A data-driven based layered rock deformation analysis method and device

Abstract

The application discloses a layered rock deformation analysis method and device based on data driving, which comprises the following steps: establishing a stress-strain database of layered rock based on test and data recognition algorithm; converting the database to the database of layered rock under a target angle by using a coordinate conversion formula of the database; constructing an over-determined equation for stress-strain data of each data point of the database, and taking a least square method solution as a balance factor matrix; finally, calculating the stress field, strain field and displacement field of the layered rock under the target angle under the loading and unloading action by using a local convex set data driving calculation method considering the noise influence of test data. The method is suitable for calculating the deformation and stress distribution of layered rock under the external loading and unloading action under different angles, and can carry out deformation analysis of layered rock under the loading and unloading action without defining the constitutive relation of layered rock, which has important significance for the research on rock mass which is a complex heterogeneous and anisotropic medium.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of layered rock deformation, and particularly relates to a layered rock deformation analysis method and device based on data driving. BACKGROUND

[0002] In nature, rocks with a layered structure account for two-thirds of the land area. Layered rock mass has obvious anisotropy characteristics, which is mainly affected by the direction of bedding plane and the mechanical properties of rock material. Correct understanding of the basic mechanical properties of layered rock mass is of great significance to the stability analysis and safety evaluation of rock mass engineering.

[0003] Before various numerical analysis methods of rock engineering are solved, the constitutive model and mechanical parameters of rock-soil material must be defined in advance, which directly affects the results of engineering stability evaluation. Due to the non-homogeneous and anisotropic characteristics of rock mass, so far there is no unified mature constitutive model that can fully and correctly represent the constitutive characteristics of various rock masses under any loading condition. The proposal and verification of new constitutive model require high intelligence and time cost, and various empirical assumptions may lead to calculation errors and decrease in model application range, while destroying the originality of test data.

[0004] In view of the above problems, the application aims to provide a numerical simulation technology without defining the constitutive model of rock mass, which can be used for deformation analysis of layered rock mass under loading and unloading, as well as excavation and reinforcement design of layered rock mass slope, foundation and underground cavern, and has important application value. SUMMARY

[0005] The application is carried out to solve the above problems, and aims to provide a layered rock deformation analysis method based on data driving, which bypasses the construction process of traditional constitutive model, directly calculates from large-scale test stress-strain data points, conservation laws and boundary conditions, greatly improves the rock mass simulation prediction ability, and the calculation results are more consistent with the actual situation.

[0006] In order to achieve the above purpose, the application adopts the following scheme:

[0007] A layered rock deformation analysis method based on data driving comprises the following steps:

[0008] Step 1: design a uniaxial compression test of a grooved rock sample, collect the strain field and loading force information in the test process, and establish an initial stress-strain database of layered rock based on a data recognition algorithm;

[0009] Step 2: convert the initial stress-strain database in step 1 to a target angle to obtain a converted stress-strain database of layered rock;

[0010] Step 3, constructing hyperstatic equations for stress-strain data of each data point in the converted stress-strain database to obtain a least square solution as a balance factor matrix;

[0011] Step 4, calculating stress field, strain field and displacement field of the target angle of the layered rock under the loading and unloading action by using a local convex set data-driven calculation method considering the influence of test data noise.

[0012] Further, the step 1 of establishing the initial stress-strain database of the layered rock based on the data recognition algorithm comprises:

[0013] Step 1: initializing mapping based on k-means algorithm;

[0014] Step 2: obtaining strain field of each unit by DIC, and calculating unit material strain according to the initial mapping result of Step 1;

[0015] Step 3: constructing a large sparse matrix, and solving the large sparse matrix based on a preconditioned bi-conjugate gradient method to calculate unit material stress and Lagrange multiplier;

[0016] Step 4: calculating stress field of each unit under X times of loading based on unit material stress and Lagrange multiplier;

[0017] Step 5: ordering the unordered point cloud by constructing a k-d tree to find the best mapping pointer of each unit under each loading condition and update it, and constructing a global target penalty function to keep the global target penalty function minimum;

[0018] Step 6: repeating Steps 2-5 until the initial mapping converges and remains unchanged.

