Three-dimensional seepage field inversion analysis method applicable to anisotropic rock masses

Abstract

This invention relates to the field of seepage inversion technology in underground engineering. To improve the accuracy of seepage tensor field generation, a three-dimensional seepage field inversion analysis method suitable for anisotropic rock masses is provided. The method maps borehole lithology, structural plane attitude, segmented water pressure, water level monitoring, groundwater inflow at the excavation face, recharge boundary, and excavation face data to three-dimensional elements to form the conditions for seepage tensor generation. Within the conditionally normalized flow affine coupling layer, retained parameters and parameters to be updated are defined. Based on the retained parameters, a transmission relationship diagram is constructed, and the random walk absorption probability and direction change are calculated. The tensor parameters are updated by embedding scale parameter and translation parameter generation steps. By matching water level and inflow response, a three-dimensional anisotropic seepage tensor field and directed connectivity state are output, thus improving the accuracy of seepage tensor field generation.

Description

Technical Field

[0001] This invention relates to the field of seepage inversion technology in underground engineering, specifically a three-dimensional seepage field inversion analysis method applicable to anisotropic rock masses. Background Technology

[0002] In underground rock masses, different lithological units, lithological interfaces, and structural planes jointly control groundwater seepage channels. After excavation activities alter the original hydraulic boundaries, the hydraulic connection from the recharge boundary to the excavation face directly affects water level changes and groundwater inflow at the excavation face. Due to the significant spatial and directional differences in permeability within the rock mass, a single permeability coefficient is insufficient to represent the main conduction direction controlled by structural planes. Engineering analysis requires establishing a three-dimensional anisotropic permeability tensor field and combining it with multi-source data such as borehole lithology, structural plane occurrence, segmented water pressure, water level monitoring, and excavation face inflow for inversion.

[0003] In existing technologies for underground engineering rock mass applications, the generation of permeability parameters, structural plane orientation constraints, and the directed connectivity relationship from the recharge boundary to the excavation face are often treated separately. The attitude of the structural plane is often used as the basis for initial value assignment or as a fixed directional condition, while the hydraulic connection from the recharge boundary to the excavation face is mostly reflected in subsequent connectivity analysis or forward modeling verification, making it difficult to incorporate into the intra-layer update process of the permeability tensor parameters. Consequently, under the combined effects of complex lithological interfaces and excavation face inflow, the generated permeability tensor field lacks a consistent expression with the main conduction direction of the structural plane and the boundary connectivity state. Summary of the Invention

[0004] To improve the accuracy of permeability tensor field generation, this invention provides a three-dimensional seepage field inversion analysis method applicable to anisotropic rock masses.

[0005] The technical solution adopted by the present invention to solve the above problems is:

[0006] Three-dimensional seepage field inversion analysis methods applicable to anisotropic rock masses include:

[0007] S1. Obtain borehole lithology, structural plane occurrence, segmented water pressure data, water level monitoring data, groundwater inflow data at the excavation face, recharge boundary and excavation face of the underground engineering rock mass to form basic data on rock mass seepage.

[0008] S2. Based on the rock mass seepage basic data, determine the lithological state, structural plane direction, local seepage constraint, and boundary relationship from the recharge boundary to the excavation face to form the conditions for the generation of the seepage tensor;

[0009] S3. Construct an affine coupling layer for conditionally normalized flow based on the permeation tensor generation conditions. Limit the generated object to three-dimensional anisotropic permeation tensor parameters constrained by the structural surface direction. Divide the permeation tensor parameters retained in this layer and the permeation tensor parameters to be updated in this layer within the affine coupling layer to form the intermediate state of the permeation tensor in this layer.

[0010] S4. Based on the retained permeability tensor parameters of this layer, the structural plane direction and the relationship between adjacent three-dimensional elements, determine the conduction relationship between the three-dimensional elements and form a three-dimensional element conduction relationship diagram.

[0011] S5. Calculate the random walk absorption probability based on the three-dimensional unit transmission relationship diagram and the boundary relationship from the supply boundary to the excavation face, form the absorption probability from the supply boundary to the excavation face, and form the change in the direction of absorption probability based on the difference in absorption probability between adjacent three-dimensional units.

[0012] S6. Input the absorption probability and the change in the direction of absorption probability from the supply boundary to the excavation face into the scale parameter and translation parameter generation stage of the affine coupling layer to form affine coupling transformation parameters, and update the permeability tensor parameters to be updated in this layer according to the affine coupling transformation parameters to obtain the candidate permeability tensor field.

[0013] S7. Calculate the water level and groundwater inflow response at the excavation face based on the candidate permeability tensor field, and match and determine the water level and groundwater inflow response at the excavation face based on the water level monitoring data and the groundwater inflow data at the excavation face, and output the three-dimensional anisotropic permeability tensor field of the rock mass and the directed connectivity state from the recharge boundary to the excavation face.

[0014] Furthermore, S1 specifically refers to:

[0015] The spatial extent of the underground engineering rock mass is discretized into a set of three-dimensional units according to a preset three-dimensional unit scale;

[0016] The borehole lithology and structural plane occurrence of the underground engineering rock mass are obtained, and the borehole lithology and structural plane occurrence are mapped to three-dimensional units according to the borehole location and the location of the structural plane measuring points to form lithological and structural spatial data.

[0017] Acquire segmented water pressure data and water level monitoring data, and map the segmented water pressure data and water level monitoring data to three-dimensional units according to the location of the water pressure test section and the location of the water level monitoring, to form hydraulic response spatial data;

[0018] Acquire groundwater inflow data, recharge boundary and excavation face at the excavation face, and form boundary inflow spatial data based on the inflow location, recharge boundary location and excavation face location;

[0019] Based on lithological structure spatial data, hydraulic response spatial data, and boundary inflow spatial data, rock mass seepage basic data are formed according to a three-dimensional unit index. The three-dimensional unit index is used to identify the corresponding unit positions of the lithological structure spatial data, hydraulic response spatial data, and boundary inflow spatial data of the same three-dimensional unit.

[0020] Furthermore, S2 specifically refers to:

[0021] Based on the borehole lithology in the rock mass seepage basic data, determine the lithology category and lithology interface location corresponding to the three-dimensional unit, and form a lithology state to limit the range of values ​​of the three-dimensional anisotropic permeability tensor parameter;

[0022] Based on the orientation of the structural planes in the rock mass seepage basic data, determine the dip direction and dip angle of the structural planes corresponding to the three-dimensional units, and convert the dip direction and dip angle of the structural planes into the structural plane directions used to define the principal conduction directions of the three-dimensional anisotropic permeability tensor parameters;

[0023] Based on the segmented water pressure data and water level monitoring data in the rock mass seepage basic data, the local seepage value range corresponding to the three-dimensional unit is determined, forming a local seepage constraint to constrain the magnitude of the three-dimensional anisotropic seepage tensor parameter;

[0024] The water level monitoring characteristics are determined based on the water level monitoring data in the rock mass seepage basic data. The water level monitoring characteristics include the reference water level and the water level change range.

[0025] The groundwater inflow data characterization of the excavation face is determined based on the groundwater inflow data of the rock mass seepage basic data. The groundwater inflow data characterization of the excavation face includes the baseline inflow value and the inflow variation range.

[0026] The boundary relationship between the recharge boundary and the excavation face is determined based on the groundwater inflow data at the excavation face, the recharge boundary, and the excavation face in the rock mass seepage basic data.

[0027] The conditions for generating the permeability tensor are formed by combining lithological state, structural plane orientation, local permeability constraint, water level monitoring characterization, groundwater inflow data characterization at the excavation face, boundary relationship from the recharge boundary to the excavation face, and the coordinates of the three-dimensional unit center.

[0028] Furthermore, S3 specifically refers to:

[0029] Based on the lithological state and local permeability constraints in the permeability tensor generation conditions, the value boundaries of the three-dimensional anisotropic permeability tensor parameters in each three-dimensional unit are determined, forming tensor value constraints.

