Deep mixed granite critical damage breakthrough discrimination method based on three-dimensional stress traction

Abstract

The present application mainly relates to the technical field of deep migmatite damage penetration discrimination, in order to solve the problem that it is unable to directly discriminate whether the deep migmatite forms critical damage penetration, the present application provides a deep migmatite critical damage penetration discrimination method based on three-dimensional stress traction, the core of which is to uniformly disperse microcracks, mineral phase interfaces, schistosity weak surfaces and vein body boundaries in the deep migmatite into a weak structure surface element set with spatial position, normal, strength and boundary distance attributes; taking the surface element close to the penetration boundary as the starting point, a candidate expansion path is constructed according to spatial proximity and consistent orientation; the traction of the ground stress on the normal and tangent of the surface element is compared with the interface strength as the path passing condition; a damage penetration chain is formed through search iteration; whether the critical penetration state is reached is judged according to the distance from the first end and the last end of the penetration chain to the penetration boundary, and a direct penetration discrimination result is output.

Description

Technical Field

[0001] This invention mainly relates to the field of deep mixed rock damage penetration discrimination technology, and in particular to a method for critical damage penetration discrimination of deep mixed rock based on three-dimensional stress traction. Background Technology

[0002] Deep migmatites exist in high-stress and complex tectonic environments, typically exhibiting various weak structures such as microfractures, mineral phase interfaces, foliation weak surfaces, and vein boundaries. Under engineering disturbances or stress redistribution, damage often occurs along these weak structures through opening, slippage, or interface debonding, gradually forming continuous spatial channels. The critical damage penetration state is directly related to the identification of local instability, fracture connectivity, and anisotropic failure in deep surrounding rocks; therefore, damage propagation analysis needs to be conducted by combining three-dimensional structure, stress state, interface strength, and penetration boundaries.

[0003] In existing technologies, the determination of damage continuity in deep rock masses is typically based on 3D scanning, digital cores, microscopic images, or numerical modeling data. First, structural information such as fractures, bedding, joints, and mineral boundaries is extracted. Then, a fracture network or weak surface network is established based on spatial proximity, directional consistency, or connectivity probability. Some methods use finite element method (FEM), discrete element method (DEM), or interface element methods to calculate stress distribution and use tensile strength, shear strength, or interface failure criteria to determine local failure. Other methods abstract weak structures into graph nodes and edges, using shortest path, seepage model, A-star search, or heuristic graph search to find possible continuity paths, and output continuity determination results based on path cost, boundary distance, or connectivity.

[0004] The main shortcomings of existing technical solutions are that geometric connectivity search and mechanical activation determination are often processed separately. Once a candidate weak structure meets the spatial distance or directional conditions, it may enter the penetration path, making it difficult to ensure that the damage propagation process is constrained by the current three-dimensional stress traction state. At the same time, normal traction, tangential traction, and interface strength conditions are not embedded in the candidate element writing and cumulative damage cost update process, which can easily form geometrically continuous but mechanically inactive damage chains, making it difficult to reflect the true penetration process of deep mixed rock weak structures under three-dimensional stress. Summary of the Invention

[0005] This invention provides a method for determining critical damage penetration in deep mixed rock based on three-dimensional stress traction, aiming to solve the problem that the actual penetration process of weak structures in deep mixed rock under three-dimensional stress cannot be accurately simulated, and it is impossible to directly determine whether deep mixed rock has formed critical penetration failure.

[0006] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows:

[0007] A method for determining critical damage penetration in deep mixed rock based on three-dimensional stress traction, the method comprising:

[0008] S1. Obtain the three-dimensional structural data of deep mixed rock, the current three-dimensional stress state, the penetration boundary and the interface strength conditions. Extract the weak structural surface from the three-dimensional structural data of deep mixed rock and divide it into local surface elements according to the preset surface element scale. Assign each local surface element a spatial position, surface element normal, interface strength conditions and distance to the penetration boundary to form a set of weak structural surface elements.

[0009] S2. Determine the initial damage leading edge surface element based on the distance from the weak structure surface element to the through boundary. Select candidate surface elements according to the spatial distance between adjacent weak structure surface elements and the included angle of the surface element normals, and generate candidate extension relationships from the damage leading edge surface element to the candidate surface element.

[0010] S3. Establish a search model based on the initial damage leading edge surface element and candidate extension relationship, take the damage leading edge surface element as extension node, take the candidate extension relationship as the source of subsequent state, take the accumulated damage cost as path record, so that the subsequent state generates damage transfer between corresponding weak structure surface elements.

[0011] S4. In the search model, based on the current three-dimensional stress state, the normal of the damage leading edge surface element, the normal of the candidate surface element, and the interface strength condition, calculate the normal traction state and tangential traction state corresponding to the candidate expansion relationship, and generate the stress-activated passage state.

[0012] S5. Using the stress-activated passage state as the damage transfer condition, generate stress-activated extension edges for candidate extension relations that satisfy the damage transfer condition, and write the corresponding candidate surface elements into the set of surface elements to be extended, while writing the cumulative damage cost into the cumulative damage cost.

[0013] S6. Determine new damage leading edge elements based on the set of elements to be expanded and the accumulated damage cost, update the expansion nodes and subsequent state sources, and form a damage penetration chain according to the continuous connection relationship of stress-activated expansion edges.

[0014] S7. Based on the distances from the first and last weak structural elements of the damage penetration chain to the penetration boundary, determine whether the damage penetration chain is connected to the penetration boundary, and output the critical damage penetration state and the three-dimensional anisotropic failure discrimination results of the deep mixed rock.

[0015] Furthermore, the extraction of weak structural surfaces from the three-dimensional structural data of deep mixed rocks, as described in S1, includes:

[0016] The three-dimensional stress state, through boundary and interface strength conditions are associated with the spatial location corresponding to the three-dimensional structural data. Microcracks, mineral phase interfaces, foliation weak surfaces and vein boundaries are extracted and segmented according to the preset surface element scale to form a local surface element sequence.

[0017] Using the geometric center of each local surface element as its spatial location, least-squares local plane fitting is performed inside the local surface element to obtain the surface element normal. The distance to the penetrating boundary is calculated based on the interface strength conditions assigned according to the spatial location, thus forming a set of weak structure surface elements.

[0018] Furthermore, S2 includes:

[0019] Weak structural elements in the set of weak structural elements whose distance from the penetrating boundary is less than or equal to a preset boundary distance threshold are identified as initial damage leading edge elements.

[0020] Based on the spatial distance between the initial damage leading edge surface element and the adjacent weak structure surface element and the angle between the surface element normals, a structural continuity determination result is generated.

[0021] Based on the structural continuity determination results, adjacent weak structural surface elements with a spatial distance not greater than a preset extension distance threshold and a surface element normal angle not greater than a preset normal angle threshold are identified as candidate surface elements, and candidate extension relationships are generated from the damage leading edge surface element to the candidate surface element.

[0022] Furthermore, the calculation of the normal traction state and the tangential traction state described in S4 includes:

[0023] The current three-dimensional stress state is projected onto the normal of the damage leading edge surface element and the normal of the candidate surface element, respectively, to obtain the three-dimensional surface element traction component of the damage leading edge surface element and the three-dimensional surface element traction component of the candidate surface element.

[0024] The normal traction value of the damage leading edge surface element and the normal traction value of the candidate surface element are calculated based on the traction components of the three-dimensional surface element to form the normal traction state;

[0025] Based on the traction components of the three-dimensional surface elements, the tangential traction values ​​of the damage leading edge surface elements and the tangential traction values ​​of the candidate surface elements are calculated to form a tangential traction state.

[0026] Furthermore, the generation of stress-activated passage state described in S4 includes:

[0027] Based on the interface strength condition, the normal traction state and the tangential traction state are constrained and determined, and the opening determination, slip determination and interface debonding determination are generated.

