A rockburst early warning method and device based on fractional-order kernel discriminant space learning

Abstract

The present invention provides a rockburst early warning method and device based on fractional-order kernel discriminant space learning, which relates to the technical field of rockburst early warning and includes: obtaining rockburst index sample training data and rockburst index sample test data, and converting the rockburst index sample training data and the rockburst index sample test data into column vectors to generate a rockburst training sample set and a rockburst test sample set; constructing a target optimization function of a fractional-order kernel discriminant space learning model according to the rockburst training sample set; optimizing and solving a kernel discriminant space projection matrix according to the target optimization function of the fractional-order kernel discriminant space learning model; applying the fractional kernel discriminant space projection matrix to a rockburst training sample set and a rockburst test sample set respectively to generate a training discriminant feature set of the rockburst training sample set and a test discriminant feature set of the rockburst test sample set; inputting the training discriminant feature set and the test discriminant feature set into a support vector basis classifier to generate a warning result of a rockburst risk.

Description

Technical Field

[0001] The present invention relates to the technical field of rockburst warning, and in particular to a rockburst warning method and device based on fractional-order kernel discriminant space learning. Background Art

[0002] Rockburst refers to the phenomenon of sudden destruction in the open rock mass in deep underground mining or areas with high tectonic stress. The reason for the occurrence is that the strain energy accumulated in the open rock mass is suddenly and violently released, causing the rock mass to undergo brittle fracture like an explosion. The impact ground pressure causes a large amount of rock to collapse, and produces a huge sound and air wave impact, which can not only destroy the mine, but also the shock wave can endanger the ground buildings.

[0003] In the field of rockburst prediction, due to the complexity of the original data, the same pattern can always obtain multiple different feature identifiers, which usually reveal different characteristics or perspectives of the same pattern, leading to high computational burden and serious distortion of warning results.

[0004] Related technologies, spatial learning is an effective method to solve the problem of high computational burden of rockburst prediction. However, due to the small number of rockburst prediction training samples or when disturbed, the intra-class divergence matrix and the inter-class divergence matrix will deviate from the true divergence matrix, reducing the accuracy of rockburst warning. Summary of the invention

[0005] The problem solved by the present invention is how to improve the accuracy of rock burst warning.

[0006] To solve the above problems, on the one hand, the present invention provides a rockburst early warning method based on fractional-order kernel discriminant space learning, comprising:

[0007] Obtain rockburst indicator sample training data and rockburst indicator sample test data, and convert the rockburst indicator sample training data and the rockburst indicator sample test data into column vectors to generate a rockburst training sample set and a rockburst test sample set;

[0008] Constructing a target optimization function of a fractional-order kernel discriminant space learning model according to the rockburst training sample set;

[0009] Optimizing and solving the nuclear discriminant space projection matrix according to the objective optimization function of the fractional-order nuclear discriminant space learning model;

[0010] Applying the fractional kernel discriminant space projection matrix to the rockburst training sample set and the rockburst test sample set respectively to generate a training discriminant feature set of the rockburst training sample set and a test discriminant feature set of the rockburst test sample set;

[0011] The training identification feature set and the test identification feature set are input into a support vector base classifier to generate an early warning result of rock burst risk.

[0012] Optionally, the objective optimization function of constructing the fractional-order kernel discriminant space learning model according to the rockburst training sample set includes:

[0013] Constructing an inter-class scatter matrix, an intra-class scatter matrix and an overall scatter matrix according to the rockburst training sample set;

[0014] Based on the inter-class scatter matrix, the intra-class scatter matrix and the overall scatter matrix, a target optimization function of a linear discriminant space model is generated;

[0015] Based on the inter-class scatter matrix and the intra-class scatter matrix, constructing a kernel inter-class scatter matrix and a kernel intra-class scatter matrix respectively;

[0016] Projecting the objective optimization function of the linear discriminant space model into the nonlinear space, and combining the inter-core scatter matrix and the intra-core scatter matrix to construct the objective optimization function of the discriminant analysis space model;

[0017] The objective optimization function of the fractional-order kernel discriminant space learning model is constructed according to the objective optimization function of the discriminant analysis space model.