[0019] Further, the formula for calculating unit material strain in Step 2 is as follows:

[0020]

[0021] wherein, is the unit material strain; ie X is the pointer for providing initial clustering, e=1, 2, 3...n is the number of finite element numerical integration points; is the integration weight of the integration point e; is the strain field of the integration point e under X times of loading condition.

[0022] Further, the large sparse matrix in Step 3 is as follows:

[0023]

[0024] Where, A is a diagonal block constant stiffness matrix, n is the total number of finite element numerical integration point number, B e is the strain conversion matrix of integration point e; C is the balance factor matrix; is the Xth loading condition, S 1 , S 2 S X are the non-diagonal block matrix under the first, second, Xth loading condition, is the integration weight of integration point e; δ e is the signal function, N mat is the number of material database generated by planning, ie X is the pointer to provide the initial clustering, M e is the mapping state of integration point e, λ 1 , λ 2 , λ X are the Lagrange multipliers under the first, second, Xth loading condition, respectively; is the element material stress; f 1 , f 2 f X are the node external force load vector array under the first, second, Xth loading condition, respectively.

[0025] The solving process using the double conjugate gradient method is: left multiplication and right multiplication of a sparse matrix by a pre-processing matrix P to balance the numerical magnitude of each row and column,

[0026]

[0027] In the formula, totalDof represents the total number of node degrees of freedom of the sampling model under various loading conditions, and S represents the independent stress components contained in each data point.

[0028] Further, the stress field formula of integration point e under the Xth loading condition calculated in Step 4 is as follows:

[0029]

[0030] Wherein, is the stress field of integration point e under the Xth loading condition.

[0031] Further, in Step 5, the unordered point cloud is ordered by constructing a k-d tree. The specific operation is: first, the data of mechanical state and material state are uniformly operated; then the material state is divided according to the dimension and the k-d tree is established by selecting the node according to the median; finally, for each kind of mechanical state, the material data point closest to it in the k-d tree is queried and the mapping pointer is recorded;

[0032] The global objective penalty function is:

[0033]

[0034] where G is the global objective penalty function; G X is the objective penalty function under the Xth loading condition, is the strain field of the integral point e under the Xth loading condition, is the element material strain;

[0035] Further, in step 2, the initial stress-strain database in step 1 is converted to the target angle based on the coordinate conversion formula, and a converted stress-strain database of the layered rock is obtained, and the coordinate conversion formula of the database is:

[0036]

[0037] where, represents the converted rock strain; represents the converted rock stress; θ is the angle between the global coordinate system and the local coordinate system; ε x is the x-direction strain; ε y is the y-direction strain; γ xy is the shear strain; σ x is the x-direction stress; σ y is the y-direction stress; τ xy is the shear stress.

[0038] Further, in step 3, by constructing the hyperstatic equation for the stress-strain data of each data point of the converted stress-strain database, the least square solution is taken as the balance factor matrix:

[0039]

[0040] In the formula, idata is the number of data points, and represent the rock stress-strain data pairs obtained by experiments, C ijkl is the least square solution balance constant matrix, N is 1, 2, 3… idata is the data point number.

[0041] Further, step 4 includes: performing linear interpolation on a plurality of stress-strain material data points closest to the mechanical state point to construct a local convex set, then projecting the mechanical state point onto the reconstructed related local convex manifold, and seeking the optimal material state search solution by the non-negative least square method;

[0042] For a given mechanical state point s z , first find the k material data points D(k, s z ) closest to it, and then use the following formula to obtain the weighting coefficient w i :

[0043]

[0044] with denotes the optimal material data point corresponding to a given mechanical state s z , then:

[0045]

[0046] where, denotes the distance between the mechanical state point s z and the i-th nearest material data point, w i is the weighting coefficient of the i-th material data point.

[0047] In another aspect, the present application provides a data-driven layered rock deformation analysis device, comprising:

[0048] A database establishment part is used to collect strain field and loading force information during the test, and establish an initial stress-strain database of layered rock based on a data recognition algorithm;

[0049] A coordinate conversion part is used to convert the initial stress-strain database to a target angle to obtain a converted stress-strain database of layered rock;

[0050] An equation construction part is used to construct a statically indeterminate equation for the stress-strain data of each data point of the converted stress-strain database, and the least square solution is used as the equilibrium factor matrix;

[0051] A calculation part is used to calculate the stress field, strain field and displacement field of the layered rock under the action of loading and unloading at the target angle by using a local convex set data-driven calculation method considering the influence of test data noise.