[0030] Based on the tensor value constraints and the structural surface orientation, the transmission correspondence between the three-dimensional anisotropic permeability tensor parameters and the structural surface orientation is determined, thus forming the structural surface orientation constraint relationship;

[0031] Based on the structural plane orientation constraint relationship, the generated object of the conditional normalized flow is set as the three-dimensional anisotropic permeation tensor parameter constrained by the structural plane orientation, thus forming a tensor generated object.

[0032] Based on the tensor-generated object, an affine coupling layer with conditionally normalized flow is constructed, and the generation target of the affine coupling layer is limited to the three-dimensional anisotropic permeation tensor parameter, thus forming the affine coupling layer generation structure.

[0033] Based on the affine coupling layer generation structure, the three-dimensional anisotropic permeation tensor parameters within the affine coupling layer are set as the tensor state to be partitioned in this layer, forming a tensor state that can participate in the partitioning within the layer.

[0034] Based on the tensor states that can participate in the intra-layer partitioning and the boundary relationship from the supply boundary to the excavation face, the parameter part used to preserve the transmission relationship between three-dimensional elements and the parameter part used to receive intra-layer updates are determined, forming the basis for tensor parameter partitioning.

[0035] Based on the tensor parameter partitioning criteria, the affine coupling layer is divided into the retained permeation tensor parameters and the permeation tensor parameters to be updated in this layer. The retained permeation tensor parameters and the permeation tensor parameters to be updated in this layer are then combined to form the intermediate state of the permeation tensor in this layer.

[0036] Furthermore, when dividing the permeation tensor parameters to be retained and the permeation tensor parameters to be updated within the affine coupling layer, the propagation direction constraint of the three-dimensional anisotropic permeation tensor parameters entering the current affine coupling layer is applied according to the structural plane direction, and the three-dimensional anisotropic permeation tensor parameters participating in the division are determined in combination with the tensor value constraint. The parameter part corresponding to the structural plane direction and used to form the propagation relationship between three-dimensional units is divided into the permeation tensor parameters to be retained in this layer, and the parameter part other than the permeation tensor parameters to be retained in this layer and used to receive the update of the affine coupling transformation parameters is divided into the permeation tensor parameters to be updated in this layer.

[0037] Furthermore, S4 specifically refers to:

[0038] Based on the intermediate state of the permeability tensor of this layer, the parameters of the retained permeability tensor of this layer are extracted, and the main conduction direction of the tensor corresponding to the parameters of the retained permeability tensor of this layer is defined according to the direction of the structural surface, thus forming the conduction capacity of the unit structural surface;

[0039] Based on the relationship between adjacent three-dimensional elements, determine the adjacent three-dimensional elements and adjacent connection directions corresponding to each three-dimensional element, and match the adjacent connection directions with the conduction capacity of the element structure surface to form the adjacent direction matching result.

[0040] The conduction strength between adjacent three-dimensional elements along the adjacent connection direction is calculated based on the adjacent direction matching results, and the conduction strength on both sides of the adjacent three-dimensional elements is constrained to form the inter-element conduction strength.

[0041] The inter-element conduction strength is compared with a preset conduction threshold, and a conduction relationship is determined between adjacent three-dimensional elements if the comparison conditions are met.

[0042] Based on the conduction relationship and adjacent connection direction, conduction edges connecting three-dimensional units are constructed, and the conduction strength between units is used as the conduction weight of the conduction edges; thus forming a three-dimensional unit conduction relationship graph for calculating the random walk absorption probability.

[0043] Further, in S4, the specific steps for forming the inter-unit conduction strength are as follows: calculate the conduction strength from any three-dimensional unit to an adjacent three-dimensional unit according to the adjacent connection direction, and calculate the conduction strength from an adjacent three-dimensional unit to any three-dimensional unit according to the opposite adjacent connection direction; match each conduction strength with the tensor principal conduction direction defined by the structural surface direction, and compare it with the preset conduction threshold respectively; if the conduction strength in any direction meets the comparison condition, form the conduction relationship between three-dimensional units along the direction that meets the comparison condition, and use the conduction strength that meets the comparison condition as the conduction weight of the corresponding conduction edge.

[0044] Furthermore, S5 specifically refers to:

[0045] Based on the boundary relationship between the supply boundary and the excavation face, the three-dimensional elements corresponding to the supply boundary and the three-dimensional elements corresponding to the excavation face are determined in the three-dimensional element transmission relationship diagram, forming the boundary calculation object of the random walk absorption probability;

[0046] Based on the transmission relationships and transmission weights in the three-dimensional unit transmission relationship diagram, the random walk transfer relationship between adjacent three-dimensional units is determined, and the random walk transfer relationship is restricted to adjacent three-dimensional units with transmission relationships, thus forming a random walk transfer constraint;

[0047] Based on the random walk transition constraints and boundary calculation objects, the three-dimensional element corresponding to the excavation face is set as the absorption calculation element, and the three-dimensional element corresponding to the supply boundary is set as the random walk initial constraint element, thus forming the absorption calculation conditions from the supply boundary to the excavation face.

[0048] Based on the absorption calculation conditions, the random walk absorption probability of each three-dimensional element reaching the absorption calculation element is calculated in the three-dimensional element transmission relationship diagram, forming the absorption probability from the supply boundary to the excavation face.

[0049] Based on the absorption probability from the supply boundary to the excavation face, the difference in absorption probability between adjacent three-dimensional elements along the direction of the conduction relationship is calculated, thus forming the difference in absorption probability between adjacent three-dimensional elements.

[0050] Based on the difference in absorption probability between adjacent three-dimensional units and the direction of conduction in the three-dimensional unit conduction diagram, the difference in absorption probability corresponding to each three-dimensional unit is assigned to the adjacent conduction direction, forming the change in absorption probability direction.

[0051] Furthermore, S6 specifically refers to:

[0052] Based on the permeability tensor parameters and permeability tensor generation conditions retained in this layer, the basic generation inputs used to generate scale parameters and translation parameters in the generation process are determined.

[0053] The absorption probability from the supply boundary to the excavation face is embedded into the basic generation input according to the correspondence of three-dimensional elements, which is the basis for the generation of the permeability tensor parameter to be updated in this layer, thus forming the absorption probability constraint input;

[0054] The transmission direction corresponding to the permeability tensor parameter to be updated in this layer is determined based on the structural plane direction, and the change in the absorption probability direction is allocated to the transmission direction to form the direction change constraint input.

[0055] Based on the absorption probability constraint input and the direction change constraint input, the scale parameters and translation parameters corresponding to the permeability tensor parameters to be updated in this layer are calculated in the scale parameter and translation parameter generation stage to form affine coupling transformation parameters.

[0056] The permeation tensor parameters to be updated in this layer are updated by performing an affine update based on the affine coupling transformation parameters, thus forming the updated permeation tensor parameters to be updated in this layer.

[0057] The updated permeability tensor parameters to be updated in this layer are combined with the permeability tensor parameters retained in this layer to form the updated three-dimensional anisotropic permeability tensor parameters;

[0058] The updated three-dimensional anisotropic permeation tensor parameters are used as the output of the affine coupling layer, and the candidate permeation tensor field is obtained based on the output of the affine coupling layer.

[0059] Furthermore, S7 specifically refers to:

[0060] Based on the candidate permeability tensor field and the boundary relationship between the recharge boundary and the excavation face, the calculation boundary of the water level and the groundwater inflow response at the excavation face is determined, thus forming the hydraulic response calculation conditions.

[0061] Based on the hydraulic response calculation conditions, seepage response calculations are performed on the candidate permeability tensor field to generate the water level and groundwater inflow response at the excavation face.

[0062] Based on water level monitoring data and groundwater inflow data at the excavation face, the matching error of water level and groundwater inflow response at the excavation face is calculated, and the matching error is compared with a preset matching threshold to form a matching judgment result.

