A deep well rock mass instability failure time accurate prediction and early warning system

Abstract

The application discloses a deep well rock mass instability damage time accurate prediction and early warning system and relates to the technical field of deep well rock mass damage early warning. The application discloses a deep well rock mass instability damage time accurate prediction and early warning system and relates to the technical field of deep well rock mass damage early warning. The application discloses a deep well rock mass instability damage time accurate prediction and early warning system and relates to the technical field of deep well rock mass damage early warning. The application discloses a deep well rock mass instability damage time accurate prediction and early warning system and relates to the technical field of deep well rock mass damage early warning. The application discloses a deep well rock mass instability damage time accurate prediction and early warning system and relates to the technical field of deep well rock mass damage early warning. The application discloses a deep well rock mass instability damage time accurate prediction and early warning system and relates to the technical field of deep well rock mass damage early warning. The application discloses a deep well rock mass instability damage time accurate prediction and early warning system and relates to the technical field of deep well rock mass damage early warning. The application discloses a deep well rock mass instability damage time accurate prediction and early warning system and relates to the technical field of deep well rock mass damage early warning. The application discloses a deep well rock mass instability damage time accurate prediction and early warning system and relates to the technical field of deep well rock mass damage early warning. The application discloses a deep well rock mass instability damage time accurate prediction and early warning system and relates to the technical field of deep well rock mass damage early warning. The application discloses a deep well rock mass instability damage time accurate prediction and early warning system and relates to the technical field of deep well rock mass damage early warning. The application discloses a deep well rock mass instability damage time accurate prediction and early warning system and relates to the technical field of deep well rock mass damage early warning. The application discloses a deep well rock mass instability damage time accurate prediction and early warning system and relates to the technical field of deep well rock mass damage early warning. The application discloses a deep well rock mass instability damage time accurate prediction and early warning system and relates to the technical field of deep well rock mass damage early warning. The application discloses a deep well rock mass instability damage time accurate prediction and early warning system and relates to the technical field of deep well rock mass damage early warning. The application disc

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of deep well rock mass failure early warning, in particular to a deep well rock mass instability failure time accurate prediction and early warning system. BACKGROUND

[0002] Deep well rock mass failure early warning technology refers to a comprehensive technology system that, in a deep well environment, identifies instability, collapse, spalling, roof fall and other failure precursors caused by stress concentration, geological structure activity, excavation disturbance and other factors in rock mass, and then predicts the failure risk in advance and issues an early warning signal to avoid or mitigate engineering disasters through the integration of monitoring, analysis, modeling and decision-making and other technical means.

[0003] The existing deep well rock mass failure early warning technology often uses acoustic emission monitoring to predict rock mass instability and failure, which is based on the characteristics of a sudden increase in event rate and energy concentration of acoustic emission signals before rock mass instability and failure. However, in the deep well environment, there are a large number of non-rock mass failure interference signals, making it difficult to effectively distinguish between effective signals and interference signals, resulting in a high false positive rate or effective signals being overwhelmed, affecting the accuracy of the early warning. Moreover, traditional acoustic emission analysis relies on a single or a small number of characteristic parameters, such as peak amplitude, ringing count, rise time, energy, etc. The data analysis and early warning decision-making are highly dependent on the experience of the operator, and the interpretation of parameters is limited to "threshold judgment". However, deep well rock mass failure is a complex process with multiple stages and mechanisms, and the acoustic emission signal characteristics at different stages differ significantly. Traditional methods are difficult to capture such multi-dimensional characteristics, which can lead to delayed early warning. For example, the patent application with publication number CN112444564A discloses a rock failure early warning method based on acoustic emission signal statistical analysis. This method uses only the ringing count to predict rock failure, which has limited accuracy and delayed early warning. Therefore, the existing deep well rock mass failure early warning technology uses manual processing and analysis of a single or a small number of characteristic parameters, which has low reliability and timeliness in early warning. SUMMARY

[0004] The present application aims to at least solve one of the technical problems in the prior art, by collecting a plurality of deep well rock samples for uniaxial compression experiments, and based on the acoustic emission sensor, real-time acquisition of acoustic emission signals to obtain original acoustic emission signal data; the original acoustic emission signal data is adaptively preprocessed, and data classification and integration are performed to obtain rock sample acoustic emission parameter data; an original failure prediction model is constructed, and model training and verification optimization are performed based on the rock sample acoustic emission parameter data to obtain a rock failure prediction model; the acoustic emission signals of the deep well rock mass are collected, and the rock failure prediction model is used to predict the instability failure time of the deep well rock mass in real time; to solve the problem that the existing deep well rock mass failure warning technology uses artificial processing and analysis of single or a small number of characteristic parameters when warning the instability failure of the rock mass through acoustic emission monitoring, and the reliability and timeliness of the warning are not high.

[0005] To achieve the above-mentioned purpose, the present application provides a deep well rock mass instability failure time accurate prediction and early warning system, comprising a test simulation module, a data processing module, a model construction module and a rock mass early warning module;

[0006] The test simulation module is used to collect a plurality of deep well rock samples for uniaxial compression experiments, and based on the acoustic emission sensor, real-time acquisition of acoustic emission signals to obtain original acoustic emission signal data;

[0007] The data processing module includes a preprocessing unit and a classification and integration unit, the preprocessing unit is used to adaptively preprocess the original acoustic emission signal data, and the classification and integration unit is used to classify and integrate the data to obtain rock sample acoustic emission parameter data;

[0008] The model construction module is used to construct an original failure prediction model, and based on the rock sample acoustic emission parameter data, model training and verification optimization are performed to obtain a rock failure prediction model;

[0009] The rock mass early warning module is used to collect acoustic emission signals of the deep well rock mass, and based on the rock failure prediction model, the instability failure time of the deep well rock mass is predicted in real time;

[0010] Further, the test simulation module is configured with a test simulation strategy, and the test simulation strategy includes:

[0011] Rock samples are collected from deep wells, and according to experimental specifications, the rock samples are processed into standard test pieces, denoted as deep well rock samples, and any one of the deep well rock samples is denoted as a first rock sample;

[0012] The acoustic emission sensor is tightly attached to the surface of the first rock sample, and the first rock sample is subjected to uniaxial compression experiment until the first rock sample fails.

[0013] Further, the test simulation strategy further includes:

[0014] During the uniaxial compression experiment, the stress of the first rock sample is monitored in real time, and the collection time is recorded as the stress change data of the first rock sample, and the acoustic emission sensor is used to monitor and record the acoustic emission signal of the first rock sample as the original acoustic emission data of the first rock sample;

[0015] The maximum stress in the stress change data of the first rock sample is marked as the limit stress of the first rock sample, and the collection time corresponding to the limit stress is recorded as the rock sample failure point of the first rock sample;

[0016] The stress change data and the original acoustic emission data of all deep well rock samples are repeatedly obtained.

