Coal rock fracture field inversion method and system capable of explaining and selecting state space architecture

Abstract

The invention discloses a coal rock fracture field inversion method and system capable of explaining and selecting a state space framework. The inversion method comprises the following steps of obtaining staged time CT detection data and full space-time acoustic emission monitoring data in a coal rock sample loading process; performing correspondence on the CT detection data at the stage moments and the whole space-time acoustic emission monitoring data according to the moments to obtain a fracture parameter data set; performing feature extraction on full-space fracture parameter data and acoustic emission monitoring data of a training set and a verification set in the fracture parameter data set based on a neural network of a selective state space architecture to obtain an inversion model; all fracture parameters in a full space-time fracture field in the loading process are obtained through full space-time acoustic emission monitoring data inversion; and inputting scalar parameter data of the test set, the inversion model and an inversion result into an SHAP interpreter to obtain an influence factor between the acoustic emission scalar parameter and the fracture parameter inversion result, thereby obtaining a fracture parameter real-time quantitative inversion mechanism model.

Description

Technical Field

[0001] This invention relates to the field of coal and rock monitoring technology, and in particular to a method and system for inverting coal and rock fracture fields with an interpretable selectable state space architecture. Background Technology

[0002] During coal mining, mining activities and other influencing factors disrupt the original stress equilibrium state of coal and rock samples. This process typically involves rapid and unstable crack propagation, which is the root cause of dynamic disasters such as rockbursts and coal and gas outbursts. Therefore, in order to gain a deeper understanding of the mechanisms of coal and rock dynamic disasters and to achieve early warning of these disasters, it is necessary to conduct in-depth research on the distribution and evolution characteristics of fracture fields.

[0003] Although existing rock mechanics theories and methods attempt to explain the evolution of coal and rock fracture fields caused by mining activities, this complex process remains a "black box" that is difficult to measure and reveal. Many challenges remain in real-time monitoring of fracture fields, accurate quantitative inversion, exploring their formation mechanisms, and revealing their development and evolution. Currently, industrial CT detection and numerical simulation methods have made some progress. These technologies can be used to reconstruct the internal structure of coal and rock samples, especially showing potential in constructing three-dimensional digital models of discontinuous structures and fracture fields. For example, patent CN108072467A discloses a method for measuring the internal stress field of discontinuous structural samples; patent CN114187423A discloses a method, electronic device, and storage medium for reconstructing surrounding rock fractures in three-dimensional simulation experiments; and patent CN114169182A discloses a method and device for reconstructing an ellipsoidal model of fractures on the surface of rock samples. Nevertheless, these methods still have some limitations. They typically rely on high-specification laboratory conditions and cumbersome model reconstruction methods, which limits their widespread application in practical engineering. Furthermore, these methods cannot provide real-time dynamic spatiotemporal evolution characteristics of fracture field distribution, which may pose certain limitations in actual production processes.

[0004] In recent years, some experimental methods combining exploration-monitoring technology and numerical simulation have been applied in laboratories. However, the current utilization of these exploration-monitoring data mainly focuses on manual analysis of their temporal variation patterns. Furthermore, relying solely on numerical simulation for fracture field inversion requires extremely high computational costs; even minor modifications to the model can potentially have significant negative impacts on the calculation process. Therefore, how to utilize exploration-monitoring data to achieve full-time and spatiotemporal fracture field inversion and overcome the limitations of traditional methods still requires further in-depth research. With the continuous advancement of science and technology, breakthroughs in neural network methods have attracted attention for their application in fracture field inversion. In particular, the use of neural networks based on selective state-space architecture and SHAP interpreters can effectively extract features from different types of data, significantly improving computational efficiency and reducing computational resource requirements compared to traditional neural network architectures. This provides the possibility of using neural network methods to achieve accurate inversion, real-time reproduction, and mechanism interpretation of loaded coal and rock fracture parameters, but currently, feasible approaches and implementation methods with specific samples are still lacking. Therefore, it is urgent to conduct actual measurements of detection and monitoring signals during the loading process of coal and rock samples, propose a coal and rock fracture field inversion method with an interpretable selective state space architecture, realize full-time and space-time fracture field inversion based on full-time and space-time acoustic emission monitoring data, and establish a real-time quantitative inversion mechanism model for fracture parameters based on acoustic emission data. This will have a significant promoting effect on revealing the "black box" evolution process of coal and rock dynamic disasters and realizing the monitoring and early warning of coal and rock dynamic disasters. Summary of the Invention

[0005] This solution addresses the problems and needs raised above by proposing an interpretable, selectable state-space architecture-based method and system for inverting coal and rock fracture fields. The method achieves the aforementioned technical objectives and brings about several other technical benefits by adopting the following technical features.

[0006] One object of this invention is to propose a coal and rock fracture field inversion method with an interpretable selected state-space architecture, comprising the following steps:

[0007] S10: CT detection sensors and acoustic emission monitoring sensors are arranged on the coal and rock sample, and the coal and rock sample is placed in the testing machine for loading test. CT detection data and full-time acoustic emission monitoring data are collected at different time points during the loading process of the coal and rock sample through the CT detection sensors and acoustic emission monitoring sensors. Among them, the CT detection data are the full-space fracture parameter data at different time points, and the acoustic emission monitoring data includes waveform data and scalar parameter data.