[0028] Based on the opening determination, slip determination, and interface debonding determination, and combined with whether the candidate expansion direction and the entry direction of the damage front form a reverse compression relationship, a stress-activated passage state is generated; when at least one of the opening determination, slip determination, and interface debonding determination is passed, and the candidate expansion direction and the entry direction of the damage front do not form a reverse compression relationship, the stress-activated passage state is set to allow passage, otherwise it is set to restrict passage.

[0029] Furthermore, the search model described in S3 is a differentiable A-star search model, which includes a weak structure element encoding layer and a candidate extension relation cost layer. The weak structure element encoding layer is used to encode the attribute information of each weak structure element into a feature vector. The candidate extension relation cost layer is used to generate the cumulative damage cost writing amount corresponding to each candidate extension relation based on the damage leading edge element encoding features, candidate element encoding features, current three-dimensional stress state, spatial distance, element normal angle, and damage leading edge entry direction.

[0030] Furthermore, the cumulative damage cost written amount is:

[0031] ;

[0032] In the formula, The amount of data written to account for accumulated damage. For the damage leading edge element index; For candidate face element index; , , and These are the non-negative dimensionless weighting coefficients for the spatial distance term, the surface element normal angle term, the penetrating boundary distance term, and the interface strength term, respectively. For the damaged leading edge element With candidate face elements Spatial distance between them; The preset extension distance threshold; For the damaged leading edge element Surface element normal and candidate surface element The angle between the normals of the surface elements; The preset normal angle threshold; It is a positive dimensionless constant used to avoid the denominator of the included angle term of the surface element being zero; Candidate face elements The distance to the boundary that runs through it; For the damaged leading edge element The distance to the boundary that runs through it; To preset the boundary distance threshold, , and Candidate face elements Tensile strength, slip strength, and interfacial debonding strength; , and These are the leading edge surface elements of the damage. Tensile strength, slip strength, and interfacial debonding strength; This is a positive stress constant used to avoid the denominator of the interfacial strength term being zero, and it uses the same stress unit as the opening strength, slip strength, and interfacial debonding strength.

[0033] Furthermore, in S5, for multiple candidate extension relationships pointing to the same candidate surface element, the generation stress activation extension edge with the minimum cumulative damage cost is selected from the allowed candidate extension relationships, the candidate surface element is written into the set of surface elements to be extended, and the cumulative damage cost is updated.

[0034] Furthermore, the formation of a damage penetration chain according to the continuous connection relationship of the stress-activated extension edge as described in S6 includes: starting from the initial damage leading edge surface element, sequentially tracing the predecessor relationship of each surface element along the confirmed stress-activated extension edge, and arranging all the traced surface elements in sequence to form a damage penetration chain from the beginning to the end.

[0035] Furthermore, the determination of whether the damage penetration chain is connected to the penetration boundary in S7 includes: calculating the spatial distances from the weak structural surface element at the beginning and end of the damage penetration chain to the penetration boundary. If both distances are not greater than the preset penetration determination threshold, the damage penetration chain is determined to be connected to the penetration boundary, and the critical damage penetration state is output as penetrated; otherwise, it is determined to be not penetrated.

[0036] Beneficial effects of the present invention

[0037] (1) The judgment results directly correspond to the actual damage phenomena in the project. This invention directly outputs the critical damage penetration state and the spatial location of the penetration chain. The judgment results can be directly verified by the fracture surface, seepage or cutting results observed on site, and have good engineering interpretability and decision-making reference value.

[0038] (2) The identification of the penetration path conforms to the real physical process of damage propagation driven by geostress. This invention embeds the three-dimensional stress traction determination as a prerequisite for the expansion path into the search process, so that the final damage penetration chain is not only structurally continuous, but also a mechanically feasible penetration path under the current stress field. The identification results more realistically reflect the anisotropic failure law of deep mixed rocks.

[0039] (3) Collaborative analysis of multiple types of weak structural surfaces within the same framework. This invention incorporates four types of weak structural surfaces into the same analysis framework through unified surface element processing and type label assignment, enabling them to participate equally in the candidate path competition during the subsequent search process, thereby realizing collaborative analysis and comprehensive discrimination of multiple types of weak surfaces.

[0040] (4) It possesses data-driven parameter learning capabilities. The weakly structured surface element encoding layer and the candidate extension relation cost layer use measured destructive data or high-precision numerical simulation results as training labels, and are trained end-to-end through backpropagation to automatically optimize the combined weights of model parameters and cost functions. Compared with empirical formulas with fixed weight configurations, this mechanism has stronger data adaptability and generalization ability, making it easier to transfer and apply under different engineering site conditions.

[0041] (5) Achieving the transition from local damage identification to critical breakthrough discrimination. This invention elevates the analysis level from local damage identification to critical breakthrough discrimination by constructing a complete array of elements from one side of the breakthrough boundary to the other and providing a clear determination of whether the breakthrough boundary is connected, which better meets the essential needs of engineering stability assessment. Attached Figure Description

[0042] Figure 1 This is a flowchart of the critical damage penetration discrimination method for deep mixed rock based on three-dimensional stress traction according to the present invention;

[0043] Figure 2 A schematic diagram of the input layer formation for a set of weakly structured surface elements;

[0044] Figure 3 This is the core discrimination flowchart of the differentiable A-star search model;

[0045] Figure 4 This is a schematic diagram illustrating the critical breakthrough determination of the output layer of the differentiable A-star search model.

[0046] Figure 5 Comparison of the state before and after the critical damage breakthrough determination with the actual breakthrough result. Detailed Implementation

[0047] This invention discretizes various types of weak structural surfaces within rock masses into surface element objects with physical and mechanical properties, establishes a differentiable search framework under the dual constraints of structural connectivity and stress activation conditions, and achieves direct determination of the critical breakthrough state through complete path tracing from the initial boundary to the final boundary. Specifically, this is manifested in three levels:

[0048] Weak structural surfaces of different origins and scales, such as microfractures, mineral phase interfaces, foliation weak surfaces, and vein boundaries, are divided into local surface elements according to a unified scale. Each surface element is assigned attributes such as spatial location, surface element normal, interface strength condition, and distance to the penetrating boundary, forming a standardized set of weak structural surface elements that can be directly processed by search algorithms. This approach solves the problem of the complex mixture of various types of weak surfaces in deep migmatites, making unified characterization difficult.

[0049] The establishment of candidate propagation paths requires the simultaneous fulfillment of two conditions: first, at the structural level, adjacent surface elements must be spatially close and oriented in the same direction, possessing a basis for geometric continuity; second, at the mechanical level, the traction force generated by the current geostress in the normal and tangential directions of the surface element must be sufficient to overcome the interface strength, satisfying at least one failure mode among opening, slip, or interface debonding. The coupling of these two conditions ensures that the search process follows both the actual structural characteristics and the physical laws of geostress-driven damage propagation, overcoming the one-sidedness of traditional methods that rely solely on geometric connectivity or only consider the stress field.

[0050] The final output is not intermediate indicators such as damage variables or crack density, but an engineering judgment result based on the spatial distribution of the damage penetration chain and the critical damage penetration state (penetrated / not penetrated). The distances from the beginning and end of the penetration chain to the penetration boundary are the sole basis for determining whether critical penetration has occurred. The judgment criteria are clear and have a direct correspondence with the fracture surface, seepage staining, or cross-sectional results observed on the engineering site.

[0051] The technical solutions in this embodiment will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0052] like Figure 1 As shown, the critical damage penetration discrimination method for deep mixed rock based on three-dimensional stress traction of the present invention includes the following steps.

