Tunnel water inrush prediction method and system based on mechanism prediction and error correction

Abstract

The present invention discloses a method and system for predicting sudden water in a tunnel based on mechanism prediction and error correction, including: real-time collection of multi-physical field information around the tunnel to determine the trend of sudden water inrush; prediction of the location of sudden water inrush disaster based on effective microseismic signals; judging whether the rock mass will be damaged under dynamic disturbance and unloading according to the geometric characteristics of cracks in the fractured rock mass and the stress state on the crack surface; obtaining a preliminary prediction probability of sudden water in the tunnel by combining the above information; using the data of the sudden water inrush trend, the location of the sudden water inrush disaster, and whether the anti-sudden water inrush rock mass is damaged, as well as the preliminary prediction probability, as inputs of the error correction model to obtain a corrected prediction probability of sudden water in the tunnel. The present invention uses a data model with a neural network as the core to perform error correction, supplements important influencing factors that cannot be reflected in the mechanism model, and makes the prediction of sudden water in the tunnel more accurate.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel water inrush prediction, and in particular to a tunnel water inrush prediction method and system based on mechanism prediction and error correction. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Disasters such as karst water and mud outbursts encountered during tunnel construction can easily cause construction delays and economic losses. Since the regional geological conditions and construction process of the project have significant uncertainties, how to reasonably predict and assess the probability of disasters and conduct effective management and prevention and control has become the key to ensuring the safe construction of underground projects.

[0004] At present, the prediction of tunnel water inrush disasters mainly focuses on analysis from the perspectives of forecast information, geological conditions, hydrological information, etc., but its accuracy and universality are relatively poor. Tunnel water inrush disasters are a complex process affected by multiple factors, and it is difficult to judge and divide them based on a single piece of information. Therefore, it is necessary to establish a judgment standard based on the comprehensive feedback of multiple information elements. It is difficult to establish an accurate tunnel water inrush prediction model only by studying the existing water inrush mechanism.

[0005] At the same time, the occurrence of sudden water burst disasters in tunnels is complex, sudden and diverse. The evolution law of the precursory multivariate information that accompanies each sudden water burst disaster is not exactly the same. The mechanism model cannot reflect all the factors that affect the sudden water burst disaster, and there is a partial error between the actual probability and the predicted probability. Summary of the invention

[0006] In order to solve the above problems, the present invention proposes a tunnel water burst prediction method and system based on mechanism prediction and error correction. Starting from the mechanism, the probability of water burst is determined, and the relevant error correction is performed through the error model to achieve accurate prediction of tunnel water burst disasters.

[0007] In some embodiments, the following technical solutions are adopted:

[0008] A tunnel water inrush prediction method based on mechanism prediction and error correction, comprising:

[0009] Collect multi-physical field information around the tunnel in real time, and determine the trend of sudden water inrush state changes based on the information;

[0010] Acquire the microseismic waveforms monitored in each water inrush channel and extract effective microseismic signals; predict the location of water inrush disaster based on the effective microseismic signals;

[0011] According to the geometric characteristics of cracks in fractured rock mass and the stress state on the crack surface, it is determined whether the rock mass will be destroyed under dynamic disturbance and unloading.

[0012] Based on the above-mentioned water inrush changing trend, water inrush disaster location and whether the anti-water inrush rock mass is damaged, the preliminary prediction probability of tunnel water inrush is obtained;

[0013] The data on the changing trend of water inrush, the location of water inrush disaster, whether the anti-water inrush rock mass is damaged, and the preliminary prediction probability are used as inputs of the error correction model to obtain a corrected tunnel water inrush prediction probability.

[0014] As a further solution, multi-physical field information around the tunnel is collected in real time, and the trend of sudden water inrush state change is determined based on the information, specifically including:

[0015] Multi-physical field information around the tunnel is collected in real time, and the multi-physical field information is processed using the principal component analysis method to obtain the correlation coefficient matrix of the multi-physical quantities. The eigenvalues ​​of the correlation coefficient matrix are used to determine the importance of each physical quantity data in the multi-physical quantities, and then the first principal component and the second principal component of the physical quantity of the sudden water inrush precursor are obtained to perform sudden water inrush pattern recognition and classification.