[0063] If the matching judgment result meets the comparison conditions, the corresponding candidate permeability tensor field is determined as the three-dimensional anisotropic permeability tensor field of the rock mass, and a directed connectivity state from the supply boundary to the excavation face is formed according to the absorption probability from the supply boundary to the excavation face.

[0064] The advantages of this invention compared to the prior art are:

[0065] 1. A six-component three-dimensional symmetric permeation tensor is used as the generation object, and the main transmission direction of the tensor is limited by the structural surface direction constraint relationship, so that the inversion result can express the directional transmission under the control of the structural surface;

[0066] 2. Dynamically divide the permeation tensor parameters of the current layer and the permeation tensor parameters to be updated within the affine coupling layer. The retained parameters are used to stably construct the three-dimensional unit transmission relationship graph, and the parameters to be updated are used to receive connectivity probability constraints, so as to avoid the structural surface transmission relationship and the parameter update process from becoming detached from each other, thereby improving the inversion accuracy.

[0067] 3. The random walk absorption probability is used to characterize the probability of reaching the excavation face from the supply boundary, and the change in the direction of the absorption probability is allocated to the translation parameter generation stage according to the transmission direction corresponding to the tensor components. This allows the boundary connectivity relationship to enter the tensor layer update process, rather than just being used as a post-processing interpretation, which significantly improves the physical consistency, anisotropy characterization ability and engineering applicability of the inversion results.

[0068] 4. After the candidate permeability tensor field is generated, the matching judgment is made by the water level response and the groundwater inflow response at the excavation face, so that the final output includes both the three-dimensional anisotropic permeability tensor field and the directed connectivity state from the recharge boundary to the excavation face. This improves the scene adaptability and deliverability of the seepage inversion results of underground engineering rock mass from a mechanism perspective. Attached Figure Description

[0069] Figure 1 Flowchart of a three-dimensional seepage field inversion analysis method applicable to anisotropic rock masses;

[0070] Figure 2 A schematic diagram illustrating the process from basic rock mass seepage data to the conditions for generating the permeability tensor;

[0071] Figure 3 This is a schematic diagram of the core process coupling conditional normalized flow and random walk absorption probability.

[0072] Figure 4 This is a schematic diagram illustrating the response calculation and result determination process for the candidate penetration tensor field. Detailed Implementation

[0073] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0074] This invention maps borehole lithology, structural plane orientation, segmented water pressure, water level monitoring, groundwater inflow at the excavation face, recharge boundary, and excavation face data to three-dimensional elements to form the conditions for permeability tensor generation. Within a conditionally normalized flow affine coupling layer, retained parameters and parameters to be updated are defined. Based on the retained parameters, a transmission relationship graph is constructed, and the random walk absorption probability and direction change are calculated. Tensor parameters are updated by embedding scale parameter and translation parameter generation steps. A three-dimensional anisotropic permeability tensor field and directed connectivity state are output through water level and inflow response matching. This achieves the unified embedding of lithological state, structural plane orientation, local permeability constraints, and directed connectivity from the recharge boundary to the excavation face into the three-dimensional anisotropic permeability tensor field inversion process, enabling the inversion results to express the structural plane control transmission and the inflow channel at the excavation face, thereby improving the accuracy of permeability tensor field generation.

[0075] like Figure 1 As shown, a three-dimensional seepage field inversion analysis method applicable to anisotropic rock masses includes:

[0076] S1. Obtain borehole lithology, structural plane occurrence, segmented water pressure data, water level monitoring data, groundwater inflow data at the excavation face, recharge boundary and excavation face of the underground engineering rock mass to form basic data on rock mass seepage.

[0077] First, the spatial extent of the underground engineering rock mass is discretized into a set of three-dimensional units according to a preset three-dimensional unit scale. , Indicates the number of three-dimensional elements. Indicates the first A three-dimensional unit, Represents a three-dimensional element index. For elements with spatial location... Point-like records, if spatial location Located in three-dimensional unit Internally, point records are bound to the three-dimensional cell index. If spatial location If it does not fall into any three-dimensional element, then use Determine the nearest 3D element. Indicates spatial location To three-dimensional unit The Euclidean distance, in Not larger than the preset space window At that time, point records are bound to the three-dimensional cell index. , This represents the distance threshold that allows spatial matching. for The corresponding three-dimensional unit.

[0078] Obtaining borehole lithology records , Indicates the number of borehole lithology records. This includes borehole location, depth range, and lithological description. Based on the borehole location and depth range, a spatial segment of the borehole lithology is determined, and this segment is then integrated with three-dimensional units. The intersection judgment is made based on the space occupied by the borehole lithology spatial line segment and the three-dimensional unit. When there is overlap, the borehole lithology record will be... Binding to 3D cell index Obtain the attitude record of structural surfaces. , Indicates the number of structural plane attitude records. This includes the location of structural plane measuring points and the attitude of the structural planes. The attitude of the structural planes is recorded based on the location of the measuring points. Bind to the corresponding 3D element index. Arrange the borehole lithology records and structural plane occurrence records bound under the same 3D element index according to their record sources to form lithological structure spatial data. , Represents three-dimensional cell index The corresponding lithological structure spatial data. The lithological structure spatial data preserves the spatial correspondence between borehole lithology and three-dimensional units, and also preserves the correspondence between the orientation of structural planes and the measurement points of three-dimensional units.

[0079] Obtain segmented water pressure records , This indicates the number of segmented pressure water records. This includes the location of the pressure test section and the pressure test results. Based on the location of the pressure test section, the spatial line segment for the pressure test is determined, and this spatial line segment is then connected to the three-dimensional unit. The intersection judgment is made based on the space occupied by the test line segment and the three-dimensional unit. When there is overlap, segmented pressure water records will be generated. Binding to 3D cell index Obtain water level monitoring records. , This indicates the number of water level monitoring records. Indicates the time of water level monitoring. This indicates the water level monitoring value. This indicates the location of the water level monitoring. (Based on the water level monitoring location...) Water level monitoring records are bound to corresponding 3D cell indices, and the records under the same 3D cell index are sorted in ascending order by monitoring time. The segmented pressure water records and water level monitoring records bound to the same 3D cell index are arranged to form hydraulic response spatial data. , Represents three-dimensional cell index The corresponding hydraulic response spatial data.

[0080] Obtain groundwater inflow data at the excavation face , This indicates the number of groundwater inflow records at the excavation face. Indicates the time when the inflow occurs. This represents the numerical value of groundwater inflow data at the excavation face. Indicates the location where the inflow occurs. Based on the location where the inflow occurs... Groundwater inflow data from the excavation face is bound to corresponding 3D element indices, and groundwater inflow data from the same 3D element index are sorted in ascending order by the time of inflow occurrence. The set of 3D element indices covered by the recharge boundary is determined based on the location of the recharge boundary. Determine the set of three-dimensional element indices covering the excavation face based on the location of the excavation face. For three-dimensional cell indexing ,like Then in the three-dimensional cell index Write the supply boundary overlay field below; if Then in the three-dimensional cell index The excavation face cover field is then written next. The groundwater inflow data, recharge boundary cover field, and excavation face cover field bound under the same 3D element index are arranged to form boundary inflow spatial data. , Represents three-dimensional cell index The corresponding boundary inflow space data.

[0081] According to the three-dimensional unit index Spatial data of lithological structure Hydraulic response spatial data Boundary inflow spatial data and the center coordinates of the three-dimensional unit Write to the same multi-field record , Represents three-dimensional cell index The corresponding rock mass seepage basic data records. For 3D cell indexes not bound to a specific record type, the corresponding empty fields are retained; records under adjacent 3D cell indexes are not used as replacements. The rock mass seepage basic data is represented as follows: It is used to determine the lithological state, structural plane orientation, local permeability constraint, and boundary relationship from the recharge boundary to the excavation face.

[0082] S2. Based on the rock mass seepage basic data, determine the lithological state, structural plane direction, local seepage constraint, and boundary relationship from the recharge boundary to the excavation face to form the conditions for the generation of the seepage tensor.

