Mine roadway deformation intelligent prediction method, equipment and medium

Abstract

The invention discloses an intelligent prediction method and device for mine roadway deformation and a medium, belongs to the technical field of geotechnical engineering testing, artificial intelligence and the Internet of Things, and aims to solve the technical problem of how to improve the density of related information of a mine roadway so as to improve the accuracy and quickness of mine roadway deformation prediction. According to the technical scheme, micro-seismic sensors and deformation sensors are arranged on the periphery of a mine roadway, and the micro-seismic sensors and the deformation sensors have the data real-time transmission function; according to the types and the positions of the micro-seismic sensors and the deformation sensors, the collected micro-seismic data and the roadway deformation data are transmitted to a database; performing data processing and feature extraction on the micro-seismic data to obtain damage characterization of the mine roadway surrounding rock; wherein the characterization of the damage of the surrounding rock of the mine roadway means that the micro-seismic data are analyzed into the fracture size, the magnitude, the energy and the fracture type.

Description

Technical Field

[0001] The present invention relates to the fields of geotechnical engineering testing, artificial intelligence and the Internet of Things, and in particular to a method, device and medium for intelligently predicting mine tunnel deformation. Background Art

[0002] In recent years, deep learning methods have achieved widespread success in many fields, inspiring more researchers to use them to predict rock deformation. Deep learning is essentially a data-driven approach. The learning process can be seen as starting from historical data, allowing the computer to find a potentially more optimal representation from the data samples that is more computationally efficient.

[0003] Currently, the most common deep learning-based rock deformation and failure prediction algorithms directly input deformation or load information during the rock failure process as time-series signals and use recurrent neural networks as deep learning models to predict future trends in rock deformation. Essentially, this algorithm infers the rock's constitutive relations from historical data, implicitly storing them in the model parameters as tensors, and using these parameters to fit the time-deformation relationship of the rock material. However, when using this strategy, due to the large number of parameters in the deep learning model and the limited information provided by a single strain data set, overfitting or underfitting are often observed. Furthermore, since these methods only use a single deformation data set, their data representation cannot reflect the internal damage of the rock.

[0004] Therefore, how to improve the density of relevant information in mine tunnels and thus improve the accuracy and speed of mine tunnel deformation prediction is a technical problem that needs to be solved urgently. Summary of the Invention

[0005] The technical task of the present invention is to provide a method, device and medium for intelligent prediction of mine tunnel deformation to solve the problem of how to increase the density of relevant information of mine tunnels, thereby improving the accuracy and speed of mine tunnel deformation prediction.

[0006] The technical task of the present invention is achieved in the following manner: a method for intelligently predicting mine tunnel deformation, the method being as follows:

[0007] Deploy microseismic sensors and deformation sensors around mine tunnels, and these sensors must be capable of real-time data transmission; the number of these sensors must be no less than eight.

[0008] According to the type and location of microseismic sensors and deformation sensors, the collected microseismic data and tunnel deformation data are transmitted to the database;

[0009] Processing and feature extraction of microseismic data to obtain damage characterization of the surrounding rock of the mine tunnel; damage characterization of the surrounding rock of the mine tunnel refers to parsing microseismic data into fracture size, magnitude, energy, and fracture type;

[0010] The damage representation of the mine tunnel surrounding rock represented by the microseismic signal is converted into a microseismic representation through the microseismic coding block. The microseismic representation and the tunnel deformation data are processed through the multi-source heterogeneous data decoder to predict the changes in mine tunnel deformation in the future. Among them, the microseismic representation reflects the damage level of the rock in the current time window.

[0011] As a preferred method, data processing and feature extraction are performed on the microseismic data to obtain damage characterization of the surrounding rock of the mine tunnel as follows:

[0012] Characterization of microseismic information: Calculate the damage characteristic parameters of the microseismic source radius r, magnitude A_s, and energy E_s respectively, and use the moment tensor decomposition method to identify the rupture type of the microseismic event, and then define the characteristic vector λ of the microseismic event;

[0013] Normalization of microseismic characteristics: The set of characteristic vectors λ is ∧, and ∧ is normalized to obtain a normalized microseismic characteristic vector sequence;

[0014] Microseismic conversion: The normalized microseismic feature vector sequence is converted into microseismic representation through a microseismic encoding block based on a long short-term memory neural network.