[0018] Optionally, the calculation formulas of the inter-class scatter matrix, the intra-class scatter matrix and the overall scatter matrix are as follows:

[0019]

[0020]

[0021] S t =S b +S w ;

[0022] Among them, the N j is the number of samples in the jth class, u j is the mean of the j-th class of samples, u j The calculation formula is u is the mean of all samples, and the calculation formula of u is X j is the set of samples of the jth class.

[0023] Optionally, the objective optimization function of the linear discriminant space model is as follows:

[0024] max W W T S b W;

[0025] WT S w W = 1;

[0026] Where W is the spatial projection matrix of the linear discriminant space model.

[0027] Optionally, the objective optimization function of the discriminant space analysis learning model is as follows:

[0028]

[0029]

[0030] in, is the kernel inter-class divergence matrix, is the intra-kernel divergence matrix, is the spatial projection matrix of the kernel discriminant space, and the relationship between the kernel discriminant space projection matrix and the discriminant space projection matrix is

[0031] Optionally, the kernel inter-class divergence matrix and the kernel intra-class divergence matrix The calculation formulas are as follows:

[0032]

[0033]

[0034] Among them, K is the kernel matrix, and the (i, j)th element of K is The calculation formula is

[0035] Optionally, constructing a target optimization function of a fractional-order kernel discriminant space learning model according to the target optimization function of the discriminant analysis space model includes:

[0036] Based on the kernel inter-class scatter matrix and the kernel intra-class scatter matrix, constructing a fractional-order inter-class scatter matrix and a fractional-order intra-class scatter matrix respectively;

[0037] Based on the fractional inter-class scatter matrix, the fractional intra-class scatter matrix and the objective optimization function of the discriminant analysis space model, the objective optimization function of the fractional-order kernel discriminant space learning model is constructed.

[0038] Optionally, the objective optimization function of the fractional-order kernel discriminant space learning model is as follows:

[0039]

[0040]

[0041] in, is the fractional order intra-class scatter matrix, is the fractional inter-class scatter matrix.

[0042] Optionally, the step of optimizing and solving the projection matrix of the kernel discriminant space according to the objective optimization function of the fractional-order kernel discriminant space learning model includes:

[0043] Constructing the Lagrange multiplier function of the target optimization function of the fractional-order kernel discriminant space learning model with respect to the kernel discriminant space projection matrix;

[0044] The derivative of the Lagrangian multiplier function with respect to the projection matrix of the kernel discriminant space is set to zero, and the eigenvector corresponding to the maximum eigenvalue of the Lagrangian multiplier function is solved;

[0045] A spatial projection matrix of the kernel discriminant space is constructed according to the feature vector.

[0046] On the other hand, the present invention also provides a rockburst early warning device based on fractional-order kernel discriminant space learning, comprising:

[0047] An information acquisition conversion module is used to acquire rockburst index sample training data and rockburst index sample test data, and convert the rockburst index sample training data and the rockburst index sample test data into column vectors to generate a rockburst training sample set and a rockburst test sample set;

[0048] A model building module, used for building a target optimization function of a fractional-order kernel discriminant space learning model according to the rockburst training sample set;

[0049] A spatial projection matrix solving module, used for optimizing and solving the nuclear discriminant space projection matrix according to the objective optimization function of the fractional-order nuclear discriminant space learning model;

[0050] A feature extraction module, used for applying the fractional kernel discriminant space projection matrix to the rockburst training sample set and the rockburst test sample set respectively, to generate a training discriminant feature set of the rockburst training sample set and a test discriminant feature set of the rockburst test sample set;

[0051] The risk warning module is used to input the training identification feature set and the test identification feature set into a support vector base classifier to generate a warning result of rock burst risk.

[0052] In summary, there are at least the following beneficial effects:

[0053] The acquired rockburst training data and rockburst index sample test data are converted into column vectors to generate rockburst training sample sets and rockburst test sample sets, so as to facilitate the subsequent data calculation and operation processing; then, with the help of the strategy of matrix decomposition features, linear matrix projection theory, fractional order theory and kernel technology, the target optimization function of the fractional order kernel discriminant space learning model is constructed from the rockburst training sample set, so that the intra-class divergence matrix and the inter-class divergence matrix in the fractional order kernel discriminant space learning model are closer to the real divergence matrix, which can reveal a more effective nonlinear manifold structure and obtain more comprehensive identification information; further, the kernel discriminant space projection matrix is ​​solved through the target optimization function of the fractional order kernel discriminant space learning model, and the kernel discriminant space projection matrix is ​​applied to the rockburst training sample set and the rockburst test sample set to obtain the training identification feature set and the test identification feature set, and then the obtained training identification feature set and the test identification feature set are input into the vector basis classifier, which can quickly obtain the identification features of the rockburst test samples, thereby obtaining accurate rockburst warning results and effectively improving the accuracy of rockburst warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 A flow chart of a rockburst early warning method based on fractional-order kernel discriminant space learning provided by an embodiment of the present invention;