[0083] Reading basic data of rock mass seepage Based on lithological structure spatial data Extracting borehole lithological record sets , Indicates binding to the three-dimensional cell index The borehole lithology records. Each borehole lithology record includes the borehole location, depth range, and lithological description. The lithological description is converted into a lithological category code according to a pre-defined lithological category table. Based on the depth range and the... Three-dimensional unit The overlap length is used to determine the three-dimensional element. The primary lithology category, secondary lithology category, and proportion of primary lithology are defined within the lithology. The primary lithology category is the lithology category with the longest overlap, the secondary lithology category is the lithology category with the second longest overlap, and the proportion of primary lithology is the percentage of the overlap length of the primary lithology category relative to the total overlap length.

[0084] When the lithology changes between adjacent depth intervals within the same borehole, the boundary between these intervals is defined as the lithological interface. If a three-dimensional unit... If there are multiple lithological interfaces within the memory, then the coordinates from the center of the three-dimensional unit should be selected. The nearest lithological interface location is determined relative to the coordinates of the three-dimensional element center. The three coordinate components are written into the lithological state. The lithological state is denoted as... , Indicates the main lithological category code, Indicates the sublithological category code, Indicates the proportion of the main lithology. , and These represent the positions of the lithological interfaces relative to the center coordinates of the three-dimensional unit. The three coordinate components. If the borehole lithology record set... If empty, then the lithological state will be... Write the zero-value vector. Lithological state. It is a six-dimensional vector used to limit the range of values ​​for the parameters of the three-dimensional anisotropic permeation tensor.

[0085] Based on lithological structure spatial data Extracting the set of structural surface attitude records , Indicates binding to the three-dimensional cell index The structural plane attitude record. Each structural plane attitude record includes the dip direction and dip angle of the structural plane. The dip direction of the structural plane is denoted as... The dip angle of the structural surface is denoted as In the engineering coordinate system, The axis indicates eastward. The axis indicates the north direction. The axis represents vertical upward; when the dip and tilt angle of the structural surface are recorded in degrees, first convert the degrees to radians. Then convert the dip and tilt angle of the structural surface to unit direction vectors. For the same set of structural plane attitude records From multiple unit direction vectors within the plane, the first unit direction vector is selected as the reference direction; all unit direction vectors whose dot product with the reference direction is less than zero are inverted; the average of the unit direction vectors after direction unification is calculated and normalized to form the structural surface direction. . , and These represent the components of the structural plane orientation along the three coordinate axes of the engineering coordinate system. If the set of structural plane attitude records... If empty, then the structural plane direction will be... Write the zero-value vector. Structural plane direction. It is a three-dimensional vector used to define the principal propagation direction of the three-dimensional anisotropic permeation tensor parameters.

[0086] Based on hydraulic response space data Extract the segmented water pressure record set Water level monitoring record set . Indicates binding to the three-dimensional cell index Segmented pressure water records, Indicates binding to the three-dimensional cell index Water level monitoring records. The permeability or permeability coefficient values ​​from the segmented pressure water records are used as pressure water conversion values, and the same three-dimensional unit index is used. The pressure water conversion values ​​are sorted in ascending order. The minimum value after sorting is taken as the lower boundary of the local permeability range, the median value after sorting is taken as the median boundary of the local permeability range, and the maximum value after sorting is taken as the upper boundary of the local permeability range. If the set of segmented pressure water records... If the value is empty, the range of local permeability values ​​will be written into the zero value vector.

[0087] Collection of water level monitoring records According to the water level monitoring time Sort in ascending order. Write the water level monitoring value corresponding to the earliest water level monitoring time as the baseline water level. The difference between the maximum and minimum water level monitoring values ​​is written as the water level change range. To form a water level monitoring characterization If the water level monitoring record set If empty, the water level monitoring characterization will be... Write the zero-value vector. Record the water level change amplitude. With preset water level fluctuation threshold Comparison, This represents the threshold used to determine whether water level fluctuations necessitate expanding the range of local seepage measurements. When the water level change... Greater than the preset water level fluctuation threshold At that time, the lower boundary will be expanded according to the preset expansion ratio. Lower it, and expand the upper boundary according to the preset expansion ratio. Adjust upwards; when the water level changes by a certain magnitude Not greater than the preset water level fluctuation threshold At this time, the lower and upper boundaries remain unchanged. Local permeability constraints are denoted as... , This represents the lower boundary of the range of local permeability values. This represents the median boundary of the range of local permeability values. This represents the upper boundary of the range of local permeability values. Local permeability constraint. It is a three-dimensional vector used to constrain the magnitude of the three-dimensional anisotropic permeation tensor parameter.

[0088] Based on boundary inflow space data Extracting groundwater inflow data set from the excavation face Supply boundary coverage field and excavation face coverage field. Indicates binding to the three-dimensional cell index Groundwater inflow data at the excavation face. The excavation face groundwater inflow data set. According to the time of inflow Sort in ascending order. The groundwater inflow data corresponding to the earliest inflow time at the excavation face is written as the baseline inflow value. The difference between the maximum and minimum inflow values ​​is written as the inflow variation range. To form a characterization of groundwater inflow data at the excavation face. If the groundwater inflow data set at the excavation face If empty, the groundwater inflow data at the excavation face will be characterized. Write the zero-value vector.

[0089] Determine the set of 3D cell indexes corresponding to the supply boundary based on the supply boundary coverage field. Determine the set of three-dimensional cell indexes corresponding to the excavation face based on the excavation face coverage field. .like Then replenish boundary markers. Write Otherwise, replenish boundary markers. Write .like Then the excavation face is marked. Write Otherwise, mark the excavation face. Write The mean coordinates of the three-dimensional element centers corresponding to the supply boundary are calculated as the supply boundary center, and the mean coordinates of the three-dimensional element centers corresponding to the excavation face are calculated as the excavation face center. The three-dimensional element center coordinates are then used as the excavation face center. The projection position along the line segment pointing from the center of the supply boundary to the center of the excavation face is normalized to... The normalized position corresponding to the center of the supply boundary is The normalized position corresponding to the center of the excavation face is Normalized positions that exceed the range of the line segment are truncated to or If the groundwater inflow data set at the excavation face If not empty, then the inflow flag is set. Write Otherwise, inflow marker Write The boundary relationship between the supply boundary and the excavation face is denoted as... , It is a four-dimensional vector used to represent the three-dimensional element corresponding to the supply boundary, the three-dimensional element corresponding to the excavation face, and the spatial correspondence between the supply boundary and the excavation face.

[0090] Lithological state Structural plane direction Local permeation constraint Water level monitoring and characterization Characterization of groundwater inflow data at the excavation face Boundary relationship between supply boundary and excavation face and the center coordinates of the three-dimensional unit A three-dimensional cell index is formed by combining fields in a fixed order. Corresponding permeation tensor generation conditions Conditions for generating the permeation tensor This is a 23-dimensional vector, which sequentially includes six dimensions of lithological state, three dimensions of structural plane orientation, three dimensions of local permeability constraint, two dimensions of water level monitoring characterization, two dimensions of groundwater inflow data characterization at the excavation face, four dimensions of boundary relationship from the recharge boundary to the excavation face, and three dimensions of coordinates of the center of the three-dimensional unit. Permeability tensor generation conditions. As input received by the conditional coding layer in the conditional normalized stream.

[0091] The process of forming the conditions for the generation of permeability tensor from basic rock mass seepage data is as follows: Figure 2As shown, lithology, structural plane occurrence, segmented water pressure data, water level monitoring data, groundwater inflow data at the excavation face, recharge boundary, and excavation face are first correlated according to the three-dimensional unit position; under a unified three-dimensional unit index, lithological state, structural plane orientation, local permeability constraints, and boundary relationships from the recharge boundary to the excavation face are further formed and combined into the permeability tensor generation conditions.

[0092] S3. Construct an affine coupling layer for conditionally normalized flow based on the permeation tensor generation conditions. Limit the generated object to three-dimensional anisotropic permeation tensor parameters constrained by the structural surface direction. Within the affine coupling layer, divide the permeation tensor parameters retained in this layer into permeation tensor parameters to be updated in this layer, forming an intermediate state of the permeation tensor in this layer.