[0015] Preferably, the microseismic information is characterized as follows:

[0016] Calculate the microseismic source radius r and magnitude A respectively s and energy E s The damage characteristic parameter is as follows:

[0017]

[0018]

[0019] Where K is Brune's constant, K = 2.34; v s represents the shear wave velocity; f c Indicates the corner frequency of the acoustic emission signal. For indoor acoustic emission tests, f c is replaced by the peak frequency of the signal, A imax is the maximum amplitude of the acoustic emission signal waveform received by the i-th microseismic sensor; r i represents the distance between the acoustic emission event and the i-th microseismic sensor, in mm; k represents the number of microseismic sensors used for amplitude calculation, E iis the absolute energy of the waveform received by the i-th microseismic sensor, obtained by integrating the waveform envelope;

[0020] According to the moment tensor decomposition method, the rupture type of the microseismic event is determined as follows:

[0021]

[0022] Among them, M1>M2>M3, M1, M2, M3 are tensors M qp The three eigenvalues ​​represent the maximum couple; when X>60%, the earthquake source is in a shear-dominant state and is judged to be shear failure; when 60%≥X≥40%, the earthquake source is in a mixed state between shear and tension; when X<40%, the earthquake source is judged to be tensile failure;

[0023] Assume that at time t, the microseismic event received by the microseismic probe is f c , through microseismic events f c The inversion of the information of magnitude l, radius d, energy E and rupture type (such as shear or tension acoustic emission event) S, then the microseismic event f c The corresponding eigenvector λ∈R 4 Characterized as λ = (l, d, E, S).

[0024] Preferably, the normalization of microseismic characteristics is as follows:

[0025] The data distribution of Λ is standardized to a standard distribution with a mean of 0 and a variance of 1. The formula is as follows:

[0026]

[0027] Among them, μ represents the translation coefficient; σ represents the scaling coefficient;

[0028] Introducing two new learnable variables, the re-translation coefficient b and the re-scaling coefficient g, further The transformation formula is as follows:

[0029]

[0030] After normalization, the acoustic emission encoding process is expressed as follows:

[0031]

[0032] Here, Λ represents the input variable, and the retranslation coefficient b and the rescaling coefficient g are learned by the gradient descent method.

[0033] Preferably, the microseismic coding is as follows:

[0034] The normalized microseismic feature vector sequence is used as input and the long short-term memory neural network is used to convert the microseismic feature vector sequence into a local microseismic latent state sequence.

[0035] Through global microseismic hidden state fusion, the local microseismic hidden state is converted into the global microseismic hidden state by adding and averaging.

[0036] Through microseismic representation conversion, the latent state is converted into microseismic representation.

[0037] More preferably, the microseismic coding block implementation method based on the long short-term memory neural network is as follows:

[0038] With is the microseismic feature tensor that has been normalized by the feature layer. When the feature vector When the previous microseismic eigenvector is processed The generated latent variable h i-1 , then in the current time step, the hidden variable h i The calculation formula is as follows:

[0039]

[0040] In order to allow the network to process microseismic feature vector sequences of arbitrary length, two artificially defined vectors representing the beginning and end of the sequence are introduced, denoted as <bos>and <eos>;in, <bos>As the initial hidden state of the long short-term memory neural network layer, it is input into the model long short-term memory neural network layer together with λ1. The vector λ1 is a zero vector with the same shape as other hidden states. <eos>Indicates that there are no other microseismic feature vectors in the time window being processed. When the microseismic coding block receives <eos>As output, <eos>Together with the last hidden state h n Input the long short-term memory neural network layer and calculate the last hidden state h n+1 , <eos>The vector is a microseismic eigenvector λ with the same shape as the input i With zero vectors of the same shape, in summary, each hidden state in the microseismic coding block is calculated using the following formula:

[0041]

[0042] Through continuous cycles, the long short-term memory neural network layer obtains a series of hidden states H = {h1,h2,…,h n+1 }, variables h1,h2,…,h n+1 They are regarded as microseismic representations in the local time window of the microseismic event. In order to obtain the microseismic event representation that can represent the intensity of the microseismic event at the global level, a global microseismic feature fusion is performed. The microseismic feature fusion first averages the hidden states of each independent local microseismic feature to obtain the global microseismic feature hidden state h global Then h global Input the multi-layer perceptron to obtain the microseismic global encoding vector R, the formula is as follows:

[0043]

[0044] R=MLP(h global ).

