Coal rock stability monitoring method and system based on acoustic signal dynamics characteristics

Abstract

The embodiment of the application provides a coal rock stability monitoring method and system based on acoustic signal dynamics, and relates to the technical field of coal rock stability monitoring. After obtaining the acoustic signal, the method segments the acoustic signal into multiple data segments by using a sliding time window. Then, nonlinear features are extracted from the data segments to obtain a comprehensive feature vector, and then the monitoring and early warning state is triggered according to the comprehensive feature vector. The comprehensive feature vector can be constructed by fusing multiple nonlinear features, and the nonlinear features are system evolution features extracted from the acoustic signal based on the nonlinear dynamics principle. The method quantifies the similarity between the current system state and the historical instability state by introducing a dynamic time warping distance, and performs a mutation detection on the real-time feature sequence corresponding to the comprehensive feature, forming a hybrid early warning monitoring combining mechanism and data driving, and improving the timeliness and accuracy of coal rock stability monitoring and early warning.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of safety monitoring and early warning of mining engineering, and in particular to a coal rock stability monitoring method and system based on acoustic signal dynamics. BACKGROUND

[0002] In mining engineering, a coal rock system refers to a geological structure composed of a coal seam and its upper and lower surrounding rocks, such as a roof and a floor. The coal seam in the coal rock system will accumulate a large amount of energy under the clamping of the upper and lower hard rock layers, and this structure exhibits unique mechanical behavior under the action of mine pressure. Since the instability of the coal rock system is the main cause of disasters such as roof collapse, rock burst, and rock pressure in the mining area, in order to prevent disasters from occurring in advance, it is necessary to monitor the stability of the coal rock system.

[0003] For example, coal rock stability monitoring can be achieved through microseismic monitoring technology. Microseismic monitoring technology can infer stress concentration areas by picking up microseismic events generated by rock mass failure. Since microseismic monitoring technology is based on microseismic events generated by rock mass failure, there are limitations such as large event positioning error and difficulty in identifying precursors of progressive failure. Therefore, stress and strain monitoring can also be performed by directly measuring the deformation of the rock mass through the embedding of deformation sensors.

[0004] However, the above-mentioned coal rock stability monitoring process is difficult to fully characterize the nonlinear characteristics and chaotic characteristics in the coal rock failure process, and in the monitoring process, the monitoring criteria are constructed depending on empirical thresholds, which need to be determined through long-term statistical results and have regional limitations, resulting in low warning timeliness and accuracy of the above-mentioned coal rock stability monitoring process. SUMMARY

[0005] Therefore, the embodiments of the present application provide a coal rock stability monitoring method and system based on acoustic signal dynamics to solve the problem of low warning timeliness and accuracy of the coal rock stability monitoring process.

[0006] According to a first aspect of the present application, a coal rock stability monitoring method based on acoustic signal dynamics is provided, which comprises:

[0007] Acquiring an acoustic signal, which is a continuous acoustic signal stream collected by an acoustic emission sensor array arranged on the surface of a coal rock mass;

[0008] Segmenting the acoustic signal using a sliding time window to obtain a plurality of data segments;

[0009] Nonlinear features are extracted from the data segment to obtain a comprehensive feature vector, which is a feature vector constructed by fusing multiple nonlinear features; the nonlinear features are system evolution features extracted from the acoustic signal based on the principle of nonlinear dynamics.

[0010] The monitoring and early warning state is triggered based on the comprehensive feature vector. The monitoring and early warning state is used to quantify the similarity between the current system state and the historical unstable state by introducing dynamic time warping distance, and to perform mutation detection in combination with the real-time feature sequence corresponding to the comprehensive feature.

[0011] In some embodiments, acquiring acoustic signals includes:

[0012] Receive the raw acoustic wave signal collected by the acoustic emission sensor array;

[0013] The original acoustic signal is amplified and then synchronously acquired and converted from analog to digital to obtain a digitized discrete signal sequence.

[0014] The discrete signal sequence is preprocessed according to a preset preprocessing term to obtain the acoustic signal. The preset preprocessing term includes bandpass filtering and normalization.

[0015] In some embodiments, the acoustic signal is segmented using a sliding time window to obtain multiple data segments, including:

[0016] Set the window parameters for the sliding time window, including the window length and window overlap ratio;

[0017] According to the window parameters, the sliding time window is moved over the continuous signal stream in the acoustic signal;

[0018] The acoustic signal within the sliding time window is extracted to obtain multiple data segments.

[0019] In some embodiments, nonlinear features are extracted from the data segment to obtain a comprehensive feature vector, including:

[0020] Calculate the optimal latency and optimal embedding dimension for the data segment;

[0021] Based on the optimal delay time and optimal embedding dimension, the phase space of the data segment is reconstructed to obtain the phase point sequence;

[0022] Based on the phase point sequence, nonlinear features are extracted in the reconstructed phase space. The nonlinear features include system complexity features, system chaos features, and system unpredictability and recursion features.

[0023] The eigenvalues ​​of the nonlinear features are combined to form the comprehensive feature vector.

[0024] In some embodiments, calculating the optimal latency and optimal embedding dimension of the data segment includes:

[0025] Get the preset maximum delay time;

[0026] Based on the maximum delay time, calculate the mutual information function of the time series corresponding to the data segment;

[0027] Calculate the local minimum point based on the mutual information function;

[0028] Extract the delay time corresponding to the local minimum point to obtain the optimal delay time;

[0029] Based on the optimal delay time, the correlation integral is calculated using the correlation integral function; the correlation integral function is a function constructed using the Grassberger-Procaccia algorithm.

[0030] The optimal embedding dimension is found based on the correlation integral. The optimal embedding dimension is the dimension that minimizes the proportion of false nearest neighbors.

