the upward deviation of the frequency band energy distribution entropy relative to the median level of intact rock mass. A larger value indicates a higher degree of energy diffusion between frequency bands and a denser distribution of scattering sources.

[0045] The magnitude balance coefficient for the two biases is calculated by taking the first term in the calibration dataset. Sample standard deviation With the second item Sample standard deviation ,make:

[0046] This ensures that the fluctuation ranges of the two items are consistent on the calibration sample, supporting the rationality of equivalent linear superposition. During the system initialization phase, the data processing module automatically calculates and stores the data based on the calibration data, which can be directly called during on-site testing and must not be manually provided. The larger the value, the deeper the compaction-type latent damage, and it will directly participate in the comprehensive scoring calculation in the next step, serving as a specific penalty item for identifying this type of damage pattern.

[0047] Through and Superimposed on the two-dimensional discriminant plane The auxiliary constraint conditions effectively separate intact rock masses and compacted, subtly damaged rock masses in high wave velocity regions from the overlapping state that cannot be distinguished by a single wave velocity index. The introduction of this provides a clear trigger for the subsequent weight switching of the comprehensive score, fundamentally blocking the transmission path that could mislead the determination of stability level due to artificially high wave velocity under high confining pressure.

[0048] S4, adaptively switches weights to calculate scores and output levels.

[0049] In an optional embodiment, all extracted state characteristics, including the wave velocity deviation index, are considered. Amplitude attenuation coefficient Normalized value of wavelet packet information entropy Normalized value of high-frequency energy ratio and the calculated compaction damage correction amount Input the comprehensive scoring model to calculate the comprehensive score of the surrounding rock stability at the detection point. Among them, the wave speed deviation index Defined as actual measurement Relative to the reference value of intact rock mass Normalized deviation:

[0050] and Ratio normalization based on the median of the calibrated data was applied respectively:

[0051] in and These are all median values ​​of the complete rock mass calibration data defined in the previous step. Since the definitions remain consistent, there is no need to introduce new symbols. After normalization, the values ​​corresponding to the complete rock mass... and All are based on 1 as a reference standard. The direction and magnitude of deviation from 1 directly reflect the degree of damage, and the physical meaning is clear.

[0052] Furthermore, the overall score adopts an adaptive weighted summation method:

[0053] Weights satisfy Its allocation is based on damage pattern markers Adaptive switching to two fixed configurations: when hour, Take a higher allocation ratio, and use wave velocity deviation as the primary criterion; when hour, Significantly reduced and Correspondingly, the frequency band energy characteristics and compaction damage correction amount are taken as the main basis, so as to avoid the high wave velocity from artificially inflating the scoring results.

[0054] Next, the two sets of weights were determined during the system calibration phase in the following way: using the damage level label of each known damage state sample in the calibration dataset as the target variable, respectively... Subset and Within the subset, calculate the absolute value of the Spearman rank correlation coefficient between each feature index and the damage level label, and denote the rank correlation coefficient as the rank correlation coefficient. The absolute value of the rank correlation coefficient of each indicator within the corresponding subset is Then the first subset The weights of each indicator are:

[0055] This relationship ensures that the sum of weights within each subset is always 1. The above calculations are automatically completed by the data processing module during the initialization phase, and the results are stored in a lookup table. During on-site testing, the results are then... Real-time index retrieval, without relying on manual experience for value assignment.

[0056] Then, The value space is divided into four stability level intervals, corresponding to stable, basically stable, anomaly warning, and severe anomaly, respectively. The boundaries of each level are determined by the calibration dataset. The three quantiles of the distribution are automatically determined: calculated from all samples in the calibration dataset. 25th percentile 50th percentile 75th percentile As three boundary lines, The range of values ​​is divided into four intervals. When When the system falls into an abnormal warning or severe abnormal range, the stability evaluation module triggers the audible and visual alarm module to issue a prompt. The result display module simultaneously displays the level judgment result, the values ​​of each indicator, and the status of compaction damage indicators for on-site personnel to refer to in real time. All test data, characteristic parameters, and evaluation records are written to the data storage module to support subsequent traceability and multi-point horizontal comparative analysis.