[0017] Further, the preprocessing unit is configured with a preprocessing strategy, and the preprocessing strategy includes:

[0018] For the acoustic emission signal of the first rock sample, denoted as the first acoustic emission signal, the average value of all data points of the first acoustic emission signal is calculated, and each data point in the first acoustic emission signal is subtracted from the average value of all data points, and after completion, the second acoustic emission signal is obtained;

[0019] The Morlet wavelet with a center frequency of k1 is selected to decompose the second acoustic emission signal into k2 scales, the modulus value of the complex wavelet coefficient of each scale is taken, and the integral of the square of the modulus value is calculated to obtain the energy corresponding to the scale, wherein k1 is the set frequency, and k2 is the set number;

[0020] The proportion of the energy of each scale to the total energy of all scales is calculated and denoted as the energy proportion, and the k3 scales with the largest energy proportion are retained and denoted as the effective scales, and k3 is the set number, k3<k2.

[0021] Further, the preprocessing strategy further includes:

[0022] For any one effective scale, denoted as the first scale, the modulus value of the complex wavelet coefficient of the first scale is taken to obtain the envelope sequence corresponding to the first scale, denoted as the first envelope sequence;

[0023] The average value of the first envelope sequence is calculated and denoted as AU, k4*AU is denoted as the peak threshold value, and the points in the first envelope sequence that exceed the peak threshold value are denoted as effective peak points, the number of effective peak points and the total number of points of the first envelope sequence are counted and denoted as BS and AS0 in order, and the sparse factor XS corresponding to the first scale is calculated, wherein XS=BS / AS0;

[0024] The standard deviation of the first envelope sequence is obtained and denoted as BU, and the suppression threshold YZ is calculated, wherein YZ=R0*BU, R0=k4*-10*(k4-1)*XS, wherein k4 is the set maximum multiple;

[0025] For the points less than the suppression threshold in the first envelope sequence, the corresponding wavelet coefficients are set to zero; for the points greater than or equal to the suppression threshold in the first envelope sequence, the corresponding wavelet coefficients are retained; the modified envelope sequence corresponding to the first scale is completed;

[0026] The modified envelope sequences corresponding to all effective scales are repeatedly obtained, for each modified envelope sequence, inverse mapping is performed using the corresponding Morlet wavelet, and all effective signal components obtained by inverse mapping are aligned along the time axis and superimposed to obtain a denoised emission signal of the first acoustic emission signal, denoted as a first denoised signal;

[0027] The denoised emission signals corresponding to the acoustic emission signals of all deep well rock samples are repeatedly obtained.

[0028] Further, the classification and integration unit is configured with a classification and integration strategy, and the classification and integration strategy comprises:

[0029] For the first denoised signal of the first rock sample, a first time interval t1 is set; according to t1, and taking the rock sample failure point as the endpoint, a plurality of time points are selected from the first denoised signal, denoted as parameter sampling points, and any one parameter sampling point is denoted as a first sampling point;

[0030] The stress size corresponding to the first sampling point is obtained, denoted as parameter 1, the amplitude and frequency of the first sampling point are obtained, denoted as parameters 2 and 3 in order;

[0031] Taking the first sampling point as the endpoint of the first denoised signal, the corresponding AE event number, ringing count, rise time, energy entropy, RA value, AE event rate and ringing event rate are obtained, denoted as parameters 4-10 in order; parameters 1-10 are denoted as parameter information of the first sampling point;

[0032] The parameter information of all parameter sampling points is repeatedly obtained and arranged in the order of time corresponding to the parameter sampling points, denoted as rock sample acoustic emission parameter data.

[0033] Further, the model construction module is configured with a model construction strategy, and the model construction strategy comprises:

[0034] Based on the long short-term memory network, an original failure prediction model is constructed, and the original failure prediction model comprises an input layer, a hidden layer and an output layer;

[0035] For the parameters 1-10 of the first de-noised signal, normalization processing is performed according to the parameter types respectively, and after completion, it is recorded as standard parameters 1-10, and any one of the standard parameters 2-9 is recorded as a standard parameter i; the standard parameter 1 and the standard parameter i of the same parameter sampling point are combined into a corresponding single-parameter feature vector, recorded as DM={standard parameter 1, standard parameter i}; and all single-parameter feature vectors are arranged in corresponding time order, recorded as single-parameter feature data;

[0036] The single-parameter feature data is divided into a single-parameter training set and a single-parameter test set according to a ratio of 7:3, the output result of the output layer is set as the stress size, the single-parameter training set is used to train the original failure prediction model, after completion, the single-parameter test set is used to test the model, and the root mean square error of the model is calculated, recorded as the RMSE corresponding to the standard parameter i;

[0037] The RMSE corresponding to all standard parameters is repeatedly obtained and arranged in order from small to large, recorded as an error sequence, and the corresponding standard parameters are reordered according to the order of the error sequence, recorded as important parameters 1-9 in turn.

[0038] Further, the model construction strategy further includes:

[0039] The combination of the standard parameter 1 and the important parameter 1-j is recorded as a parameter combination j, where j∈[1, 9], and all parameter combinations are obtained;

[0040] For the parameter combination j, the standard parameter 1 and the important parameter 1-j of the same parameter sampling point are combined into a corresponding parameter feature vector, recorded as XM={standard parameter 1, important parameter 1, …, important parameter j}; and all parameter feature vectors are arranged in corresponding time order, recorded as the jth combination feature data;

[0041] The jth combination feature data is divided into a combination training set and a combination test set according to a ratio of 7:3, the combination training set is used to train the original failure prediction model, after completion, the jth prediction model is obtained, the combination test set is used to test the jth prediction model, and the determination coefficient, the average determination error, the root mean square error and the average absolute percentage error of the model are calculated, recorded as the evaluation parameters of the parameter combination j.

[0042] Further, the model construction strategy further includes:

[0043] The evaluation parameters of all parameter combinations are repeatedly obtained, and the average determination error, the root mean square error and the average absolute percentage error of all parameter combinations are respectively arranged in order from large to small, recorded as error sequence 1, error sequence 2 and error sequence 3 in order, and the determination coefficients of all parameter combinations are arranged in order from small to large, recorded as error sequence 4;

[0044] respectively, and the position sequence numbers of the evaluation parameters of the parameter combination j in the corresponding error sequence 1-4 are obtained and accumulated, and the total score of the parameter combination j is recorded as;

[0045] The total scores of all parameter combinations are repeatedly obtained, and the parameter combination with the maximum total score is recorded as the predicted parameter combination of the first rock sample, the prediction model corresponding to the predicted parameter combination is marked as the rock failure prediction model of the first rock sample, and the average absolute percentage error corresponding to the rock failure prediction model is recorded as PE;

[0046] The predicted parameter combination, the rock failure prediction model and PE of the deep rock sample are repeatedly obtained.