[0031] In some embodiments, extracting nonlinear features in the reconstructed phase space based on the phase point sequence includes:

[0032] In the reconstructed phase space, the nearest neighbor is found for each phase point to obtain adjacent point pairs, which include the phase point and the nearest neighbor.

[0033] Track the exponential separation distance of the adjacent point pairs over time;

[0034] Calculate the average value of all phase points based on the separation distance;

[0035] Generate the maximum Lyapunov exponent, which is the slope of the linear region of the curve of the average value changing over time;

[0036] The chaotic characteristics of the system are generated based on the sensitive dependence of the maximum Lyapunov exponent quantization system on initial conditions.

[0037] In some embodiments, extracting nonlinear features in the reconstructed phase space based on the phase point sequence includes:

[0038] The original sequence of the phase point sequence is coarsened to construct new sequences under different scale factors;

[0039] Calculate the permutation entropy of the new sequence;

[0040] Plot the average percentage error curve with the scale factor as the horizontal axis and the permutation entropy as the vertical axis;

[0041] Based on the average percentage error curve, the permutation entropy under a first preset number of scale factors is selected to constitute the system complexity feature.

[0042] In some embodiments, extracting nonlinear features in the reconstructed phase space based on the phase point sequence includes:

[0043] Calculate the recursive matrix based on the reconstructed phase space;

[0044] A recursion rate is generated based on the recursion matrix, whereby the recursion rate is the proportion of recursive points in the recursion matrix; the recursion rate is used to reflect the overall recurrence probability of the system state.

[0045] Construct a recursion graph based on the recursion matrix;

[0046] Calculate the proportion of the length of the diagonal structure in the recursive graph to the total number of recursive points to generate deterministic parameters;

[0047] The unpredictability and recursiveness characteristics of the system are constructed based on the recursion rate and the deterministic parameters.

[0048] In some embodiments, triggering a monitoring and early warning status based on the comprehensive feature vector includes:

[0049] A stability criterion is constructed based on the comprehensive feature vector. The stability criterion includes multiple dynamic parameters, including the reconstructed phase space, the Lyapunov exponent, and the correlation dimension.

[0050] Set logical judgment conditions, which include primary warning conditions and advanced warning conditions;

[0051] If the comprehensive feature vector satisfies the primary warning conditions, first monitoring information is generated; the primary warning conditions include the Lyapunov index being greater than the index threshold within a second preset number of time sliding windows, and the growth rate of the correlation dimension exceeding the growth rate threshold compared to the previous stage.

[0052] If the comprehensive feature vector satisfies the advanced early warning conditions, second monitoring information is generated; the advanced early warning conditions include that the dynamic time warping distance between the real-time feature sequence corresponding to the comprehensive feature vector and the instability precursor sequence is less than a set distance threshold.

[0053] According to a second aspect of this application, a coal and rock stability monitoring system based on acoustic signal dynamics characteristics is provided, the system comprising:

[0054] The signal acquisition module is used to acquire acoustic signals, which are continuous acoustic signal streams collected by an array of acoustic emission sensors arranged on the surface of the coal and rock mass.

[0055] The signal segmentation module is used to segment the acoustic signal using a sliding time window to obtain multiple data segments;

[0056] The feature extraction module is used to extract nonlinear features from the data segment to obtain a comprehensive feature vector, wherein the comprehensive feature vector is a feature vector constructed by fusing multiple nonlinear features; the nonlinear features are system evolution features extracted from the acoustic signal based on the principle of nonlinear dynamics.

[0057] The dynamic early warning module is used to trigger a monitoring and early warning state based on the comprehensive feature vector. The monitoring and early warning state is used to quantify the similarity between the current system state and the historical instability state by introducing a dynamic time warp distance, and to perform mutation detection in combination with the real-time feature sequence corresponding to the comprehensive feature.

[0058] According to a third aspect of this application, a computer device is provided, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor executes the program to implement the above-described method for monitoring coal and rock stability based on acoustic signal dynamics characteristics.

[0059] According to a fourth aspect of this application, a storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the above-described method for monitoring coal and rock stability based on acoustic signal dynamics characteristics.

[0060] By employing the above technical solutions, this application provides a method and system for monitoring coal and rock stability based on the dynamic characteristics of acoustic signals. The method, after acquiring acoustic signals, segments the signals using a sliding time window to obtain multiple data segments. Nonlinear features are then extracted from these data segments to obtain a comprehensive feature vector, which is then used to trigger a monitoring and early warning state. The comprehensive feature vector can be constructed by fusing multiple nonlinear features, which are system evolution features extracted from the acoustic signals based on nonlinear dynamics principles. The method introduces dynamic time warping distance to quantify the similarity between the current system state and historical instability states, and combines this with real-time feature sequences corresponding to the comprehensive features to perform mutation detection, forming a hybrid early warning monitoring system that combines mechanism and data-driven approaches, thereby improving the timeliness and accuracy of coal and rock stability monitoring and early warning.

[0061] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below. Attached Figure Description

[0062] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0063] Figure 1 A schematic diagram of the coal and rock stability monitoring method based on acoustic signal dynamics characteristics provided in this application embodiment;

[0064] Figure 2 This is a schematic diagram of the feature extraction process provided in an embodiment of this application;

[0065] Figure 3 This is a schematic diagram of the process for extracting chaotic features of the system provided in an embodiment of this application;

[0066] Figure 4 This is a schematic diagram of the process for extracting system complexity features provided in an embodiment of this application;

[0067] Figure 5 This is a schematic diagram of the recursive feature extraction process provided in the embodiments of this application;

[0068] Figure 6 This is a schematic diagram of the trigger monitoring and early warning process provided in the embodiments of this application;

[0069] Figure 7 A schematic diagram of the structure of a coal and rock stability monitoring system based on acoustic signal dynamics characteristics provided in this application embodiment. Detailed Implementation

[0070] The present application will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the embodiments of the present application can be combined with each other.