[0052] Compared with the prior art, the present application has the following beneficial effects:

[0053] 1) The present application can realize stress-strain analysis of layered rock under external loading and unloading, as well as excavation and reinforcement design of layered rock slope, foundation and underground chamber, without defining the constitutive relationship of the layered rock, which is a complex heterogeneous, discontinuous, nonlinear and anisotropic medium, and has important application value.

[0054] 2) The present application proposes a method for generating a stress-strain database of layered rock. Through digital image full-field measurement and external force loading boundary conditions, the stress components of the mechanical state under each loading condition and the stress-strain database of layered rock under all loading conditions can be identified without using constitutive equations.

[0055] 3) The application uses the coordinate conversion formula of the database to broaden the use range of the database. For the deformation problem of layered rock at different angles, multiple tests are not required, and only the layered rock database at a certain angle is converted to a suitable angle through the coordinate conversion formula.

[0056] 4) The application directly uses the stress-strain database to drive simulation calculation, and uses the local convex set data-driven algorithm to reduce the influence of test noise. Compared with the traditional numerical simulation analysis of rock mass, the error caused by empirical modeling is overcome, and the originality of the test data is preserved, so that the actual situation can be better simulated. BRIEF DESCRIPTION OF DRAWINGS

[0057] Figure 1 The flowchart of the layered rock deformation analysis method based on data driving related to the application;

[0058] Figure 2 The 0-degree layered rock size chart related to the embodiment of the application;

[0059] Figure 3 The stress scatter plot and its projection chart generated by training related to the embodiment of the application;

[0060] Figure 4 The database coordinate conversion principle diagram related to the embodiment of the application;

[0061] Figure 5 The 30-degree layered rock size chart related to the embodiment of the application;

[0062] Figure 6 The simplified mechanics and finite element mesh diagram of the 30-degree layered rock related to the embodiment of the application;

[0063] Figure 7 The x-direction displacement comparison cloud chart and the relative error chart of the representative points: (a) is the x-direction displacement measured by DIC; (b) is the x-direction displacement calculated by LCDDCM; (c) is the relative error chart of the x-direction;

[0064] Figure 8 The y-direction displacement comparison cloud chart and the relative error chart of the representative points: (a) is the y-direction displacement measured by DIC; (b) is the y-direction displacement calculated by LCDDCM; (c) is the relative error chart of the y-direction. DETAILED DESCRIPTION

[0065] In order to facilitate those skilled in the art to understand and implement the application, the application will be further described in detail below in combination with the drawings and the embodiments of the application. It should be understood that the embodiments described herein are only used to illustrate and explain the application, and are not used to limit the application.

[0066] As Figure 1 shown, the present application provides a data-driven-based layered rock deformation analysis method, comprising the following steps:

[0067] Step 1, collecting material stress-strain data under partial loading conditions during the test, and establishing an initial stress-strain database of layered rock under all loading conditions based on data recognition algorithm;

[0068] In step 1, by designing the test and collecting the loading force information and the corresponding strain field information during the test, combining the data recognition algorithm (DDI) to train the stress field information under various loading conditions, and the complete initial material stress-strain database under all loading conditions. In the past, when obtaining the stress field, it mostly depends on the assumption of uniform stress field, or in the case of non-uniform stress field, it depends on the inverse parameter identification method of the constitutive model. This method can directly obtain the stress-strain database without relying on the constitutive equation.

[0069] Step 1, based on the data recognition algorithm, establishes an initial stress-strain database of layered rock under all loading conditions, including:

[0070] Step 1: Initialize mapping based on k-means algorithm;

[0071] Step 2: Calculate the unit material strain according to the initial mapping result of Step 1;

[0072]

[0073] Where, is the unit material strain; ie X is the pointer to provide the initial cluster, e = 1, 2, 3... n is the finite element numerical integration point number; is the integration weight of integration point e; is the strain field of the Xth loading condition of integration point e.