[0055] Figure 2 A flowchart of another rockburst early warning method based on fractional-order kernel discriminant space learning provided by an embodiment of the present invention;

[0056] Figure 3 A flowchart of S102 in another rockburst early warning method based on fractional-order kernel discriminant space learning provided in an embodiment of the present invention;

[0057] Figure 4 A structural schematic diagram of a rockburst early warning device based on fractional-order kernel discriminant space learning provided in an embodiment of the present invention.

[0058] Explanation of the accompanying drawings: 501, information acquisition and conversion module; 502, model building module; 503, space projection matrix solution module; 504, feature extraction module; 505, risk warning module. DETAILED DESCRIPTION

[0059] In order to make the above-mentioned objects, features and advantages of the present invention more clearly understood, the following Figure 1-4 The specific embodiments of the present invention are described in detail.

[0060] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein.

[0061] In the description of this specification, the description with reference to the terms "embodiment", "some embodiments" and "optional embodiments" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or implementation are included in at least one embodiment or exemplary implementation of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or implementation. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or implementations in a suitable manner.

[0062] like Figure 1 and Figure 2 As shown, an embodiment of the present invention provides a rockburst warning method based on fractional-order kernel discriminant space learning, which can be applied to rockburst warning in underground mining environments, wherein the execution subject of the method can be a rockburst warning device based on analytical-order kernel discriminant space learning, and is assisted by other input and output devices. The rockburst warning device of the present invention converts and generates training identification feature sets and test identification feature sets based on the acquired rockburst index sample training data and rockburst index sample test data, and at the same time uses the strategy of matrix decomposition features, linear matrix projection theory, fractional-order theory and kernel technology to facilitate the input of the generated training identification feature sets and test identification feature sets into the support vector basis classifier, thereby obtaining accurate rockburst warning results and effectively improving the accuracy of rockburst warning.

[0063] The following will be combined with specific embodiments. Figure 2 The processing flow shown is described in detail, and the content can be as follows:

[0064] S101, obtaining rockburst indicator sample training data and rockburst indicator sample test data, and converting the rockburst indicator sample training data and the rockburst indicator sample test data into column vectors to generate a rockburst training sample set and a rockburst test sample set.

[0065] In implementation, the rockburst warning device obtains rockburst indicator sample training data and rockburst indicator sample test data. The rockburst indicator sample training data is a data set used to train the rockburst indicator model, which usually contains a series of sample data. Each sample data contains some features and corresponding rockburst indicators. The features here can be geological, geophysical, geochemical and other data related to rockburst, while the rockburst indicator is an assessment of the degree of rockburst based on actual observations or simulation results. The rockburst indicator sample test data is a data set used to evaluate the performance of the rockburst indicator model. The rockburst indicator sample test data usually contains some sample data similar to the training data, but the model has not seen it during the training process. These sample data are input into the trained rockburst indicator model and compared with the actual observations or simulation results, which can evaluate the prediction ability and accuracy of the model on unknown data.

[0066] The rockburst warning device converts the acquired rockburst indicator sample training data into a column vector to generate a rockburst training sample set. In this embodiment, the rockburst training sample set is X = [x1, x2, ..., x m ]∈R n×m , where n is the sample dimension and m is the number of training samples;

[0067] The rockburst warning device converts the acquired rockburst index sample test data into a column vector to generate a rockburst test sample set. In this embodiment, the rockburst test sample set is Where n is the sample dimension, m is the number of training samples, is the number of test samples, x i is the i-th training sample of X, for The i-th test sample of y i is the training sample x i The class label of .

[0068] S102, constructing a target optimization function of a fractional-order kernel discriminant space learning model according to the rockburst training sample set.