[0057] Thus, through The dual-weight adaptive switching of the trigger condition enables the comprehensive scoring model to actively avoid the misleading effect of artificially high wave velocity on the scoring results when facing high confining pressure compaction-type latent damage. It converts the quantitative deviation of frequency domain scattering characteristics into scoring penalties, ensuring that the stability level judgment results have reliable distinguishing ability under two completely different physical action modes: macroscopic cracking damage and compaction-type latent damage. This achieves the adaptive conversion from a single wave velocity-dominated system to a multi-dimensional frequency domain characteristic-dominated system.

[0058] like Figure 2 As shown, it can be observed that the three types of states exhibit a clustered distribution pattern with good separation in the joint feature space of P-wave velocity and wavelet packet information entropy. Intact rock masses are clustered in the high wave velocity and low information entropy region; macroscopically cracked rock masses are distributed in the low wave velocity region; and although the high confining pressure compacted latent damage rock masses highly overlap with the intact rock masses on the wave velocity axis, they are significantly shifted to the high value region on the information entropy axis, thus achieving effective separation on the two-dimensional joint feature plane. This intuitively presents the core discrimination logic of the method of this invention for identifying latent damage using frequency domain joint features.

[0059] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for evaluating the stability of mine surrounding rock based on elastic wave response characteristics, characterized in that, include: S1, continuously apply elastic wave excitation to the surface of the surrounding rock to be tested, filter the effective response signals based on adaptive signal-to-noise ratio and perform aligned superposition and averaging to obtain the enhanced response signal, and extract the measured longitudinal wave velocity and amplitude attenuation coefficient. S2, perform wavelet packet decomposition on the enhanced response signal, extract the energy of each frequency band node based on the preset decomposition level, and calculate the high-frequency energy ratio and the information entropy of the wavelet packet frequency band node; S3. Based on the measured longitudinal wave velocity and the proportion of high-frequency energy, a two-dimensional feature discrimination plane is constructed. The information entropy is combined to identify the latent damage state of high confining pressure compaction to generate damage mode markers. When the damage is determined to be latent damage, the damage index correction amount is calculated. By combining information entropy to identify latent damage states caused by high confining pressure compaction, damage pattern markers are generated, including: Determine whether the measured longitudinal wave velocity is higher than the preset ratio of the reference value of the wave velocity of intact rock mass; If it is higher than the preset ratio, then it is further determined whether the proportion of high-frequency energy is lower than the first quantile value and whether the information entropy is higher than the second quantile value. If the proportion of high-frequency energy and the information entropy simultaneously meet the above conditions, it is determined to be a latent damage state of high confining pressure compaction, and a damage mode marker marked as latent damage is generated. When a lesion is identified as latent, a correction to the lesion index is calculated, including: The downward deviation of the proportion of high-frequency energy relative to the median level of the intact rock mass was obtained. The upward deviation of the information entropy relative to the median level of the intact rock mass; Based on the pre-calculated magnitude balance coefficient, the downward deviation magnitude and the upward deviation magnitude are equivalently linearly superimposed to finally obtain the damage index correction amount; S4 inputs the measured longitudinal wave velocity deviation index, amplitude attenuation coefficient, normalized value of information entropy, normalized value of high frequency energy ratio, and damage index correction amount into the comprehensive scoring model. Based on the damage mode indicator, the weight configuration is adaptively switched to calculate the comprehensive score, and the surrounding rock stability level is output based on the comprehensive score.

2. The method for evaluating the stability of mine surrounding rock based on elastic wave response characteristics according to claim 1, characterized in that, The enhanced response signal is obtained by selecting effective response signals based on adaptive signal-to-noise ratio and aligning, superimposing, and averaging them, including: The root mean square amplitude of the response signal acquired during the first excitation before the arrival of the first wave is calculated as the background noise baseline. The signal-to-noise ratio is obtained by extracting the peak value of the first arrival envelope of each subsequent excitation and calculating the ratio with the background noise baseline value. Excitation stops when the number of valid excitations that meet the signal-to-noise ratio threshold reaches a preset lower limit. The selected valid response signals are then time-domain aligned, superimposed, and averaged using the energy accumulation method to finally obtain the enhanced response signal.