[0008] S20: The CT detection data and all-time acoustic emission monitoring data at different moments during the loading process of the coal and rock samples are correlated according to time to obtain the fracture parameter dataset, and the fracture parameter dataset is divided into training set, validation set and test set;

[0009] S30: Based on a selective state-space architecture, a neural network extracts features from the phased time-space fracture parameter data, corresponding time-waveform data, and corresponding time-scalar parameter data of the fracture parameter dataset in the training and validation sets, to obtain a real-time quantitative inversion model of fracture parameters; the real-time quantitative inversion model of fracture parameters inverts each fracture parameter in the full-time-space fracture field during loading by using full-time-space acoustic emission monitoring data, thus obtaining the evolution process of the full-time-space fracture field parameters of the coal and rock sample under loading;

[0010] S40: Input the spatiotemporal acoustic emission scalar parameter data of the fracture parameter dataset into the SHAP interpreter, along with the spatiotemporal fracture parameter inversion results obtained from the real-time quantitative inversion model of fracture parameters, and the real-time quantitative inversion model of fracture parameters. This will yield the influence factor between the acoustic emission scalar parameters and the fracture parameter inversion results, and ultimately obtain the real-time quantitative inversion mechanism model of fracture parameters based on the acoustic emission scalar parameters.

[0011] Furthermore, the coal and rock fracture field inversion method with an interpretable selected state space architecture according to the present invention may also have the following technical features:

[0012] In one example of the present invention, in step S30, the feature extraction of the phased time full-space fracture parameter data, corresponding time waveform data, and corresponding time scalar parameter data of the fracture parameter dataset based on the selective state-space architecture neural network includes the following steps:

[0013] S31: The acoustic emission scalar parameters are embedded into the waveform data through the data coding layer to obtain parameter-enhanced acoustic emission data, and the parameter-enhanced acoustic emission features AEf are extracted. se ;

[0014] S32: The encoded data is concatenated to the feature extraction layer B through the residual connection layer to obtain the feature AE. b middle;

[0015] S33: Enhance acoustic emission characteristics AEf by using the first Normalization function. se Regularization is performed to obtain the feature AE n ;

[0016] S34: Regularized Feature AE n Feature extraction AE is obtained through the residual connection formed by the second linear feature extraction layer A and the second activation function layer. a+s The feature AE spliced ​​into the output of the selective state space architecture sssm Get AE SSSM ;

[0017] S35: Feature AE output by the first linear feature extraction layer A a Higher-dimensional features (AE) are obtained through feature extraction via convolutional layers and a first activation function layer. conv+s ;

[0018] S36: Gated computation of high-dimensional features is performed using a selective state-space architecture, and then concatenated with the features extracted by the convolutional layer and the first activation function layer to obtain feature AE. SSSM ;

[0019] S37: Linear feature extraction layer B extracts features from the concatenated features and concatenates the extracted features with the features of the encoded data obtained by the residual connection layer to obtain feature AEf. b+se ;

[0020] S38: Regularize the concatenated features using the second Normalization function to obtain feature AEf. n ;

[0021] S39: The features extracted by the fully connected layer are used for calculation to obtain the crack parameter F. N .

[0022] In one example of the present invention, in step S36, the selective state-space architecture sample is implemented as follows:

[0023] First, a state-space architecture is constructed using continuous state-space equations, the expressions of which are as follows:

[0024] h t =Ah t-1 +Bx t

[0025] y t =Ch t

[0026] Among them, h st-1 Represents the system state; h t Represents the updated system state; matrix A represents the state transition matrix; matrix B represents the input gating matrix; matrix C represents the output mapping matrix; x t The data represents the current state of the input architecture.

[0027] Secondly, the state-space architecture is subjected to discrete data adaptation using discrete state-space equations; the equation for the state transition matrix A is as follows:

[0028]

[0029] In the formula, n is the row index related to the orthogonal basis functions in the polynomial space, and k is the column index related to the orthogonal basis functions in the polynomial space.

[0030] In one example of the present invention, the method further includes: randomly initializing the state transition matrix A and using a low-rank matrix obtained from matrix decomposition, the equation of which is as follows:

[0031]

[0032] Where P and Q are both vectors of length N; V represents the eigenvector matrix; V * Λ represents the transpose matrix; Λ represents the eigenvalue matrix; * represents the projection operation.

[0033] In one example of the present invention, the expression of the discrete state-space equation is as follows:

[0034]

[0035] In the formula, H k and Y k These represent the memory control unit and the output control unit, respectively, for selecting the current state in the state-space architecture; h k-1 h0 represents the memory control unit that selects the previous state in the state space architecture; h0 represents the initialization state of the memory control unit; x k x0 represents the data of the current state input architecture; x0 represents the data of the initial state input architecture. All are matrices after discrete adaptation of A, B, and C.

[0036] In one example of the invention, the matrix It is obtained from the following derivation formula:

[0037]

[0038] In the formula, Δ represents the discretized nonlinear function, which discretizes the continuous state; Ι represents the identity matrix.

[0039] In one example of the present invention, in step S40, the expression formula of the SHAP interpreter is as follows:

[0040]

[0041] in, The value represents the influence factor; M represents the number of input acoustic emission scalar parameters. This represents the set of all permutations and combinations of M features. f represents the set of all feature combinations preceding feature k in permutation R. x (S)=E(f(X)|Xs =x S ) represents the conditional expectation given a known feature subset S; i represents the index of the influence factor.

[0042] In one example of this invention, the influence factor, acoustic emission scalar parameter, and fracture parameter obtained through the SHAP interpreter constitute a mechanism model for real-time quantitative inversion of fracture parameters, and its expression equation is as follows:

[0043]

[0044] Where MM represents the mechanism model for real-time quantitative inversion of fracture parameters; F N Represents the fracture parameter; AE si This represents the scalar parameter of acoustic emission.