[0045] As a preferred method, the damage representation of the mine roadway surrounding rock represented by the microseismic signal is converted into a microseismic representation through the microseismic encoding block, and the microseismic representation and roadway deformation data are processed through the multi-source heterogeneous data decoder to predict the change of mine roadway deformation in the future period as follows:

[0046] Assume that within the time window (t-Δt,t), the microseismic sensor detects microseismic events e1, e2,…, e n At the same time, the deformation sensor monitors the continuous rock deformation data and uses the equal-interval sampling method to obtain the deformation sequence s1, s2, ..., s m , where s i ∈R;

[0047] Transform the sequence s1,s2,…,s m Written in the form of a vector, it is recorded as S1=(s1,s2,…,s m ); where the subscript of the vector S represents the sequence number of the time window, and the subscript 1 indicates that the corresponding time window is the time window being monitored;

[0048] Once the microseismic events e1, e2,…, e n The microseismic encoding block is converted into a microseismic representation R, and the deformation vectors S2, S3, ... S in the next several time windows are predicted using the microseismic representation R and the deformation vector S1. I ,…S N .

[0049] More preferably, once microseismic events e1, e2,…, e n The microseismic encoding block is converted into a microseismic representation R, and the deformation vectors S2, S3, ... S in the next several time windows are predicted using the microseismic representation R and the deformation vector S1. I ,…S N The details are as follows:

[0050] The detected deformation vector S1 and the microseismic representation R are input into the long short-term memory neural network layer to obtain the deformation sequence of the next time window and hidden state h1;

[0051] The long short-term memory neural network cyclically predicts the deformation vector in time window i and hidden state h i-1 And input it together with the microseismic representation R into the long short-term memory neural network to predict the deformation vector S of the i+1th time window i+1 and hidden state h i Until the long short-term memory neural network thinks that the information in the next time window cannot be predicted based on the current information and outputs the end flag <eos>.

[0052] An electronic device comprising: a memory and at least one processor;

[0053] Wherein, the memory stores a computer program;

[0054] The at least one processor executes the computer program stored in the memory, so that the at least one processor executes the above-mentioned intelligent prediction method for mine tunnel deformation.

[0055] A computer-readable storage medium stores a computer program, which can be executed by a processor to implement the above-mentioned intelligent prediction method for mine tunnel deformation.

[0056] The mine tunnel deformation intelligent prediction method, device and medium of the present invention have the following advantages:

[0057] (1) The present invention achieves the fusion of multi-dimensional heterogeneous data such as microseismic and surrounding rock deformation by re-encoding microseismic information, thereby improving information density and providing more useful information for surrounding rock deformation prediction, thereby achieving the goal of faster and more accurate prediction of tunnel deformation;

[0058] (2) The present invention realizes the establishment of a microseismic signal encoder by deploying and collecting data on multiple monitoring devices such as mine tunnel strain and microseismic, and then uses a multi-heterogeneous data decoder to predict mine tunnel deformation, etc.;

[0059] (3) The present invention characterizes the microseismic information during the damage process of the tunnel surrounding rock into a damage state representation vector through a microseismic coding block. Combined with strain monitoring data, an autoencoder network with a long short-term memory network as the backbone is used to achieve mine tunnel deformation prediction that simultaneously considers multi-source heterogeneous information such as external strain and internal damage (microseismic information). The present invention is mainly used in the fields of smart mines and one-network unified management.