[0071] In this embodiment of the application, coal and rock are mineral materials, referring to the geological structure composed of coal seams and the surrounding rocks above and below them in mining engineering. The surrounding rocks generally refer to the materials surrounding the coal seam, which may include natural environmental materials, such as adjacent rock layers, or artificial building materials, such as the roof and floor of the coal mining area.

[0072] When coal and rock reach a certain material scale, they can form a coal-rock system in the mining environment. A coal-rock system can include a coal seam and rock strata located above and below the coal seam. Because the coal seam in the coal-rock system accumulates a large amount of energy under the pressure of the hard rock strata above and below, the coal-rock system will exhibit unique physical characteristics, such as stress behavior, under the action of mining pressure.

[0073] The energy accumulated within the coal-rock system can affect the stability of the entire coal mining area. When the coal-rock system becomes unstable, it can lead to disasters such as roof collapse, rock bursts, and rock bursts in the mining area. Therefore, in order to prevent disasters from occurring in advance, it is necessary to monitor the stability of the coal-rock system.

[0074] In some embodiments, coal and rock stability monitoring can be achieved through microseismic monitoring technology. Microseismic monitoring technology can infer stress concentration areas by detecting microseismic events generated by rock mass fracturing. For example, when using microseismic monitoring technology to monitor coal and rock stability, data and field investigations of the mining area can be collected first. Based on the collected data and field investigation results, sensors can be deployed considering factors such as the geological conditions of the monitoring area and the impact range of mining activities to comprehensively and accurately monitor the microseismic activity of the coal and rock mass.

[0075] Then, by acquiring microseismic signals from the coal and rock mass in real time, seismic methods are used to locate the seismic source. Based on the arrival time and wave velocity of the seismic waves received by the sensors, the spatial location of the microseismic event is calculated. Signal analysis is then performed on the located microseismic events, analyzing characteristics such as the amplitude, frequency, duration, and energy of the seismic waves to monitor the deformation and failure of the coal and rock mass. Based on the distribution characteristics, signal characteristics, and source location of the microseismic events, the stability of the coal and rock mass is assessed, and timely early warning information is issued based on the stability assessment results.

[0076] However, microseismic monitoring technology, which is based on microseismic events generated by rock mass fracturing, has limitations such as large event location errors and difficulty in identifying precursors of gradual failure. Therefore, in some embodiments, stress and strain monitoring can also be carried out by directly measuring the deformation of the rock mass by embedding deformation sensors.

[0077] For example, when monitoring coal and rock stability through stress and strain, stress and strain sensors can be installed in the monitoring mining area, ensuring close contact between the sensors and the coal and rock mass. The sensors can collect stress and strain data at a set frequency, recording the time, location, and environmental conditions of the data collection. The collected data can then be analyzed for stress and strain variations, such as analyzing trends and observing whether stress and strain change over time or during mining activities. Furthermore, stress and strain curves can be used to determine the deformation and failure of the coal and rock mass, thereby assessing its stability and issuing early warnings based on the stress and strain data.

[0078] However, when monitoring the stability of coal and rock through stress and strain, it is necessary to bury stress and strain sensors to measure rock deformation, which makes the monitoring process susceptible to environmental interference and difficult to achieve large-area coverage.

[0079] To improve monitoring coverage and reduce environmental interference, acoustic emission technology can be applied to coal and rock stability monitoring in some embodiments. For example, when using acoustic emission technology for coal and rock stability monitoring, acoustic emission sensors can send sound waves to the coal and rock system and collect the detection signals of the sound waves from the coal and rock system. Then, based on parameters such as acoustic emission event counts and energy release in the detection signals, the degree of rock mass damage can be determined, and the stability of the coal and rock mass can be assessed. When a decrease in the stability of the coal and rock mass is detected, an early warning signal can be issued in a timely manner. However, when using acoustic emission technology for coal and rock stability monitoring, analyzing only parameters such as acoustic emission event counts and energy release is insufficient for extracting features from nonlinear and non-stationary signals.

[0080] It is evident that the coal and rock stability monitoring process described in the above embodiments is insufficient to fully characterize the nonlinear and chaotic features of the coal and rock failure process. Furthermore, the monitoring criteria rely on empirical thresholds, which need to be determined through long-term statistical results and have regional limitations, resulting in low early warning timeliness and accuracy of the above coal and rock stability monitoring process.

[0081] To address the issues of low timeliness and accuracy in early warning during coal and rock stability monitoring, this application provides a coal and rock stability monitoring method based on the dynamic characteristics of acoustic signals in some embodiments. This method can extract system evolution characteristics from acoustic signals using nonlinear dynamics methods, and construct stability criteria based on dynamic parameters such as phase space reconstruction, Lyapunov exponent, and correlation dimension. This enables early and accurate early warning of coal and rock system instability, thereby improving the safety of mining operations.

[0082] The method can be applied to electronic devices with data processing capabilities. These electronic devices include, but are not limited to, computers, servers, mobile terminals, smart wearable devices, mine monitoring instruments, mine or tunnel safety devices, and industrial control computers. For ease of description, this application embodiment uses an electronic device as the execution subject of the method. It should be understood that the method can also be applied to other types of execution subjects, which are not illustrated in this application embodiment. Figure 1 As shown, the method includes:

[0083] S101, Acquire acoustic signals.

[0084] When monitoring the stability of a coal-rock system, electronic equipment can first acquire acoustic signals, which are continuous acoustic signal streams collected by an array of acoustic emission sensors arranged on the surface of the coal-rock mass. Therefore, before monitoring, multiple acoustic emission sensors need to be arranged on the surface of the coal-rock mass in the mining area, forming an acoustic emission sensor array, based on the geological data, mining history and current status of the mining area, as well as the results of on-site exploration.