[0045] Another object of the present invention is to provide an interpretable, selectable state-space architecture-based fracture field inversion system for coal and rock samples, comprising:

[0046] The information acquisition module is configured to deploy CT detection sensors and acoustic emission monitoring sensors on coal and rock samples, and place the coal and rock samples in a testing machine for loading tests. The CT detection sensors and acoustic emission monitoring sensors acquire CT detection data and full-time acoustic emission monitoring data at different stages during the loading process of the coal and rock samples. Among them, the CT detection data is the full-space fracture parameter data at different stages, and the acoustic emission monitoring data includes waveform data and scalar parameter data.

[0047] The data partitioning module is configured to map the CT detection data and all-time acoustic emission monitoring data at different stages during the loading process of coal and rock samples according to time, obtain the fracture parameter dataset, and divide the fracture parameter dataset into training set, validation set and test set;

[0048] The full-time-space fracture parameter quantitative inversion module is configured to extract features from the phased full-space fracture parameter data, corresponding waveform data, and corresponding scalar parameter data of the fracture parameter dataset based on a selective state-space architecture neural network, thereby obtaining a real-time quantitative inversion model of fracture parameters. The real-time quantitative inversion model of fracture parameters inverts each fracture parameter in the full-time-space fracture field during loading by using full-time-space acoustic emission monitoring data, thus obtaining the evolution process of the full-time-space fracture field parameters of the coal and rock sample under loading.

[0049] The inversion mechanism model module is configured to input the spatiotemporal acoustic emission scalar parameter data of the fracture parameter dataset into the SHAP interpreter, the spatiotemporal fracture parameter inversion results obtained from the real-time quantitative inversion model of fracture parameters, and the real-time quantitative inversion model of fracture parameters, to obtain the influence factor between the acoustic emission scalar parameters and the fracture parameter inversion results, and then obtain the real-time quantitative inversion mechanism model of fracture parameters based on the acoustic emission scalar parameters.

[0050] In one example of the present invention, the all-temporal-space fracture parameter quantitative inversion module includes: a data encoding layer, a first normalization function layer, a linear feature extraction layer A, a convolutional layer, a first activation function layer, a selective state space architecture, a linear feature extraction layer B, a second normalization function layer, and a fully connected layer, connected in series; wherein, the output of the data encoding layer and the output of the linear feature extraction layer B are connected through a residual connection layer, and the output of the first normalization function layer and the output of the selective state space architecture are connected in series with the second linear feature extraction layer A and the second activation function layer;

[0051] The data encoding layer is configured to embed acoustic emission scalar parameters into waveform data to obtain parameter-enhanced acoustic emission data, and to extract parameter-enhanced acoustic emission features AEf. se ;

[0052] The residual connection layer is configured to concatenate the encoded data to the feature extraction layer B to obtain feature AE. b middle;

[0053] The first Normalization function layer is configured to enhance the acoustic emission feature AEf of the parameters. se Regularization is performed to obtain the feature AE n ;

[0054] The residual connection consisting of the second linear feature extraction layer A and the second activation function is configured to perform feature extraction to obtain AE. a+s The feature AE spliced ​​into the output of the selective state space architecture sssm Get AE SSSM ;

[0055] The convolutional layer and the first activation function layer are configured to extract the feature AE output by the first linear feature extraction layer A. a Feature extraction is performed to obtain higher-dimensional features (AE). conv+s ;

[0056] The selective state-space architecture is configured to perform gated computation on high-dimensional features and concatenate them with the features extracted by the convolutional layer and the first activation function layer to obtain feature AE. SSSM ;

[0057] The linear feature extraction layer B is configured to extract features from the concatenated features, and then concatenate the extracted features with the features of the encoded data obtained by the residual connection layer to obtain feature AEf. b+se ;

[0058] The second Normalization function layer is configured to regularize the concatenated features to obtain feature AEf. n ;

[0059] The fully connected layer is configured to perform calculations on the extracted features to obtain the fracture parameter F. N .

[0060] Compared with the prior art, the present invention has the following beneficial effects:

[0061] 1. This invention focuses on using neural network methods to invert the spatiotemporal fracture field of loaded coal and rock samples through probe-monitoring information from loading tests. It combines a neural network based on a selective state-space architecture with a SHAP interpreter to achieve full spatiotemporal reproduction of the fracture field of loaded coal and rock samples, revealing a real-time quantitative inversion mechanism model for fracture parameters. Compared to traditional numerical simulation and neural network methods, this spatiotemporal fracture field inversion method based on interpretable deep learning has higher computational efficiency and accuracy, effectively overcoming the limitations of traditional methods.

[0062] 2. The real-time quantitative inversion model for fracture parameters in this invention is obtained by extracting spatiotemporal features from the data in the fracture parameter dataset using a neural network based on a selective state-space architecture. A feature extraction method for fracture parameter data and acoustic emission data during the loading process of coal and rock samples is proposed. Based on the proposed spatiotemporal feature extraction method, full-spatiotemporal fracture parameters can be inverted using full-spatiotemporal acoustic emission data obtained in the laboratory during the loading process of coal and rock samples.

[0063] 3. The model inversion mechanism in this invention extracts features from the acoustic emission scalar parameter data and the inversion results of fracture parameters in all time and space using an interpreter, thereby obtaining the influencing factors in the mechanism model. The interpreter can obtain a real-time quantitative inversion mechanism model of fracture parameters during the loading process of coal and rock samples through acoustic emission data and model prediction results, revealing the black box process of fracture parameter inversion during the loading process of coal and rock samples.

[0064] The preferred embodiments of the invention will be described in more detail below with reference to the accompanying drawings, so as to facilitate an understanding of the features and advantages of the invention. Attached Figure Description

[0065] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. The drawings are merely illustrative of some embodiments of the present invention and are not intended to limit the scope of the present invention to all embodiments.