[0016] As a further solution, the multi-physical field information around the tunnel specifically includes: stress, displacement, osmotic pressure, temperature and apparent resistivity.

[0017] As a further solution, the microseismic waveform monitored in each sudden water surge channel is obtained, and effective microseismic signals are extracted, specifically including: filtering and denoising the microseismic waveform recorded in each sudden water surge channel by wavelet packet decomposition to achieve separation and extraction of effective microseismic signals.

[0018] As a further solution, the location of sudden water disaster is predicted based on the effective microseismic signal, specifically including:

[0019] Based on the extracted effective microseismic signals, the STA / LTA algorithm is used to pick up the first arrival time of the effective microseismic signal in each sudden water channel, and the first arrival time is substituted into the set target equation for solution to obtain the spatial position of the earthquake source.

[0020] As a further solution, based on the geometric characteristics of the cracks in the fractured rock mass and the stress state on the crack surface, it is determined whether the rock mass will be damaged under dynamic disturbance and unloading, including:

[0021] Determine the critical water pressure at which cracks undergo compression and shear failure. When the pore water pressure on the crack surface exceeds the critical water pressure, the rock mass will undergo compression and shear failure.

[0022] As a further solution, the error correction model is a deep neural network model, and the training process of the error correction model includes:

[0023] The error correction model is trained using historically acquired data series on the changing trend of water inrush status, water inrush disaster location data and water inrush channel quantity data, geometric characteristics of cracks in fractured rock mass, stress state on crack surface and data on whether the rock mass is damaged, as well as preliminary prediction data on the probability of water inrush in tunnels, as input data of the error correction model.

[0024] In other embodiments, the following technical solutions are adopted:

[0025] A tunnel water inrush prediction system based on mechanism prediction and error correction, comprising:

[0026] A water inrush state trend prediction module is used to collect multi-physical field information around the tunnel in real time and determine the change trend of the water inrush state based on the information;

[0027] A water inrush disaster location prediction module is used to obtain the microseismic waveforms monitored in each water inrush channel and extract effective microseismic signals; and predict the location of the water inrush disaster based on the effective microseismic signals;

[0028] The rock mass failure prediction module is used to determine whether the rock mass will be destroyed under dynamic disturbance and unloading according to the geometric characteristics of the cracks in the fractured rock mass and the stress state on the crack surface;

[0029] The initial probability prediction module for sudden water inrush is used to obtain the initial prediction probability of sudden water inrush in the tunnel by comprehensively considering the sudden water inrush change trend, the location of sudden water inrush disaster and whether the anti-sudden water inrush rock mass is damaged;

[0030] The probability error correction module is used to use the data on the sudden water flow change trend, the location of the sudden water flow disaster, whether the anti-sudden water flow rock mass is damaged, and the preliminary prediction probability as inputs to the error correction model to obtain a corrected tunnel sudden water flow prediction probability.

[0031] In other embodiments, the following technical solutions are adopted:

[0032] A terminal device includes a processor and a memory, wherein the processor is used to implement various instructions; the memory is used to store multiple instructions, and the instructions are suitable for the processor to load and execute the above-mentioned tunnel sudden water inrush prediction method based on mechanism prediction and error correction.

[0033] In other embodiments, the following technical solutions are adopted:

[0034] A computer-readable storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded by a processor of a terminal device and executing the above-mentioned tunnel water inrush prediction method based on mechanism prediction and error correction.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] (1) Starting from the mechanism aspect, the present invention forms a mechanism model for tunnel water inrush prediction through water inrush state trend prediction, water inrush disaster location prediction and rock mass damage prediction modules, obtains the preliminary prediction probability of tunnel water inrush, and then establishes an error correction model. Error correction is performed through a data model with a neural network as the core, which supplements the important influencing factors that cannot be reflected in the mechanism model, making the tunnel water inrush prediction more accurate.

[0037] (2) Considering that the process of sudden water inrush in a tunnel is an extremely complex process, the current mechanism model has a certain simplification process. The relationship between the influencing factors and the errors that cannot be reflected in the model is relatively complex and cannot be described by establishing a mathematical model through a certain relationship. The present invention obtains and analyzes the precursor information of sudden water inrush, combines microseismic information to locate the rock fracture position and the water inrush channel, and establishes a tunnel sudden water inrush mechanism prediction model based on the rock fracture criterion. This solves the problem that the traditional prediction method only considers multi-physical field information warning. Compared with the previous methods, it includes more influencing factors in the sudden water in the tunnel, thereby making the mechanism model closer to the actual situation.