[0085] For example, an acoustic emission sensor array can include multiple piezoelectric acoustic emission (AE) sensors. Upon activation of monitoring, the piezoelectric acoustic emission materials, such as lead zirconium titanate (PZT), within the sensors can rapidly release internal energy to generate transient elastic waves when subjected to external stress. Utilizing the piezoelectric effect, the mechanical energy of the acoustic emission waves is converted into an electrical signal, thereby enabling the detection of the acoustic emission signal.

[0086] In order to acquire acoustic signals, electronic devices can send data acquisition requests to acoustic emission sensor arrays. After receiving the data acquisition request, the acoustic emission sensor arrays can detect acoustic signals in the monitoring area, obtain raw acoustic signals, and feed back the raw acoustic signals to the electronic devices so that the electronic devices can obtain a continuous acoustic signal stream within a certain monitoring period.

[0087] In some embodiments, the electronic device can perform signal processing on the raw acoustic wave signal fed back by the acoustic emission sensor array to convert the raw acoustic wave signal into an acoustic signal that the controller can directly process. When the device acquires the acoustic signal, it can first receive the raw acoustic wave signal collected by the acoustic emission sensor array and amplify the raw acoustic wave signal to obtain an amplified signal. Then, the amplified signal formed after amplification of the raw acoustic wave signal is synchronously acquired and analog-to-digital converted to obtain a digitized discrete signal sequence.

[0088] For example, electronic equipment can receive raw acoustic emission (AE) signals acquired in real time from a piezoelectric acoustic emission (AE) sensor array arranged on the surface of a coal or rock mass, within a frequency response range of 20 kHz to 1 MHz. A preamplifier with a gain of 40-60 dB is then used to amplify the raw acoustic signal, resulting in an amplified signal. A high-speed data acquisition card with a sampling rate fs ≥ 2 MHz and a resolution ≥ 16 bits is then used to synchronously acquire and perform analog-to-digital conversion on the amplified signal, yielding a digitized discrete signal sequence. s ( n ), where n = 1, 2, ..., N, and N is the number of sampling points.

[0089] The discrete signal sequence is then preprocessed according to preset preprocessing terms to obtain the acoustic signal. These preset preprocessing terms include bandpass filtering and normalization. For example, to obtain the digitized discrete signal sequence... s ( n After that, discrete signal sequences can be processed. s ( n Preprocessing is performed, including bandpass filtering and normalization.

[0090] When performing bandpass filtering, electronic devices can use a 4th-order Butterworth filter to retain the 100kHz-800kHz main frequency components of the signal, remove low-frequency mechanical noise and high-frequency electromagnetic interference, and output the filtered signal. x ( n During normalization, the filtered signal can be processed. x ( n Perform maximum and minimum value normalization to obtain a dimensionless signal. x' ( n This is used as an acoustic signal for subsequent analysis. Among them, the dimensionless signal... x' ( n It can be calculated using the following formula:

[0091] ;

[0092] in, x' ( n Acoustic signals are dimensionless signals and can contain multiple normalized signal values ​​arranged in chronological order. x ( n ) represents the filtered signal; max( x ) and min ( x The filtered signals are shown below. x ( n The maximum and minimum values ​​of the sequence.

[0093] S102. The acoustic signal is segmented using a sliding time window to obtain multiple data segments.

[0094] After acquiring the acoustic signal, it can be segmented into multiple data segments. This is achieved by using a sliding time window to segment the acoustic signal, resulting in multiple data segments. Segmenting the acoustic signal into multiple data segments enhances the local features of the signal, facilitating feature extraction from each local segment. Furthermore, by dividing a long signal into multiple shorter data segments, it is possible to observe the changes in the signal at different time points more precisely, improving the resolution of signal analysis.

[0095] In some embodiments, when performing segmentation processing of the acoustic signal using a sliding time window to obtain multiple data segments, window parameters of the sliding time window can be set first. These window parameters include window length and window overlap rate. Then, according to the window parameters, the sliding time window is moved along the continuous signal stream in the acoustic signal. During the movement of the sliding time window, the acoustic signal within the sliding time window can be extracted to obtain multiple data segments.

[0096] For example, to perform real-time continuous monitoring, the window length in the window parameters can be set to L, such as L = 8192 points, with a window overlap rate of 50%. Then, the electronic device can use a sliding time window of length L and an overlap rate of 50% to monitor the acoustic signal. x' ( n The corresponding continuous signal stream is segmented to obtain multiple data segments. Then the first... k The data segment within a time window can be represented as:

[0097] X k ={ x k ' (1), x k ' (2), ..., x k ' ( L )};

[0098] in, X k Indicates the first k Data segments within a time window; x k ' (1), x k ' (2), ..., x k ' ( L ) respectively represent the first k The first to last time window L One signal value; L This represents the window length. As can be seen, after converting the acoustic signal into multiple data segments, each data segment will have its features extracted independently.

[0099] S103. Extract nonlinear features from the data segment to obtain a comprehensive feature vector.

[0100] After obtaining multiple data segments through segmentation, nonlinear feature extraction can be performed on each data segment to extract nonlinear features. A comprehensive feature vector is then constructed based on these extracted nonlinear features. This comprehensive feature vector is a feature vector constructed by fusing multiple nonlinear features. The nonlinear features are system evolution characteristics extracted from acoustic signals based on nonlinear dynamics principles.

[0101] In some embodiments, when an electronic device extracts nonlinear features from a data segment to obtain a composite feature vector, it may first calculate the optimal delay time and optimal embedding dimension of the data segment. The optimal delay time and optimal embedding dimension can be calculated using the correlation coefficient method (CC).

[0102] When calculating the optimal delay time and optimal embedding dimension of a data segment, the parameters can be defined using the correlation coefficient method, that is, for each data segment... X k You can set the maximum delay time. τ max and embedding dimension range m and selecting multiple neighborhood radii r i Among them, the neighborhood radius r i = i ×0.5 σ , σ It is the standard deviation of the time series. i It can take integers such as 1, 2, 3, 4, etc.