[0066] Figure 1 This is a flowchart of a coal and rock sample fracture field inversion method based on an interpretable selective state space architecture according to an embodiment of the present invention;

[0067] Figure 2 This is a schematic diagram of the arrangement of coal and rock samples in the testing machine according to an embodiment of the present invention;

[0068] Figure 3 This is a schematic diagram of the structure of a real-time quantitative inversion model for fracture parameters according to an embodiment of the present invention.

[0069] List of reference numerals in the attached diagram:

[0070] Testing machine 100;

[0071] Loading device 110;

[0072] Acoustic emission sensor 120;

[0073] Coupler 130;

[0074] Preamplifier 140;

[0075] X-ray detector 150;

[0076] X-ray source 160;

[0077] 200 coal and rock samples;

[0078] Lateral direction X;

[0079] The vertical direction is Y.

[0080] Sample Implementation Method

[0081] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. The same reference numerals in the drawings represent the same parts. It should be noted that the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0082] Unless otherwise defined, the technical or scientific terms used herein shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms “first,” “second,” and similar terms used in this patent application specification and claims do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, “an” or “a” and similar terms do not necessarily indicate a quantity limitation. Terms such as “comprising” or “including” mean that the element or object preceding the word encompasses the element or object listed following the word and its equivalents, without excluding other elements or objects. Terms such as “connected” or “linked” are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as “upper,” “lower,” “left,” and “right” are used only to indicate relative positional relationships; these relative positional relationships may change accordingly when the absolute position of the described object changes.

[0083] According to a first aspect of the present invention, an interpretable selected state-space architecture method for inverting coal and rock fracture fields is provided, such as... Figure 1 As shown, it includes the following steps:

[0084] S10: CT detection sensors and acoustic emission monitoring sensors are placed on the coal and rock sample, and the sample is placed in a testing machine for loading tests. The CT detection sensors and acoustic emission monitoring sensors collect CT detection data at different stages of the loading and fracturing process of the coal and rock sample, and full-time acoustic emission monitoring data. The CT detection data includes full-space fracture parameter data at each stage, and the acoustic emission monitoring data includes waveform data and scalar parameter data. (The sample location is as follows...) Figure 2As shown, the testing machine 100 includes: a loading device 110, an acoustic emission sensor 120, a coupler 130, a preamplifier 140, an X-ray detector 150, an X-ray source 160, and a test specimen 170; wherein, the loading device 110, acoustic emission sensor 120, coupler 130, preamplifier 140, X-ray detector 150, and X-ray source 160 are all installed in the test specimen 170, and the coal and rock sample 200 is disposed in the loading device 110. Acoustic emission sensors 120 are installed on both sides of the coal and rock sample 200 in the transverse direction X, wherein the acoustic emission sensors 120 are electrically connected to the coupler 130 and the preamplifier 140, and X-ray detectors 150 and X-ray sources 160 are respectively disposed on both sides of the coal and rock sample 200 in the longitudinal direction Y, wherein the transverse direction X and the longitudinal direction Y are perpendicular to each other. The working process is as follows: The loading device 110 loads the coal and rock sample 200. During this process, the X-ray source 160 emits X-rays and the X-ray detector 150 receives the X-rays, thereby obtaining CT detection data during the fracturing process of the coal and rock sample 200. At the same time, the acoustic emission sensor 120 obtains all-time and all-space acoustic emission monitoring data.

[0085] S20: The CT detection data and all-time acoustic emission monitoring data at different moments during the loading process of the coal and rock samples are correlated according to time to obtain the fracture parameter dataset, and the fracture parameter dataset is divided into training set, validation set and test set;

[0086] S30: Based on a selective state-space architecture, a neural network extracts features from the phased time-space fracture parameter data, corresponding time-waveform data, and corresponding time-scalar parameter data of the fracture parameter dataset in the training and validation sets, to obtain a real-time quantitative inversion model of fracture parameters; the real-time quantitative inversion model of fracture parameters inverts each fracture parameter in the full-time-space fracture field during loading by using full-time-space acoustic emission monitoring data, thus obtaining the evolution process of the full-time-space fracture field parameters of the coal and rock sample under loading;

[0087] S40: Input the spatiotemporal acoustic emission scalar parameter data of the fracture parameter dataset into the SHAP interpreter, along with the spatiotemporal fracture parameter inversion results obtained from the real-time quantitative inversion model of fracture parameters, and the real-time quantitative inversion model of fracture parameters. This will yield the influence factor between the acoustic emission scalar parameters and the fracture parameter inversion results, and ultimately obtain the real-time quantitative inversion mechanism model of fracture parameters based on the acoustic emission scalar parameters.

[0088] This method for inverting the fracture field of coal and rock samples focuses on using neural networks to invert the spatiotemporal fracture field of loaded coal and rock samples through probe-monitoring information from loading and fracturing tests. It combines a neural network based on a selective state-space architecture with a SHAP interpreter to achieve full spatiotemporal reproduction of the fracture field of loaded coal and rock samples, revealing a real-time quantitative inversion mechanism model for fracture parameters. Compared to traditional numerical simulation and neural network methods, this spatiotemporal fracture field inversion method based on interpretable deep learning has higher computational efficiency and accuracy, effectively overcoming the limitations of traditional methods.

[0089] The real-time quantitative inversion model for fracture parameters in this coal and rock sample fracture field inversion method is obtained by extracting spatiotemporal features from the fracture parameter dataset using a neural network based on a selective state-space architecture. A feature extraction method for fracture parameter data and acoustic emission data during coal and rock sample loading is proposed. Based on the proposed spatiotemporal feature extraction method, full-spatiotemporal fracture parameters can be inverted using full-spatiotemporal acoustic emission data obtained in the laboratory during coal and rock sample loading.