[0053] S1. Identification of weak structural surfaces and establishment of weak structural surface element sets

[0054] Weak structure surface identification and weak structure surface element set establishment, as follows Figure 2 As shown:

[0055] The three-dimensional structural data of deep migmatite is represented as a spatial sampling sequence in a unified three-dimensional coordinate system. . For spatial sampling quantity, For spatial sampling index, For the first Each spatial sample. Including spatial location ( , and The first Each spatial sample consists of three coordinate components in a unified three-dimensional coordinate system, grayscale or density values, mineral phase labels, void labels, foliation direction labels, and vein labels. Void labels are obtained by comparing grayscale or density values ​​with a preset void threshold, mineral phase labels are obtained by mineral phase segmentation, foliation direction labels are obtained by identifying the direction of local thin layers, and vein labels are obtained by segmenting vein regions.

[0056] The current three-dimensional stress state is represented as a stress sampling sequence. , The number of stress samples. For stress sampling index, For the first One stress sampling location, , respectively, represent the six stress components.

[0057] The through boundary is represented as a boundary sampling set. , To connect the boundaries, For the number of boundary samples, For boundary sampling index, For the first A boundary sampling point or the center of a boundary patch.

[0058] The interface strength condition is represented as a strength condition sampling sequence. , The number of samples for intensity conditions. For intensity condition sampling index, For the first Sampling locations under intensity conditions, , indicating the first The opening strength, slip strength, and interface debonding strength are corresponding to the sampling locations under each strength condition.

[0059] Register the three-dimensional structural data of deep mixed rock, the current three-dimensional stress state, the continuity boundary, and the interface strength conditions to the same three-dimensional coordinate system. For any spatial location... ,exist Selecting from and The stress sampling location with the smallest distance If the distance is less than or equal to a preset spatial matching threshold, then The corresponding current three-dimensional stress state is associated with When the current three-dimensional stress state is the global stress input, the same current three-dimensional stress state will be associated with all spatial locations. For any spatial location ,exist Selecting from and Sampling location with minimum distance for intensity conditions If the distance is less than or equal to a preset spatial matching threshold, then The corresponding interface strength condition is associated with For any spatial location ,Establish With the boundary The distance lookup relationship is used to determine the distance from microcracks, mineral phase interfaces, foliation weak surfaces, and vein boundaries to through boundaries.

[0060] From spatial sampling sequence When extracting weak structural surfaces, local connectivity is first defined. Locally connected surfaces labeled as fractures with an aperture not less than a preset aperture threshold are extracted as microfractures; local interfaces formed between adjacent sampled mineral phase labels are extracted as mineral phase interfaces; continuous thin bedding planes with foliation directions meeting preset alignment conditions and thicknesses not exceeding a preset foliation thickness threshold are extracted as foliation weak surfaces; and local interfaces adjacent to vein labels and surrounding rock labels are extracted as vein boundaries. The extraction results are then segmented according to a preset surface element scale to form a local surface element sequence. ,in, For the first A local surface element, The number of local face elements. This is a local surface element index.

[0061] local elements The geometric center is determined as the spatial position. Set the internal sampling points Perform least-squares local plane fitting, and determine the unit normal direction of the fitted plane as the surface element normal. , , and For surface element normal The three directional components. When the fitting produces two opposite normals, the sampling point closest to the through boundary is selected. point to The direction angle between the candidate unit normal direction and the nearest through-boundary sampling point is less than or equal to the preset direction threshold; point to If the directional angle is greater than a preset directional threshold, the candidate unit normal direction will be reversed. The surface element normal is the normal direction of the local surface where the weak structure surface element is located.

[0062] For each local element When assigning interface strength conditions, based on spatial location exist Select the nearest intensity condition sampling location ,Will corresponding Assign local surface elements When the interface strength conditions are given in the type table of microcracks, mineral phase interfaces, foliation weak surfaces, and vein boundaries, they are determined according to the local surface element. The source type is selected to correspond to the interface strength condition, and a formation is formed. The interface strength condition assignment involves applying the interface strength condition to the corresponding weak structural elements of microcracks, mineral phase interfaces, foliation weak surfaces, and vein boundaries. When determining the distance based on the through boundary, the following steps are taken: arrive The spatial distance to the nearest boundary sampling point or the nearest boundary patch is determined as the distance to the through boundary. .

[0063] The local surface element whose spatial location, surface element normal, interface strength condition, and distance to the penetrating boundary have been assigned values. Convert to weak structure surface element Each weak structure surface element Forming sixteen-dimensional input features , For local surface elements The area scale; when microcracks are extracted as linear crack center surfaces. The length dimension along the direction of crack propagation; For local surface elements The opening or interface thickness; This is a four-dimensional type indicator (the component corresponding to the source type is 1, otherwise 0). (And...) The associated current three-dimensional stress state is denoted as , The stress components are six-dimensional. All weak structural elements are considered. This is used to form a set of weakly structured surface elements. .

[0064] S2. Initial Leading Edge Surface Determination and Candidate Extension Path Construction

[0065] For each weak structure element According to spatial location To the boundary The spatial distance between the nearest boundary sampling point or the nearest boundary patch is used to generate the boundary distance. When weak structural surface elements Distance already carried to the through boundary At that time, As boundary distance .

[0066] Set the preset boundary distance threshold. Select initial damage leading edge elements from the set of weakly structured elements. A preset boundary distance threshold is used. The spatial resolution and preset pixel scale of the deep mixed rock 3D structure data are pre-defined and remain unchanged in the same discrimination task. When the boundary distance... Not greater than the preset boundary distance threshold At that time, weak structural surface elements The initial damage leading edge elements are identified. All initial damage leading edge elements are represented as follows: ,in, This is the initial set of surface elements at the leading edge of the damage. This represents the initial number of surface elements at the leading edge of the damage. For the damage leading edge element index, For the first Each initial damage leading edge element. The sixteen-dimensional input features of the source weakly structured surface elements are preserved.

[0067] For each initial damage leading edge element In the set of weakly structured surface elements Excluding the initial damage leading edge element Weakly structured elements with the same element number are separated by a preset expansion distance threshold. Select adjacent weakly structured surface elements for the neighborhood search radius. Preset the expansion distance threshold. The preset element size and allowable spacing between weak structures are given in advance and remain unchanged in the same discrimination task. The first... The set of adjacent weak structure elements corresponding to each initial damage leading edge element is represented as follows: ,in, For adjacent weak structure surface element indexes, For the first Adjacent weak structure surface elements Adjacent weak structure surface elements The face element number, For the initial damage leading edge element The face element number.

[0068] For each adjacent weak structure element Based on the initial damage leading edge surface element Spatial location and adjacent weak structural elements Spatial location, generating spatial distance Spatial distance It is obtained by taking the square root of the sum of the squares of the differences between the three coordinate components. Based on the initial damage leading edge element... normal to the surface element and adjacent weak structure surface elements The normal of the element generates the angle between the normals of the element. Surface element normal angle The angle is obtained by using the inverse cosine of the inner product of the normals of two unit surface elements; when the normals of the two unit surface elements have opposite directions, the smaller angle between the normals of the two unit surface elements is taken as the angle of the surface element normal. .

[0069] Set the preset normal angle threshold. Preset normal angle threshold The allowable structural plane orientation deviation is predetermined and remains constant throughout the same discrimination task. When the spatial distance... Not greater than the preset extension distance threshold And the angle between the normals of the face elements Not greater than the preset normal angle threshold At that time, a structural continuity determination result with the value "pass" is generated. When spatial distance Greater than the preset extension distance threshold Or the angle between the normals of the face elements Greater than the preset normal angle threshold At that time, a structural continuity judgment result with the value of "failed" is generated. The structural continuity determination result is based on the spatial distance and the included angle between the surface element normals to determine whether the initial damage leading edge surface element and the adjacent weak structural surface element have the potential to form a candidate extension relationship.