[0038] Other features and advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a flow chart of a method for predicting sudden water inrush in a tunnel based on mechanism prediction and error correction in an embodiment of the present invention;

[0040] FIG2( a ) is a schematic diagram of the compression and shear state of the crack surface in an embodiment of the present invention;

[0041] FIG. 2( b ) is a schematic diagram of the tensile shear state of the crack surface in an embodiment of the present invention. DETAILED DESCRIPTION

[0042] It should be noted that the following detailed descriptions are illustrative and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0043] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0044] Embodiment 1

[0045] In one or more embodiments, a tunnel water inrush prediction method based on mechanism prediction and error correction is disclosed, referring to Figure 1 , specifically including the following process:

[0046] Step (1): collecting multi-physical field information around the tunnel in real time, and determining the trend of sudden water inrush state changes based on the information;

[0047] Specifically, through the deployment of a variety of sensors in the tunnel, multi-physical field information around the tunnel is collected in real time, mainly including: stress, displacement, seepage pressure, temperature and apparent resistivity.

[0048] By using the principal component analysis method to process the multi-physical parameter information obtained from monitoring, the connection between the multi-physical parameter information can be obtained; that is, there may be information overlap or even redundant information between the multiple information. Typical information parameters can be used to replace redundant information parameters to highlight the key effective information, and the typical information can also represent all the original information.

[0049] In this embodiment, the process of performing principal component analysis on multi-physical parameter information specifically includes:

[0050] ① Standardization of original monitoring data

[0051] The multi-physical parameter monitoring data form a matrix sequence: {microseismic, displacement, stress, seepage pressure, temperature, apparent resistivity}. The arrangement of the matrix row data corresponds to each other according to the time coordinate, so as to establish the sample data matrix X.

[0052] X=(x ij ) n×p

[0053] Where n is the number of samples of monitoring data, p is the number of indicators of physical parameters, i = 1, 2, ..., n, j = 1, 2, ..., p. ij Represents the jth physical parameter index value of the i-th monitoring sample. The sample element matrix is ​​standardized as follows:

[0054]

[0055] in,

[0056] Thus, the standardized sample data matrix is ​​obtained:

[0057] Z=(z ij ) n×p i=1,2,…,nj=1,2,…,p.

[0058] ②Correlation coefficient matrix of multiple physical quantities

[0059] The covariance matrix is ​​the correlation coefficient matrix R, where k = 1, 2, ..., p. jk is the correlation coefficient between indicator parameter j and indicator parameter k.

[0060] R=(r jk ) p×p

[0061]

[0062] or

[0063] r ij =E{[z i -E(z i )][z j -E(z j )]}

[0064] Right now:

[0065]

[0066] have:

[0067] r ij =1

[0068] r jk =r kj

[0069] The correlation coefficient matrix value of multi-physical quantities is between -1 and 1, that is, -1≤r≤1. Its properties are

[0070] When r>0, it means that the two physical quantities are positively correlated, and when r<0, the two physical quantities are negatively correlated.

[0071] When |r|=1, it means that the two physical quantities are completely linearly correlated, that is, a functional relationship.

[0072] When r = 0, it means there is no linear correlation between the two physical quantities. When 0 < |r| < 1, it means there is a certain linear relationship between the two physical quantities.

[0073] The relationship between physical quantities is usually divided according to the value range of |r|: |r|<0.4 is a low linear correlation; 0.4≤|r|<0.7 is a significant linear correlation; 0.7≤|r|<1 is a high linear correlation.

[0074] ③Eigenvalues ​​of the correlation coefficient matrix

[0075] Through the characteristic equation |R-λE|=0, P non-negative characteristic roots λ are calculated g (g=1,2,…,p), according to λ g The order of magnitude (λ1≥λ2≥λ p ≥0) to obtain the eigenvector I g (I g1 ,I g2 ,…,I gp ), and the processed indicator variable is transformed into: F g =I g1 Z1+I g2 Z2+…+I gp Z p , where F p is the pth principal component.