[0103] Next, obtain the preset maximum delay time, and calculate the mutual information function of the time series corresponding to the data segment based on the maximum delay time. Then, based on the mutual information function, calculate the local minimum point, and obtain the optimal delay time by extracting the delay time corresponding to the local minimum point.

[0104] For example, for a given maximum delay time τ max It can calculate the mutual information function of time series. I ( τ ), and based on the mutual information function I ( τ Calculate the correlation statistic, including for each delay time t and embedding dimension m. S ( m , r i , t Then calculate the average statistic. That is, the average value of the correlation statistic is:

[0105]

[0106] And the maximum deviation of the calculated statistic ,Right now:

[0107]

[0108] Where, ΔS ( m , t )yes S ( m , r i , t )about r The maximum deviation. This is determined by finding the point corresponding to the first local minimum. τ This is the optimal delay time, which is achieved by finding the maximum deviation. The first local minimum or average statistic The first zero point, corresponding to t This is the optimal delay time. τ .

[0109] After calculating the optimal delay time, the correlation integral can be calculated based on the optimal delay time using the correlation integral function, which can be constructed using the Grassberger-Procaccia algorithm. Then, the optimal embedding dimension is found based on the correlation integral; this optimal embedding dimension is the dimension that minimizes the proportion of false nearest neighbors.

[0110] For example, the Grassberger-Procaccia algorithm can be used to calculate the correlated integral function. C ( r ),Right now:

[0111]

[0112] in, C ( r () is an integral function; M It is the length of the time series; r The radius of the hypersphere; Y i and Y j Let Θ be two points in phase space; ) is the Heaviside step function, when x When ≥0, Θ( x =1 if ) otherwise 0; Indicates taking Y i and Y j The Euclidean norm.

[0113] Based on the constructed correlation integral function, the correlation dimension can be defined. D 2, represented as:

[0114]

[0115] In actual calculations, we can use C(r)ln r ln C ( r In a log-log plot, find a linear scale-free region; the slope of this region is the correlation dimension. D The estimated value of 2.

[0116] Therefore, based on the optimal delay time τ It can calculate the correlation integral. C ( m , r And by finding the dimension that minimizes the proportion of false nearest neighbors. m The optimal embedding dimension is determined by setting the maximum embedding dimension. m max For each embedding dimension m Calculate the correlation statistic S ( m , r i , t And calculate the average statistic, i.e.:

[0117] ;

[0118] By searching S corr ( t The local minimum of ) is the corresponding t, which is the optimal embedding window. τ w And then according to τ w =( m 1) τ Calculate the optimal embedding dimension m .

[0119] like Figure 2 As shown, after calculating the optimal delay time and optimal embedding dimension of the data segment, the phase space of the data segment can be reconstructed based on the optimal delay time and optimal embedding dimension to obtain the phase point sequence.

[0120] For example, the phase space of the data segment can be reconstructed using Takens' embedding theorem to obtain the phase point sequence:

[0121]

[0122] in, τ Indicates the optimal delay time; x' ( i () indicates the original time series at time point i The value; m Indicates the optimal embedding dimension;T Indicates the transpose operation; i =1, 2, ..., M ; M = L ( m 1) τ This represents the total number of phase points in phase space. Y i It is m A phase point in a given dimension represents the state of a dynamical system at a certain moment.

[0123] Based on the phase point sequence, nonlinear features are extracted from the reconstructed phase space, and the eigenvalues ​​of the nonlinear features are combined into a comprehensive feature vector. These nonlinear features include system complexity features, system chaos features, and system unpredictability and recursion features.

[0124] System chaos characteristics are used to quantify the system's sensitivity to initial conditions and are a direct criterion for system stability. For example... Figure 3 As shown, regarding the chaotic characteristics of the system, in some embodiments, when the electronic device extracts nonlinear features in the reconstructed phase space based on the phase point sequence, it can find the nearest neighbor for each phase point in the reconstructed phase space to obtain adjacent point pairs. The adjacent point pairs include both the phase point and its nearest neighbor.

[0125] Next, the separation distance of adjacent point pairs is tracked exponentially over time, and the average value of all phase points is calculated based on the separation distance. Then, the maximum Lyapunov exponent is generated based on the average value, and the sensitivity of the system to initial conditions is quantified based on the maximum Lyapunov exponent to generate the system's chaotic characteristics. Here, the maximum Lyapunov exponent is the slope of the linear region of the curve showing the change of the average value over time.

[0126] For example, using Rosenstein's small data method, it is possible to obtain data for each phase point in the reconstructed phase space. Y j Find the nearest neighbor Y j' , satisfy | j - j' |> T a T a If the average period is , then the initial distance is d j (0)= min || Y j - Y j' Then, by tracking the exponential growth of each adjacent point pair over time... i Separation distanced j ( i ),in i =1, 2, ..., min( M j , M j' Then calculate the average value of all phase points j, i.e.:

[0127] ;

[0128] in, This indicates that for all phase points j Find the average, Δ t This is the sampling time interval. y ( i The slope of the linear region of the curve that changes with time is the maximum Lyapunov exponent. λ 1. It can be seen that when the maximum Lyapunov exponent... λ A value of 1 > 0 indicates that the coal-rock system is chaotic, and the larger the value of the maximum Lyapunov exponent, the more unstable and unpredictable the coal-rock system is.