[0070] The result of the structural continuity determination is passed. Corresponding adjacent weak structure elements The candidate surface element is identified. For each candidate surface element, a candidate extension relation is generated, pointing from the damage leading edge surface element to the candidate surface element. Each candidate extension relationship Includes the surface element number of the damage leading edge surface element, the surface element number of the candidate surface element, the spatial position of the damage leading edge surface element, the spatial position of the candidate surface element, the normal of the damage leading edge surface element, the normal of the candidate surface element, and the spatial distance. Surface element normal angle Boundary distance from candidate surface element to through boundary and structural continuity determination results .

[0071] Each candidate extended relation The corresponding structure fields are represented as follows . For candidate extended relations The structure fields, For the initial damage leading edge element The sixteen-dimensional input features Candidate face elements The sixteen-dimensional input features. Candidate extension relations. Only the damage leading edge surface element and the candidate surface element have the possibility of structural continuity. Candidate extension relationship. It is not considered a connection relationship that has already met the damage transfer conditions, does not generate stress-activated expansion edges, does not trigger the writing of candidate surface elements into the set of surface elements to be expanded, and does not trigger the update of cumulative damage cost.

[0072] S3. Establishment of a differentiable A-star search model

[0073] like Figure 3 As shown, the initial damage leading edge element is... The first expansion node of the differentiable A-star search model was determined. ( (Index to the damage leading edge element). Based on the initial damage leading edge element. The element number forms the starting item of the path record. The starting item of the path record. Includes initial damage leading edge elements The element number and initial cumulative damage cost are set to zero, and the starting item of the path record is... The element number does not include candidate elements. The path record contains path information used to write the damage transfer and corresponding cumulative damage cost between the determined weak structure elements.

[0074] Based on candidate extension relationships The damage leading edge element number recorded in the data is used for the first extended node. Matching the successor state source from the damage leading edge element to the candidate element. The damage leading edge element number is compared with the initial damage leading edge element. Candidate extension relationship with consistent element number Write and will be written Candidate extension relation The source of the successor state to be generated is defined as the candidate extension relation of the candidate surface element selected by the differentiable A-star search model during successor state generation. The source of the successor state to be generated is the candidate extension relation of the candidate surface element that has not undergone stress activation and passability state determination and has not yet been generated during the successor state generation and cumulative damage cost writing stages.

[0075] Construct a differentiable A-star search model that uses damage leading edge elements as expansion nodes, candidate expansion relationships as the source of subsequent states, and cumulative damage cost as the path record. Differentiable A-star search model include:

[0076] Weak structure surface element coding layer , These are the weights and biases for the weakly structured surface-level encoding layer. The weakly structured surface-level encoding layer receives sixteen-dimensional input features. The sequence passes through two fully connected layers with identical 32 neurons each. Each neuron is connected to all neurons in the previous layer. Each neuron includes weights, biases, and rectified linear units, outputting 32-dimensional weakly structured surface element encoded features. , to Thirty-two coding components are generated for the weak structure surface element coding layer.

[0077] Candidate extended relation cost layer , The weights and biases of the cost layer for each candidate extension relation. Damage leading edge element 32-dimensional weakly structured surface element encoding features and candidate surface elements Thirty-two weakly structured surface element encoding features and candidate extension relationships The associated six-dimensional current three-dimensional stress state, the one-dimensional spatial distance between the damage leading edge surface element and the candidate surface element, the one-dimensional angle between the normal of the damage leading edge surface element and the normal of the candidate surface element, and the three-dimensional damage leading edge entry direction are combined to form the internal features of the seventy-five-dimensional candidate extension relationship. The current three-dimensional stress state in the six-dimensional model includes three normal stress components and three shear stress components. The direction of entry of the three-dimensional damage front is represented as follows: , For the initial damage leading edge element The corresponding direction of entry of the damage front, , and These are the three directional components of the damage leading edge entry direction. The three-dimensional damage leading edge entry direction is determined at the initial damage leading edge element based on the normal of the penetrating boundary; when the penetrating boundary is given by boundary sampling points, the unit direction pointing from the nearest boundary sampling point to the spatial position of the initial damage leading edge element is used. The candidate extension relation cost layer receives the internal features of the seventy-five-dimensional candidate extension relation. It passes through a fully connected layer of 64 neurons, a fully connected layer of 32 neurons, and an output layer of one neuron in sequence.

[0078] Based on the spatial location, normal direction, interface strength condition, and distance to the penetration boundary of the damage leading edge surface element and candidate surface element in the weak structure surface element set, calculate the cumulative damage cost writing amount corresponding to the candidate expansion relationship. The output layer of a neuron in the candidate extended relation cost layer generates a one-dimensional single-step cost value, which is calculated using the cumulative damage cost write amount. It is represented and dimensionlessly combined according to the spatial distance term, the surface element normal angle term, the penetrating boundary distance term, and the interface strength term:

[0079] ;

[0080] in, The non-negative dimensionless weight of the spatial distance term. The non-negative dimensionless weight of the angle term of the surface element normal. For the non-negative dimensionless weight of the boundary distance term, The non-negative dimensionless weights for the interface strength item are set to the four preset dimensionless weights when no configuration parameters are specified. ; For the damaged leading edge element With candidate face elements Spatial distance between them; The preset extension distance threshold; For the damaged leading edge element Surface element normal and candidate surface element The angle between the normals of the surface elements; The preset normal angle threshold; It is a positive dimensionless constant used to avoid the denominator of the included angle term of the surface element being zero; Candidate face elements The distance to the boundary that runs through it; For the damaged leading edge element The distance to the boundary that runs through it; To preset the boundary distance threshold, , and Candidate face elements Tensile strength, slip strength, and interfacial debonding strength; , and These are the leading edge surface elements of the damage. Tensile strength, slip strength, and interfacial debonding strength; This is a positive stress constant used to avoid the denominator of the interfacial strength term being zero, and it uses the same stress unit as the opening strength, slip strength, and interfacial debonding strength.

[0081] Starting item based on path record and cumulative damage cost write volume Build search state Search status This includes the current damage leading edge element, candidate extension relationships, damage leading edge entry direction, path record, and cumulative damage cost. The search state is an intermediate state of a differentiable A-star search model containing the current damage leading edge element, candidate extension relationships, and cumulative damage cost. The cumulative damage cost is the damage cost accumulated along the path record. The amount of cumulative damage cost written before generating the stress-activated passage state is [amount missing]. Only as a candidate extended relation The amount to be written is retained, and the cumulative damage cost is not written.

[0082] Search status Each candidate extended relationship Pointing to candidate face element Set as the object to be written. Associate the object to be written with the candidate expansion relationship. Binding enables candidate extension relationships In the differentiable A-star search model, the state change corresponds to the expansion of the damage leading edge surface element into the candidate surface element. The damage transfer between weak structure surface elements is a state change of the expansion of the damage leading edge surface element into the candidate surface element.

[0083] The operation of writing candidate face elements into the set of face elements to be expanded is limited to the subsequent state generation and cumulative damage cost writing stages. The subsequent state generation and cumulative damage cost writing stages are configured as search states. The update process. Candidate extension relationships. In the differentiable A-star search model, it is limited to the source of successor states, and is a candidate extension relation. Corresponding candidate face elements Limited to objects to be written; candidate extension relations before generating stress-activated pass states. Without writing path records, the cumulative damage cost is written. The cumulative damage cost is not written; when the damage transfer condition is met in the stress-activated passage state, the candidate extension relation is... Generate stress-activated extended edges and write the accumulated damage cost into the quantity. The cumulative damage cost is written in. This leads to the differentiable A-star search model. .