[0060] (4) The present invention provides more useful information and improves information density by deploying multiple sensors, thereby achieving more accurate prediction of tunnel deformation through microseismic coding blocks based on long-short-term memory networks;

[0061] (5) The present invention constructs a microseismic coding block based on a long short-term memory network. By encoding additional information, the data representation obtained can be considered as an equivalent representation of the damage variable of the rock material at the current moment, thereby providing a mine tunnel deformation prediction with greater advantages in calculation accuracy, speed and operability. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] The present invention will be further described below with reference to the accompanying drawings.

[0063] Attachment Figure 1 This is a flowchart of the intelligent prediction method for mine tunnel deformation;

[0064] Attachment Figure 2 This is a schematic diagram of the intelligent prediction method for mine tunnel deformation;

[0065] Attachment Figure 3 This is a schematic diagram of the microseismic encoder architecture;

[0066] Attachment Figure 4 Schematic diagram of the architecture of a multi-source heterogeneous data decoder. DETAILED DESCRIPTION

[0067] The method, device and medium for intelligent prediction of mine tunnel deformation of the present invention are described in detail below with reference to the accompanying drawings and specific embodiments.

[0068] Example 1:

[0069] As attached Figure 1 As shown, this embodiment provides a method for intelligent prediction of mine tunnel deformation, which is specifically as follows:

[0070] S1. Deploy microseismic sensors and deformation sensors around mine tunnels, and these sensors must be capable of real-time data transmission. The number of microseismic sensors must be no less than eight.

[0071] S2. transmitting the collected microseismic data and tunnel deformation data to a database according to the types and locations of the microseismic sensors and deformation sensors;

[0072] S3. Processing and feature extraction of microseismic data to obtain damage characterization of the surrounding rock of the mine tunnel; wherein damage characterization of the surrounding rock of the mine tunnel refers to parsing the microseismic data into fracture size, magnitude, energy, and fracture type;

[0073] S4. The damage representation of the mine tunnel surrounding rock represented by the microseismic signal is converted into a microseismic representation through the microseismic coding block. The microseismic representation and the tunnel deformation data are processed through the multi-source heterogeneous data decoder to predict the changes in mine tunnel deformation in the future. Among them, the microseismic representation reflects the damage level of the rock in the current time window.

[0074] As attached Figure 2 As shown, in step S3 of this embodiment, the microseismic data is processed and features are extracted to obtain damage representation of the surrounding rock of the mine tunnel as follows:

[0075] S301, microseismic information characterization: calculate the microseismic source radius r, magnitude A s and energy E s The characteristic parameters of the damage are used to identify the rupture type of the microseismic event according to the moment tensor decomposition method, and then the characteristic vector λ of the microseismic event is defined;

[0076] S302, normalizing microseismic features: the set of eigenvectors λ is Λ, Λ is normalized to obtain a normalized microseismic feature vector sequence;

[0077] S303, microseismic conversion: converting the normalized microseismic feature vector sequence into a microseismic representation through a microseismic encoding block based on a long short-term memory neural network.

[0078] The microseismic information characterization in step S301 of this embodiment is specifically as follows:

[0079] S30101. Calculate the microseismic source radius r and magnitude A respectively. s and energy E s The damage characteristic parameter is as follows:

[0080]

[0081] Where K is Brune's constant, K = 2.34; v s represents the shear wave velocity; f c Indicates the corner frequency of the acoustic emission signal. For indoor acoustic emission tests, f c is replaced by the peak frequency of the signal, A imax is the maximum amplitude of the acoustic emission signal waveform received by the i-th microseismic sensor; r i represents the distance between the acoustic emission event and the i-th microseismic sensor, in mm; k represents the number of microseismic sensors used for amplitude calculation, E i is the absolute energy of the waveform received by the i-th microseismic sensor, obtained by integrating the waveform envelope;

[0082] S30102. According to the moment tensor decomposition method, the rupture type of the microseismic event is determined using the following formula:

[0083]

[0084] Among them, M1>M2>M3, M1, M2, M3 are tensors M qp The three eigenvalues ​​represent the maximum couple; when X>60%, the earthquake source is in a shear-dominant state and is judged to be shear failure; when 60%≥X≥40%, the earthquake source is in a mixed state between shear and tension; when X<40%, the earthquake source is judged to be tensile failure;

[0085] S30103. Assume that at time t, the microseismic event received by the microseismic probe is f c , through microseismic events f c The inversion of the information of magnitude l, radius d, energy E and rupture type (such as shear or tension acoustic emission event) S, then the microseismic event f c The corresponding eigenvector λ∈R 4 Characterized as λ = (l, d, E, S).