[0069] In implementation, the rockburst warning device defines the inter-class divergence matrix, intra-class divergence matrix and total divergence matrix based on the rockburst training sample set, and then constructs the target optimization function of the linear discriminant space model. With the help of the strategy of matrix decomposition features, linear matrix projection theory, fractional order theory and kernel technology, the intra-class divergence matrix and the inter-class divergence matrix in the fractional order kernel discriminant space learning model are closer to the real divergence matrix, which can reveal a more effective nonlinear manifold structure.

[0070] S103, optimizing and solving the projection matrix of the kernel discriminant space according to the target optimization function of the fractional-order kernel discriminant space learning model.

[0071] In implementation, the rockburst warning device combines the Lagrangian function of the objective optimization function of the fractional-order kernel discriminant space learning model to solve the kernel discriminant space projection matrix in the generalized eigenvalue analysis split.

[0072] S104, applying the score kernel discriminant space projection matrix to the rockburst training sample set and the rockburst test sample set respectively, to generate a training discriminant feature set of the rockburst training sample set and a test discriminant feature set of the rockburst test sample set.

[0073] In the implementation, the rockburst early warning device applies the score kernel discriminant space projection matrix to the rockburst training sample set and the rockburst test sample set respectively. In this embodiment, the training discriminant feature set of the rockburst training sample set X is Z = [z1, z2, ..., z m ]∈R d×m , test sample set The test discriminant feature set is in, is the i-th training discriminant feature of Z, for The i-th test discriminant feature of The spatial projection matrix for the kernel discriminant space. for The transposed matrix of .

[0074] S105, inputting the training identification feature set and the test identification feature set into a support vector basis classifier to generate a warning result of rock burst risk.

[0075] During implementation, the rockburst warning device inputs the training identification features of the training identification feature set and the test identification features of the test identification feature set into the support vector basis classifier respectively, and the support vector basis classifier classifies and compares the training identification features of the training identification feature set with the test identification features of the test identification feature set.

[0076] The training discriminant features in the training discriminant feature set include rockburst warning level labels, which are divided into no risk, weak risk, medium risk, and strong risk. After the test discriminant features in the test discriminant feature set are input into the support vector basis classifier, the classification warning results of rockburst risk are obtained, namely no risk, weak risk, medium risk, and strong risk.

[0077] Optionally, in step S102, there is also Figure 3 The specific process is as follows:

[0078] S202, constructing an inter-class scatter matrix, an intra-class scatter matrix, and an overall scatter matrix according to the rockburst training sample set.

[0079] In implementation, based on a classification goal, the distance between different categories is as far as possible, and the distance between data in the same category is as close as possible. To this end, the rock burst prediction device constructs an inter-class scatter matrix, an intra-class scatter matrix and a total scatter matrix from the rock burst training sample set.

[0080] Optionally, the calculation formulas of the between-class scatter matrix, the within-class scatter matrix and the overall scatter matrix are as follows:

[0081]

[0082]

[0083] S t =S b +S w ;

[0084] Among them, N j is the number of samples in the jth class, u j is the mean of the j-th class of samples, u j The calculation formula is u is the mean of all samples, and the calculation formula of u is X j is the set of samples of the jth class.

[0085] S203, based on the inter-class scatter matrix and the intra-class scatter matrix, construct a kernel inter-class scatter matrix and a kernel intra-class scatter matrix respectively.

[0086] In the implementation, the between-class scatter matrix is ​​the representation of the between-class scatter matrix in the kernel space. Therefore, the kernel between-class scatter matrix is ​​an extended form of the between-class scatter matrix, which is used to deal with nonlinear problems. Similarly, the kernel within-class scatter matrix is ​​an extended form of the within-class scatter matrix, which is used to deal with nonlinear problems.

[0087] Optionally, the kernel inter-class divergence matrix and the intra-kernel divergence matrix The calculation formulas are as follows:

[0088]

[0089]

[0090] Among them, K is the kernel matrix, and the (i, j)th element of K is The calculation formula is

[0091] In the implementation, the kernel inter-class scatter matrix is ​​transformed into and the intra-kernel divergence matrix They are divided into the following formulas:

[0092]

[0093]

[0094] Where V is The eigenvector matrix of express The diagonal matrix composed of the eigenvalues ​​of and Q is The eigenvector matrix of The diagonal matrix composed of the eigenvalues ​​of and

[0095] S204, projecting the target optimization function of the linear discriminant space model into the nonlinear space, and combining the inter-core scatter matrix and the intra-core scatter matrix to construct the target optimization function of the discriminant analysis space model.