[0093] Reading the conditions for generating the permeation tensor The parameters of the three-dimensional anisotropic permeability tensor are denoted as... The six components represent the three-dimensional symmetric permeability tensor in... Axial direction, Axial direction, Axial direction, Planar coupling direction Planar coupling direction and Independent parameters in the planar coupling direction.

[0094] Based on lithological conditions Reading the 3D cell index The corresponding primary lithology category, secondary lithology category, primary lithology percentage, and lithological interface location are defined. A pre-defined lithological permeability value table is used to determine the allowable intervals for the six tensor components corresponding to the primary and secondary lithology categories. This table includes four fields: lithology category code, tensor component labels, allowable lower bound, and allowable upper bound. The set of tensor component labels is denoted as... Based on the proportion of the primary lithology, the allowable intervals for the primary lithology category and the allowable intervals for the secondary lithology category are merged. During interval merging, the lower bound is weighted according to the proportion of the primary lithology to calculate the allowable lower bounds for both the primary and secondary lithology categories, and the upper bound is weighted according to the proportion of the primary lithology to calculate the allowable upper bounds for both the primary and secondary lithology categories. When the lithological interface is located within a three-dimensional unit, [the following will be implemented]. , and The allowable interval expansion is the union of the allowable intervals of the corresponding coupling components of the primary lithology category and the secondary lithology category. , and Preserve the interval fusion results. Implement local penetration constraints. As a three-dimensional unit index The local value boundaries are determined by intersecting the allowable intervals of the tensor components after lithological fusion with the local value boundaries. When an intersection exists, it is used as the value boundary for the corresponding tensor component; when no intersection exists, the lower and upper boundaries given by the local permeability constraints are used as the value boundaries for the corresponding tensor component. Tensor value constraints are formed by the value boundaries of the six tensor components. , Represents three-dimensional cell index Inner Allowable lower bounds for each tensor component Represents three-dimensional cell index Inner The allowed upper bound of each tensor component.

[0095] According to the direction of the structural plane Establish the transmission correspondence between the parameters of the three-dimensional anisotropic permeability tensor and the orientation of the structural plane. correspond The shaft transmission component will correspond The shaft transmission component will correspond The shaft transmission component will correspond Planar coupled conduction components, correspond Planar coupled conduction components, correspond Planar coupled transport components. Each tensor component is assigned a unit direction label, which indicates the spatial direction in which the tensor component participates in the transport relationship calculation. , and The unit direction labels are taken respectively , and , , and The unit direction labels are taken respectively , and Align the unit direction label with the structural plane direction. Cosine matching of the included angle is performed, and the tensor components corresponding to the principal propagation directions of the tensor are determined according to the matching results. The structural surface orientation constraint relationship is formed by the six tensor components, six unit direction labels, and the correspondence between the principal propagation directions of the tensor. The structural surface orientation constraint relationship is used to confine the principal propagation directions of the three-dimensional anisotropic permeability tensor parameters to the spatial directions controlled by the structural surface orientation.

[0096] Construct a conditionally normalized flow. The conditionally normalized flow consists of a conditional coding layer and eight affine coupling layers. The conditional coding layer is denoted as... , This represents the parameters of the conditional coding layer. The conditional coding layer generates conditions using a 23D permeation tensor. As input, the data first passes through a first fully connected layer to generate a 64-dimensional intermediate representation. This first fully connected layer consists of 64 neurons, each fully fully connected to a 23-dimensional permeation tensor generation conditional fully connected layer, and includes weights, biases, and linear rectified units. Then, a second fully connected layer is used to generate the 64-dimensional permeation tensor generation conditional representation. The second fully connected layer consists of sixty-four neurons, each fully connected to the sixty-four-dimensional output of the first fully connected layer, and includes weights, biases, and linear rectified units. The permeation tensor generates a conditional representation. Eight affine coupling layers are input. The eight affine coupling layers, the conditional coding layer, the intra-layer partitioning stage, and the parameter update interface together form the affine coupling layer generation structure.

[0097] The object generating the conditionally normalized flow is set as the three-dimensional anisotropic permeability tensor parameter constrained by the structural surface direction. Conditional normalized flow uses six-dimensional fundamental random variables. As the starting point for generation, the six-dimensional fundamental random variable Obtained by sampling from the standard normal fundamental distribution. Based on tensor value constraints. Six-dimensional fundamental random variables Mapped to the input tensor state of the first affine coupling layer During mapping, for each component of the fundamental random variable... Cut off, below Components written higher than Components written ,lie in The original value of the component is retained; the truncated component is then processed according to... Linear position mapping within Make the input tensor state The six components fall within the corresponding tensor value constraints. The eight affine coupling layers are arranged according to their layer numbers. Each affine coupling layer is configured sequentially to receive six-dimensional three-dimensional anisotropic permeation tensor parameters. and the generation condition of the 64-dimensional permeation tensor Furthermore, the generation target of the affine coupling layer is limited to the updated six-dimensional anisotropic permeation tensor parameters.

[0098] In the Within the first affine coupling layer, the second layer will be entered. The six-dimensional three-dimensional anisotropic permeation tensor parameters of the affine coupling layer Set the state of the tensor to be partitioned in this layer as the current tensor state. Then, set the current tensor state to be partitioned with the tensor value constraints. Structural surface directional constraints and boundary relationships from the replenishment boundary to the excavation face. Synchronous input layer partitioning. The layer partitioning process first follows the tensor value constraints. Boundary preservation is performed on the tensor state to be partitioned in this layer, below tensor components according to Participation in the division, higher than tensor components according to Participating in the division, located in The tensor components within the layer retain their original values ​​and participate in the partitioning. The six tensor components after boundary preservation, the original component positions of the six tensor components, the value boundaries corresponding to the six tensor components, and the unit direction labels corresponding to the six tensor components together form the tensor state that can participate in the intra-layer partitioning.

[0099] Based on the tensor states that can participate in intra-layer partitioning and the boundary relationship from the supply boundary to the excavation face. The tensor parameter partitioning criteria are determined. Based on the supply boundary markers in the boundary relationships from the supply boundaries of all 3D elements to the excavation face, the 3D elements corresponding to the supply boundaries are determined; based on the excavation face markers in the boundary relationships from the supply boundaries of all 3D elements to the excavation face, the 3D elements corresponding to the excavation face are determined. The average center coordinates of the 3D elements corresponding to the supply boundaries are taken as the supply boundary center, and the average center coordinates of the 3D elements corresponding to the excavation face are taken as the excavation face center. When the supply boundary center and the excavation face center do not coincide, the unit vector pointing from the supply boundary center to the excavation face center is taken as the spatial direction from the supply boundary to the excavation face. When the center of the supply boundary coincides with the center of the excavation face, The spatial direction of the supply boundary pointing towards the excavation face To ensure that the permeability tensor parameters are retained in this layer while simultaneously reflecting the structural plane orientation control, the spatial orientation from the recharge boundary to the excavation face, and the tensor value constraints, in the first... The parameters used to preserve the transmission relationships between 3D elements are selected in each affine coupling layer according to the following rules:

[0100] ,

[0101] in, Represents three-dimensional cell index In the The three tensor component labels in each affine coupling layer are used to retain the permeation tensor parameters of this layer. Represents the set of labels for tensor components. Select the labels of the three tensor components with the largest values ​​within the parentheses; Indicates the label of the tensor component; Indicates by , , , , and A set consisting of six tensor component labels; Indicates an interval indicator quantity, in fall into Time to take ,exist Not fallen Time to take ; Tensor value constraints The Middle Allowed lower bounds for each tensor component; Tensor value constraints The Middle The allowed upper bound of each tensor component; Indicates the first The input of the affine coupling layer is the first A three-dimensional anisotropic permeation tensor parameter; Indicates the structural plane orientation matching weight; This indicates the spatial direction matching weight of the supply boundary pointing towards the excavation face; This represents the relative positional weight within the constraints of tensor value selection; , and All are dimensionless non-negative numbers, and , and The sum of ; Indicates the first Unit direction label for each tensor component; Indicates the first The unit direction label of each tensor component and the direction of the structural surface. The inner product; Indicates vector transpose; Indicates the spatial direction in which the supply boundary points to the excavation face; Indicates the first The inner product of the unit direction label of each tensor component and the spatial direction of the supply boundary pointing to the excavation face; This indicates that only the values ​​projected in the forward direction are retained; Indicates the relationship with the first Each tensor component has a pre-defined positive value with the same dimension to prevent the denominator from being zero. If multiple tensor components have the same value within parentheses, then they are classified according to... , , , , , The preset component order is selected. The remaining three tensor component labels are formed. , Represents three-dimensional cell index In the The labels of the three tensor components in each affine coupling layer are divided into the permeation tensor parameters to be updated in this layer.