[0076] ④ Determine the importance of each physical quantity data in multiple physical quantities.

[0077] The characteristic root λ obtained from the characteristic equation g , according to its size, it can characterize the contribution of each physical parameter to the monitoring target. When the sum of the characteristic root variances corresponding to some physical parameters is ≥85%, it means that these physical parameters can characterize the main characteristics of all physical parameters. The calculation formula of the cumulative variance is as follows:

[0078]

[0079]

[0080] After the principal component analysis is completed, when the characteristic variance sum contribution rate of the physical quantity is greater than 85%, that is, W ≥ 85%, it indicates that the physical quantity is used to reflect the meaning represented by all physical quantities. At the same time, the first principal component and the second principal component are divided according to the size of the characteristic root.

[0081] In this embodiment, different water inrush types correspond to different first principal components and second principal components, and a tunnel water inrush pattern matching model is established based on the difference in the order of the principal components. The model is divided into two levels. The first level uses the principal component analysis method to determine the information of the physical quantity indicators according to the principle that the cumulative variance contribution rate is greater than 85%, and these physical quantity indicators are divided into primary or secondary information; the second level uses the factor analysis method to weight the principal components with their respective variance contribution rates, and completes the final principal component evaluation through linear weighted summation to achieve different water inrush pattern recognition and classification. Table 1 gives a list of the relationships between different water inrush types and their precursor information.

[0082] Table 1 Water inrush types and their precursor information

[0083]

[0084] Step (2): obtaining the microseismic waveform monitored in each water inrush channel and extracting effective microseismic signals; predicting the location of the water inrush disaster based on the effective microseismic signals;

[0085] In this embodiment, the microseismic waveform recorded in each water surge channel is filtered and denoised by wavelet packet decomposition, thereby achieving separation and extraction of effective microseismic signals.

[0086] The STA / LTA algorithm is used to pick up the first arrival time of the signal in a single channel, and the first arrival time is brought into the target equation for solution, and the spatial position of the earthquake source is solved using the Newton iteration method. By monitoring and counting microseismic events, the rupture position of the anti-burst rock mass and the water inrush channel can be accurately judged, thereby realizing the prediction of the location of the sudden water disaster.

[0087] In this embodiment, in the process of processing microseismic signals using the STA / LTA method, two sliding windows, a long time window and a short time window, are set respectively. STA and LTA refer to the average amplitude or energy of the signal in the short time window and the long time window, respectively. LTA contains more sampling data that can reflect the average state of the signal in a longer period of time, and LTA mainly reflects the change law of background noise. The amount of data contained in STA is less than that of LTA, and it is used to characterize the transient changes of the signal. When a microseismic event does not occur, the background noise is dominant, and the ratio of STA to LTA does not change much; when a microseismic event is about to occur, the transformed waveform curve shows a phenomenon of instantaneous increase in amplitude. Finally, the purpose of microseismic event identification and initial arrival picking is achieved by setting a threshold to count the "spikes" with larger signal amplitudes.

[0088] In this embodiment, the target equation is specifically:

[0089]

[0090] Among them, (x, y, z) is the coordinate of the earthquake source point P, (x i,y i ,z i )(i=1,2,3……) is the sensor M i (i=1,2,3……) spatial coordinates. Sensor M1 is the reference sensor, and R1 is the distance from the source P to the sensor M1. i The difference between the distance from M1 to P and the distance from M1 to P is d1. 1i =d i -d1. The travel time difference between the sensors is τ i1 , sensor M i Distance difference d from sensor M1 1i =τ i1 C, C is the propagation speed of stress wave in rock mass. Substituting the spatial coordinates of known sensors, stress wave velocity and travel time difference data between sensors into the target equation group, the spatial coordinates (x, y, z) of microseismic source P can be obtained.

[0091] By monitoring and counting microseismic events, the location of rock mass rupture and water inrush channels can be accurately determined, thereby predicting the location of water inrush disasters.