[0047] Further, the rock mass early warning module comprises a rock mass early warning strategy, and the rock mass early warning strategy comprises:

[0048] The stress size and the acoustic emission signal of the deep well rock mass are collected, and are recorded as the current stress size and the current acoustic emission signal respectively, the current acoustic emission signal is adaptively preprocessed to obtain a current denoising signal;

[0049] According to the first time interval, a plurality of time points are uniformly selected from the current denoising signal, and parameters corresponding to the predicted parameter combination are obtained at the time points, recorded as current parameters, all current parameters and the current stress size are normalized according to the parameter type, and are combined into a corresponding parameter feature vector, and then the corresponding rock failure prediction model is input according to the type of the deep well rock mass, to obtain stress prediction data at a future time;

[0050] The stress size of each time point in the future predicted is multiplied by (1+PE) to obtain the corrected stress size of each time point in the future, and the corrected stress sizes are sequentially searched according to time sequence, the first corrected stress size greater than or equal to the corresponding limit stress is recorded as the instability failure point, and the prediction time of the deep well rock mass to occur instability failure is obtained according to the time point corresponding to the instability failure point, and a prediction warning is performed.

[0051] The beneficial effects of the present application are as follows: the present application collects a plurality of deep well rock samples for uniaxial compression experiment, and based on the acoustic emission sensor, real-time acoustic emission signals are collected to obtain original acoustic emission signal data; the original acoustic emission signal data is adaptively preprocessed, and data classification and integration are performed to obtain rock sample acoustic emission parameter data; an original failure prediction model is constructed, and model training and verification optimization are performed based on the rock sample acoustic emission parameter data to obtain a rock failure prediction model; the acoustic emission signal of the deep well rock mass is collected, and the instability failure time of the deep well rock mass is real-time predicted based on the rock failure prediction model; the present application can automatically process and analyze various characteristic parameters of the acoustic emission signal without relying on manual operation, and the reliability and timeliness of the early warning are improved;

[0052] The application sets an adaptive threshold denoising for each scale by Morlet wavelet decomposition and the sparse factor of the decomposed scale, and has the advantages that the physical characteristics of the acoustic emission signal are used to realize adaptive denoising, the threshold of the denoising processing can be automatically adjusted, the robustness is improved while the denoising effect is ensured, the single prediction ability of each parameter is obtained quantitatively by training the model with a single parameter, sorting by RMSE, then combining parameters step by step to train the model and evaluate multiple indexes, the error caused by subjective selection is avoided, the best parameter combination is selected, the error caused by invalid and redundant parameters is avoided, and the reliability of the early warning is improved. BRIEF DESCRIPTION OF DRAWINGS

[0053] Figure 1 It is a principle block diagram of the system of the application;

[0054] Figure 2 It is a step flow chart of the method of the application;

[0055] Figure 3 It is a pretreatment strategy flow chart of the application;

[0056] Figure 4 It is a structural schematic diagram of the electronic device of the application. DETAILED DESCRIPTION

[0057] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative efforts belong to the protection scope of the application.

[0058] Embodiment 1, please refer to Figure 1 As shown in the figure, the application provides a deep well rock mass instability failure time accurate prediction and early warning system, which comprises a test simulation module, a data processing module, a model construction module and a rock mass early warning module;

[0059] The test simulation module is used for collecting a plurality of deep well rock samples for uniaxial compression experiments, and collecting acoustic emission signals in real time based on acoustic emission sensors to obtain original acoustic emission signal data;

[0060] The test simulation module is configured with a test simulation strategy, and the test simulation strategy comprises: collecting rock samples from deep wells, and processing the rock samples into standard test pieces according to experimental specifications, denoted as deep well rock samples, and denoting any one of the deep well rock samples as a first rock sample; processing different natural rocks into standard test pieces with the same geometric size and surface morphology, which can eliminate the test deviation caused by size and shape, ensure uniform stress during loading, and avoid eccentricity, and the morphology of the standard test piece is generally a cylinder;

[0061] The acoustic emission sensor is closely attached to the surface of the first rock sample, and a uniaxial compression experiment is performed on the first rock sample until the first rock sample is destroyed; the close attachment is to maximize the signal transmission efficiency, ensure the authenticity of the signal characteristics, reduce energy attenuation, and at the same time reduce environmental interference and improve the signal-to-noise ratio;

[0062] The stress of the first rock sample is monitored and collected in real time during the uniaxial compression experiment, and the collection time is recorded as the stress change data of the first rock sample, and at the same time the acoustic emission sensor is used to monitor and record the acoustic emission signal of the first rock sample, which is recorded as the original acoustic emission data of the first rock sample;

[0063] The maximum stress in the stress change data of the first rock sample is marked as the limit stress of the first rock sample, and the collection time corresponding to the limit stress is recorded as the rock sample destruction point of the first rock sample; the limit stress is used as a recognition tag for subsequent rock mass destruction;

[0064] The stress change data and the original acoustic emission data of all deep well rock samples are repeatedly obtained;

[0065] In the specific implementation process, the uniaxial compression test of the rock sample is one of the most basic and most commonly used tests in the field of rock mechanics. By applying a slowly increasing pressure in the axial direction of the rock sample, the entire process from stress to deformation to final destruction of the rock sample is observed to obtain the basic mechanical properties and failure rules of the rock. The core is to simulate the mechanical response of the rock under uniaxial stress, which provides key parameters for geotechnical engineering design, geological disaster warning, etc. According to the actual application scene, other failure mechanics experiments can also be selected.

[0066] The data processing module includes a preprocessing unit and a classification and integration unit. The preprocessing unit is used for adaptive preprocessing of the original acoustic emission signal data, and the classification and integration unit is used for data classification and integration to obtain rock sample acoustic emission parameter data.

[0067] The preprocessing unit is configured with a preprocessing strategy, which includes: referring to Figure 3 For the acoustic emission signal of the first rock sample, denoted as the first acoustic emission signal, the average value of all data points of the first acoustic emission signal is calculated, and each data point in the first acoustic emission signal is subtracted from the average value of all data points. After completion, the second acoustic emission signal is obtained. The purpose is to remove the direct current component of the first acoustic emission signal, i.e. the constant offset in the signal. In the deep well environment, the acoustic emission sensor may be affected by temperature drift and cable vibration, resulting in baseline drift or transient large amplitude noise. These interferences are unrelated to microcracks, and removing them can avoid interference with subsequent decomposition;

[0068] The Morlet wavelet with a center frequency of k1 is selected to decompose the second acoustic emission signal into k2 scales, the modulus value of the complex wavelet coefficient of each scale is taken, and the integral of the square of the modulus value is calculated to obtain the energy corresponding to the scale, wherein k1 is a set frequency, and k2 is a set number; in the embodiment, k1 = 200 kHz, because the main frequency of the micro-crack acoustic emission of the deep well rock sample is mostly in 100-400 kHz, k2 = 3, k2 is generally 3-5, and k1 and k2 can be flexibly set according to the actual application scene;

[0069] The proportion of the energy of each scale in the total energy of all scales is calculated and recorded as an energy proportion, the k3 scales with the largest energy proportions are reserved and recorded as effective scales, k3 is a set number, and k3 < k2; in the embodiment, k3 = 2; the frequency band in which the effective signal is most concentrated is focused, and the low-energy noise scale is automatically removed.