[0102] Based on the criteria for tensor parameter partitioning, in the... Within each affine coupling layer, the permeation tensor parameter of this layer is preserved. And the penetration tensor parameters to be updated in this layer This layer retains the permeation tensor parameters. For a three-dimensional vector, by The three corresponding tensor components are arranged in a preset component order. This layer retains the penetration tensor parameters, which are not directly updated by the current affine coupling transformation parameters within the current affine coupling layer, and serve as the source of tensor parameters for the transmission relationship between 3D units. The penetration tensor parameters to be updated in this layer... For a three-dimensional vector, by The three corresponding tensor components are arranged in a preset component order; the permeability tensor parameters to be updated in this layer are used to receive updates to the affine coupling transformation parameters and to bear the absorption probability constraint from the supply boundary to the excavation face. The permeability tensor parameters to be retained in this layer are... The permeation tensor parameters to be updated in this layer The original component locations of the six tensor components, and the constraints on tensor values. Combined with the structural plane orientation constraints, the intermediate state of the permeability tensor of this layer is formed. Each affine coupling layer re-executes the propagation direction constraints and 3D-plus-3D partitioning based on the 3D anisotropic permeation tensor parameters entering the current affine coupling layer, so that the intermediate state of the permeation tensor of this layer is updated with the output of the affine coupling layer.

[0103] S4. Based on the permeability tensor parameters retained in this layer, the orientation of the structural plane and the relationship between adjacent three-dimensional units, determine the conduction relationship between the three-dimensional units and form a three-dimensional unit conduction relationship diagram.

[0104] In the Index of 3D unit processing within an affine coupling layer Intermediate state of the permeability tensor of this layer ,in Indicates the layer number of the affine coupling layer. Represents a three-dimensional cell index. This indicates the number of three-dimensional units. Based on the intermediate state of the permeability tensor of this layer. Extract the permeability tensor parameters of this layer Read the permeation tensor parameters of this layer. The corresponding tensor component label set For each tensor component label Read the first Unit direction label of each tensor component Read the first The tensor component at the th The values ​​in an affine coupling layer and read the tensor value constraints. The Middle Allowed lower bounds for tensor components and allow upper bound .when Below At that time, As the first The values ​​of each tensor component involved in the calculation of the transmission relationship; when Higher than At that time, As the first The values ​​of each tensor component involved in the calculation of the transmission relationship; when lie in Inside, will As the first The numerical values ​​of each tensor component involved in the propagation relation calculation. The numerical values ​​after boundary preservation are denoted as... , Indicates the first The permeability value of the permeation tensor parameter is retained in each layer.

[0105] According to the direction of the structural plane Preserve the permeability tensor parameters for this layer The three tensor components are directionally defined. Each unit direction label is then assigned. With structural plane direction The absolute value of the cosine of the included angle is used as the structural surface orientation matching value. The larger the structural surface orientation matching value, the more significant the first... The closer a tensor component is to the propagation direction controlled by the structural plane orientation, the more likely it is to be the main propagation direction of the tensor. If multiple tensor components have the same structure plane orientation matching value, then select from the tensor components with the same structure plane orientation matching value. The largest tensor component; if If they are still the same, then follow , , , , , The preset component order is selected. The three... Three unit direction labels Matching values ​​of the three structural planes and the principal propagation direction of the tensor. Combining to form the conductive capacity of the unit structure surface Surface conductivity of unit structure Characterizing the three-dimensional unit index In the The ability of an affine coupling layer to conduct along the structural plane to adjacent three-dimensional units.

[0106] Based on the center coordinates of the three-dimensional unit Determine the relationship between adjacent 3D elements. Place the 3D elements in... Axial direction, Axial direction and The element dimensions in the axial direction are denoted as follows: , and , , and Determined when the spatial extent of the underground engineering rock mass is discretized into three-dimensional elements. Indexing any three-dimensional element. This will satisfy the condition that the center coordinates differ by only one unit scale in one coordinate axis direction, and the coordinate differences in the other two coordinate axes are not greater than the preset coordinate tolerance. 3D cell index The adjacent 3D cell indexes are determined to form a set of adjacent 3D cells. The preset coordinate tolerance Used to eliminate coordinate deviations caused by grid discretization and coordinate recording precision. For each , the center coordinates of the three-dimensional unit Pointing to the coordinates of the three-dimensional unit center The unit vector is determined as the adjacent connection direction. .

[0107] Adjacent connection directions Conductivity of unit structure surface Perform direction matching to generate adjacent direction matching results. Specifically, for each , unit direction label Adjacent connection direction The absolute value of the cosine of the included angle is used as the adjacent connection direction matching value, and the adjacent connection direction matching value and the structural surface direction matching value are jointly written into the adjacent direction matching result. Adjacent direction matching results Represents three-dimensional cell index The permeability tensor parameters of this layer are preserved along the adjacent connection direction. The degree to which the conduction is simultaneously matched with the direction of the structural plane and the direction of the adjacent connection.

[0108] Based on the matching results of adjacent directions Calculate the three-dimensional element index Pointer to adjacent 3D cell index Conductivity To simultaneously incorporate the numerical values ​​of the permeation tensor parameters of this layer, the structural plane orientation constraints, the adjacent connection orientation constraints, and the boundary positions within the tensor value constraints into the inter-element conduction relationship, the conduction intensity... Determine according to the following formula:

[0109] ,

[0110] in, This indicates the summation of the three tensor components in the retained permeability tensor parameter of this layer; Indicates the label of the tensor component; Indicates the first The absolute quantity of the conductivity of the permeation tensor parameter retained in each layer; Indicates the structural plane orientation matching weight; Indicates the matching weight of adjacent connection directions; This represents the relative positional weight within the constraints of tensor value selection; , and All are dimensionless non-negative numbers, and , and The sum of ; Indicates the first Unit direction label for each tensor component; Indicates vector transpose; Represents three-dimensional cell index The direction of the structural surface; Indicates the first The inner product of the unit direction label of each tensor component with the direction of the structural surface; Indicates the first Structure surface orientation matching value for each tensor component; Indicates the first The dot product of the unit direction label of each tensor component and the adjacent connected direction; Indicates the first The adjacent connection direction matching value of each tensor component; Indicates the relationship with the first Each tensor component has a pre-defined positive number with the same dimension; This represents the forward projection of the structural plane direction onto the adjacent connection direction; This represents the absolute value operation; This represents the directional conduction modulation coefficient, determined jointly by the structural plane direction and the adjacent connection direction. Conduction intensity. The dimensions of the dimensionality are consistent with those of the three-dimensional anisotropic permeability tensor parameters. The same conduction strength formula is used, and the three-dimensional element index is... Replace with adjacent 3D cell index Index adjacent 3D units Replace with 3D cell index Calculate the index of adjacent 3D elements Pointing to the three-dimensional unit index Conductivity .