[0092] Step (3): Based on the geometric characteristics of the cracks in the fractured rock mass and the stress state on the crack surface, determine whether the rock mass will be damaged under dynamic disturbance and unloading;

[0093] Specifically, the stress state on the crack surface includes: ground stress, water pressure, dynamic load, angle between crack and maximum principal stress, etc. The compression shear state and tension shear state of crack surface are shown in Figure 2(a) and Figure 2(b), respectively. Among them, the length of the water-bearing crack is 2a, and it is affected by the near-zone ground stress σ1 and σ3. The angle between the crack and the ground stress σ1 is α. The crack is affected by the pore water pressure Pw, and the crack is also affected by the stress wave P wave (or SV wave).

[0094] In this embodiment, based on the rock mass compression and shear failure criterion, it is determined whether the anti-burst rock mass will be destroyed under the action of dynamic disturbance and unloading, thereby inducing the occurrence of sudden water disasters.

[0095] ① When the influence of dynamic wave is not considered, the critical water pressure P for the crack to undergo compression and shear failure is c :

[0096]

[0097] Where: is the internal friction angle on the crack surface, K IIC It is the value of mode II fracture toughness of rock in compression state.

[0098] ② When considering the influence of dynamic waves, the critical water pressure P for the crack to undergo compression and shear failure is c :

[0099]

[0100] Where: is the dynamic stress intensity factor; τ is the shear stress; c is the cohesion of the crack surface.

[0101] When the pore water pressure P w Exceeding the calculated value P c When the rock mass is subjected to compression and shear failure.

[0102] Step (4): Based on the above-mentioned water inrush change trend, water inrush disaster location, and whether the anti-water inrush rock mass is damaged, a preliminary prediction probability of tunnel water inrush is obtained;

[0103] In this embodiment, the preliminary prediction probability of tunnel water inrush is a combination of the three data prediction results obtained above. The preliminary prediction results can be obtained through human experience or directly through artificial intelligence means, which can be selected by those skilled in the art according to actual needs.

[0104] Step (5): using the data on the changing trend of the sudden water flow, the location of the sudden water flow disaster, whether the anti-sudden water flow rock mass is damaged, and the preliminary prediction probability as inputs to the error correction model to obtain a corrected tunnel sudden water flow prediction probability.

[0105] In this embodiment, in order to effectively compensate for the errors predicted by the sudden water burst mechanism model, machine learning is used as the core, and an error model is established for the errors generated by the mechanism prediction through a large amount of past tunnel monitoring data and sudden water burst cases. That is, the changing law of the multi-physical field information of tunnel sudden water burst, the strength of the anti-sudden water rock mass and microseismic events, and the probability of sudden water burst predicted by the mechanism model are used as the input layer, and the data model parameters are optimized using database sample data. The corrected probability of sudden water burst is used as the output layer, and a nonlinear mapping model based on a neural network is established as a prediction error correction model for sudden water burst in tunnels.

[0106] In this embodiment, firstly, the tunnel water inrush samples in the previous database are classified into: training samples, verification samples, and test samples. The neural network model includes an input layer, a hidden layer, and an output layer. The number of units in the input layer depends on the number of features, including the judgment results of the mechanism model, the change law of the stress field, the displacement field, and the seepage field, as well as the number of microseismic events and the strength of the anti-water inrush rock mass. The number of units in the output layer depends on the number of categories of the results in the training set, which mainly refers to the probability of water inrush prediction.

[0107] First, the parameters are randomly initialized to the minimum value close to zero, and then forward propagation is performed. i , we can calculate y(xi ), the result is in the form of a vector, and then the cost function J(θ) is written and calculated. The partial derivative of J(θ) with respect to θ is calculated by the back propagation algorithm. The value of J(θ) is further minimized by combining the gradient descent method or the conjugate gradient method with the back propagation algorithm to obtain a local optimal solution. Through the above-mentioned nonlinear mapping model based on neural network, the prediction data of the mechanism model is further optimized and corrected, so as to obtain a more accurate prediction result of tunnel sudden water disaster.

[0108] In order to more accurately predict the probability of sudden water inrush in tunnels, this embodiment combines the prediction of the sudden water inrush mechanism with the error correction model, establishes a tunnel sudden water inrush prediction method driven by both the mechanism prediction and the error model, and combines the output obtained from the mechanism prediction with the error prediction output of the data model to obtain a more accurate prediction output result, thereby realizing the status tracking and prediction and early warning of sudden water in tunnels, and providing guidance for safe production and construction of tunnels.