[0070] For any one effective scale, recorded as a first scale, the modulus value of the complex wavelet coefficient of the first scale is taken to obtain an envelope sequence corresponding to the first scale, recorded as a first envelope sequence;

[0071] The mean value of the first envelope sequence is calculated and recorded as AU, k4*AU is recorded as a peak threshold value, the points in the first envelope sequence that are greater than the peak threshold value are recorded as effective peak points, the number of the effective peak points and the total number of points in the first envelope sequence are counted and recorded as BS and AS0 respectively, and the sparse factor XS corresponding to the first scale is calculated, wherein XS = BS / AS0; the sparse factor represents the proportion of the effective peak points in the total signal length, and quantifies the intensity of the transient pulse in the signal; when XS is small, it indicates that the signal is mainly continuous noise, and the transient pulse is few and sparse; when XS is large, it indicates that the signal is mainly transient pulse, and the micro-crack activity is frequent and the pulse is dense;

[0072] XS provides a quantitative basis for subsequent threshold adjustment: avoids the problem that the traditional fixed threshold value excessively denoises the sparse signal and insufficiently denoises the dense signal, and makes the denoising strategy adapt to the dynamic changes of the signal.

[0073] The standard deviation of the first envelope sequence is obtained and recorded as BU, and the suppression threshold value YZ is calculated, wherein YZ = R0*BU, R0 = k4*-10*(k4-1)*XS, wherein k4 is a set maximum multiple; in the embodiment, k4 = 3, which can be flexibly set; the envelope amplitude of the continuous noise is generally low and stable, and the amplitude of the weak pulse may be close to the noise, and by dynamically adjusting the threshold value through XS, a balance between noise suppression and pulse preservation can be found;

[0074] For the points in the first envelope sequence less than the suppression threshold, the corresponding wavelet coefficients are set to zero; for the points in the first envelope sequence greater than or equal to the suppression threshold, the corresponding wavelet coefficients are retained; the modified envelope sequence corresponding to the first scale is completed; that is, the pulse components exceeding the threshold are retained, and the noise below the threshold is suppressed;

[0075] Repeat the process of obtaining the modified envelope sequence corresponding to all effective scales, for each modified envelope sequence, use the corresponding Morlet wavelet for inverse mapping, and align all inverse mapping effective signal components on the time axis, superimpose and synthesize to obtain the denoised emission signal of the first acoustic emission signal, denoted as the first denoised signal;

[0076] Repeat the process of obtaining the denoised emission signal corresponding to the acoustic emission signal of all deep well rock samples.

[0077] The classification integration unit is configured with a classification integration strategy, which includes: for the first denoised signal of the first rock sample, setting the first time interval as t1; according to t1, and taking the rock sample failure point as the endpoint, selecting multiple time points from the first denoised signal, denoted as parameter sampling points, and any one parameter sampling point is denoted as a first sampling point; in this embodiment, t1 = 0.1s, a smaller t1 provides more fine-grained timing information, but increases the sample dimension and computational complexity; a larger t1 is more coarse but more robust, and can be flexibly set;

[0078] Obtain the stress size corresponding to the first sampling point, denoted as parameter 1, obtain the amplitude and frequency of the first sampling point, denoted as parameters 2 and 3 in order, which reflect the intensity and fracture type of a single event;

[0079] Taking the first sampling point as the endpoint of the first denoised signal, obtain the corresponding AE event number, ringing count, rise time, energy entropy, RA value, AE event rate and ringing event rate, denoted as parameters 4-10 in order; for depicting event density, waveform shape and energy distribution, these usually change significantly before failure; parameters 1-10 are denoted as parameter information of the first sampling point; for the selection of parameters, changes or additions can be made according to the actual application scenario;

[0080] Repeat the process of obtaining the parameter information of all parameter sampling points, and arrange them in the order of the time corresponding to the parameter sampling points, denoted as rock acoustic emission parameter data;

[0081] In the specific implementation process, the oscillation characteristics of the Morlet wavelet match the time-frequency characteristics of the acoustic emission transient pulse, which can effectively capture the start and end time and amplitude change of the pulse; the envelope of the transient pulse shows a steep peak in a short time, while the envelope of the continuous noise shows a smooth fluctuation over a long period of time, and after decomposition, the two can be distinguished intuitively in the envelope domain.

[0082] The model construction module is configured to construct an original damage prediction model, and train and verify and optimize the model based on rock sample acoustic emission parameter data to obtain a rock damage prediction model.

[0083] The model construction module is configured with a model construction strategy, which includes: constructing an original damage prediction model based on a long short-term memory network, the original damage prediction model including an input layer, a hidden layer and an output layer; the long short-term memory network is specially designed to learn long-term and short-term dependencies of sequence data, such as the cumulative effect of crack from initiation to convergence, which is more suitable for predicting damage time than traditional static regression, which is strongly related to historical behavior.

[0084] For the parameters 1-10 of the first denoised signal, the normalization processing is performed according to the parameter types respectively. Different parameters have different dimensions and large differences in value ranges. Normalization can avoid errors caused by too large numerical differences and too large changes. After completion, it is recorded as standard parameters 1-10. Any one of the standard parameters 2-9 is recorded as a standard parameter i. The standard parameter 1 and the standard parameter i of the same parameter sampling point are combined into a corresponding single-parameter feature vector, recorded as DM={standard parameter 1, standard parameter i}. And all single-parameter feature vectors are arranged in corresponding time order, recorded as single-parameter feature data. The independent prediction ability of each standard parameter is evaluated one by one. By pairing stress with a single standard parameter, the long short-term memory network is trained and the error is calculated, which can quantify the prediction ability brought by each standard parameter alone.

[0085] The single-parameter feature data is divided into a single-parameter training set and a single-parameter test set according to a ratio of 7:3. The output result of the output layer is set as the stress size. The single-parameter training set is used to train the original damage prediction model. After completion, the single-parameter test set is used to test the model, and the root mean square error of the model is calculated, recorded as the RMSE corresponding to the standard parameter i. RMSE is used as a single standard parameter prediction ability measure. The smaller it is, the stronger the correlation between the standard parameter and the stress. The division ratio of the single-parameter training set and the single-parameter test set can be flexibly set.