[0111] Consistent constraints are applied to the conduction strength on both sides of adjacent three-dimensional elements to form the inter-element conduction strength. For the same adjacent three-dimensional element pair... ,Will With preset conduction threshold Compare and With preset conduction threshold Comparison. Preset conduction threshold. A preset conduction threshold is used to determine whether the conduction intensity between elements constitutes a conduction relationship between three-dimensional elements. The dimensions of are consistent with those of conductivity. When Greater than or equal to the preset conduction threshold At that time, determine the three-dimensional cell index. Pointer to adjacent 3D cell index The comparison conditions are met, and As a three-dimensional unit index Pointer to adjacent 3D cell index Inter-unit conduction strength; when Greater than or equal to the preset conduction threshold At that time, determine the index of adjacent three-dimensional units. Pointing to the three-dimensional unit index The comparison conditions are met, and As an index of adjacent three-dimensional units Pointing to the three-dimensional unit index The inter-unit conduction strength. If the conduction strength in both directions is less than the preset conduction threshold... Then a three-dimensional cell index will not be formed. index of adjacent 3D cells The transmission relationship between them.

[0112] Construct conductive edges connecting the 3D elements in the directions where the comparison conditions are met. Index the 3D elements. Pointer to adjacent 3D cell index The conduction edge is denoted as The corresponding inter-element conduction strength is written into the conduction weight of the conduction edge. Index adjacent 3D units. Pointing to the three-dimensional unit index The conduction edge is denoted as The corresponding inter-element conduction strength is written into the conduction weight of the conduction edge. The conduction weight of the conduction edge represents the conduction strength between the corresponding 3D elements along the direction of the conduction edge.

[0113] The first element consists of all three-dimensional element indices, all propagating edges that satisfy the comparison conditions, and the propagation weights of the propagating edges. Three-dimensional unit conduction relationship diagram of an affine coupling layer Three-dimensional unit conduction relationship diagram Each conductive edge retains its conductive direction, adjacent connection direction, inter-element conductive strength, and conductive weight, which are used to calculate the random walk absorption probability.

[0114] S5. Calculate the random walk absorption probability based on the three-dimensional unit transmission relationship diagram and the boundary relationship from the supply boundary to the excavation face, form the absorption probability from the supply boundary to the excavation face, and form the change in the direction of absorption probability based on the difference in absorption probability between adjacent three-dimensional units.

[0115] In the Reading the three-dimensional unit conduction relationship diagram within an affine coupling layer Supply boundary marker field and excavation face marking field Supply boundary marker field Derived from the boundary relationship between the supply boundary and the excavation face, used to represent the three-dimensional element index. Is it a 3D element corresponding to the supply boundary? Excavation face marker field Derived from the boundary relationship between the supply boundary and the excavation face, used to represent the three-dimensional element index. Is it a three-dimensional element corresponding to the excavation face?

[0116] According to the supply boundary marker field Determine the set of 3D element indices corresponding to the supply boundary. , will satisfy 3D cell index writing According to the excavation face marking field Determine the set of three-dimensional element indices corresponding to the excavation face. , will satisfy 3D cell index writing .Will and They are jointly determined as the boundary calculation objects for the random walk absorption probability.

[0117] Based on the three-dimensional unit conduction relationship diagram The transitive edges in the random walk determine the transition relationships. For any transitive edge... If the conduction edge If it exists, then random walks can be indexed by three-dimensional cells. Transfer to adjacent 3D cell index If the conduction edge If it does not exist, then random walks are not allowed to be indexed by 3D cells. Transfer to adjacent 3D cell index Indexing the three-dimensional elements The outward adjacency set is composed of the indices of the adjacent 3D elements corresponding to all outward propagation edges. , conduct the edge The transmission weight is denoted as The random walk transition constraint is limited to random walks occurring only when... The transfer occurs internally; the absorption calculation unit no longer extends the random walk transfer relationship outward, and the absorption probability from the supply boundary of the absorption calculation unit to the excavation face is fixed at... .

[0118] Based on the random walk initial constraint unit Transmission relationship diagram along three-dimensional unit The edges in the path are traversed using breadth-first search. Breadth-first search starts from... Starting with the 3D cell index, the visited 3D cell indices are written to the access queue, and the adjacent sets are visited sequentially according to the direction of the propagation edges. The 3D cell indexes in the access queue are updated until no more 3D cell indexes are added. The 3D cell indexes reachable from the supply boundary are written to the reachable 3D cell index set. Will belong to The reachability marker corresponding to the three-dimensional unit index Write It will not belong to The reachability marker corresponding to the three-dimensional unit index Write Reachable marker This is a dimensionless numerical value used to ensure that the absorption probability from the supply boundary to the excavation face iterates only within the graph region reachable by the initial constraint cell of the random walk. The initial absorption probability of the absorption calculation cell is written into... Write the initial absorption probability of the non-absorbing computational unit into Based on the random walk transition constraints, absorption calculation units, random walk initial constraint units, and reachability markers, the absorption probability from the supply boundary to the excavation face is calculated iteratively according to the following formula:

[0119] ,

[0120] In the formula, Indicates the first 3D unit index in an affine coupling layer In the The absorption probability from the supply boundary to the excavation face after the next iteration; Indicates the index of the absorption probability iteration number; Represents three-dimensional cell index The excavation face marker, with a value of or ; Represents three-dimensional cell index In the The reachability flag in an affine coupling layer has a value of or ; Represents the absorption probability iterative relaxation coefficient. It is a dimensionless numerical value and its range is ; Indicates the first 3D unit index in an affine coupling layer In the The absorption probability from the supply boundary to the excavation face after the next iteration; Represents three-dimensional cell index In the Outward adjacent set in an affine coupling layer; Indicates the first Conducting edges in an affine coupling layer The transmission weights; Indicates the first Adjacent 3D unit indices in an affine coupling layer In the The absorption probability from the supply boundary to the excavation face after the next iteration; This represents a pre-defined positive number with the same dimension as the transit weights, used to ensure the denominator is not zero when the outgoing adjacent set is empty or the sum of the outgoing transit weights is zero. The numerator and denominator have the same dimension, and the fraction is a dimensionless numerical value. , , , and All are dimensionless values, therefore This is a dimensionless probability value.

[0121] During iterative calculation, from Start incrementing. After each iteration, compare the absolute difference between the absorption probabilities of two adjacent iterations corresponding to all 3D unit indices; when the maximum absolute difference is not greater than the preset absorption probability convergence threshold... Iteration stops when the number of iterations reaches the preset maximum number of iterations. Iteration stops when the threshold for absorbing probability convergence is set. For dimensionless values, the maximum number of iterations is preset. It is a positive integer. After stopping the iteration, the result of the last iteration is... Write as a three-dimensional cell index The absorption probability from the supply boundary to the excavation face .

[0122] Based on the absorption probability from the supply boundary to the excavation face Transmission relationship diagram along three-dimensional unit The direction of conduction is used to calculate the difference in absorption probability between adjacent 3D elements. For elements with conduction edges... The adjacent three-dimensional elements are indexed. absorption probability Subtract the three-dimensional unit index absorption probability , obtain along the conduction edge Difference in absorption probability between adjacent three-dimensional units in the direction . The value is dimensionless; a positive value indicates that the absorption calculation cell is closer along the conduction edge direction, a negative value indicates that the absorption calculation cell is further away along the conduction edge direction, and a zero value indicates that the absorption probability does not change along the conduction edge direction.

[0123] Index each 3D unit The corresponding absorption probability differences are assigned to adjacent conduction directions. Six directional labels are used for adjacent conduction directions. ,in and Indicates along the engineering coordinate system Positive and negative directions of the axis and Indicates along the engineering coordinate system Positive and negative directions of the axis and Indicates along the engineering coordinate system Positive and negative axes. Based on the transmission edge. The preserved adjacent connection directions will absorb probability differences. Write the corresponding direction label. (If the 3D element index...) If there is no conduction edge in a certain adjacent conduction direction, then the change in the absorption probability direction in the corresponding direction is written into... If multiple conduction edges exist within the same adjacent conduction direction, the differences in absorption probabilities within the same adjacent conduction direction are weighted and averaged according to the conduction weights, and the weighted average result is written into the corresponding direction label. The change in absorption probability direction is formed by the differences in absorption probabilities across six adjacent conduction directions. . It is a six-dimensional vector, with each of its six components representing a three-dimensional cell index. In the Along an affine coupling layer , , , , and The change in the absorption probability direction of six adjacent conduction directions.