[0074] Step 3: Construct a large sparse matrix and solve the large sparse matrix based on the preconditioned double conjugate gradient method to calculate the unit material stress and Lagrange multiplier;

[0075] The large sparse matrix is:

[0076]

[0077] A is the diagonal block constant stiffness matrix, n is the total number of finite element numerical integration points, B e is the strain conversion matrix of integration point e; C is the balance factor matrix; is the Xth loading condition of any, S1 S 2 S X These are the off-diagonal block matrices under the 1st, 2nd, and Xth loading conditions, respectively. The integral weight at integration point e; δ e Let N be a signal function. mat For the planned number of material databases, ie X To provide pointers to the initial clusters, M e Let λ be the mapping state of the integration point e. 1 , λ 2 , λ X These are the Lagrange multipliers under the 1st, 2nd, and Xth loading conditions, respectively; For the element material stress; f 1 f 2 f X These are the nodal external force load vector arrays under the 1st, 2nd, and Xth loading conditions, respectively.

[0078] Step 4: Calculate the stress field under the X loading conditions based on the element material stress and Lagrange multipliers;

[0079]

[0080] in, Let be the stress field under the Xth loading condition at integration point e.

[0081] Step 5: By constructing a kd-tree, the disordered point cloud is ordered, the optimal mapping pointer of each unit under each loading condition is found and updated, and a global objective penalty function is constructed to keep the global objective penalty function at its minimum.

[0082]

[0083] Where G is the global objective penalty function; G X Let X be the target penalty function under the loading condition of the Xth loading cycle. Let e ​​be the strain field under the Xth loading condition at the integration point. For the strain of the unit material;

[0084] Step 6: Repeat Steps 2-5 until the initialization mapping converges and remains unchanged.

[0085] In Step 1, initialize the mapping ie. X In this case, using an initial mapping based on the k-means algorithm instead of the original random initial mapping significantly improves the convergence efficiency of the data identification algorithm. Specifically, applying the k-means algorithm to strain fields... In the middle, according to the Euclidean distance squared metric, The data is divided into several clusters with the least strain difference, and in order to improve the distribution effect, secondary distribution is used and the number of data in each cluster is ensured to be approximately the same.

[0086] In Step 3, when solving large sparse matrices, the preconditioned double conjugate gradient method is used instead of the original left division to obtain better convergence results. The large sparse matrix has a large difference in magnitude between its elements, which can lead to a serious ill-conditioned matrix and even singularity. An unreasonable solution will result in large calculation errors and time-consuming. The specific operation is as follows: multiply the sparse matrix by a preconditioning matrix P on the left and right to balance the numerical magnitude of each row and column; then use the double conjugate gradient method to solve.

[0087]

[0088] In the formula, totalDof represents the total number of node degrees of freedom of the sampling model under various loading conditions, and S represents the independent stress components contained in each data point (1 for one-dimensional problems, 3 for two-dimensional problems, and 6 for three-dimensional problems); N mat The number of material databases generated for planning.

[0089] In Step 5, a best mapping pointer ie X is needed for each element under each loading condition, which is very expensive to calculate. By constructing a k-d tree, the unordered point cloud is ordered, which can achieve faster and more efficient retrieval operations compared to the conventional traversal method. The specific operation is as follows: first, normalize the data of the mechanical state and the material state, which is Then, divide the material state by dimension alternately (select the axis as the partition surface alternately with the depth of the tree) and select the node according to the median to establish the k-d tree; finally, query the nearest material data point in the k-d tree for each mechanical state and record ie X .

[0090] Step 2: Convert the initial stress-strain database in Step 1 to the target angle to obtain the converted stress-strain database of layered rock;

[0091] In Step 2, for the deformation problem of layered rock under various angles, separate experiments are not needed to obtain the stress-strain database under the corresponding angle, but only the coordinate conversion formula of the database is needed for conversion. The coordinate conversion formula of the database is:

[0092]

[0093] where, represents the converted rock strain; Represents the transformed rock stress; θ is the angle between the global and local coordinate systems (clockwise is positive); ε x Strain in the x-direction; ε y Strain in the y-direction; γ xy For shear strain; σ x The stress is in the x-direction; σ y Stress in the y-direction; τ xy This is shear stress.

[0094] Step 3: Construct statically indeterminate equations for the stress-strain data of each data point in the transformed stress-strain database, and use the least squares solution as the equilibrium factor matrix.

[0095] In step 3, statically indeterminate equations are constructed using stress-strain data from each data point in the database, and the least squares solution is used as the equilibrium factor matrix:

[0096]

[0097] In the formula, idata represents the number of data points. and This represents the rock stress-strain data obtained from the experiment, C ijkl is the equilibrium constant matrix of the least squares solution, N is 1, 2, 3…idata is the data point number.