[0090] The model inversion mechanism in this coal and rock sample fracture field inversion method extracts features from acoustic emission scalar parameter data and full-time fracture parameter inversion results using an interpreter, obtaining the influencing factors in the mechanism model. The interpreter can obtain a real-time quantitative inversion mechanism model of fracture parameters during the loading process of coal and rock samples through acoustic emission data and model prediction results, revealing the black box process of fracture parameter inversion during the loading process of coal and rock samples.

[0091] In one example of the present invention, such as Figure 3 As shown, in step S30, the feature extraction of the phased time full-space fracture parameter data, corresponding time waveform data, and corresponding time scalar parameter data of the fracture parameter dataset based on the selective state-space architecture neural network includes the following steps:

[0092] S31: The acoustic emission scalar parameters are embedded into the waveform data through the data coding layer to obtain parameter-enhanced acoustic emission data, and the parameter-enhanced acoustic emission features AEf are extracted. se ;

[0093] S32: The encoded data is concatenated to the feature extraction layer B through the residual connection layer to obtain the feature AE. b middle;

[0094] S33: Enhance acoustic emission characteristics AEf by using the first Normalization function. sn Regularization is performed to obtain the feature AE b ;

[0095] S34: Regularized Feature AEn Feature extraction AE is obtained through the residual connection formed by the second linear feature extraction layer A and the second activation function layer. a+s The feature AE spliced ​​into the output of the selective state space architecture sssm Get AE SSSM ;

[0096] S35: Feature AE output by the first linear feature extraction layer A a Higher-dimensional features (AE) are obtained through feature extraction via convolutional layers and a first activation function layer. conv+s ;

[0097] S36: Gated computation of high-dimensional features is performed using a selective state-space architecture, and then concatenated with the features extracted by the convolutional layer and the first activation function layer to obtain feature AE. SSSM ;

[0098] S37: Linear feature extraction layer B extracts features from the concatenated features and concatenates the extracted features with the features of the encoded data obtained by the residual connection layer to obtain feature AEf. b+se ;

[0099] S38: Regularize the concatenated features using the second Normalization function to obtain feature AEf. n ;

[0100] S39: The features extracted by the fully connected layer are used for calculation to obtain the crack parameter F. N .

[0101] In one example of the present invention, in step S36, the selective state-space architecture sample is implemented as follows:

[0102] First, a state-space architecture is constructed using continuous state-space equations, the expressions of which are as follows:

[0103] h t =Ah t-1 +Bx t

[0104] y t =Ch t

[0105] Among them, h t-1 Represents the system state; h t Represents the updated system state; matrix A represents the state transition matrix; matrix B represents the input gating matrix; matrix C represents the output mapping matrix; x t The data represents the current state of the input architecture.

[0106] Secondly, the state-space architecture is subjected to discrete data adaptation using discrete state-space equations; the equation for the state transition matrix A is as follows:

[0107]

[0108] In the formula, n is the row index related to the orthogonal basis functions in the polynomial space, and k is the column index related to the orthogonal basis functions in the polynomial space.

[0109] In one example of the present invention, to simplify the calculation process, the method further includes: randomly initializing the state transition matrix A and using a low-rank matrix after matrix decomposition, the equation of which is as follows:

[0110]

[0111] Where P and Q are both vectors of length N; V represents the eigenvector matrix; V * Λ represents the transpose matrix; Λ represents the eigenvalue matrix; * represents the projection operation.

[0112] In one example of the present invention, the expression of the discrete state-space equation is as follows:

[0113]

[0114] In the formula, h k and y k These represent the memory control unit and the output control unit, respectively, for selecting the current state in the state-space architecture; h k-1 h0 represents the memory control unit that selects the previous state in the state space architecture; h0 represents the initialization state of the memory control unit; x k x0 represents the data of the current state input architecture; x0 represents the data of the initial state input architecture. All are matrices after discrete adaptation of A, B, and C.

[0115] All are matrices after discrete adaptation of A, B, and C.

[0116] In one example of the invention, the matrix It is obtained from the following derivation formula:

[0117]

[0118] In the formula, Δ represents the discretized nonlinear function, which discretizes the continuous state; Ι represents the identity matrix.

[0119] In one example of the present invention, in step S40, the expression formula of the SHAP interpreter is as follows:

[0120]

[0121] in, The value represents the influence factor; M represents the number of input acoustic emission scalar parameters. This represents the set of all permutations and combinations of M features. f represents the set of all feature combinations preceding feature k in permutation R. x (S)=E(f(X)|X s =x S ) represents the conditional expectation given a known feature subset S; i represents the index of the influence factor.

[0122] In one example of this invention, the influence factor, acoustic emission scalar parameter, and fracture parameter obtained through the SHAP interpreter constitute a mechanism model for real-time quantitative inversion of fracture parameters, and its expression equation is as follows:

[0123]

[0124] Where MM represents the mechanism model for real-time quantitative inversion of fracture parameters; F N Represents the fracture parameter; AE si This represents the scalar parameter of acoustic emission.

[0125] According to a second aspect of the present invention, an interpretable selected state-space architecture coal and rock sample fracture field inversion system comprises:

[0126] The information acquisition module is configured to deploy CT detection sensors and acoustic emission monitoring sensors on coal and rock samples, and place the coal and rock samples in a testing machine for loading tests. The CT detection sensors and acoustic emission monitoring sensors acquire CT detection data and full-time acoustic emission monitoring data at different stages during the loading process of the coal and rock samples. Among them, the CT detection data is the full-space fracture parameter data at different stages, and the acoustic emission monitoring data includes waveform data and scalar parameter data.