[0084] S4. Three-dimensional stress traction determination and stress activation passage state generation

[0085] Search model for differentiable A-star In the subsequent state generation and cost writing stages, candidate extension relations are received. .according to The damage leading edge element number and candidate element number are recorded in the weak structure element set. Reading the leading edge surface element of the damage and candidate face elements Based on the leading edge surface element of the damage Extract the normal of the damage leading edge element , , and These are the three directional components of the normal to the leading edge surface element. Based on the candidate surface element... Extracting candidate surface normals , , and These are the three directional components of the candidate surface element normal. Read the candidate extension relation. Associated current three-dimensional stress state Three The components are normal stress components, three in total. The component is the shear stress component. When the damage leading edge element... and candidate face elements When each of the six stress components is associated with the current three-dimensional stress state, the average value of each component is taken to form... When the current three-dimensional stress state is the overall stress input, the overall six-dimensional current three-dimensional stress state is taken as... .

[0086] Based on the surface element of the leading edge of the damage Interface strength conditions and candidate surface elements Based on the interface strength condition, candidate extension relationships are generated. Corresponding interface strength conditions , For tensile strength, For slip strength, This represents the interfacial debonding strength. (Generation) At that time, for the damaged leading edge element and candidate face elements The tensile strength, slip strength, and interfacial debonding strength were compared item by item, and the strength with the larger value in each item was selected as the candidate extension relationship. The corresponding strength. By taking the largest value for each item, the candidate extended relationship is made more consistent. It has a unified three-dimensional interface strength condition when entering the three-dimensional stress traction judgment.

[0087] The direction of entry into the damage front is represented as... , , and These are the three directional components of the damage leading edge entry direction. The damage leading edge entry direction of the initial damage leading edge element is determined based on the normal of the penetrating boundary; for damage leading edge elements with already generated stress-activated extension edges, the direction of the already generated stress-activated extension edges is used as the damage leading edge entry direction. The current three-dimensional stress state is then considered in six dimensions. 3D damage leading edge surface element normal 3D candidate surface normal Three-dimensional interface strength conditions and the direction of entry of the three-dimensional damage front Composition of 18-dimensional decision input Perform three-dimensional stress traction determination. When calculating the surface element traction components, normal traction state, and tangential traction state, the same traction decomposition rule is used to project the current three-dimensional stress state onto the surface element directions corresponding to the damage leading edge surface element normal and the candidate surface element normal, respectively:

[0088] ;

[0089] in, The three-dimensional surface element traction component is formed by the current three-dimensional stress state acting on the normal of the damage leading edge surface element; The three-dimensional surface element traction component is formed by the current three-dimensional stress state acting on the normal of the candidate surface element; For the normal direction of the leading edge element of the damage; For candidate element normals; For candidate extended relations The corresponding two-dimensional normal traction state; For candidate extended relations The corresponding two-dimensional tangential traction state; This is a transpose operation. All stress components in the formula use the same stress unit, and the normal to the damage leading edge element is... and candidate element normal All are dimensionless unit vectors, in normal traction state. and tangential traction state All values ​​are in stress units. If the value in the square root operation is less than zero due to floating-point rounding, the value in the square root operation will be set to zero.

[0090] Will The first component is determined as the normal traction value of the damage leading edge element. ,Will The second component is determined as the normal traction value of the candidate surface element. A positive normal traction value indicates tension, while a negative normal traction value indicates compression. The first component is determined as the tangential traction value of the damage leading edge element. ,Will The second component is determined as the tangential traction value of the candidate surface element. The tangential traction value is the non-negative amplitude of the force component within the local tangential plane.

[0091] According to the interface strength condition For normal traction state and tangential traction state Perform constraint determination. and The normal traction value of the tensioning action is respectively with Comparison; when the normal traction value of any tension action is not less than At that time, the opening determination value is passed. Otherwise, a failure to open the property will be generated. .

[0092] Will and respectively with Comparison; when any tangential traction value is not less than At that time, a slip decision with a value of "pass" is generated. Otherwise, a slip decision with a value of "failed" will be generated. According to the normal traction state Tangential traction state and interface strength conditions Generate interface de-adhesion judgment When the normal traction value of the tensioning action on any surface element is not less than... Or, the tangential traction value of any surface element is not less than At that time, a value of "passing" is generated for the interface de-adhesion determination. Otherwise, a failure to pass the interface de-adhesion judgment will be generated. .

[0093] According to the opening judgment Slip determination Interface debonding determination and the direction of entry of the leading edge of the damage Generate stress-activated passage state The damaged leading edge element will be Spatial location pointing to candidate surface element The unit direction is denoted as the candidate expansion direction. Candidate expansion directions Based on the surface element of the leading edge of the damage Spatial location and candidate surface element The spatial location is determined first by forming surface elements at the leading edge of the damage. Pointing to candidate face elements The three-dimensional direction vector is then normalized. A preset reverse angle threshold is set. , An angle threshold used to identify whether a candidate propagation direction is reversed relative to the direction of entry into the lesion front. Candidate propagation direction relative to the direction of entry of the damage front edge The angle between the two unit direction vectors is obtained by the inner product of the inverse cosine.

[0094] If open judgment Slip determination Debonding determination of interface At least one of them is passed, and the candidate expansion direction relative to the direction of entry of the damage front edge If a reverse compression relationship is not formed, the stress will activate the passage state. Set to allow passage; if open, determine Slip determination Debonding determination of interface All were rejected, or were candidate expansion directions. relative to the direction of entry of the damage front edge If a reverse compression relationship is formed, the stress will activate the passage state. Set to restricted passage. The reverse containment relationship is determined jointly by the directional angle and the normal traction value, when the candidate expansion direction... relative to the direction of entry of the damage front edge The included angle is greater than the preset reverse included angle threshold. and normal traction state When both normal traction values ​​are in a compressive effect, it is determined that a reverse compressive relationship exists.

[0095] Normal pulling state Tangential traction state Opening Judgment Slip determination Interface debonding determination and stress-activated passage state Write candidate extension relationship The state field forms a candidate expansion relationship bound to a stress-activated pass state. The stress-activated pass state is formed based on the normal traction state, tangential traction state, and interface strength condition, and is used to determine whether the candidate expansion relationship allows the writing of the set of surface elements to be expanded and the cumulative damage cost. The bound candidate expansion relationship remains within the subsequent state generation and cumulative damage cost writing stages of the differentiable A-star search model. Whether a candidate surface element is written into the set of surface elements to be expanded and whether the cumulative damage cost is written are both determined by the stress-activated pass state. This serves as the basis for judgment.

[0096] S5. Confirmation of effective expansion path and advancement of damage front.

[0097] In the subsequent state generation and cumulative damage cost writing stages of the differentiable A-star search model, candidate extension relations are received. and stress-activated passage state Based on candidate extension relations The damage leading edge element number and candidate element number are used to read the damage leading edge element from the weak structure element set. and candidate face elements And form candidate extension relationships bound to stress-activated passage states. .

[0098] Set the damage transfer condition to stress-activated pass state. The value is "allow passage". For each bound candidate extension relationship with stress-activated passage state,... In candidate face elements Before writing to the set of elements to be expanded, activate the pass-through state for stress. Compare with damage transfer conditions. When stress activates the passage state. When the value is set to allow passage, candidate extension relationships for stress-activated passage states will be bound. Record candidate extension relationships that satisfy the damage transfer conditions; when stress activates the pass state When the value is set to restrict passage, candidate extension relationships will be bound to the stress-activated passage state. Recorded as candidate extension relationships that do not meet the damage transfer criteria.

[0099] Candidate extension relationships that meet the damage transfer criteria Determine the leading edge surface element of the damage To candidate face elements Damage transfer between weak structural elements and generation of stress-activated extension edges. . For the damaged leading edge element Pointing to candidate face elements Stress-activated extended edge. This includes the damage leading edge element number, candidate element number, stress activation passage state, candidate propagation direction, and cumulative damage cost corresponding to the candidate propagation relationship. The stress activation propagation edge number is generated by combining the damage leading edge element number and the candidate element number, and remains unique in one search state update of the differentiable A-star search model.