[0098] Step 4: The stress field, strain field and displacement field of the layered rock at the target angle under loading and unloading are calculated using a local convex set data-driven calculation method that takes into account the influence of experimental data noise.

[0099] In step 4, in order to reduce the impact of data noise caused by material samples, measurement methods, test environment or human factors during the experiment, a local convex set data-driven calculation method with k=5 was adopted to replace the traditional distance minimization data-driven calculation method. This method can significantly reduce the impact of data noise and enhance the robustness and convergence stability of data-driven calculation.

[0100] Distance-minimum data-driven computation methods use spatial distance as the unit of measurement, extracting material properties from a single data point for each element. When data noise exists in the material database, it can easily amplify the noise from a single data point, thus affecting computational accuracy. To mitigate this impact, a local convex set data-driven computation method with k=5 is employed. This method first constructs a local convex set by performing linear interpolation on the few stress-strain material data points closest to the mechanical state point. Then, it projects the mechanical state point onto the reconstructed relevant local convex manifold and seeks the optimal material state search solution using non-negative least squares.

[0101] For a given mechanical state point s z, first find the nearest k material data points D(k, s z ), and then find the weighted coefficient w i :

[0102]

[0103] Let D(k, s ) represent the optimal material data point corresponding to the given mechanical state s z , then:

[0104]

[0105] where D(i, s ) represents the i-th nearest material data point to the mechanical state point s z , and w i is the weighted coefficient of the i-th material data point.

[0106] The LCDDCM uses the idea of local convexity to significantly reduce the impact of data noise, enhancing the robustness and convergence stability in data-driven computing. To ensure the effectiveness of linear approximation, in principle, the number of selected nearest points k should be greater than the expected dimension d of the underlying manifold (k = 5 is selected). It is worth noting that the stress and strain values in the database differ greatly, and in order to ensure the correctness of the solution, it is necessary to perform a normalization operation on the stress and strain before solving, which is

[0107] Embodiment 1

[0108] In this embodiment, the data-driven layered rock deformation analysis method provided by the present application is used for calculation, and the specific steps are as follows:

[0109] Step 1: A 0-degree layered rock sample with a length, width, and thickness of 100x100x10mm is cut from a parent rock sample, and the rock sample is notched to increase the non-uniformity of the displacement field, as shown in Figure 2 . A speckle field is prepared on the surface of the rock sample, and a rock mass test (RMT) system is used for uniaxial compression. In the test process, the RMT's built-in pressure sensor collects loading force information, and a commercial digital image correlation (DIC) system measures the rock sample's surface strain field. By collecting the loading force information and the corresponding strain field information, and combining the data recognition algorithm (DDI), the stress-strain database of the rock sample material can be trained. During the uniaxial compression of the rock sample, a total of N X (N X =100) times of loading force and strain field information are selected. The DIC surface analysis area is divided into a triangular grid with N elem =19408, and the total number of calculated mechanical states is N mecha=N elem ×N X =1940800. To ensure good performance of the DDI algorithm, the target generation rate selected in this training was 1%, and the final number of material stress-strain databases obtained from the training was N. * =N mecha / 100 = 19408. Figure 3 The trained stress scatter plot and its projections in various directions are shown.

[0110] Step 2: Use the database's coordinate transformation formula to convert it to a database of layered rocks at the target angle (30 degrees). The principle of database coordinate transformation is as follows... Figure 4 As shown (taking stress components as an example). The coordinate transformation formula in the database is:

[0111]

[0112] in, This represents the transformed rock strain; Represents the transformed rock stress; θ is the angle between the global and local coordinate systems (clockwise is positive); ε x Strain in the x-direction; ε y Strain in the y-direction; γ xy For shear strain; σ x The stress is in the x-direction; σ y Stress in the y-direction; τ xy This is shear stress.

[0113] Step 3: Construct statically indeterminate equations using stress-strain data from each data point in the 30-degree database, and use the least squares solution as the equilibrium factor matrix:

[0114]

[0115] In the formula, idata represents the number of data points. and This represents the rock stress-strain data obtained from the experiment, C ijkl This is the equilibrium constant matrix of the least squares solution.