[0086] The microseismic feature normalization in step S302 of this embodiment is specifically as follows:

[0087] S30201. Normalize the data distribution of Λ to a standard distribution with a mean of 0 and a variance of 1. The formula is as follows:

[0088]

[0089] Among them, μ represents the translation coefficient; σ represents the scaling coefficient;

[0090] S30202, introduce two new learnable variables, the re-translation coefficient b and the re-scaling coefficient g, to further The transformation formula is as follows:

[0091]

[0092] S30203. After normalization, the acoustic emission encoding process is expressed as follows:

[0093]

[0094] Here, Λ represents the input variable, and the retranslation coefficient b and the rescaling coefficient g are learned by the gradient descent method.

[0095] The microseismic coding in step S303 of this embodiment is specifically as follows:

[0096] S30301. Taking the normalized microseismic feature vector sequence as input, and using a long short-term memory neural network to convert the microseismic feature vector sequence into a local microseismic latent state sequence;

[0097] S30302. By fusion of the global microseismic hidden state, the local microseismic hidden state is converted into the global microseismic hidden state by adding and averaging.

[0098] S30303. Convert the latent state into a microseismic representation through microseismic representation conversion.

[0099] As attached Figure 3 As shown, the implementation method of the microseismic coding block based on the long short-term memory neural network in this embodiment is as follows:

[0100] (1) With is the microseismic feature tensor that has been normalized by the feature layer. When the feature vector When the previous microseismic eigenvector is processed The generated latent variable h i-1 , then in the current time step, the hidden variable h i The calculation formula is as follows:

[0101]

[0102] (2) In order to allow the network to process microseismic feature vector sequences of arbitrary length, two artificially defined vectors representing the beginning and end of the sequence are introduced, denoted as <bos>and <eos>;in, <bos>As the initial hidden state of the long short-term memory neural network layer, it is input into the model long short-term memory neural network layer together with λ1. The vector λ1 is a zero vector with the same shape as other hidden states. <eos>Indicates that there are no other microseismic feature vectors in the time window being processed. When the microseismic coding block receives <eos>As output, <eos>Together with the last hidden state h n Input the long short-term memory neural network layer and calculate the last hidden state h n+1 , <eos>The vector is a microseismic eigenvector λ with the same shape as the input i With zero vectors of the same shape, in summary, each hidden state in the microseismic coding block is calculated using the following formula:

[0103]

[0104] (3) Through continuous cycles, the long short-term memory neural network layer obtains a series of hidden states H = {h1,h2,…,h n+1 }, variables h1,h2,…,h n+1 They are regarded as microseismic representations in the local time window of the microseismic event. In order to obtain the microseismic event representation that can represent the intensity of the microseismic event at the global level, a global microseismic feature fusion is performed. The microseismic feature fusion first averages the hidden states of each independent local microseismic feature to obtain the global microseismic feature hidden state h global Then h global Input the multi-layer perceptron to obtain the microseismic global encoding vector R, the formula is as follows:

[0105]

[0106] R=MLP(h global ).

[0107] As attached Figure 4 As shown, in step S4 of this embodiment, the damage representation of the mine roadway surrounding rock represented by the microseismic signal is converted into a microseismic representation through the microseismic encoding block. The microseismic representation and the roadway deformation data are processed through the multi-source heterogeneous data decoder to predict the change of mine roadway deformation in the future period as follows:

[0108] S401: Assume that within the time window (t-Δt, t), the microseismic sensor detects microseismic events e1, e2, ..., e n At the same time, the deformation sensor monitors the continuous rock deformation data and uses the equal-interval sampling method to obtain the deformation sequence s1, s2, ..., s m , where s i ∈R;

[0109] S402, transform the sequence s1, s2, ..., s m Written in the form of a vector, it is recorded as S1=(s1,s2,…,s m ); where the subscript of the vector S represents the sequence number of the time window, and the subscript 1 indicates that the corresponding time window is the time window being monitored;

[0110] S403, once microseismic events e1, e2,…, e n The microseismic encoding block is converted into a microseismic representation R, and the deformation vectors S2, S3, ... S in the next several time windows are predicted using the microseismic representation R and the deformation vector S1. I ,…S N .