[0109] Embodiment 2

[0110] In one or more embodiments, a tunnel water inrush prediction system based on mechanism prediction and error correction is disclosed, comprising:

[0111] A water inrush state trend prediction module is used to collect multi-physical field information around the tunnel in real time and determine the change trend of the water inrush state based on the information;

[0112] A water inrush disaster location prediction module is used to obtain the microseismic waveforms monitored in each water inrush channel and extract effective microseismic signals; and predict the location of the water inrush disaster based on the effective microseismic signals;

[0113] The rock mass failure prediction module is used to determine whether the rock mass will be destroyed under dynamic disturbance and unloading according to the geometric characteristics of the cracks in the fractured rock mass and the stress state on the crack surface;

[0114] The initial probability prediction module for sudden water inrush is used to obtain the initial prediction probability of sudden water inrush in the tunnel by comprehensively considering the sudden water inrush change trend, the location of sudden water inrush disaster and whether the anti-sudden water inrush rock mass is damaged;

[0115] The probability error correction module is used to use the data on the sudden water flow change trend, the location of the sudden water flow disaster, whether the anti-sudden water flow rock mass is damaged, and the preliminary prediction probability as inputs to the error correction model to obtain a corrected tunnel sudden water flow prediction probability.

[0116] It should be noted that the specific implementation methods of the above modules have been described in Example 1 and will not be described in detail here.

[0117] Embodiment 3

[0118] In one or more embodiments, a terminal device is disclosed, including a server, the server including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the tunnel water inrush prediction method based on mechanism prediction and error correction in Example 1 when executing the program. For the sake of brevity, it will not be described here.

[0119] It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general-purpose processors, digital signal processors DSP, application-specific integrated circuits ASIC, off-the-shelf programmable gate arrays FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0120] The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0121] In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in a processor or an instruction in the form of software.

[0122] Embodiment 4

[0123] In one or more embodiments, a computer-readable storage medium is disclosed, in which a plurality of instructions are stored, wherein the instructions are suitable for being loaded by a processor of a terminal device and executing the tunnel water inrush prediction method based on mechanism prediction and error correction described in Example 1.