[0100] Based on stress activation of extended edge Extract stress to activate extended edges Corresponding candidate face elements Candidate face elements Write to the set of face elements to be expanded and candidate face elements The damaged leading edge elements are identified as those to be expanded. The set of elements to be expanded. China targets candidate face elements Save the candidate surface element number, the preceding damage leading edge surface element number, the stress-activated extension edge number, the candidate extension direction, and the current cumulative damage cost, so that the differentiable A-star search model can select new extension nodes according to the damage leading edge state update method.

[0101] Based on stress activation of extended edge Corresponding candidate extension relationship Read the one-dimensional single-step cost value or cumulative damage cost written amount generated by the candidate extended relation cost layer. Read the damage leading edge element. Corresponding existing cumulative damage cost The cost of damage Write the cumulative damage cost and form candidate surface elements. The corresponding updated cumulative damage cost.

[0102] When multiple candidate extension relationships point to the same candidate element In the case of candidate face elements with the same number and candidate face elements being The candidate extension relationships of the bound stress-activated passage state are composed of the candidate surface element relationship group. For the same candidate surface element relationship group For each candidate extension relationship, the passage state is first activated based on stress. Generate access markers, and then based on the damage leading edge elements Existing cumulative damage cost and damage cost This generates an updated cumulative damage cost. For the same candidate surface element relationship group... If at least one pass flag has a value of 1, select the stress-activated extension edge corresponding to the smallest updated cumulative damage cost from the allowed candidate extension relations:

[0103] ;

[0104] in, For candidate extended relations The passage flag has a value of 1 indicating that passage is permitted and a value of 0 indicating that passage is restricted. For candidate extended relations Pointing to candidate face elements The updated cumulative damage cost formed at that time; For the damaged leading edge element The cumulative damage cost that has already been written; For candidate extended relations The corresponding damage cost; To be in the same candidate surface element relationship group The damage leading edge element index that minimizes the updated cumulative damage cost value; To select from the damage leading edge element index that meets the conditions The index with the smallest value; To bind candidate extension relationships for stressed-activated passage states; For the same candidate surface element relationship group; Candidate face elements Write to the set of face elements to be expanded The cumulative damage cost used at that time; Candidate face elements Write to the set of face elements to be expanded The stress-activated extended edge is used at that time; For the damaged leading edge element Pointing to candidate face elements The stress activates the extended edge. In the formula... , , and All use the same damage cost unit; when the cumulative damage cost write amount is configured as dimensionless, , , and All values ​​are dimensionless. If multiple damage leading edge element indices correspond to the same minimum... The surface element with the smallest surface element number was selected as the leading edge surface element. If the candidate surface element is related to the group There are no candidate extension relationships with a pass flag value of one, so no execution is performed. Select and group the candidate face elements together. All candidate extension relationships were treated as not meeting the damage transfer condition.

[0105] according to Same candidate element Write to the set of face elements to be expanded and in accordance with Update candidate face elements In the set of face elements to be expanded The cumulative damage cost in the set of elements to be expanded. There are already candidate face elements in the middle. When recording, the set of face elements to be expanded will be... Candidate face elements The preceding damage leading edge element number, stress-activated extension edge number, and current cumulative damage cost are replaced with and The corresponding data; when the set of face elements to be expanded There are no candidate face elements. When recording, in the set of face elements to be expanded New candidate elements in China The record.

[0106] Candidate extension relationships that do not meet the damage transfer criteria Determine candidate face elements Not satisfied with writing to the set of face elements to be expanded The conditions are given, and extended restricted relations are generated. Extending Restricted Relations Save the damage leading edge element number, candidate element number, stress activation passage status, and confined marker. Extend the confined relation. No stress-activated extended edges are generated, and they are not written into the set of elements to be extended. No damage cost is triggered and written to the accumulated damage cost. Candidate face elements are written to the set of face elements to be expanded. Both the cumulative damage cost update and the candidate extension relation that satisfies the damage transfer condition in the stress-activated passage state are limited.

[0107] S6. Iterative generation of the damage-connected chain

[0108] Receiver stress-activated extended edge in differentiable A-star search model Extending Restricted Relationships Set of surface elements to be expanded and cumulative damage cost table ,in, For the first The cumulative damage cost corresponding to each weak structure surface element. This represents the number of weakly structured surface elements. The weakly structured surface elements that are not written to the path record correspond to... Set to preset maximum cost , This is a preset value that is greater than the cumulative damage cost allowed for any write operation.

[0109] Differentiable A-star search model reads current path record and the source of the current and successor states . Record the current path. This represents the number of records in the current path. Index for path record items, For the first Each path record item This is the source of subsequent state that has not yet been filtered out and written. Each path record item... It includes the damage leading edge element number, candidate element number, stress-activated extension edge number, candidate extension direction, damage cost, and cumulative damage cost after writing. Each subsequent state source includes the candidate extension relationship number, damage leading edge element number, candidate element number, and a marker to be generated.

[0110] Based on stress activation of extended edge The candidate extension relation number, damage leading edge element number, and candidate element number stored in the database are used to locate and generate stress-activated extension edges. Candidate extension relation The generated stress will activate the extended edge. Candidate extension relation Candidate extension relationships identified as writable path records Candidate extension relation Only when stress is generated can the extended edge be activated. Access to path records is only allowed after that. Candidate extension relationships for which stress-activated extension edges are not generated are not included in the path record. .

[0111] Based on the candidate extension relationship of the writable path record Stress activation extended edge Corresponding damage leading edge element and candidate face elements Write path record This creates a path record containing damage transfer information between weakly structured surface elements. A new path record item is added. Write damage leading edge element face number, candidate face Surface element number, stress-activated extended edge The edge number, derived from the damaged leading edge element Pointing to candidate face elements Candidate propagation direction, damage cost and candidate face elements Corresponding cumulative damage cost after update . For candidate extended relations Generate stress-activated extended edges The single damage cost is written as the cumulative damage cost at the time. Candidate face elements Write path record The corresponding cumulative damage cost. Read the cumulative damage cost table. Middle damage leading edge surface element Corresponding cumulative damage cost ,Will With damage cost Add them together to get and the cumulative damage cost table Candidate face elements The corresponding entry is updated to .

[0112] Based on extended restricted relations The damage leading edge element number and candidate element number are stored in the database to locate the source of the current successor state. The corresponding candidate extension relationships that do not meet the damage transfer conditions are identified. The located candidate extension relationships are then moved from the current successor state source. Screening and expanding restricted relations The restricted tags are updated to those that have been filtered out. Extend the restricted relationships. No stress-activated extended edges are generated, and no path records are written. The cumulative damage cost table is not triggered. Updated. Path records have been written. Candidate extension relationships of writable path records From the current successor state source The ungenerated parts are removed to avoid the same candidate extension relationship from participating in the state update repeatedly.

[0113] Based on the set of face elements to be expanded and cumulative damage cost table From writing to the set of face elements to be expanded New damage leading edge elements were identified from the candidate elements. Specifically, read the set of face elements to be expanded. Candidate surfaces marked as to be expanded but not yet written to the expansion node are ranked according to their position in the cumulative damage cost table. The corresponding cumulative damage cost values ​​are sorted from smallest to largest, and the candidate element with the smallest cumulative damage cost value is selected as... When the cumulative damage cost values ​​of multiple candidate face elements are the same, the candidate face element with the smallest face element number value is selected as the candidate face element. .Will Write to the extended node and update the element number, the preceding damage leading edge element number, the damage leading edge entry direction, and the accumulated damage cost within the extended node. The damage leading edge entry direction is recorded using the write path. The stress activates the extension direction of the edge.