[0127] The data partitioning module is configured to map the CT detection data and all-time acoustic emission monitoring data at different stages during the loading process of coal and rock samples according to time, obtain the fracture parameter dataset, and divide the fracture parameter dataset into training set, validation set and test set;

[0128] The full-time-space fracture parameter quantitative inversion module is configured to extract features from the phased full-space fracture parameter data, corresponding waveform data, and corresponding scalar parameter data of the fracture parameter dataset based on a selective state-space architecture neural network, thereby obtaining a real-time quantitative inversion model of fracture parameters. The real-time quantitative inversion model of fracture parameters inverts each fracture parameter in the full-time-space fracture field during loading by using full-time-space acoustic emission monitoring data, thus obtaining the evolution process of the full-time-space fracture field parameters of the coal and rock sample under loading.

[0129] The inversion mechanism model module is configured to input the spatiotemporal acoustic emission scalar parameter data of the fracture parameter dataset into the SHAP interpreter, the spatiotemporal fracture parameter inversion results obtained from the real-time quantitative inversion model of fracture parameters, and the real-time quantitative inversion model of fracture parameters, to obtain the influence factor between the acoustic emission scalar parameters and the fracture parameter inversion results, and then obtain the real-time quantitative inversion mechanism model of fracture parameters based on the acoustic emission scalar parameters.

[0130] This coal and rock sample fracture field inversion system focuses on using neural network methods to invert the full-temporal and spatiotemporal fracture field of loaded coal and rock samples through probe-monitoring information from loading and fracturing tests. It combines a neural network based on a selective state-space architecture with a SHAP interpreter to achieve full-temporal and spatiotemporal reproduction of the fracture field of loaded coal and rock samples, revealing a real-time quantitative inversion mechanism model for fracture parameters. Compared with traditional numerical simulation and neural network methods, this coal and rock spatiotemporal fracture field inversion method based on interpretable deep learning has higher computational efficiency and accuracy, effectively overcoming the limitations of traditional methods.

[0131] The real-time quantitative inversion model of fracture parameters in this coal and rock sample fracture field inversion system is obtained by extracting spatiotemporal features from the data in the fracture parameter dataset using a neural network based on a selective state-space architecture. A feature extraction method for fracture parameter data and acoustic emission data during the loading process of coal and rock samples is proposed. Based on the proposed spatiotemporal feature extraction method, full-spatiotemporal fracture parameters can be inverted using full-spatiotemporal acoustic emission data obtained in the laboratory during the loading process of coal and rock samples.

[0132] The model inversion mechanism in this coal and rock sample fracture field inversion system extracts features from acoustic emission scalar parameter data and full-time fracture parameter inversion results using an interpreter, obtaining the influencing factors in the mechanism model. The interpreter can obtain a real-time quantitative inversion mechanism model of fracture parameters during the coal and rock sample loading process using acoustic emission data and model prediction results, revealing the black-box process of fracture parameter inversion during loading.

[0133] In one example of the present invention, such as Figure 3As shown, the all-temporal-space fracture parameter quantitative inversion module includes: a data encoding layer, a first normalization function layer, a linear feature extraction layer A, a convolutional layer, a first activation function layer, a selective state space architecture, a linear feature extraction layer B, a second normalization function layer, and a fully connected layer, connected in series. The output of the data encoding layer is connected to the output of the linear feature extraction layer B via a residual connection layer, and the output of the first normalization function layer is connected to the output of the selective state space architecture via the second linear feature extraction layer A and the second activation function layer. For example, both the first and second activation function layers use the SiLU activation function.

[0134] The data encoding layer is configured to embed acoustic emission scalar parameters into waveform data to obtain parameter-enhanced acoustic emission data, and to extract parameter-enhanced acoustic emission features AEf. se ;

[0135] The residual connection layer is configured to concatenate the encoded data to the feature extraction layer B to obtain feature AE. b middle;

[0136] The first Normalization function layer is configured to enhance the acoustic emission feature AEf of the parameters. se Regularization is performed to obtain the feature AE n ;

[0137] The residual connection consisting of the second linear feature extraction layer A and the second activation function is configured to perform feature extraction to obtain AE. a+s The feature AE spliced ​​into the output of the selective state space architecture sssm Get AE SSSM ;

[0138] The convolutional layer and the first activation function layer are configured to extract the feature AE output by the first linear feature extraction layer A. a Feature extraction is performed to obtain higher-dimensional features (AE). conv+s ;

[0139] The selective state-space architecture is configured to perform gated computation on high-dimensional features and concatenate them with the features extracted by the convolutional layer and the first activation function layer to obtain feature AE. SSSM ;

[0140] The linear feature extraction layer B is configured to extract features from the concatenated features, and then concatenate the extracted features with the features of the encoded data obtained by the residual connection layer to obtain feature AEf. b+se ;

[0141] The second Normalization function layer is configured to regularize the concatenated features to obtain feature AEf. n ;

[0142] The fully connected layer is configured to perform calculations on the extracted features to obtain the fracture parameter F. N .

[0143] In other words, the feature extraction of full-space fracture parameter data, corresponding waveform data, and corresponding scalar parameter data at different time points from the training and validation sets of the fracture parameter dataset based on a selective state-space architecture includes the following steps:

[0144] The acoustic emission scalar parameters are embedded into the waveform data through a data encoding layer to obtain parameter-enhanced acoustic emission data, and the parameter-enhanced acoustic emission features AEf are extracted. se ;

[0145] The encoded data is concatenated to the feature extraction layer B through a residual connection layer to obtain the feature AE. b middle;

[0146] The acoustic emission characteristics AEf are enhanced by the first Normalization function. se Regularization is performed to obtain the feature AE n ;

[0147] Regularized features AE n Feature extraction AE is obtained through the residual connection formed by the second linear feature extraction layer A and the second activation function layer. a+s The feature AE spliced ​​into the output of the selective state space architecture sssm Get AE SSSM ;

[0148] The feature AE output by the linear feature extraction layer A a Higher-dimensional features (AE) are obtained through feature extraction via convolutional layers and a first activation function layer. conv+s ;

[0149] High-dimensional features are gated using a selective state-space architecture and then concatenated with features extracted from convolutional layers and the first activation function layer to obtain feature AE. SSSM ;

[0150] Linear feature extraction layer B extracts features from the concatenated features and concatenates the extracted features with the features of the encoded data obtained by the residual connection layer to obtain feature AEf. b+se ;

[0151] The concatenated features are regularized using the second Normalization function to obtain the feature AEf. n ;

[0152] The features extracted by the fully connected layer are used for calculation to obtain the crack parameter F. N .