[0124] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A tunnel water inrush prediction method based on mechanism prediction and error correction, characterized in that: include: Collect multi-physical field information around the tunnel in real time, and determine the trend of sudden water inrush state changes based on the information; The real-time collection of multi-physical field information around the tunnel and determination of the change trend of the sudden water inrush state based on the information specifically include: Collect multi-physical field information around the tunnel in real time, process the multi-physical field information using the principal component analysis method, obtain the correlation coefficient matrix of the multi-physical quantities, determine the importance of each physical quantity data in the multi-physical quantities using the eigenvalue of the correlation coefficient matrix, and then obtain the first principal component and the second principal component of the physical quantity of the sudden water inrush precursor, and perform sudden water inrush pattern recognition and classification; Acquire the microseismic waveform monitored in each water burst channel and extract the effective microseismic signal; the acquiring the microseismic waveform monitored in each water burst channel and extracting the effective microseismic signal specifically includes: filtering and denoising the microseismic waveform recorded in each water burst channel by wavelet packet decomposition to achieve separation and extraction of effective microseismic signals; Predicting the location of sudden water disaster based on the effective microseismic signal; Predicting the location of sudden water disaster based on the effective microseismic signal specifically includes: Based on the extracted effective microseismic signals, the STA / LTA algorithm is used to pick up the first arrival time of the effective microseismic signal in each sudden water channel, and the first arrival time is brought into the set target equation for solution to obtain the spatial position of the earthquake source; the target equation is specifically: Among them, (x, y, z) is the coordinate of the earthquake source point P, (x i ,y i ,z i )(i=1,2,3……) is the sensor M i (i=1,2,3……) spatial coordinates; sensor M1 is the reference sensor, R1 is the distance from the source P to the sensor M1; sensor M i The difference between the distance from M1 to P and the distance from M1 to P is d1. 1i =d i -d1; the travel time difference between sensors is τ i1 , sensor M i Distance difference d from sensor M1 1i =τ i1 C, C is the propagation speed of stress wave in rock mass; Substituting the spatial coordinates of known sensors, stress wave velocity and travel time difference data between sensors into the target equation group, the spatial coordinates (x, y, z) of microseismic source P can be obtained; According to the geometric characteristics of the cracks in the fractured rock mass and the stress state on the crack surface, it is determined whether the rock mass will be destroyed under the dynamic disturbance and unloading action; According to the geometric characteristics of the cracks in the fractured rock mass and the stress state on the crack surface, it is determined whether the rock mass will be destroyed under the dynamic disturbance and unloading action, specifically including: Determine the critical water pressure at which cracks fail in compression and shear. When the pore water pressure on the crack surface exceeds the critical water pressure, the rock mass will fail in compression and shear. Based on the above-mentioned water inrush state change trend, water inrush disaster location and whether the anti-water inrush rock mass is damaged, the preliminary prediction probability of tunnel water inrush is obtained; The data on the changing trend of the water inrush state, the location of the water inrush disaster, whether the anti-water inrush rock mass is damaged, and the preliminary predicted probability are used as inputs of the error correction model to obtain a corrected prediction probability of tunnel water inrush; the error correction model is specifically as follows: the changing law of the multi-physical field information of tunnel water inrush, the strength of the anti-water inrush rock mass and microseismic events, and the probability of tunnel water inrush predicted by the mechanism model are used as input layers, the data model parameters are optimized using database sample data, the corrected probability of water inrush is used as the output layer, and a nonlinear mapping model based on a neural network is established as a prediction error correction model for tunnel water inrush; The error correction model is a deep neural network model, and the training process of the error correction model includes: The error correction model is trained by using the historically acquired data series of water inrush state change trend, water inrush disaster location data and water inrush channel quantity data, the geometric characteristics of cracks in the fractured rock mass, the stress state on the crack surface and whether the rock mass is damaged, and the preliminary prediction data of the probability of water inrush in the tunnel as the input data of the error correction model. Specifically, the parameters are first randomly initialized to the minimum value close to zero, and then forward propagation is performed. For the input x i , we can calculate y(x i ), the result is in the form of a vector, and then the cost function J(θ) is calculated. The partial derivative of J(θ) with respect to θ is calculated by the back-propagation algorithm, and the value of J(θ) is further minimized by combining the gradient descent method or the conjugate gradient method with the back-propagation algorithm to obtain the local optimal solution.

2. A method for predicting sudden water inrush in a tunnel based on mechanism prediction and error correction as claimed in claim 1, characterized in that: The multi-physical field information around the tunnel specifically includes: stress, displacement, seepage pressure, temperature and apparent resistivity.

3. A tunnel water inrush prediction system based on mechanism prediction and error correction, using the prediction method according to any one of claims 1-2, characterized in that: include: A water inrush state trend prediction module is used to collect multi-physical field information around the tunnel in real time and determine the change trend of the water inrush state based on the information; A water inrush disaster location prediction module is used to obtain the microseismic waveforms monitored in each water inrush channel and extract effective microseismic signals; and predict the location of the water inrush disaster based on the effective microseismic signals; The rock mass failure prediction module is used to determine whether the rock mass will be destroyed under dynamic disturbance and unloading according to the geometric characteristics of the cracks in the fractured rock mass and the stress state on the crack surface; The initial probability prediction module for sudden water inrush is used to obtain the initial prediction probability of sudden water inrush in the tunnel by comprehensively considering the above-mentioned sudden water inrush state change trend, sudden water inrush disaster location and whether the anti-sudden water inrush rock mass has been damaged; The probability error correction module is used to use the data on the change trend of the sudden water state, the location of the sudden water disaster, whether the anti-sudden water rock mass is damaged, and the preliminary predicted probability as inputs to the error correction model to obtain a corrected tunnel sudden water prediction probability.

4. A terminal device, comprising a processor and a memory, wherein the processor is used to implement each instruction; and the memory is used to store multiple instructions, characterized in that: The instructions are suitable for being loaded by a processor and executing the tunnel water inrush prediction method based on mechanism prediction and error correction as described in any one of claims 1-2.

5. A computer-readable storage medium storing a plurality of instructions, characterized in that: The instructions are suitable for being loaded by a processor of a terminal device and executing the tunnel water inrush prediction method based on mechanism prediction and error correction as described in any one of claims 1-2.

Images