[0114] Based on the updated extended nodes and the updated successor state source The candidate extension relationships pointing from the damage leading edge surface element to the candidate surface element are matched. During matching, the damage leading edge surface element number in the extension node is used as the search key, and the updated successor state source is used for matching. The selected elements are those with consistent damage leading edge element numbers, those not filtered out by extended restricted relations, and those not written into the path record. The candidate extension relations are used as the successor state source for the differentiable A-star search model. If the updated successor state source... If no candidate extension relationship matches the extension node, the extension node is marked as an extension node with no available successor state source.

[0115] Based on path records Cumulative Damage Cost Table and the updated successor state source Continuity connection determination is performed on stress-activated extended edges. Path recording is then performed. Each stress-activated extension edge is considered a directed connection from the damage-leading surface element to the candidate surface element, and the connection is made according to the rule that the candidate surface element number of the previous stress-activated extension edge is consistent with the damage-leading surface element number of the next stress-activated extension edge. The damage-leading surface element number corresponding to the current extension node is used as the chain tail surface element number, and this is recorded in the path record. The system searches for path records where the candidate element number matches the chain tail element number. The found stress-activated extension edge is written into the chain record. The damage leading edge element number from the found path record is then used as the new chain tail element number to continue the search until a path record start item is found or no preceding path record item exists. The chain records are then arranged according to the connection direction from the damage leading edge element to the candidate element, forming a damage-through chain. ,in, The number of stress-activated extension edges in the damaged through-chain. To activate the extended edge index for stress in the damaged through-chain, For the first in the damage-through chain Strip stress activates the extended edge. Damage penetrates the chain. The adjacent stress-activated extension edges satisfy the condition that the candidate element number of the previous stress-activated extension edge is consistent with the damage leading edge element number of the next stress-activated extension edge.

[0116] S7. Critical Breakthrough State Judgment and Result Output

[0117] Read the set of weakly structured surface elements from the output layer of the differentiable A-star search model. Damage through chain , Through-boundary, Preset through-judgment threshold and cumulative damage cost table . It is a set of weakly structured surface elements. For the first A weak structural surface element The number of weak structure surface elements. This refers to a damage chain formed by continuously connecting stress-activated extension edges from the damage leading edge surface element to the candidate surface element. For damaged through-chain The first in Strip stress activates extended edges, To activate the extended edge index for stress in the damaged through-chain, For damaged through-chain The number of stress-activated extended edges in the middle, This is a preset connectivity threshold used to determine whether the boundary connection distance meets the connectivity boundary conditions. This is a table of cumulative damage costs stored according to the weak structure element numbering. For the first The cumulative damage cost corresponds to each weak structure element. The output layer is a regularized output layer with no trainable weights. The output layer does not regenerate candidate extension relationships, does not re-execute the 3D stress traction judgment, and does not re-call the weak structure element encoding layer and the candidate extension relationship cost layer. The output layer is based on the already formed damage continuity chain. Perform a boundary connection determination.

[0118] when When the value is greater than zero, the damage will penetrate the chain. The first stress-activated extension edge The damaged leading edge surface element was identified as the weak structural surface element at the head end. Damage to the entire chain The last stress-activated extension edge The candidate surface element was determined to be the terminal weak structure surface element. Each stress-activated extension edge Save the damage leading edge element number and candidate element number, and then perform a search on the weak structure element set based on the saved element number. The weak structural surface element at the beginning was obtained by matching. and terminal weak structural elements .when When the value equals zero, the critical damage is considered to be in a continuous state. Set to a numerical value Accumulated damage cost across the damage chain Set to a numerical value And write invalid padding values ​​to all positions in the element number sequence of the damaged through-chain surface. .

[0119] exist When the value is greater than zero, based on the weak structural surface element at the beginning. At the center position, calculate the weak structural surface element at the beginning. Spatial distance to the boundary Based on the weak structural elements at the ends At the center location, calculate the terminal weak structure element. Spatial distance to the boundary When the through boundary is given in the form of triangular facets, calculate the Euclidean distance from the center of the weak structural element to each boundary triangular facet. If the foot of the perpendicular from the center of the weak structural element to the plane of the boundary triangular facet is inside the facet, use the foot of the perpendicular distance; if the foot of the perpendicular is outside the facet, use the minimum Euclidean distance from the center of the weak structural element to the three sides and three vertices of the facet. When the through boundary is given in the form of boundary sampling points, calculate the Euclidean distance from the center of the weak structural element to each boundary sampling point and take the minimum Euclidean distance. and Constituting boundary connection distance .

[0120] like Figure 4 As shown, the boundary is connected by distance. In With preset penetration determination threshold Compare and generate head-end join tags. .when Less than or equal to At that time, Set to a numerical value Otherwise, Set to a numerical value Connect the boundary distances In With preset penetration determination threshold Compare and generate terminal join tags. .when Less than or equal to At that time, Set to a numerical value Otherwise, Set to a numerical value Connect the first end marker. and end connection mark Composition of the through boundary connection result .

[0121] Based on the connection results of the through boundary Determine the damage to the through chain Does it connect to the boundary? If so... Greater than zero, start-end connection mark For numerical values End connection mark For numerical values Then the critical damage will be in a continuous state. Set to a numerical value numerical values To connect and penetrate the boundary. If Equal to zero, start-end connection mark For numerical values Or end-connection mark For numerical values Then the critical damage will be in a continuous state. Set to a numerical value numerical values This is an unconnected boundary. To represent the damaged through chain Whether the critical damage penetration state has reached the critical condition of connecting with the penetration boundary. When the value is greater than zero, the weak structural element at the end will be... In the cumulative damage cost table The corresponding cumulative damage cost is determined as the cumulative damage cost of the damage-through-chain mechanism. .

[0122] Based on the critical damage penetration state Damage through chain And the cumulative damage cost of the damage-through-chain Three-dimensional anisotropic failure discrimination results of deep migmatite were generated. . The result vector is 258 dimensions. . The first element in the sequence of damaged through-chain elements Each element is numbered. This is the sequence index of the element number of the damaged through-chain surface, with a value range of [value range missing]. to According to the damage through-chain For continuously connected stress-activated extension edges, the damage-through-chain element number sequence is written in the order from the first weak structural element to the last weak structural element; when the number of damage-through-chain element numbers is less than 256, the remaining positions are filled with invalid fill values. When the number of surface element numbers in the damage-through-chain path exceeds 256, retain the first 255 surface element numbers and the last weak structural surface element number in the direction from the first weak structural surface element to the last weak structural surface element. Critical damage-through state. Occupying one dimension, the damage-through-chain surface element number sequence occupies 256 dimensions, and the cumulative damage cost of the damage-through-chain is... Occupy one dimension.

[0123] Figure 5 It provides a specific comparison between the state before and after the critical damage breakthrough determination and the actual breakthrough result. Figure 5 The left side shows the damage penetration chain before reaching the critical penetration stage. Although the chain has extended to a certain extent, the beginning and end are not yet connected to the penetration boundary (the distance is about 1-3 mm). The right side shows the damage penetration chain connecting to the penetration boundary after a slight increase in stress (the distance from the beginning and end to the penetration boundary is 0.5-1 mm less), reaching the critical damage penetration state. This corresponds to the end-face seepage, fracture penetration surface, or sectioning results, proving that the critical damage penetration discrimination method for deep mixed rocks based on three-dimensional stress traction described in this invention can identify the moment and state of reaching the critical penetration stage, rather than just judging whether local cracks exist. Therefore, it is more suitable for the criticality discrimination of deep mixed rock penetration failure, and the results are easier to verify with actual failure phenomena.