[0111] In step S403 of this embodiment, once the microseismic events e1, e2, ..., e n The microseismic encoding block is converted into a microseismic representation R, and the deformation vectors S2, S3, ... S in the next several time windows are predicted using the microseismic representation R and the deformation vector S1. I ,…S N The details are as follows:

[0112] S40301. The detected deformation vector S1 and the microseismic representation R are input into the long short-term memory neural network layer to obtain the deformation sequence of the next time window. and hidden state h1;

[0113] S40302, the long short-term memory neural network cyclically predicts the deformation vector in time window i and hidden state h i-1 And input it together with the microseismic representation R into the long short-term memory neural network to predict the deformation vector S of the i+1th time window i+1 and hidden state h i Until the long short-term memory neural network thinks that the information in the next time window cannot be predicted based on the current information and outputs the end flag <eos>.

[0114] Example 3:

[0115] This embodiment also provides an electronic device, including: a memory and at least one processor;

[0116] wherein the memory stores computer-executable instructions;

[0117] The at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the intelligent prediction method for mine tunnel deformation in any embodiment of the present invention.

[0118] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor may be a microprocessor or any conventional processor, etc.

[0119] The memory can be used to store computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area. The program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created based on the use of the terminal, etc. In addition, the memory can also include high-speed random access memory and non-volatile memory, such as a hard disk, internal memory, a plug-in hard disk, a smart memory card (SMC), a secure digital (SD) card, a flash memory card, at least one disk storage period, a flash memory device, or other volatile solid-state memory devices.

[0120] Example 4:

[0121] This embodiment further provides a computer-readable storage medium storing a plurality of instructions, which are loaded by a processor to cause the processor to execute the method for intelligently predicting mine tunnel deformation according to any embodiment of the present invention. Specifically, a system or device equipped with a storage medium can be provided, wherein the storage medium stores software program code that implements the functions of any of the above-described embodiments, and a computer (or CPU or MPU) of the system or device can read and execute the program code stored in the storage medium.

[0122] In this case, the program code itself read from the storage medium can realize the function of any one of the above-mentioned embodiments, and thus the program code and the storage medium storing the program code constitute part of the present invention.

[0123] Examples of storage media for providing program code include floppy disks, hard disks, magneto-optical disks, optical disks (e.g., CD-ROMs, CD-Rs, CD-RWs, DVD-ROMs, DVD-RYMs, DVD-RWs, DVD+RWs), magnetic tapes, non-volatile memory cards, and ROMs. Alternatively, the program code may be downloaded from a server computer via a communications network.

[0124] In addition, it should be clear that the functions of any of the above embodiments can be achieved not only by executing the program code read by the computer, but also by enabling the operating system operating on the computer to complete part or all of the actual operations based on the instructions of the program code.

[0125] In addition, it can be understood that the program code read from the storage medium is written into the memory provided in the expansion board inserted into the computer or into the memory provided in the expansion unit connected to the computer, and then based on the instructions of the program code, the CPU installed on the expansion board or expansion unit is enabled to perform part or all of the actual operations, thereby realizing the functions of any of the above embodiments.