[0096] In implementation, the rockburst warning device projects the target optimization function of the linear discriminant space model into the nonlinear space, thereby optimizing and constructing the target optimization function of the discriminant analysis space model in the nonlinear space to facilitate the processing of nonlinear problems and obtain better classification performance.

[0097] Optionally, the objective optimization function of the discriminant space analysis learning model is as follows:

[0098]

[0099]

[0100] in, is the kernel inter-class divergence matrix, is the intra-kernel divergence matrix, is the spatial projection matrix of the kernel discriminant space, and the relationship between the kernel discriminant space projection matrix and the discriminant space projection matrix is

[0101] Furthermore, the objective optimization function of the discriminant space analysis learning model can be decomposed as follows:

[0102]

[0103]

[0104] S205 , projecting the target optimization function of the linear discriminant space model into the nonlinear kernel space to generate the target optimization function of the discriminant analysis space model.

[0105] In implementation, the rockburst warning device projects the target optimization function of the linear space model into the nonlinear space, and can form a discriminant analysis space learning model and a corresponding target optimization function in a higher-dimensional linearly analyzable kernel space.

[0106] Optionally, in step S205, there is also the following processing method, and the specific processing flow is as follows:

[0107] S301, based on the kernel inter-class scatter matrix and the kernel intra-class scatter matrix, construct a fractional-order inter-class scatter matrix and a fractional-order intra-class scatter matrix respectively.

[0108] In the implementation, the rockburst warning device uses the fractional order theory to construct a fractional order inter-class scatter matrix and the fractional intra-class scatter matrix The fractional inter-class scatter matrix and the fractional intra-class scatter matrix They are as follows:

[0109]

[0110]

[0111] Among them, 0≤β≤1 represents a fraction, and V is The eigenvector matrix of express The diagonal matrix composed of the eigenvalues ​​of and 0≤α≤1 represents a fraction, Q is The eigenvector matrix of The diagonal matrix composed of the eigenvalues ​​of and

[0112] S302, constructing a target optimization function of a fractional-order kernel discriminant space learning model based on the fractional inter-class scatter matrix, the fractional intra-class scatter matrix and the target optimization function of the discriminant analysis space model.

[0113] In implementation, the rockburst warning device uses the objective optimization function of the fractional-order kernel discriminant space learning model to perform feature extraction and classification in the nonlinear space, which can map the data more flexibly to better handle nonlinear problems.

[0114] Optionally, the objective optimization function of the fractional-order kernel discriminant space learning model is as follows:

[0115]

[0116]

[0117] in, is the fractional order intra-class scatter matrix, is the fractional inter-class scatter matrix.

[0118] Optionally, in step 103, there is also the following processing method, and the specific operation flow is as follows:

[0119] S401, constructing the Lagrange multiplier function of the target optimization function of the fractional-order kernel discriminant space learning model with respect to the kernel discriminant space projection matrix.

[0120] In the implementation, the spatial projection matrix of the kernel discriminant space is constructed for the objective optimization function of the fractional-order kernel discriminant space learning model The Lagrange multiplier function

[0121]

[0122] in, is the Lagrange multiplier.

[0123] S402, setting the derivative of the Lagrange multiplier function with respect to the projection matrix of the kernel discriminant space to zero, and solving the eigenvector corresponding to the maximum eigenvalue of the Lagrange multiplier function.

[0124] In implementation, for The derivative of is set to 0, and we get:

[0125]

[0126] S403, constructing a spatial projection matrix of the kernel discriminant space according to the eigenvector.

[0127] In the implementation, the optimization problem of the fractional-order kernel discriminant space learning model is transformed into the following generalized eigenvalue decomposition problem:

[0128]

[0129] In the generalized eigenvalue decomposition problem, the Lagrange multiplier φ is also called the eigenvalue, and the spatial projection matrix of the kernel discriminant space is It is composed of the eigenvectors corresponding to the first d largest eigenvalues, that is, Where d is the projection parameter of the projection direction, w i is the eigenvector corresponding to the i-th largest eigenvalue.