[0129] System complexity characteristics can measure the randomness and complexity of a time series across multiple time scales. Since a significant increase in entropy indicates increased disorder and decreased stability in the coal-rock system, system complexity characteristics can be generated by calculating entropy. That is, for example... Figure 4 As shown, in some embodiments, when extracting nonlinear features in the reconstructed phase space based on phase point sequences, the original phase point sequence can be coarsened to construct new sequences under different scale factors. Then, the permutation entropy of the new sequences is calculated, and an average percentage error curve is plotted with the scale factor on the horizontal axis and the permutation entropy on the vertical axis. Based on the average percentage error curve, the permutation entropy under a first preset number of scale factors is selected to constitute the system complexity feature.

[0130] For example, by analyzing the original sequence corresponding to the data segment X k By coarsening, different scale factors can be constructed. δ The new sequence below z (δ) , new sequence z (δ) It can be represented as:

[0131]

[0132] in, z j(δ) Indicates the first j A new sequence, j =1, 2, ..., L / δ ; δ Indicates the scale factor; x' ( i ) indicates that the original time series is at the 1st epoch. i The values ​​at each time point; L is the window length.

[0133] For each coarse-grained new sequence z (δ) It can calculate the permutation entropy of a new sequence. PE ( δ ). and using the scale factor δ The horizontal axis is... PE ( δ Plot the MPE curve with the first 5 scales ( ) as the vertical axis. δ The entropy values ​​of the permutation of (=1, 2, 3, 4, 5) can form the feature vector MPE of the system complexity characteristics. PE 1, PE 2, PE 3, PE 4, PE 5).

[0134] The unpredictability and recursiveness of a system can be used to reflect the unpredictability and recursiveness of a coal-rock system. The recursiveness can be quantified according to the dynamic transformation of the coal-rock system. Therefore, such as... Figure 5 As shown, in some embodiments, the recurrence characteristics of system states can be analyzed using recurrence graphs to reveal the dynamic transitions of the system. Specifically, when extracting nonlinear features from the reconstructed phase space based on phase point sequences, a recurrence matrix can be calculated first based on the reconstructed phase space, and a recurrence rate can be generated based on the recurrence matrix. The recurrence rate represents the proportion of recurrence points in the recurrence matrix and reflects the overall recurrence probability of the system states.

[0135] Then, a recursion graph is constructed based on the recursion matrix, and the proportion of the length of the diagonal structure in the recursion graph to the total number of recursion points is calculated to generate deterministic parameters. Then, the unpredictability and recursiveness characteristics of the system are constructed based on the recursion rate and deterministic parameters.

[0136] For example, electronic devices can calculate recursion matrices based on the reconstructed phase space. R ,Right now:

[0137]

[0138] in, Y i andY j Let them be two points in phase space; Θ ( ) is the Heaviside step function; i , j =1, 2, ..., M ; It is a preset distance threshold.

[0139] Next, calculate the recursion rate (RR), which is the proportion of recursive points in the recursion matrix.

[0140]

[0141] And calculate the deterministic (DET) parameter, which is the proportion of the length of the diagonal structure representing deterministic behavior in the recursion graph to the total number of recursive points, i.e.:

[0142]

[0143] in, P ( ) is of length The number of diagonal structures. It can be seen that the recurrence rate (RR) can reflect the overall recurrence probability of the system state, and the deterministic parameter DET will decrease before the coal-rock system becomes unstable. Therefore, by combining the recurrence rate and the deterministic parameter, an eigenvector of the unpredictability and recurrence characteristics of the system can be constructed.

[0144] According to the feature extraction method shown in the above embodiments, in the first... k Data for each time window X k After extracting multiple nonlinear features, the electronic device can combine the calculated feature values ​​into a comprehensive feature vector. F k Furthermore, the feature vectors corresponding to laboratory coal and rock failure test data and field measured data can be stored in a historical database. D This forms a knowledge base for building hybrid early warning models.

[0145] The laboratory coal and rock failure test data and the field measured data can be labeled with stability states such as "stable," "critical," and "instable" based on the experimental and field measured results. Therefore, the hybrid early warning model can be a machine learning model, meaning that the machine learning model is trained using data from the knowledge base to obtain the hybrid early warning model.

[0146] S104. Trigger monitoring and early warning status based on comprehensive feature vector.

[0147] After extracting nonlinear features from the data segment to obtain a comprehensive feature vector, a monitoring and early warning state can be triggered based on the comprehensive feature vector. This monitoring and early warning state is used to quantify the similarity between the current system state and historical instability states by introducing a dynamic time warp distance, and to perform mutation detection in conjunction with the real-time feature sequence corresponding to the comprehensive features.

[0148] like Figure 6 As shown, in some embodiments, when triggering a monitoring and early warning state based on the comprehensive feature vector, a stability criterion can be constructed first based on the comprehensive feature vector. The stability criterion includes multiple dynamic parameters, including the reconstructed phase space, the Lyapunov exponent, and the correlation dimension.

[0149] Further logical judgment conditions are set, including primary warning conditions and advanced warning conditions. The primary warning conditions include a Lyapunov exponent greater than an exponent threshold within a second preset number of time sliding windows, and a growth rate exceeding a growth rate threshold for the correlation dimension compared to the previous stage. The advanced warning conditions include a dynamic time warping distance between the real-time feature sequence corresponding to the comprehensive feature vector and the instability precursor sequence being less than a set distance threshold.

[0150] By comparing the stability criterion with the logical judgment condition, it can be determined whether the comprehensive feature vector meets the primary warning condition and the advanced warning condition. When the Lyapunov exponent is greater than the exponent threshold within a second preset number of time sliding windows, and the growth rate of the correlation dimension compared to the previous stage exceeds the growth rate threshold, it can be determined that the comprehensive feature vector meets the primary warning condition. At this time, the first monitoring information can be generated to indicate that the primary warning (yellow warning) has been triggered.

[0151] For example, the exponential threshold β set in the primary early warning conditions is a threshold greater than 0, and the growth rate threshold is... γ %. If three consecutive time windows meet the requirements λ 1 k >β and D 2( k Growth exceeding the previous stage γ If the percentage reaches a certain threshold, a primary warning (yellow warning) will be triggered, indicating that the coal and rock system may be entering a critical state.