[0126] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.< / eos> < / eos> < / eos> < / eos> < / eos> < / bos> < / eos> < / bos> < / eos> < / eos> < / eos> < / eos> < / eos> < / bos> < / eos> < / bos>

Claims

1. A method for intelligent prediction of mine tunnel deformation, characterized in that: The method is as follows: Deploy microseismic sensors and deformation sensors around mine tunnels, and these sensors and deformation sensors have the function of real-time data transmission; According to the type and location of microseismic sensors and deformation sensors, the collected microseismic data and tunnel deformation data are transmitted to the database; Processing and feature extraction of microseismic data to obtain damage characterization of the surrounding rock of the mine tunnel; damage characterization of the surrounding rock of the mine tunnel refers to parsing microseismic data into fracture size, magnitude, energy, and fracture type; The damage representation of the mine tunnel surrounding rock represented by the microseismic signal is converted into a microseismic representation through the microseismic coding block. The microseismic representation and the tunnel deformation data are processed through the multi-source heterogeneous data decoder to predict the changes in mine tunnel deformation in the future. Among them, the microseismic representation reflects the damage level of the rock in the current time window.

2. The intelligent prediction method for mine tunnel deformation according to claim 1, characterized in that: The microseismic data is processed and features are extracted to obtain damage characterization of the surrounding rock of the mine tunnel as follows: Characterization of microseismic information: Calculate the microseismic source radius r and magnitude A respectively s and energy E s The characteristic parameters of the damage are used to identify the rupture type of the microseismic event according to the moment tensor decomposition method, and then the characteristic vector λ of the microseismic event is defined; Microseismic characteristics Normalization: The set of eigenvectors λ is Λ, and Λ is normalized to obtain a normalized microseismic eigenvector sequence; Microseismic conversion: The normalized microseismic feature vector sequence is converted into microseismic representation through a microseismic encoding block based on a long short-term memory neural network.

3. The intelligent prediction method for mine tunnel deformation according to claim 2, characterized in that: The microseismic information characterization is as follows: Calculate the microseismic source radius r and magnitude A respectively s and energy E s The damage characteristic parameter is as follows: Where K is Brune's constant, K = 2.34; v s represents the shear wave velocity; f c Indicates the corner frequency of the acoustic emission signal. For indoor acoustic emission tests, f c is replaced by the peak frequency of the signal, A imax is the maximum amplitude of the acoustic emission signal waveform received by the i-th microseismic sensor; r i represents the distance between the acoustic emission event and the i-th microseismic sensor, in mm; k represents the number of microseismic sensors used for amplitude calculation, E i is the absolute energy of the waveform received by the i-th microseismic sensor, obtained by integrating the waveform envelope; According to the moment tensor decomposition method, the rupture type of the microseismic event is determined as follows: Among them, M1>M2>M3, M1, M2, M3 are tensors M qp The three eigenvalues ​​represent the maximum couple; when X>60%, the earthquake source is in a shear-dominant state and is judged to be shear failure; when 60%≥X≥40%, the earthquake source is in a mixed state between shear and tension; when X<40%, the earthquake source is judged to be tensile failure; Assume that at time t, the microseismic event received by the microseismic probe is f c , through microseismic events f c The information S of magnitude l, radius d, energy E and rupture type is inverted, then the microseismic event f c The corresponding eigenvector λ∈R 4 Characterized as λ = (l, d, E, S).

4. The intelligent prediction method for mine tunnel deformation according to claim 2, characterized in that: The normalization of microseismic characteristics is as follows: The data distribution of Λ is standardized to a standard distribution with a mean of 0 and a variance of 1. The formula is as follows: Among them, μ represents the translation coefficient; σ represents the scaling coefficient; Introducing two new learnable variables, the re-translation coefficient b and the re-scaling coefficient g, further The transformation formula is as follows: After normalization, the acoustic emission encoding process is expressed as follows: Here, Λ represents the input variable, and the retranslation coefficient b and the rescaling coefficient g are learned by the gradient descent method.

5. The intelligent prediction method for mine tunnel deformation according to claim 2, characterized in that: The microseismic conversion is as follows: The normalized microseismic feature vector sequence is used as input and the long short-term memory neural network is used to convert the microseismic feature vector sequence into a local microseismic latent state sequence. Through global microseismic hidden state fusion, the local microseismic hidden state is converted into the global microseismic hidden state by adding and averaging. Through microseismic representation conversion, the latent state is converted into microseismic representation.