[0130] In summary, the rockburst warning device learns the fractional-order kernel discriminant space based on the intrinsic geometric structure and discriminant information of the rockburst index sample training data and the rockburst index sample test data, which effectively improves the class separation of the discriminant features; at the same time, with the help of fractional-order theory and kernel technology, the small sample problem is solved and the intra-class divergence matrix and the inter-class divergence matrix are closer to the real divergence matrix, which can reveal more effective nonlinear manifold structures, obtain more comprehensive discriminant information, and effectively improve the accuracy of rockburst warning; further, through theoretical derivation, the analytical solution of the fractional-order kernel discriminant space model is obtained, which can quickly obtain the discriminant features of the rockburst test samples, thereby obtaining accurate rockburst warning results.

[0131] Based on the above, the present invention also provides a rockburst early warning device based on fractional-order kernel discriminant space learning, such as Figure 4 As shown, including:

[0132] The information acquisition conversion module 501 is used to obtain rockburst index sample training data and rockburst index sample test data, and convert the rockburst index sample training data and rockburst index sample test data into column vectors to generate a rockburst training sample set and a rockburst test sample set;

[0133] A model building module 502 is used to build a target optimization function of a fractional-order kernel discriminant space learning model according to a rockburst training sample set;

[0134] A spatial projection matrix solving module 503, used for optimizing and solving the nuclear discriminant space projection matrix according to the target optimization function of the fractional-order nuclear discriminant space learning model;

[0135] A feature extraction module 504 is used to apply the fractional kernel discriminant space projection matrix to the rockburst training sample set and the rockburst test sample set respectively to generate a training discriminant feature set of the rockburst training sample set and a test discriminant feature set of the rockburst test sample set;

[0136] The risk warning module 505 is used to input the training identification feature set and the test identification feature set into the support vector basis classifier to generate a warning result of the rock burst risk.

[0137] It should be noted that the rockburst warning device based on fractional-order kernel identification space learning provided in the above embodiment and the rockburst warning method based on fractional-order kernel identification space learning provided in the above embodiment belong to the same concept, wherein the specific manner in which each module and unit performs the operation has been described in detail in the method embodiment and will not be repeated here. In practical applications, the rockburst warning method based on fractional-order kernel identification space learning provided in the above embodiment can allocate the above functions to different functional modules as needed, that is, divide the internal structure of the system into different functional modules to complete all or part of the functions described above, and this is not limited here.

[0138] Although the present invention is disclosed as above, the protection scope of the present invention is not limited thereto. Those skilled in the art may make various changes and modifications without departing from the spirit and scope of the present invention, and these changes and modifications will fall within the protection scope of the present invention.

Claims

1. A rockburst early warning method based on fractional-order kernel discriminant space learning, characterized in that: include: Obtain rockburst indicator sample training data and rockburst indicator sample test data, and convert the rockburst indicator sample training data and the rockburst indicator sample test data into column vectors to generate a rockburst training sample set and a rockburst test sample set; Constructing a target optimization function of a fractional-order kernel discrimination space learning model according to the rockburst training sample set; Optimizing and solving the nuclear discriminant space projection matrix according to the objective optimization function of the fractional-order nuclear discriminant space learning model; Applying the fractional kernel discriminant space projection matrix to the rockburst training sample set and the rockburst test sample set respectively to generate a training discriminant feature set of the rockburst training sample set and a test discriminant feature set of the rockburst test sample set; Inputting the training identification feature set and the test identification feature set into a support vector machine classifier to generate an early warning result of rock burst risk; Wherein, the objective optimization function of constructing the fractional-order kernel discriminant space learning model according to the rockburst training sample set includes: Constructing an inter-class scatter matrix, an intra-class scatter matrix and an overall scatter matrix according to the rockburst training sample set; Based on the inter-class scatter matrix, the intra-class scatter matrix and the overall scatter matrix, a target optimization function of a linear discriminant space model is generated; Based on the inter-class scatter matrix and the intra-class scatter matrix, constructing a kernel inter-class scatter matrix and a kernel intra-class scatter matrix respectively; Projecting the objective optimization function of the linear discriminant space model into the nonlinear space, and combining the inter-core scatter matrix and the intra-core scatter matrix to construct the objective optimization function of the discriminant analysis space model; Constructing a target optimization function of a fractional-order kernel discriminant space learning model according to the target optimization function of the discriminant analysis space model; The objective optimization function of the fractional-order kernel discriminant space learning model is as follows: in, is the fractional order intra-class scatter matrix, is the fractional inter-class scatter matrix; Fractional inter-class scatter matrix and the fractional intra-class scatter matrix They are as follows: in, is the kernel inter-class divergence matrix, is the intra-core divergence matrix, 0≤β≤1 represents a score, V is The eigenvector matrix of express The diagonal matrix composed of the eigenvalues ​​of and 0≤α≤1 represents a fraction, Q is The eigenvector matrix of The diagonal matrix composed of the eigenvalues ​​of and 2. The rockburst early warning method based on fractional-order kernel discriminant space learning according to claim 1 is characterized in that: The calculation formulas of the inter-class scatter matrix, the intra-class scatter matrix and the overall scatter matrix are as follows: S t =S b +S w ; Among them, the N j is the number of samples in the jth class, u j is the mean of the j-th sample, u j The calculation formula is u is the mean of all samples, and the calculation formula of u is X j is the set of samples of the jth class.