Claims

1. A method for determining critical damage penetration in deep mixed rock based on three-dimensional stress traction, characterized in that the method... include: S1. Obtain the three-dimensional structural data of deep mixed rock, the current three-dimensional stress state, the penetration boundary and the interface strength conditions. Extract the weak structural surface from the three-dimensional structural data of deep mixed rock and divide it into local surface elements according to the preset surface element scale. Assign each local surface element a spatial position, surface element normal, interface strength conditions and distance to the penetration boundary to form a set of weak structural surface elements. S2. Determine the initial damage leading edge surface element based on the distance from the weak structure surface element to the through boundary. Select candidate surface elements according to the spatial distance between adjacent weak structure surface elements and the included angle of the surface element normals, and generate candidate extension relationships from the damage leading edge surface element to the candidate surface element. S3. Establish a search model based on the initial damage leading edge surface element and candidate extension relationship, take the damage leading edge surface element as extension node, take the candidate extension relationship as the source of subsequent state, take the accumulated damage cost as path record, so that the subsequent state generates damage transfer between corresponding weak structure surface elements. S4. In the search model, based on the current three-dimensional stress state, the normal of the damage leading edge surface element, the normal of the candidate surface element, and the interface strength condition, calculate the normal traction state and tangential traction state corresponding to the candidate expansion relationship, and generate the stress-activated passage state. The calculation of the normal traction state and the tangential traction state includes: The current three-dimensional stress state is projected onto the normal of the damage leading edge surface element and the normal of the candidate surface element, respectively, to obtain the three-dimensional surface element traction component of the damage leading edge surface element and the three-dimensional surface element traction component of the candidate surface element. The normal traction value of the damage leading edge surface element and the normal traction value of the candidate surface element are calculated based on the traction components of the three-dimensional surface element to form the normal traction state; Based on the traction components of the three-dimensional surface elements, the tangential traction values ​​of the damage leading edge surface elements and the tangential traction values ​​of the candidate surface elements are calculated respectively to form the tangential traction state; The generation of stress-activated passage states includes: Based on the interface strength condition, the normal traction state and the tangential traction state are constrained and determined, and the opening determination, slip determination and interface debonding determination are generated. Based on the opening determination, slip determination, and interface debonding determination, and combined with whether the candidate expansion direction and the damage front entry direction form a reverse compression relationship, a stress-activated passage state is generated; when at least one of the opening determination, slip determination, and interface debonding determination is passed, and the candidate expansion direction and the damage front entry direction do not form a reverse compression relationship, the stress-activated passage state is set to allow passage, otherwise it is set to restrict passage. S5. Using the stress-activated passage state as the damage transfer condition, generate stress-activated extension edges for candidate extension relations that satisfy the damage transfer condition, and write the corresponding candidate surface elements into the set of surface elements to be extended, while writing the cumulative damage cost into the cumulative damage cost. S6. Determine new damage leading edge elements based on the set of elements to be expanded and the accumulated damage cost, update the expansion nodes and subsequent state sources, and form a damage penetration chain according to the continuous connection relationship of stress-activated expansion edges. S7. Based on the distances from the first and last weak structural elements of the damage penetration chain to the penetration boundary, determine whether the damage penetration chain is connected to the penetration boundary, and output the critical damage penetration state and the three-dimensional anisotropic failure discrimination results of the deep mixed rock.

2. The method for determining critical damage penetration in deep mixed rock based on three-dimensional stress traction according to claim 1, characterized in that, The extraction of weak structural surfaces from the three-dimensional structural data of deep mixed rocks, as described in S1, includes: The three-dimensional stress state, through boundary and interface strength conditions are associated with the spatial location corresponding to the three-dimensional structural data. Microcracks, mineral phase interfaces, foliation weak surfaces and vein boundaries are extracted and segmented according to the preset surface element scale to form a local surface element sequence. Using the geometric center of each local surface element as its spatial location, least-squares local plane fitting is performed inside the local surface element to obtain the surface element normal. The distance to the penetrating boundary is calculated based on the interface strength conditions assigned according to the spatial location, thus forming a set of weak structure surface elements.

3. The method for determining critical damage penetration in deep mixed rock based on three-dimensional stress traction according to claim 1, characterized in that, S2 include: Weak structural elements in the set of weak structural elements whose distance from the penetrating boundary is less than or equal to a preset boundary distance threshold are identified as initial damage leading edge elements. Based on the spatial distance between the initial damage leading edge surface element and the adjacent weak structure surface element and the angle between the surface element normals, a structural continuity determination result is generated. Based on the structural continuity determination results, adjacent weak structural surface elements with a spatial distance not greater than a preset extension distance threshold and a surface element normal angle not greater than a preset normal angle threshold are identified as candidate surface elements, and candidate extension relationships are generated from the damage leading edge surface element to the candidate surface element.

4. The method for determining critical damage penetration in deep mixed rock based on three-dimensional stress traction according to claim 1, characterized in that, The search model described in S3 is a differentiable A-star search model, which includes a weak structure element encoding layer and a candidate extended relation cost layer; the weak structure element encoding layer is used to encode the attribute information of each weak structure element into a feature vector; Candidate The extended relation cost layer is used to generate the cumulative damage cost writing amount corresponding to each candidate extended relation based on the damage leading edge surface element encoding features, candidate surface element encoding features, current three-dimensional stress state, spatial distance, surface element normal angle, and damage leading edge entry direction.

5. The method for determining critical damage penetration in deep mixed rock based on three-dimensional stress traction according to claim 4, characterized in that, The cumulative damage cost written amount is: ； In the formula, The amount of data written to account for accumulated damage. For the damage leading edge element index; For candidate face element index; , , and These are the non-negative dimensionless weighting coefficients for the spatial distance term, the surface element normal angle term, the penetrating boundary distance term, and the interface strength term, respectively. For the damaged leading edge element With candidate face elements Spatial distance between them; The preset extension distance threshold; For the damaged leading edge element Surface element normal and candidate surface element The angle between the normals of the surface elements; The preset normal angle threshold; It is a positive dimensionless constant used to avoid the denominator of the included angle term of the surface element being zero; Candidate face elements The distance to the boundary that runs through it; For the damaged leading edge element The distance to the boundary that runs through it; To preset the boundary distance threshold, , and Candidate face elements Tensile strength, slip strength, and interfacial debonding strength; , and These are the leading edge surface elements of the damage. Tensile strength, slip strength, and interfacial debonding strength; This is a positive stress constant used to avoid the denominator of the interfacial strength term being zero, and it uses the same stress unit as the opening strength, slip strength, and interfacial debonding strength.

6. The method for determining critical damage penetration in deep mixed rock based on three-dimensional stress traction according to claim 1, characterized in that, In S5, for multiple candidate extension relationships pointing to the same candidate element, the generation stress activation extension edge with the minimum cumulative damage cost is selected from the allowed candidate extension relationships. The candidate element is then written into the set of elements to be extended and the cumulative damage cost is updated.

7. The method for determining critical damage penetration in deep mixed rock based on three-dimensional stress traction according to claim 1, characterized in that, The formation of a damage penetration chain according to the continuous connection relationship of the stress-activated extension edge as described in S6 includes: starting from the initial damage leading edge surface element, sequentially tracing the predecessor relationship of each surface element along the confirmed stress-activated extension edge, and arranging all the traced surface elements in sequence to form a damage penetration chain from the beginning to the end.

8. The method for determining critical damage penetration in deep mixed rock based on three-dimensional stress traction according to claim 1, characterized in that, The determination of whether the damage penetration chain is connected to the penetration boundary as described in S7 includes: calculating the spatial distances from the weak structural surface element at the beginning and end of the damage penetration chain to the penetration boundary. If both distances are not greater than the preset penetration determination threshold, the damage penetration chain is determined to be connected to the penetration boundary, and the critical damage penetration state is output as "connected". Otherwise, it is determined to be "not connected".

Images