6. The intelligent prediction method for mine tunnel deformation according to claim 2, characterized in that: The implementation method of the microseismic coding block based on the long short-term memory neural network is as follows: With is the microseismic feature tensor that has been normalized by the feature layer. When the feature vector When the previous microseismic eigenvector is processed The generated latent variable h i-1 , then in the current time step, the hidden variable h i The calculation formula is as follows: Two artificially defined vectors representing the beginning and end of the sequence are introduced, denoted as <bos>and <eos>;in, <bos>As the initial hidden state of the long short-term memory neural network layer, it is input into the model long short-term memory neural network layer together with λ1. The vector λ1 is a zero vector with the same shape as other hidden states. <eos>Indicates that there are no other microseismic feature vectors in the time window being processed. When the microseismic coding block receives <eos>As output, <eos>Together with the last hidden state h n Input the long short-term memory neural network layer and calculate the last hidden state h n+1 , <eos>The vector is a microseismic eigenvector λ with the same shape as the input i With zero vectors of the same shape, in summary, each hidden state in the microseismic coding block is calculated using the following formula:< / eos> < / eos> < / eos> < / eos> < / bos> < / eos> < / bos> Through continuous cycles, the long short-term memory neural network layer obtains a series of hidden states H = {h1,h2,…,h n+1 }, variables h1,h2,…,h n+1 They are regarded as microseismic representations in the local time window of the microseismic event. In order to obtain the microseismic event representation that can represent the intensity of the microseismic event at the global level, a global microseismic feature fusion is performed. The microseismic feature fusion first averages the hidden states of each independent local microseismic feature to obtain the global microseismic feature hidden state h global Then h global Input the multi-layer perceptron to obtain the microseismic global encoding vector R, the formula is as follows: R=MLP(h global )。 7. The intelligent prediction method for mine tunnel deformation according to claim 1, characterized in that: The damage representation of the mine roadway surrounding rock represented by the microseismic signal is converted into a microseismic representation through the microseismic coding block. The microseismic representation and roadway deformation data are processed through the multi-source heterogeneous data decoder to predict the changes in mine roadway deformation in the future as follows: Assume that within the time window (t-Δt,t), the microseismic sensor detects microseismic events e1, e2,…, e n At the same time, the deformation sensor monitors the continuous rock deformation data and uses the equal-interval sampling method to obtain the deformation sequence s1, s2, ..., s m , where s i ∈R; Transform the sequence s1,s2,…,s m Written in the form of a vector, it is recorded as S1=(s1,s2,…,s m ); where the subscript of the vector S represents the sequence number of the time window, and the subscript 1 indicates that the corresponding time window is the time window being monitored; Once the microseismic events e1, e2,…, e n The microseismic encoding block is converted into a microseismic representation R, and the deformation vectors S2, S3, ... S in the next several time windows are predicted using the microseismic representation R and the deformation vector S1. I ,…S N .

8. The intelligent prediction method for mine tunnel deformation according to claim 7, characterized in that: Once the microseismic events e1, e2,…, e n The microseismic encoding block is converted into a microseismic representation R, and the deformation vectors S2, S3, ... S in the next several time windows are predicted using the microseismic representation R and the deformation vector S1. I ,…S N The details are as follows: The detected deformation vector S1 and the microseismic representation R are input into the long short-term memory neural network layer to obtain the deformation sequence of the next time window and hidden state h1; The long short-term memory neural network cyclically predicts the deformation vector in time window i and hidden state h i-1 And input it together with the microseismic representation R into the long short-term memory neural network to predict the deformation vector S of the i+1th time window i+1 and hidden state h i Until the long short-term memory neural network thinks that the information in the next time window cannot be predicted based on the current information and outputs the end flag <eos> 。< / eos> 9. An electronic device, characterized in that: Includes: memory and at least one processor; Wherein, the memory stores a computer program; The at least one processor executes the computer program stored in the memory, so that the at least one processor executes the mine tunnel deformation intelligent prediction method according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which can be executed by a processor to implement the intelligent prediction method for mine tunnel deformation according to any one of claims 1 to 8.

Images