3. The method for evaluating the stability of mine surrounding rock based on elastic wave response characteristics according to claim 1, characterized in that, Extract the measured P-wave velocity and amplitude attenuation coefficient, including: Based on the initial arrival and take-off point of the enhanced response signal, the propagation time difference of the elastic wave at the known transmission and reception distance is calculated to obtain the measured longitudinal wave velocity. The peak value of the first arrival envelope of the enhanced response signal is extracted synchronously, and the ratio of the peak value of the first arrival envelope to the excitation reference amplitude is calculated. The degree of energy loss is calculated based on the ratio, and the amplitude attenuation coefficient is finally obtained.

4. The method for evaluating the stability of mine surrounding rock based on elastic wave response characteristics according to claim 1, characterized in that, Based on a preset number of decomposition layers, the energy of each frequency band node is extracted to calculate the proportion of high-frequency energy and the information entropy of wavelet packet frequency band nodes, including: The minimum bandwidth that meets the resolution requirements is determined based on the signal sampling frequency and the excitation center frequency, and the number of decomposition layers is calculated. The node energy is calculated for the reconstructed signals of all frequency band nodes under the decomposition layer, and the high frequency band and low frequency band are divided by the excitation center frequency. The high-frequency energy percentage is obtained by calculating the ratio of the sum of the high-frequency band energies to the total signal energy. Then, the Shannon entropy of the discrete distribution of the energy percentages of each node is calculated, and finally, the information entropy of the wavelet packet band nodes is obtained.

5. The method for evaluating the stability of mine surrounding rock based on elastic wave response characteristics according to claim 1, characterized in that, The weight configuration is adaptively switched based on the damage mode flag to calculate the overall score, including: The normalized deviation of the measured longitudinal wave velocity from the reference value of the intact rock mass is calculated to obtain the wave velocity deviation index, and the normalized values ​​of information entropy and high-frequency energy ratio are calculated respectively. Determine whether the damage pattern markers indicate a latent damage state; If the damage is latent, the weight of the wave velocity deviation index is reduced, and the weights of the normalized value of the high-frequency energy ratio and the damage index correction are increased. The weighted sum is then calculated based on the weighted configuration after the switch, and the final comprehensive score is obtained.

6. The method for evaluating the stability of mine surrounding rock based on elastic wave response characteristics according to claim 1, characterized in that, The surrounding rock stability level is output based on the comprehensive score, including: Obtain multiple quantile values ​​of the comprehensive score distribution in the calibration dataset as boundary lines; The range of values ​​for the comprehensive score is divided into multiple stability level intervals using boundary lines; Determine the stability level range into which the comprehensive score falls, and output the corresponding surrounding rock stability level.

7. The method for evaluating the stability of mine surrounding rock based on elastic wave response characteristics according to claim 5, characterized in that, The adaptive switching of weight configuration based on damage mode flags also includes: During the system calibration phase, the damage level label of each known damage state sample in the calibration dataset is used as the target variable; Within subsets of different damage pattern markers, the absolute values ​​of Spearman rank correlation coefficients between each feature index and the damage level label are calculated. The weighting of feature indicators within each subset is determined by normalizing the absolute value of the Spearman rank correlation coefficient.

8. The method for evaluating the stability of mine surrounding rock based on elastic wave response characteristics according to claim 6, characterized in that, After outputting the corresponding surrounding rock stability level, it also includes: When the stability level of the surrounding rock falls into the abnormal warning or severe abnormal range, the audible and visual alarm module is triggered to issue a prompt. The system simultaneously displays the grade determination results, the values ​​of each indicator, and the status of compaction damage indicators, and writes the test data and evaluation records into the data storage module.

Images