3. The rockburst early warning method based on fractional-order kernel discriminant space learning according to claim 2 is characterized in that: The objective optimization function of the linear discriminant space model is as follows: max W W T S b W; s.t. W T S w W=1; Where W is the spatial projection matrix of the linear discriminant space model.

4. The rockburst early warning method based on fractional-order kernel discriminant space learning according to claim 3 is characterized in that: The objective optimization function of the discriminant space analysis learning model is as follows: in, is the kernel inter-class divergence matrix, is the intra-kernel divergence matrix, is the spatial projection matrix of the kernel discriminant space, and the relationship between the kernel discriminant space projection matrix and the discriminant space projection matrix is 5. The rockburst early warning method based on fractional-order kernel discriminant space learning according to claim 4 is characterized in that: The kernel inter-class scatter matrix and the kernel intra-class divergence matrix The calculation formulas are as follows: Among them, K is the kernel matrix, and the (i, j)th element of K is The calculation formula is 6. The rockburst early warning method based on fractional-order kernel discriminant space learning according to claim 5 is characterized in that: The step of constructing a target optimization function of a fractional-order kernel discriminant space learning model according to the target optimization function of the discriminant analysis space model comprises: Based on the kernel inter-class scatter matrix and the kernel intra-class scatter matrix, constructing a fractional-order inter-class scatter matrix and a fractional-order intra-class scatter matrix respectively; Based on the fractional inter-class scatter matrix, the fractional intra-class scatter matrix and the objective optimization function of the discriminant analysis space model, the objective optimization function of the fractional-order kernel discriminant space learning model is constructed.

7. The rockburst early warning method based on fractional-order kernel discriminant space learning according to claim 1 is characterized in that: The step of optimizing and solving the projection matrix of the nuclear discriminant space according to the objective optimization function of the fractional-order nuclear discriminant space learning model comprises: Constructing the Lagrange multiplier function of the target optimization function of the fractional-order kernel discriminant space learning model with respect to the kernel discriminant space projection matrix; The derivative of the Lagrangian multiplier function with respect to the projection matrix of the kernel discriminant space is set to zero, and the eigenvector corresponding to the maximum eigenvalue of the Lagrangian multiplier function is solved; A spatial projection matrix of the kernel discriminant space is constructed according to the feature vector.

8. A rockburst warning device based on fractional-order kernel discriminant space learning, used to execute the rockburst warning method based on fractional-order kernel discriminant space learning according to any one of claims 1 to 7, characterized in that: include: An information acquisition conversion module is used to acquire rockburst index sample training data and rockburst index sample test data, and convert the rockburst index sample training data and the rockburst index sample test data into column vectors to generate a rockburst training sample set and a rockburst test sample set; A model building module, used for building a target optimization function of a fractional-order kernel discriminant space learning model according to the rockburst training sample set; A spatial projection matrix solving module, used for optimizing and solving the nuclear discriminant space projection matrix according to the objective optimization function of the fractional-order nuclear discriminant space learning model; A feature extraction module, used for applying the fractional kernel discriminant space projection matrix to the rockburst training sample set and the rockburst test sample set respectively, to generate a training discriminant feature set of the rockburst training sample set and a test discriminant feature set of the rockburst test sample set; The risk warning module is used to input the training identification feature set and the test identification feature set into a support vector machine classifier to generate a warning result of rock burst risk.

Images