[0116] Step 4: Cut a layered rock sample with dimensions of 100×100×10mm (length×width×thickness) from the same parent rock sample, such as... Figure 5 As shown. A speckle field was prepared on the rock sample surface, and uniaxial compression was performed on the rock sample using a rock mass testing (RMT) system. During the test, the displacement field on the rock sample surface was measured using a commercial digital image correlation (DIC) system. Based on existing techniques, a... Figure 6The shown layered rock numerical calculation model contains 4400 triangular elements. The loading problem of the rock sample is solved by using the above generated 30-degree layered rock database and the local convex set data-driven calculation method (LCDDCM) k=5 considering the noise effect of test data, and the data-driven results are compared with the test results obtained by DIC to verify the accuracy and reliability of the method.

[0117] Figure 7 The x-direction displacement contrast cloud and relative error map of the representative points are shown. Figure 8 The y-direction displacement contrast cloud and relative error map of the representative points are shown. It can be seen that the displacement fields measured by LCDDCM and DIC are smooth and similar in variation. At the same time, the displacement results obtained by using the LCDDCM method are more obvious and more symmetrical than the DIC results, because the data-driven results also satisfy the balance and compatibility constraints, and can better represent the stress-strain relationship of layered rock under loading and unloading.

[0118] Embodiment 2

[0119] The embodiment provides a layered rock deformation analysis device based on data driving, comprising:

[0120] A database establishing part is configured to collect material stress-strain data under partial loading conditions in a test process, and establish an initial stress-strain database of layered rock under all loading conditions based on a data recognition algorithm;

[0121] A coordinate conversion part is configured to convert the initial stress-strain database to a target angle to obtain a converted stress-strain database of layered rock;

[0122] An equation constructing part is configured to construct a statically indeterminate equation for stress-strain data of each data point of the converted stress-strain database, and take a least square solution as a balance factor matrix;

[0123] A calculation part is configured to calculate a stress field, a strain field and a displacement field of layered rock under loading and unloading at a target angle by using a local convex set data-driven calculation method considering the noise effect of test data.

[0124] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical range disclosed in the present application, which should be covered within the protection scope of the present application.

[0125] It should be understood that parts not elaborated in the specification are all prior art.

[0126] It should be understood that the above description is merely a detailed explanation of the preferred embodiments and is not intended to limit the patent protection scope of the present application. Any modification or alternation made by those skilled in the art without departing from the scope of the present application shall fall within the patent protection scope of the present application. The patent protection scope of the present application shall be subject to the appended claims.

Claims

1. A data-driven method for analyzing the deformation of layered rocks, characterized in that, Includes the following steps: Step 1: Design a uniaxial compression test for grooved rock samples, collect strain field and loading force information during the test, and establish an initial stress-strain database for layered rocks based on data recognition algorithms; include: Step 1: Initialize the mapping based on the k-means algorithm; Step 2: Obtain the strain field of each element through DIC, and calculate the strain of the element material based on the initial mapping results of Step 1; Step 3: Construct a large sparse matrix and solve the large sparse matrix based on the preprocessed double conjugate gradient method to calculate the element material stress and Lagrange multipliers; Step 4: Calculate the stress of each element based on the element material stress and Lagrange multipliers. X Stress field under secondary loading conditions; Step 5: By constructing a kd-tree, the disordered point cloud is ordered, the optimal mapping pointer of each unit under each loading condition is found and updated, and a global objective penalty function is constructed to keep the global objective penalty function at its minimum. Step 6: Repeat Steps 2-5 until the initial mapping converges and remains unchanged; Step 2: Convert the initial stress-strain database from Step 1 to the target angle to obtain the transformed stress-strain database of the layered rock; Step 3: Construct statically indeterminate equations for the stress-strain data of each data point in the transformed stress-strain database, and use the least squares solution as the equilibrium factor matrix. Step 4: Calculate the stress, strain, and displacement fields of the layered rock at the target angle under loading and unloading operations using a local convex set data-driven calculation method that considers the influence of experimental data noise. This includes: Linear interpolation is performed on the few stress-strain material data points closest to the mechanical state point to construct a local convex set. Then, the mechanical state point is projected onto the reconstructed relevant local convex manifold, and the optimal material state search solution is sought by non-negative least squares method. For a given mechanical state point First find the one closest to it. material data points The weighting coefficients can be calculated using the following formula. : use Represents a given mechanical state The corresponding optimal material data point, then: in, Represents the distance from the mechanical state point The most recent i One material data point, For the first i Weighting coefficients for each material data point.