[0152] When the dynamic time warping distance between the real-time feature sequence corresponding to the comprehensive feature vector and the instability precursor sequence is less than a set distance threshold, it can be determined that the comprehensive feature vector meets the advanced warning conditions. At this time, a second monitoring information can be generated to indicate that an advanced warning (red warning) has been triggered.

[0153] For example, it can be calculated from the current time... M The feature vector sequence of the window {Fk-M+1 , ..., F k} and all reference sequences {F} marked as "critical" and "instable" in the historical database DD r The Dynamic Time Warping (DTW) distance between two sequences is used to effectively align and measure the similarity between two sequences of different lengths and with a phase difference. If the DTW distance between the current sequence and a precursor sequence of instability is less than a set threshold... d th If this occurs, a high-level warning (red warning) will be immediately triggered, indicating that the coal and rock system is repeating its historical instability pattern.

[0154] By applying the technical solutions of the above embodiments, the coal and rock stability monitoring based on the dynamic characteristics of acoustic signals described in the above embodiments can abandon the concept of treating acoustic signals as isolated events, and instead treat acoustic signals as the output of the evolution of a nonlinear dynamic system. By inverting the intrinsic dynamic characteristics of the system through phase space reconstruction theory, a multi-dimensional, multi-scale nonlinear characteristic parameter system can be constructed, including features such as system complexity, system chaos, system unpredictability, and system recursion. This system can jointly characterize the stable state of the coal and rock system from different perspectives. The method can overcome the limitations of a single index and quantify the similarity between the current system state and the historical unstable state by introducing dynamic time warping (DTW) distance. Combined with the mutation detection of real-time feature sequences, a hybrid early warning model combining mechanism and data-driven approaches is formed, which significantly improves the accuracy and robustness of the coal and rock system stability monitoring criteria.

[0155] In some embodiments, as a specific implementation of the coal and rock stability monitoring method based on acoustic signal dynamics characteristics in the above embodiments, some embodiments of this application also provide a coal and rock stability monitoring system based on acoustic signal dynamics characteristics, such as... Figure 7 As shown, the system includes:

[0156] The signal acquisition module is used to acquire acoustic signals, which are continuous acoustic signal streams collected by an array of acoustic emission sensors arranged on the surface of the coal and rock mass.

[0157] The signal segmentation module is used to segment the acoustic signal using a sliding time window to obtain multiple data segments;

[0158] The feature extraction module is used to extract nonlinear features from the data segment to obtain a comprehensive feature vector. The comprehensive feature vector is a feature vector constructed by fusing multiple nonlinear features. The nonlinear features are system evolution features extracted from acoustic signals based on the principle of nonlinear dynamics.

[0159] The dynamic early warning module is used to trigger the monitoring and early warning status based on the comprehensive feature vector. The monitoring and early warning status is used to quantify the similarity between the current system state and the historical instability state by introducing dynamic time warping distance, and to perform mutation detection in combination with the real-time feature sequence corresponding to the comprehensive features.

[0160] By applying the technical solutions of the above embodiments, the coal and rock stability monitoring system based on acoustic signal dynamics provided in the above embodiments can acquire acoustic signals through a signal acquisition module, and then the signal segmentation module uses a sliding time window to segment the acoustic signals to obtain multiple data segments. The feature extraction module then extracts nonlinear features from the data segments to obtain a comprehensive feature vector. Finally, the dynamic early warning module triggers a monitoring and early warning state based on the comprehensive feature vector. The comprehensive feature vector can be constructed by fusing multiple nonlinear features, and the nonlinear features are system evolution features extracted from the acoustic signals based on nonlinear dynamics principles. The system introduces dynamic time warping distance to quantify the similarity between the current system state and historical instability states, and combines this with real-time feature sequences corresponding to the comprehensive features to perform mutation detection, forming a hybrid early warning monitoring system that combines mechanism and data-driven approaches, thereby improving the timeliness and accuracy of coal and rock stability monitoring and early warning.

[0161] It should be noted that other corresponding descriptions of the functional units involved in the coal and rock stability monitoring system based on acoustic signal dynamics provided in the embodiments of this application can be found in the corresponding descriptions in the coal and rock stability monitoring method based on acoustic signal dynamics provided in the above embodiments, and will not be repeated here.

[0162] This application also provides a computer device, specifically a personal computer, server, network device, etc. The computer device includes a bus, processor, memory, and communication interface, and may also include input / output interfaces and a display device. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the computer device stores location information. The network interface of the computer device is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements the steps in the various method embodiments.

[0163] Those skilled in the art will understand that the structure of the computer device described above is only a partial structure related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. A specific computer device may include more or fewer components, or combine certain components, or have different component arrangements.

[0164] In one embodiment, a computer-readable storage medium is also provided, which may be non-volatile or volatile, having stored thereon a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0165] In one embodiment, a computer program product is also provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0166] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0167] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the methods described above.

[0168] Any references to memory, database, or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc.

[0169] Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can take many forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0170] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchain. The processors involved in the embodiments provided in this application may be, but are not limited to, general-purpose processors, graphics processors, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc.