[0086] The RMSE corresponding to all standard parameters is repeatedly obtained and arranged in order from small to large, recorded as an error sequence. The corresponding standard parameters are reordered according to the order of the error sequence, recorded as important parameters 1-9 in turn.

[0087] The combination of the standard parameter 1 and the important parameters 1-j is denoted as parameter combination j, wherein j ∈ [1, 9], all parameter combinations are obtained; that is, gradually accumulated from a single important parameter to multiple important parameters; for example, parameter combination 1 is standard parameter 1 + important parameter 1, parameter combination 2 is standard parameter 1 + important parameter 1 + important parameter 2, and parameter combination 4 is standard parameter 1 + important parameter 1 + important parameter 2 + important parameter 3 + important parameter 4; through gradual accumulation, it can help to identify which characteristic parameters truly bring gains, and avoid blindly adding redundant and noise features;

[0088] For parameter combination j, the standard parameter 1 and the important parameters 1-j of the same parameter sampling point are combined into a corresponding parameter feature vector, denoted as XM = {standard parameter 1, important parameter 1, …, important parameter j}; and all parameter feature vectors are arranged in corresponding time order, denoted as the jth combination feature data;

[0089] The jth combination feature data is divided into a combination training set and a combination test set according to a ratio of 7:3, the original damage prediction model is trained using the combination training set, and after training, the jth prediction model is obtained, the jth prediction model is tested using the combination test set, and the determination coefficient, the average determination error, the root mean square error, and the average absolute percentage error of the model are calculated, denoted as the evaluation parameters of parameter combination j; starting from the most important feature, the less important features are gradually added, and the marginal improvement of each step on the model performance is observed, which facilitates finding effective and non-redundant parameter combinations;

[0090] The evaluation parameters of all parameter combinations are repeatedly obtained, and the average determination error, the root mean square error, and the average absolute percentage error of all parameter combinations are respectively arranged in descending order, denoted as error sequence 1, error sequence 2, and error sequence 3, and the determination coefficients of all parameter combinations are arranged in ascending order, denoted as error sequence 4; the smaller the average determination error, the root mean square error, and the average absolute percentage error, the better the prediction performance of the model, and the larger the determination coefficient, the better the prediction performance of the model;

[0091] The position sequence numbers of the evaluation parameters of parameter combination j in the corresponding error sequences 1-4 are respectively obtained and accumulated, denoted as the overall score of parameter combination j; the position numbers of the four different sorting results are added up, which is used to quantify the prediction performance of different parameter combinations, form a selection standard, and avoid selecting a combination with extremely excellent in one indicator but poor in other indicators;

[0092] The overall score of all parameter combinations is repeatedly obtained, and the parameter combination with the maximum overall score is recorded as the predicted parameter combination of the first rock sample, the prediction model corresponding to the predicted parameter combination is marked as the rock failure prediction model of the first rock sample, and the average absolute percentage error corresponding to the rock failure prediction model is recorded as PE; PE is used to modify the output results of the model subsequently to reduce the risk of false negatives in early warning and improve the timeliness of early warning;

[0093] The predicted parameter combination, the rock failure prediction model and PE of all deep rock samples are repeatedly obtained;

[0094] In the specific implementation process, the input layer is used to receive the original sequence data and convert it into a format that can be processed by the long short-term memory network, the number of units and the number of layers of the hidden layer can be flexibly adjusted according to the actual application scenario to match the input complexity and sample size, and the output layer is used to convert the time sequence features output by the hidden layer into results that meet the task objectives.

[0095] The rock mass early warning module is used to collect the acoustic emission signals of the deep well rock mass, and to predict the real-time instability and failure time of the deep well rock mass based on the rock failure prediction model;

[0096] The rock mass early warning module includes a rock mass early warning strategy, which includes: collecting the stress size and acoustic emission signals of the deep well rock mass, respectively recorded as the current stress size and the current acoustic emission signal, and performing adaptive preprocessing on the current acoustic emission signal to obtain the current denoising signal; The time of stress collection and acoustic emission signal collection should be synchronized for subsequent sampling based on time point extraction parameters and combination;

[0097] According to the first time interval, a plurality of time points are uniformly selected from the current denoising signal, and parameters corresponding to the predicted parameter combination are obtained at the time points, recorded as current parameters, all current parameters and current stress sizes are normalized according to the parameter types, and combined into corresponding parameter feature vectors, and then the corresponding rock failure prediction model is input according to the type of the deep well rock mass to obtain stress prediction data at future time; That is, the same parameter set as the predicted parameter combination is used; Normalization processing keeps the numerical scale consistent to match the corresponding model, avoiding prediction errors caused by scale drift.

[0098] The stress size of each future time point predicted is multiplied by (1+PE) to obtain the corrected stress size of each future time point, and the corrected stress size is retrieved in time sequence, the first corrected stress size greater than or equal to the corresponding limit stress is recorded as the instability and failure point, the prediction time of the deep well rock mass instability and failure is obtained according to the time point corresponding to the instability and failure point, and the prediction and early warning are carried out; The stress size is multiplied by (1+PE) to make up for the average percentage underestimation in the history of the model, aiming to reduce the risk of false negatives and improve the timeliness of prediction and early warning;

[0099] In the implementation process, the PE is the average absolute percentage error of the output result of the model obtained by statistical analysis when the model is tested, and the average absolute percentage error of the corresponding model of the model is used instead of the average absolute percentage error of other models.

[0100] Embodiment 2, please refer to Figure 2 As shown in the figure, the application provides a deep well rock mass instability failure time accurate prediction and early warning method, including the following steps:

[0101] Step S1, collect a plurality of deep well rock samples for uniaxial compression experiment, and collect acoustic emission signals in real time based on acoustic emission sensors to obtain original acoustic emission signal data; step S1 includes the following sub-steps:

[0102] Step S101, collect rock samples from deep wells, and process the rock samples into standard test pieces according to experimental specifications, denoted as deep well rock samples, and any one of the deep well rock samples is denoted as a first rock sample;

[0103] Step S102, tightly adhere the acoustic emission sensor to the surface of the first rock sample, and perform uniaxial compression experiment on the first rock sample until the first rock sample fails.

[0104] Step S103, monitor and collect the stress of the first rock sample in real time during the uniaxial compression experiment, record the collection time, denoted as the stress change data of the first rock sample, and simultaneously monitor and record the acoustic emission signals of the first rock sample using the acoustic emission sensor, denoted as the original acoustic emission data of the first rock sample;

[0105] Step S104, mark the maximum stress in the stress change data of the first rock sample as the limit stress of the first rock sample, and record the collection time corresponding to the limit stress as the rock sample failure point of the first rock sample;

[0106] Step S105, repeat the acquisition of the stress change data and the original acoustic emission data of all deep well rock samples.