2. The data-driven layered rock deformation analysis method according to claim 1, characterized in that: The formula for calculating the strain of the unit material in Step 2 is as follows: in, For the strain of the unit material; To provide pointers to the initial cluster, Numbering of finite element numerical integration points; For integration points e The integral weight; For integration points e No. X Strain field under secondary loading conditions.

3. The data-driven layered rock deformation analysis method according to claim 2, characterized in that: The large sparse matrix in Step 3 is: Where A is the constant stiffness matrix of the diagonal block, and n is the total number of finite element numerical integration points. For integration points e The strain transformation matrix; C This is the balance factor matrix; For any of the first X Sub-loading conditions, , These are the off-diagonal block matrices under the 1st, 2nd, and Xth loading conditions, respectively. For integration points e The integral weight; For signal functions, The number of material databases to be generated, To provide pointers to the initial cluster, The mapping state of integration point e, , , They are respectively the 1st, 2nd, and 3rd times. X Lagrange multipliers under sub-loading conditions; For element material stress; , They are respectively the 1st, 2nd, and 3rd times. X Vector array of nodal external force loads under secondary loading conditions; The solution process using the biconjugate gradient method involves left-multiplying and right-multiplying the sparse matrix by a preprocessing matrix. This is to balance the magnitude of the values ​​in each row and column. In the formula, This represents the total number of node degrees of freedom of the sampled model under various loading conditions. This represents the independent stress components contained in each data point.

4. The data-driven layered rock deformation analysis method according to claim 3, characterized in that: The integral point is calculated in Step 4. e No. X The stress field formula under the secondary loading condition is as follows: in, For integration points e No. X Stress field under secondary loading conditions.

5. The data-driven layered rock deformation analysis method according to claim 4, characterized in that: In Step 5, a kd-tree is constructed to arrange the disordered point cloud into an ordered one. The specific operations are as follows: First, the data of mechanical state and material state are uniformized; then, the material state is divided according to the dimension and the nodes are selected according to the median to build a kd-tree; finally, for each mechanical state, the nearest material data point is queried in the kd-tree and the mapping pointer is recorded. The global objective penalty function is: in, The global objective penalty function; For the first X The target penalty function under the sub-loading condition For the integration point e, the first... X Strain field under secondary loading conditions The strain is the unit material strain.

6. The data-driven layered rock deformation analysis method according to claim 1, characterized in that: In step 2, the initial stress-strain database from step 1 is transformed to the target angle based on the coordinate transformation formula to obtain the transformed stress-strain database of the layered rock. The coordinate transformation formula for the database is: in, This represents the transformed rock strain; This represents the transformed rock stress; The angle between the global coordinate system and the local coordinate system; Strain in the x-direction; Strain in the y-direction; For shear strain; Stress in the x-direction; Stress in the y-direction; This is shear stress.

7. The data-driven layered rock deformation analysis method according to claim 1, characterized in that: In step 3, statically indeterminate equations are constructed using the stress-strain data of each data point in the transformed stress-strain database, and the least squares solution is used as the equilibrium factor matrix: ； In the formula, idata represents the number of data points. and This represents the rock stress-strain data pairs obtained from the experiment. C ijkl The equilibrium constant matrix is ​​the least squares solution. N 1, 2, 3…idata represents the data point number.

8. A data-driven layered rock deformation analysis device, characterized in that, include: The database establishment department is used to collect strain field and loading force information during the test, and to establish an initial stress-strain database for layered rocks based on data recognition algorithms. The coordinate transformation unit is used to transform the initial stress-strain database to the target angle to obtain the transformed stress-strain database of layered rock. The equation construction section is used to construct statically indeterminate equations from the stress-strain data of each data point in the transformed stress-strain database, and the least squares solution is used as the equilibrium factor matrix. The computational unit is used to calculate the stress, strain, and displacement fields of layered rock at the target angle under loading and unloading operations using a local convex set data-driven computational method that takes into account the noise of experimental data. The data-driven layered rock deformation analysis device is used to perform the steps in the method according to any one of claims 1-7.

Images