[0153] The foregoing description, with reference to preferred embodiments, details the exemplary implementation of the coal and rock fracture field inversion method and system with interpretable selectable state space architecture proposed in this invention. However, those skilled in the art will understand that various modifications and alterations can be made to the above-described sample embodiments without departing from the concept of this invention, and various combinations can be made to the various technical features and structures proposed in this invention without exceeding the protection scope of this invention, which is determined by the appended claims.

Claims

1. A method for inverting coal and rock fracture fields with an interpretable selected state-space architecture, characterized in that, Includes the following steps: S10: CT detection sensors and acoustic emission monitoring sensors are arranged on the coal and rock sample, and the coal and rock sample is placed in the testing machine for loading test. CT detection data and full-time acoustic emission monitoring data are collected at different time points during the loading process of the coal and rock sample through the CT detection sensors and acoustic emission monitoring sensors. Among them, the CT detection data are the full-space fracture parameter data at different time points, and the acoustic emission monitoring data includes waveform data and scalar parameter data. S20: The CT detection data and all-time acoustic emission monitoring data at different moments during the loading process of the coal and rock samples are correlated according to time to obtain the fracture parameter dataset, and the fracture parameter dataset is divided into training set, validation set and test set; S30: Based on a selective state-space architecture, a neural network extracts features from the phased time-space fracture parameter data, corresponding time-waveform data, and corresponding time-scalar parameter data of the fracture parameter dataset in the training and validation sets, to obtain a real-time quantitative inversion model of fracture parameters; the real-time quantitative inversion model of fracture parameters inverts each fracture parameter in the full-time-space fracture field during loading by using full-time-space acoustic emission monitoring data, thus obtaining the evolution process of the full-time-space fracture field parameters of the coal and rock sample under loading; S40: Input the spatiotemporal acoustic emission scalar parameter data of the fracture parameter dataset into the SHAP interpreter, along with the spatiotemporal fracture parameter inversion results obtained from the real-time quantitative inversion model of fracture parameters, and the real-time quantitative inversion model of fracture parameters. This will yield the influence factor between the acoustic emission scalar parameters and the fracture parameter inversion results, and ultimately obtain the real-time quantitative inversion mechanism model of fracture parameters based on the acoustic emission scalar parameters.

2. The coal and rock fracture field inversion method with an interpretable selectable state-space architecture according to claim 1, characterized in that, In step S30, the feature extraction of the phased time full-space fracture parameter data, corresponding time waveform data, and corresponding time scalar parameter data of the fracture parameter dataset based on the selective state-space architecture neural network includes the following steps: S31: The acoustic emission scalar parameters are embedded into the waveform data through the data coding layer to obtain parameter-enhanced acoustic emission data, and the parameter-enhanced acoustic emission features AEf are extracted. se ; S32: The encoded data is concatenated to the feature extraction layer B through the residual connection layer to obtain the feature AE. b middle; S33: Enhance acoustic emission characteristics AEf by using the first Normalization function. sn Regularization is performed to obtain the feature AE n ; S34: Regularized Feature AE n Feature extraction AE is obtained through the residual connection formed by the second linear feature extraction layer A and the second activation function layer. a+s The feature AE spliced ​​into the output of the selective state space architecture sssm Get AE SSSM ; S35: Feature AE output by the first linear feature extraction layer A a Higher-dimensional features (AE) are obtained through feature extraction via convolutional layers and a first activation function layer. conv+s ; S36: Gated computation of high-dimensional features is performed using a selective state-space architecture, and then concatenated with the features extracted by the convolutional layer and the first activation function layer to obtain feature AE. SSSM ; S37: Linear feature extraction layer B extracts features from the concatenated features and concatenates the extracted features with the features of the encoded data obtained by the residual connection layer to obtain feature AEf. b+se ; S38: Regularize the concatenated features using the second Normalization function to obtain feature AEf. n ; S39: The features extracted by the fully connected layer are used for calculation to obtain the crack parameter F. N .

3. The coal and rock fracture field inversion method with an interpretable selected state-space architecture according to claim 2, characterized in that, In step S36, the selective state-space structure fixture sample implementation process is as follows: First, a state-space architecture is constructed using continuous state-space equations, the expressions of which are as follows: h t =Ah t-1 +Bx t the t =Ch t Among them, h t-1 Represents the system state; h t Represents the updated system state; matrix A represents the state transition matrix; matrix B represents the input gating matrix; matrix C represents the output mapping matrix; x t The data represents the current state of the input architecture. Secondly, the state-space architecture is subjected to discrete data adaptation using discrete state-space equations; the equation for the state transition matrix A is as follows: In the formula, n is the row index related to the orthogonal basis functions in the polynomial space, and k is the column index related to the orthogonal basis functions in the polynomial space.

4. The coal and rock fracture field inversion method with an interpretable selected state-space architecture according to claim 3, characterized in that, This also includes: randomly initializing the state transition matrix A and using a low-rank matrix obtained from matrix decomposition. The equation for the low-rank matrix is ​​as follows: Where P and Q are both vectors of length N; V represents the eigenvector matrix; V * Λ represents the transpose matrix; Λ represents the eigenvalue matrix; * represents the projection operation.