[0107] Step S2, adaptively pre-process the original acoustic emission signal data, and classify and integrate the data to obtain rock sample acoustic emission parameter data; step S2 includes the following sub-steps:

[0108] Step S201, for the acoustic emission signal of the first rock sample, denoted as the first acoustic emission signal, calculate the average value of all data points of the first acoustic emission signal, and subtract the average value of all data points from each data point in the first acoustic emission signal, to obtain a second acoustic emission signal after completion;

[0109] Step S202, selecting a Morlet wavelet with a center frequency of k1 to decompose the second acoustic emission signal into k2 scales, taking the modulus value of the complex wavelet coefficient of each scale, and then calculating the integral of the square of the modulus value to obtain the energy of the corresponding scale, wherein k1 is a set frequency, and k2 is a set number;

[0110] Step S203, calculating the proportion of the energy of each scale in the total energy of all scales, denoted as the energy proportion, retaining the k3 scales with the largest energy proportion, denoted as the effective scales, and k3 is a set number, k3 < k2.

[0111] Step S204, for any one effective scale, denoted as the first scale, taking the modulus value of the complex wavelet coefficient of the first scale to obtain the envelope sequence corresponding to the first scale, denoted as the first envelope sequence;

[0112] Step S205, calculating the mean value of the first envelope sequence, denoted as AU, taking k4*AU as the peak threshold value, taking the points in the first envelope sequence that exceed the peak threshold value as the effective peak points, and counting the number of effective peak points and the total number of points in the first envelope sequence, denoted as BS and AS0 respectively, and calculating the sparsity factor XS corresponding to the first scale, wherein XS = BS / AS0;

[0113] Step S206, obtaining the standard deviation of the first envelope sequence, denoted as BU, and calculating the suppression threshold YZ, wherein YZ = R0*BU, R0 = k4*-10*(k4-1)*XS, and k4 is a set maximum multiple;

[0114] Step S207, for the points in the first envelope sequence that are less than the suppression threshold, setting the corresponding wavelet coefficient to zero; for the points in the first envelope sequence that are greater than or equal to the suppression threshold, retaining the corresponding wavelet coefficient; and obtaining the modified envelope sequence corresponding to the first scale after completion;

[0115] Step S208, repeating the process of obtaining the modified envelope sequence corresponding to all effective scales, for each modified envelope sequence, performing inverse mapping using the corresponding Morlet wavelet, and aligning all inverse mapping obtained effective signal components on the time axis to synthesize the denoised emission signal of the first acoustic emission signal, denoted as the first denoised signal;

[0116] Step S209, repeating the process of obtaining the denoised emission signal corresponding to the acoustic emission signal of all deep well rock samples.

[0117] Step S210, for the first denoised signal of the first rock sample, setting the first time interval as t1; selecting multiple time points from the first denoised signal according to t1 and taking the rock sample failure point as the endpoint, denoted as the parameter sampling point, and denoting any one parameter sampling point as the first sampling point;

[0118] Step S211, the stress size corresponding to the first sampling point is obtained, denoted as parameter 1, the amplitude and frequency of the first sampling point are obtained, denoted as parameters 2 and 3 in order;

[0119] Step S212, taking the first sampling point as the end point of the first denoising signal, the corresponding AE event number, ringing count, rise time, energy entropy, RA value, AE event rate and ringing event rate are obtained, denoted as parameters 4-10 in order; parameters 1-10 are denoted as the parameter information of the first sampling point;

[0120] Step S213, repeat the parameter information of all parameter sampling points, and arrange them in the order of time corresponding to the parameter sampling points, denoted as rock sample acoustic emission parameter data.

[0121] Step S3, construct an original damage prediction model, and train and verify the model based on the rock sample acoustic emission parameter data to obtain a rock damage prediction model; step S3 includes the following sub-steps:

[0122] Step S301, based on the long short-term memory network, an original damage prediction model is constructed, which includes an input layer, a hidden layer and an output layer;

[0123] Step S302, for parameters 1-10 of the first denoising signal, the parameters are normalized according to the parameter type, and after completion, the standard parameters 1-10 are denoted as standard parameters 2-9, any one of the standard parameters is denoted as a standard parameter i; the standard parameter 1 and the standard parameter i of the same parameter sampling point are combined into a corresponding single parameter feature vector, denoted as DM={standard parameter 1, standard parameter i}; and all single parameter feature vectors are arranged in the order of time corresponding to the single parameter feature vectors, denoted as single parameter feature data;

[0124] Step S303, the single parameter feature data is divided into a single parameter training set and a single parameter test set according to a ratio of 7:3, the output result of the output layer is set as the stress size, the original damage prediction model is trained by using the single parameter training set, after completion, the model is tested by using the single parameter test set, and the root mean square error of the model is calculated, denoted as the RMSE corresponding to the standard parameter i;

[0125] Step S304, repeat the RMSE corresponding to all standard parameters, and arrange them in order from small to large, denoted as an error sequence, reorder the corresponding standard parameters according to the order of the error sequence, denoted as important parameters 1-9.

[0126] Step S305, the combination of the standard parameter 1 and the important parameter 1-j is denoted as a parameter combination j, where j∈[1, 9], all parameter combinations are obtained;

[0127] Step S306, for parameter combination j, the standard parameter 1 and the important parameter 1-j of the same parameter sampling point are combined into a corresponding parameter feature vector, denoted as XM={standard parameter 1, important parameter 1, …, important parameter j}; and all parameter feature vectors are arranged in corresponding time sequence, denoted as the jth combination feature data;

[0128] Step S307, the jth combination feature data is divided into a combination training set and a combination test set according to a 7:3 ratio, the original failure prediction model is trained by using the combination training set, and after the training is completed, the jth prediction model is obtained, the jth prediction model is tested by using the combination test set, and the determination coefficient, the average determination error, the root mean square error and the average absolute percentage error of the model are calculated, denoted as the evaluation parameters of parameter combination j.

[0129] Step S308, the evaluation parameters of all parameter combinations are repeatedly obtained, and the average determination error, the root mean square error and the average absolute percentage error of all parameter combinations are arranged in descending order, respectively, denoted as error sequence 1, error sequence 2 and error sequence 3, and the determination coefficients of all parameter combinations are arranged in ascending order, denoted as error sequence 4;

[0130] Step S309, the position sequence number of the evaluation parameters of parameter combination j in the corresponding error sequence 1-4 is obtained respectively, and is accumulated, denoted as the overall score of parameter combination j;

[0131] Step S310, the overall scores of all parameter combinations are repeatedly obtained, and the parameter combination with the maximum overall score is denoted as the prediction parameter combination of the first rock sample, the prediction model corresponding to the prediction parameter combination is marked as the rock failure prediction model of the first rock sample, and the average absolute percentage error of the rock failure prediction model is denoted as PE;

[0132] Step S311, the prediction parameter combination, the rock failure prediction model and PE of all deep rock samples are repeatedly obtained.