[0171] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0172] The above embodiments are merely illustrative of several implementation methods of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for monitoring coal and rock stability based on acoustic signal dynamics, characterized in that, The method includes: Acquire acoustic signals, wherein the acoustic signals are a continuous acoustic signal stream collected by an array of acoustic emission sensors arranged on the surface of the coal and rock mass; The acoustic signal is segmented using a sliding time window to obtain multiple data segments; Extracting nonlinear features from the data segment to obtain a comprehensive feature vector, wherein the comprehensive feature vector is a feature vector constructed by fusing multiple nonlinear features; the nonlinear features are system evolution features extracted from the acoustic signal based on the principle of nonlinear dynamics; extracting nonlinear features from the data segment to obtain the comprehensive feature vector includes: calculating the optimal delay time and optimal embedding dimension of the data segment; reconstructing the phase space of the data segment according to the optimal delay time and optimal embedding dimension to obtain a phase point sequence; extracting nonlinear features from the reconstructed phase space based on the phase point sequence, wherein the nonlinear features include system complexity features, system chaos features, and system unpredictability and recursion features; and combining the feature values ​​of the nonlinear features into the comprehensive feature vector; Extracting nonlinear features in the reconstructed phase space based on the phase point sequence includes: coarsening the original sequence of the phase point sequence to construct new sequences under different scale factors; calculating the permutation entropy of the new sequences; plotting an average percentage error curve with the scale factor as the horizontal axis and the permutation entropy as the vertical axis; and selecting the permutation entropy under a first preset number of scale factors based on the average percentage error curve to constitute the system complexity feature. Extracting nonlinear features from the reconstructed phase space based on the phase point sequence includes: calculating a recursion matrix based on the reconstructed phase space; generating a recursion rate based on the recursion matrix, wherein the recursion rate is the proportion of recursive points in the recursion matrix; the recursion rate is used to reflect the overall recurrence probability of the system state; constructing a recursion graph based on the recursion matrix; calculating the proportion of the length of the diagonal structure in the recursion graph to the total number of recursive points to generate deterministic parameters; and constructing a feature vector of the unpredictability and recursiveness features of the system by combining the recursion rate and the deterministic parameters. The monitoring and early warning state is triggered based on the comprehensive feature vector. The monitoring and early warning state is used to quantify the similarity between the current system state and the historical unstable state by introducing dynamic time warping distance, and to perform mutation detection in combination with the real-time feature sequence corresponding to the comprehensive feature.

2. The method according to claim 1, characterized in that, Acquiring acoustic signals, including: Receive the raw acoustic wave signal collected by the acoustic emission sensor array; The original acoustic signal is amplified and then synchronously acquired and converted from analog to digital to obtain a digitized discrete signal sequence. The discrete signal sequence is preprocessed according to a preset preprocessing term to obtain the acoustic signal. The preset preprocessing term includes bandpass filtering and normalization.

3. The method according to claim 1, characterized in that, The acoustic signal is segmented using a sliding time window to obtain multiple data segments, including: Set the window parameters for the sliding time window, including the window length and window overlap ratio; According to the window parameters, the sliding time window is moved over the continuous signal stream in the acoustic signal; The acoustic signal within the sliding time window is extracted to obtain multiple data segments.

4. The method according to claim 1, characterized in that, Calculating the optimal latency and optimal embedding dimension of the data segment includes: Get the preset maximum delay time; Based on the maximum delay time, calculate the mutual information function of the time series corresponding to the data segment; Calculate the local minimum point based on the mutual information function; Extract the delay time corresponding to the local minimum point to obtain the optimal delay time; Based on the optimal delay time, the correlation integral is calculated using the correlation integral function; the correlation integral function is a function constructed using the Grassberger-Procaccia algorithm. The optimal embedding dimension is found based on the correlation integral. The optimal embedding dimension is the dimension that minimizes the proportion of false nearest neighbors.

5. The method according to claim 1, characterized in that, Extracting nonlinear features from the reconstructed phase space based on the phase point sequence includes: In the reconstructed phase space, the nearest neighbor is found for each phase point to obtain adjacent point pairs, which include the phase point and the nearest neighbor. Track the exponential separation distance of the adjacent point pairs over time; Calculate the average value of all phase points based on the separation distance; Generate the maximum Lyapunov exponent, which is the slope of the linear region of the curve of the average value changing over time; The chaotic characteristics of the system are generated based on the sensitive dependence of the maximum Lyapunov exponent quantization system on initial conditions.

6. The method according to claim 1, characterized in that, Triggering a monitoring and early warning status based on the comprehensive feature vector includes: A stability criterion is constructed based on the comprehensive feature vector. The stability criterion includes multiple dynamic parameters, including the reconstructed phase space, the Lyapunov exponent, and the correlation dimension. Set logical judgment conditions, which include primary warning conditions and advanced warning conditions; If the comprehensive feature vector satisfies the primary warning conditions, first monitoring information is generated; the primary warning conditions include the Lyapunov index being greater than the index threshold within a second preset number of time sliding windows, and the growth rate of the correlation dimension exceeding the growth rate threshold compared to the previous stage. If the comprehensive feature vector satisfies the advanced early warning conditions, second monitoring information is generated; the advanced early warning conditions include that the dynamic time warping distance between the real-time feature sequence corresponding to the comprehensive feature vector and the instability precursor sequence is less than a set distance threshold.

7. A coal and rock stability monitoring system based on acoustic signal dynamics characteristics, characterized in that, The system is applied to the method according to any one of claims 1-6; the system comprises: The signal acquisition module is used to acquire acoustic signals, which are continuous acoustic signal streams collected by an array of acoustic emission sensors arranged on the surface of the coal and rock mass. The signal segmentation module is used to segment the acoustic signal using a sliding time window to obtain multiple data segments; The feature extraction module is used to extract nonlinear features from the data segment to obtain a comprehensive feature vector, wherein the comprehensive feature vector is a feature vector constructed by fusing multiple nonlinear features; the nonlinear features are system evolution features extracted from the acoustic signal based on the principle of nonlinear dynamics. The dynamic early warning module is used to trigger a monitoring and early warning state based on the comprehensive feature vector. The monitoring and early warning state is used to quantify the similarity between the current system state and the historical instability state by introducing a dynamic time warp distance, and to perform mutation detection in combination with the real-time feature sequence corresponding to the comprehensive feature.

Images