5. The coal and rock fracture field inversion method with an interpretable selected state-space architecture according to claim 3, characterized in that, The expression for the discrete state-space equation is as follows: In the formula, h k and y k These represent the memory control unit and the output control unit, respectively, for selecting the current state in the state-space architecture; h k-1 h0 represents the memory control unit that selects the previous state in the state space architecture; h0 represents the initialization state of the memory control unit; x k x0 represents the data of the current state input architecture; x0 represents the data of the initial state input architecture. All are matrices after discrete adaptation of A, B, and C.

6. The coal and rock fracture field inversion method with an interpretable selected state-space architecture according to claim 5, characterized in that, matrix It is obtained from the following derivation formula: In the formula, Δ represents the discretized nonlinear function, which discretizes the continuous state; Ι represents the identity matrix.

7. The coal and rock fracture field inversion method with an interpretable selectable state-space architecture according to claim 1, characterized in that, In step S40, the expression formula of the SHAP interpreter is as follows: in, The value represents the influence factor; M represents the number of input acoustic emission scalar parameters. This represents the set of all permutations and combinations of M features. f represents the set of all feature combinations preceding feature k in permutation R. x (S)=E(f(X)|X s =M S ) represents the conditional expectation given a known feature subset S; i represents the influence factor index.

8. The coal and rock fracture field inversion method with an interpretable selected state-space architecture according to claim 7, characterized in that, The influence factor, acoustic emission scalar parameter, and fracture parameter obtained through the SHAP interpreter constitute the mechanism model for real-time quantitative inversion of fracture parameters, and its expression equation is as follows: Where M represents the mechanism model for real-time quantitative inversion of fracture parameters; F N Represents the fracture parameter; AE si This represents the scalar parameter of acoustic emission.

9. A coal and rock fracture field inversion system with an interpretable selectable state-space architecture, characterized in that, include: The information acquisition module is configured to deploy CT detection sensors and acoustic emission monitoring sensors on coal and rock samples, and place the coal and rock samples in a testing machine for loading tests. The CT detection sensors and acoustic emission monitoring sensors acquire CT detection data and full-time acoustic emission monitoring data at different stages during the loading process of the coal and rock samples. Among them, the CT detection data is the full-space fracture parameter data at different stages, and the acoustic emission monitoring data includes waveform data and scalar parameter data. The data partitioning module is configured to map the CT detection data and all-time acoustic emission monitoring data at different stages during the loading process of coal and rock samples according to time, obtain the fracture parameter dataset, and divide the fracture parameter dataset into training set, validation set and test set; The full-time-space fracture parameter quantitative inversion module is configured to extract features from the phased full-space fracture parameter data, corresponding waveform data, and corresponding scalar parameter data of the fracture parameter dataset based on a selective state-space architecture neural network, thereby obtaining a real-time quantitative inversion model of fracture parameters. The real-time quantitative inversion model of fracture parameters inverts each fracture parameter in the full-time-space fracture field during loading by using full-time-space acoustic emission monitoring data, thus obtaining the evolution process of the full-time-space fracture field parameters of the coal and rock sample under loading. The inversion mechanism model module is configured to input the spatiotemporal acoustic emission scalar parameter data of the fracture parameter dataset into the SHAP interpreter, the spatiotemporal fracture parameter inversion results obtained from the real-time quantitative inversion model of fracture parameters, and the real-time quantitative inversion model of fracture parameters, to obtain the influence factor between the acoustic emission scalar parameters and the fracture parameter inversion results, and then obtain the real-time quantitative inversion mechanism model of fracture parameters based on the acoustic emission scalar parameters.

10. The coal and rock fracture field inversion system with an interpretable selectable state-space architecture according to claim 9, characterized in that, The all-temporal-space fracture parameter quantitative inversion module includes: a data encoding layer, a first normalization function layer, a linear feature extraction layer A, a convolutional layer, a first activation function layer, a selective state space architecture, a linear feature extraction layer B, a second normalization function layer, and a fully connected layer, connected in series. The output of the data encoding layer is connected to the output of the linear feature extraction layer B via a residual connection layer, and the output of the first normalization function layer is connected to the output of the selective state space architecture via the second linear feature extraction layer A and the second activation function layer. The data encoding layer is configured to embed acoustic emission scalar parameters into waveform data to obtain parameter-enhanced acoustic emission data, and to extract parameter-enhanced acoustic emission features AEf. se ; The residual connection layer is configured to concatenate the encoded data to the feature extraction layer B to obtain feature AE. b middle; The first Normalization function layer is configured to enhance the acoustic emission feature AEf of the parameters. se Regularization is performed to obtain the feature AE n ; The residual connection consisting of the second linear feature extraction layer A and the second activation function is configured to perform feature extraction to obtain AE. a+s The feature AE spliced ​​into the output of the selective state space architecture sssm Get AE SSSM ; The convolutional layer and the first activation function layer are configured to extract the feature AE output by the first linear feature extraction layer A. a Feature extraction is performed to obtain higher-dimensional features (AE). conv+s ; The selective state-space architecture is configured to perform gated computation on high-dimensional features and concatenate them with the features extracted by the convolutional layer and the first activation function layer to obtain feature AE. SSSM ; The linear feature extraction layer B is configured to extract features from the concatenated features, and then concatenate the extracted features with the features of the encoded data obtained by the residual connection layer to obtain feature AEf. b+se ; The second Normalization function layer is configured to regularize the concatenated features to obtain feature AEf. n ; The fully connected layer is configured to perform calculations on the extracted features to obtain the fracture parameter F. N .

Images