[0133] Step S4, the acoustic emission signals of the deep well rock mass are collected, and the instability failure time of the deep well rock mass is predicted in real time based on the rock failure prediction model; step S4 includes the following substeps:

[0134] Step S401, the stress size and the acoustic emission signal of the deep well rock mass are collected, and are denoted as the current stress size and the current acoustic emission signal, respectively, the current acoustic emission signal is adaptively preprocessed to obtain a current denoising signal;

[0135] Step S402, according to the first time interval, a plurality of time points are uniformly selected from the current de-noising signal, and parameters corresponding to the prediction parameter combination are obtained at the time points, denoted as current parameters, all current parameters and the current stress size are normalized according to the parameter type, and combined into a corresponding parameter feature vector, and then the corresponding rock failure prediction model is input according to the type of deep well rock mass, and the stress prediction data at the future time is obtained.

[0136] Step S403, the stress size of each time point in the future is multiplied by (1+PE) to obtain the corrected stress size of each time point in the future, and the corrected stress size is retrieved in time sequence, the first corrected stress size greater than or equal to the corresponding limit stress is recorded as the instability failure point, the prediction time of the deep well rock mass instability failure is obtained according to the time point corresponding to the instability failure point, and the prediction and early warning are carried out.

[0137] Embodiment 3, please refer to Figure 4 As shown in the figure, Figure 4 An electronic device can include a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory complete mutual communication through the communication bus. The memory stores computer readable instructions, and the processor can call the instructions in the memory, when the computer readable instructions are executed by the processor, the steps in a kind of deep well rock mass instability failure time accurate prediction and early warning method are run, to realize the following functions: a plurality of deep well rock samples are collected to carry out uniaxial compression experiment, and based on acoustic emission sensor, real-time acoustic emission signal is collected, and original acoustic emission signal data is obtained;Original acoustic emission signal data is adaptively preprocessed, and data classification integration is carried out, to obtain rock sample acoustic emission parameter data;Original failure prediction model is constructed, and model training and verification optimization are carried out based on rock sample acoustic emission parameter data, to obtain rock failure prediction model;Acoustic emission signal of deep well rock mass is collected, and the instability failure time of deep well rock mass is predicted in real time based on rock failure prediction model.

[0138] In addition, the logic instructions in the memory described above can be implemented in the form of a software functional unit and sold or used as an independent product, and can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part of the prior art that contributes to the present application or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the method described in various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0139] In embodiment 4, the present application also provides a computer readable storage medium, and the present application provides a storage medium having a computer program stored thereon. The computer program is executed by a processor to run the steps of the above deep well rock mass instability failure time accurate prediction and early warning method to realize the following functions: collecting a plurality of deep well rock samples for uniaxial compression experiment, and collecting acoustic emission signals in real time based on acoustic emission sensors to obtain original acoustic emission signal data; adaptively pre-processing the original acoustic emission signal data, and classifying and integrating the data to obtain rock sample acoustic emission parameter data; constructing an original failure prediction model, and training and verifying and optimizing the model based on the rock sample acoustic emission parameter data to obtain a rock failure prediction model; collecting acoustic emission signals of the deep well rock mass, and predicting the instability failure time of the deep well rock mass in real time based on the rock failure prediction model.

[0140] Through the description of the above embodiments, the embodiments of the present application can be provided as a method, a system or a computer program product. Based on such understanding, the above technical solutions essentially or the part of the prior art that contributes to the present application can be embodied in the form of a software product. The computer software product can be stored in a computer readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the method described in each embodiment or some part of the embodiment.

[0141] In the embodiments of the present application, it should be understood that the disclosed system or method can be implemented in other ways. The embodiments described above are only illustrative, for example, the division of modules or units is only a logical function division, and other division manners can be used in actual implementation, for example, a plurality of modules or units can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed each other can be through some communication interface, the indirect coupling or communication connection between the system, the module and the unit can be electrical, mechanical or other forms.

[0142] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A deep well rock mass instability failure time accurate prediction and early warning system, characterized in that, The test simulation module, the data processing module, the model construction module, and the rock mass early warning module are included. The test simulation module is used for collecting various deep well rock samples for uniaxial compression experiments and collecting acoustic emission signals in real time based on acoustic emission sensors to obtain original acoustic emission signal data. The data processing module includes a preprocessing unit and a classification integration unit. The model construction module is used for constructing an original failure prediction model and training and verifying and optimizing the model based on rock sample acoustic emission parameter data to obtain a rock failure prediction model. The rock mass early warning module is used for collecting acoustic emission signals of deep well rock mass and predicting the instability and failure time of deep well rock mass in real time based on the rock failure prediction model. The model construction module is configured with a model construction strategy, which includes: Based on the long short-term memory network, an original failure prediction model is constructed, which includes an input layer, a hidden layer, and an output layer. For parameters 1-10 of the first denoised signal, the parameters are normalized according to the parameter types, and after the normalization, the parameters are recorded as standard parameters 1-10. Any one of the standard parameters 2-9 is recorded as a standard parameter i. The standard parameter 1 and the standard parameter i of the same parameter sampling point are combined into a corresponding single-parameter feature vector, which is recorded as DM={standard parameter 1, standard parameter i}. All single-parameter feature vectors are arranged in corresponding time order, which is recorded as single-parameter feature data. The single-parameter feature data is divided into a single-parameter training set and a single-parameter test set according to a ratio of 7:

3. The output result of the output layer is set as the stress size. The original failure prediction model is trained using the single-parameter training set. After the training, the model is tested using the single-parameter test set. The root mean square error of the model is calculated, which is recorded as the RMSE corresponding to the standard parameter i. The RMSEs corresponding to all standard parameters are repeatedly obtained and arranged in ascending order, which is recorded as an error sequence. The corresponding standard parameters are reordered according to the error sequence, which are sequentially recorded as important parameters 1-9. The combination of the standard parameter 1 and the important parameter 1-j is recorded as a parameter combination j, where j∈[1, 9]. All parameter combinations are obtained. For the parameter combination j, the standard parameter 1 and the important parameter 1-j of the same parameter sampling point are combined into a corresponding parameter feature vector, which is recorded as XM={standard parameter 1, important parameter 1, …, important parameter j}. All parameter feature vectors are arranged in corresponding time order, which is recorded as the jth combination feature data. The jth combination feature data is divided into a combination training set and a combination test set according to a ratio of 7:

3. The original failure prediction model is trained using the combination training set. After the training, the jth prediction model is obtained. The jth prediction model is tested using the combination test set. The determination coefficient, the average determination error, the root mean square error, and the average absolute percentage error of the model are calculated, which are recorded as the evaluation parameters of the parameter combination j. The evaluation parameters of all parameter combinations are repeatedly obtained, and the average determination error, the root mean square error, and the average absolute percentage error of all parameter combinations are respectively arranged in descending order, and are sequentially recorded as error sequence 1, error sequence 2, and error sequence 3; the determination coefficients of all parameter combinations are arranged in ascending order, and are recorded as error sequence 4; The position sequence numbers of the evaluation parameters of the parameter combination j in the corresponding error sequences 1-4 are respectively obtained, and are accumulated, and are recorded as the overall score of the parameter combination j; The overall scores of all parameter combinations are repeatedly obtained, and the parameter combination with the maximum overall score is recorded as the prediction parameter combination of the first rock sample, the prediction model corresponding to the prediction parameter combination is marked as the rock failure prediction model of the first rock sample, and the average absolute percentage error corresponding to the rock failure prediction model is recorded as PE; The prediction parameter combination, the rock failure prediction model, and PE of each deep well rock sample are repeatedly obtained.

2. The system of claim 1, wherein the system is characterized by, The test simulation module is configured with a test simulation strategy, and the test simulation strategy comprises: Rock samples are collected from a deep well, and the rock samples are processed into standard test pieces according to experimental specifications, and are recorded as deep well rock samples; any one of the deep well rock samples is recorded as a first rock sample; The acoustic emission sensor is tightly attached to the surface of the first rock sample, and a uniaxial compression experiment is performed on the first rock sample until the first rock sample is damaged.

3. The system of claim 2, wherein the system is characterized by, The test simulation strategy further comprises: The stress of the first rock sample is monitored and collected in real time during the uniaxial compression experiment, and the collection time is recorded as the stress change data of the first rock sample, and the acoustic emission sensor is used to monitor and record the acoustic emission signal of the first rock sample at the same time, and is recorded as the original acoustic emission data of the first rock sample; The maximum stress in the stress change data of the first rock sample is marked as the ultimate stress of the first rock sample, and the collection time corresponding to the ultimate stress is recorded as the rock sample failure point of the first rock sample; The stress change data and the original acoustic emission data of all deep well rock samples are repeatedly obtained.

4. The system of claim 3, wherein the system is characterized by, The preprocessing unit is configured with a preprocessing strategy, and the preprocessing strategy comprises: For the acoustic emission signal of the first rock sample, recorded as the first acoustic emission signal, the average value of all data points of the first acoustic emission signal is calculated, and each data point in the first acoustic emission signal is subtracted by the average value of all data points, and after completion, the second acoustic emission signal is obtained; The Morlet wavelet with a center frequency of k1 is selected to decompose the second acoustic emission signal into k2 scales, the modulus value of the complex wavelet coefficient of each scale is taken, and the integral of the square of the modulus value is calculated to obtain the energy of the corresponding scale, wherein k1 is the set frequency, and k2 is the set number; The proportion of the energy of each scale in the total energy of all scales is calculated, recorded as the energy proportion, and the k3 scales with the largest energy proportion are retained, recorded as the effective scales, k3 is the set number, and k3 < k2.

5. The system of claim 4, wherein the system is characterized by, The preprocessing strategy further comprises: For any one of the effective scales, recorded as the first scale, the modulus value of the complex wavelet coefficient of the first scale is taken to obtain the envelope sequence corresponding to the first scale, recorded as the first envelope sequence; Calculate the mean of the first envelope sequence, denoted as AU, and take k4*AU as the peak threshold value. Points in the first envelope sequence that exceed the peak threshold value are denoted as valid peak points. Count the number of valid peak points and the total number of points in the first envelope sequence, denoted as BS and AS0 respectively. Calculate the sparsity factor XS corresponding to the first scale, where XS = BS / AS0; Obtain the standard deviation of the first envelope sequence, denoted as BU, and calculate the suppression threshold value YZ, where YZ = R0*BU, and R0 = k4*-10*(k4-1)*XS, where k4 is the maximum multiple set; For points in the first envelope sequence that are less than the suppression threshold value, set the corresponding wavelet coefficients to zero. For points in the first envelope sequence that are greater than or equal to the suppression threshold value, retain the corresponding wavelet coefficients. Complete the modified envelope sequence corresponding to the first scale; Repeat the process of obtaining the modified envelope sequence corresponding to all effective scales. For each modified envelope sequence, perform inverse mapping using the corresponding Morlet wavelet, and align all inverse mapping results along the time axis. Superimpose and combine to obtain the denoised emission signal of the first acoustic emission signal, denoted as the first denoised signal. Repeat the process of obtaining the denoised emission signal corresponding to the acoustic emission signal of all deep well rock samples.

6. The system of claim 5, wherein the system is characterized by, The classification and integration unit is configured with a classification and integration strategy, which includes: For the first denoised signal of the first rock sample, set the first time interval as t1. According to t1, and with the rock sample failure point as the endpoint, select multiple time points from the first denoised signal, denoted as parameter sampling points. Denote any one parameter sampling point as the first sampling point. Obtain the stress size corresponding to the first sampling point, denoted as parameter 1. Obtain the amplitude and frequency of the first sampling point, denoted as parameters 2 and 3 respectively. Take the first sampling point as the endpoint of the first denoised signal, and obtain the corresponding AE event number, ringing count, rise time, energy entropy, RA value, AE event rate, and ringing event rate, denoted as parameters 4-10 respectively. Record parameters 1-10 as the parameter information of the first sampling point. Repeat the process of obtaining the parameter information of all parameter sampling points and arrange them in the order of their corresponding time, denoted as rock acoustic emission parameter data.

7. The system according to claim 6, wherein the system is characterized by, The rock mass early warning module includes a rock mass early warning strategy, which includes: Collect the stress size and acoustic emission signal of the deep well rock mass, denoted as the current stress size and the current acoustic emission signal respectively. Perform adaptive preprocessing on the current acoustic emission signal to obtain the current denoised signal; According to the first time interval, uniformly select multiple time points from the current denoised signal, and obtain the corresponding parameters of the prediction parameter combination at the time points, denoted as the current parameters. Normalize all current parameters and the current stress size according to the parameter type, and combine them into the corresponding parameter feature vector. Then, input the corresponding rock failure prediction model according to the type of the deep well rock mass to obtain the stress prediction data at the future time. The predicted stress size of each time point in the future is multiplied by (1+PE) to obtain the corrected stress size of each time point in the future, the corrected stress sizes are sequentially retrieved according to time sequence, the first corrected stress size greater than or equal to the corresponding limit stress is recorded as an instability failure point, the predicted time of instability failure of the deep well rock mass is obtained according to the time point corresponding to the instability failure point, and a prediction and early warning is performed.

Images