[0124] S6. Input the absorption probability and the change in the direction of absorption probability from the supply boundary to the excavation face into the scale parameter and translation parameter generation stage of the affine coupling layer to form affine coupling transformation parameters, and update the permeability tensor parameters to be updated in this layer according to the affine coupling transformation parameters to obtain the candidate permeability tensor field.

[0125] In the Reading the intermediate state of the permeation tensor within each affine coupling layer Conditions for generating the permeation tensor Absorption probability from the supply boundary to the excavation face Change in the direction of absorption probability and structural plane direction .

[0126] Conditions for generating the permeation tensor Input Conditional Coding Layer The generation condition of the 64-dimensional permeation tensor is obtained. Conditional coding layer Includes two fully connected layers. This represents the weights and biases in the conditional coding layer. The first fully connected layer consists of sixty-four neurons, each connected to a twenty-three-dimensional permeation tensor generating conditional layers. The first fully connected layer includes weights, biases, and linear rectified units. The second fully connected layer consists of sixty-four neurons, each fully connected to the sixty-four-dimensional output of the first fully connected layer, and includes weights, biases, and linear rectified units. The permeation tensor generates a conditional representation. The sixty-four components are jointly encoded by lithological state, structural plane orientation, local permeability constraint, water level monitoring characterization, groundwater inflow data characterization at the excavation face, boundary relationship from the recharge boundary to the excavation face, and the coordinates of the three-dimensional unit center.

[0127] Representing the generation conditions of the 64-dimensional permeation tensor Preservation of permeability tensor parameters in the three-dimensional local layer Assemble them in a fixed order to form the basic generation input. Basic input generation This is a 67-dimensional vector; the first 64 dimensions represent the generation conditions of the permeation tensor, and the last three dimensions represent the parameters of the permeation tensor retained in this layer. Basic generation input. It is used to provide the scale parameter and translation parameter generation process with the tensor propagation state that has not been directly updated in the current affine coupling layer and the condition information corresponding to the three-dimensional unit.

[0128] The absorption probability from the supply boundary to the excavation face According to the three-dimensional unit index Embedded basic generation input The parameters of the permeation tensor to be updated in this layer are generated based on this, forming the absorption probability constraint input. Absorption probability constraint input It is a 68-dimensional vector. During the generation of scale and translation parameters, when generating the scale parameters corresponding to the penetration tensor parameters to be updated at this layer, the same 3D cell index is read. The absorption probability from the supply boundary to the excavation face .

[0129] According to the direction of the structural plane The set of tensor component labels corresponding to the penetration tensor parameters to be updated in this layer. This determines the propagation direction corresponding to the penetration tensor parameters to be updated in this layer. according to , , , , , Arrange the preset components in order to obtain the first... The tensor component labels corresponding to the penetration tensor parameters to be updated in this layer. ,in This indicates the group index of the penetration tensor parameters to be updated in this layer. Read tensor component labels. Corresponding unit direction label Form a direction label set by combining six adjacent conduction direction labels. and will Each direction label The corresponding unit vector is denoted as For each direction label Calculate the unit direction label With unit vector The absolute value of the inner product, and the direction of the structural plane are calculated. In unit vector The orthogonal projection on; add the absolute value of the inner product to the orthogonal projection to obtain the first... The permeation tensor parameters to be updated in this layer are in the direction label. The direction label with the highest direction match value is determined as the [number]. The propagation direction corresponding to the permeation tensor parameter to be updated in this layer. When multiple direction labels have the same direction matching value, the result is determined by... , , , , , The fixed sequence determines the direction of conduction.

[0130] Change in the direction of absorption probability In the direction of conduction The same component is assigned to the first The permeation tensor parameters to be updated in this layer form the input for directional change constraints. .in, Indicates the first Each permeation tensor parameter to be updated in this layer receives the component of the change in absorption probability direction. Each permeation tensor parameter to be updated in this layer only receives the change in absorption probability direction in the corresponding conduction direction.

[0131] Input the absorption probability constraint and the complete six-dimensional absorption probability directional change spliced ​​into a 74-dimensional coupled input 74-dimensional coupled input The first sixty-four dimensions are the conditions for generating the permeation tensor. Dimensions 65 to 67 represent the retained permeability tensor parameters for this layer; dimension 68 represents the absorption probability from the recharge boundary to the excavation face; and dimensions 69 to 74 represent the directional changes in the absorption probability along six adjacent conduction directions. Directional change constraint input. Used to limit the 74-dimensional coupled input The intragroup connection positions of the six components of the change in the direction of absorption probability.

[0132] The generation process of scale parameters and translation parameters is denoted as... , This represents the weights and biases in the scaling and translation parameter generation process. With 74-dimensional coupled input The system consists of two fully connected layers and one output layer as input. The first fully connected layer contains 72 neurons, divided into three groups of 24 neurons each, based on the 3D permeation tensor parameters to be updated in this layer. The second fully connected layer also contains 72 neurons, maintaining the three-group grouping. The output layer contains six neurons and does not have linear rectified units.

[0133] In the generation of scale and translation parameters In the middle, the first The group of twenty-four neurons reads the sixty-four-dimensional permeation tensor to generate a conditional representation. The first component arranged according to the preset component order The permeability tensor parameters of this layer are retained, and the absorption probability from the replenishment boundary to the excavation face is also considered. and the The directional change constraint input corresponding to the permeation tensor parameter to be updated in this layer. 74-dimensional coupled input The corresponding number in the middle The group retains the permeation tensor parameter components and the uncorresponding propagation direction in this layer. The component of the direction change of absorption probability, in the first... No effective connections are established within the group of 24 neurons. The first, second, and third groups of 24 neurons each generate three intra-group hidden representations of the penetration tensor parameters to be updated in this layer. The first three neurons of the output layer generate the scaling parameters. The last three neurons in the output layer generate translation parameters. Scale parameters For a three-dimensional dimensionless vector, the translation parameter is... This is a three-dimensional vector; the units of the translation parameters are consistent with the units of the permeation tensor parameters to be updated in this layer. Scale parameters... Translation parameters Together they form the parameters of the six-dimensional affine coupling transformation. .

[0134] Based on affine coupling transformation parameters For the penetration tensor parameters to be updated in this layer Perform an affine update. For the... The parameters of the current layer's penetration tensor to be updated will be... The current layer's permeation tensor parameter to be updated is multiplied by the first... Individual scale parameters The index value, plus the first Translation parameters , and get the updated number The current layer has three permeability tensor parameters to be updated. The three updated components are then arranged in the original order of the current layer's permeability tensor parameters to form the updated permeability tensor parameters. .

[0135] Update the permeation tensor parameters of this layer. Preserving the permeability tensor parameters of this layer The original component positions of the six tensor components are backfilled and combined to form the updated three-dimensional anisotropic permeability tensor parameters. Updated 3D anisotropic permeability tensor parameters It is a six-dimensional vector, with its six components corresponding to... , , , , and Six tensor component labels. When At that time, The input layer number is The affine coupling layer and the corresponding intermediate state of the permeation tensor of this layer are re-formed; when At that time, the index corresponding to all three-dimensional elements Form a candidate penetration tensor field.

[0136] Figure 3 This invention demonstrates the core process of coupling conditionally normalized flow with random walk absorption probability. After the permeation tensor generation conditions enter the affine coupling layer, they are divided into a retained portion and a portion to be updated. The retained portion, combined with the structural plane orientation and the relationship between adjacent 3D elements, constructs a 3D element transmission relationship diagram and calculates the absorption probability and the change in the absorption probability direction. The absorption probability and the change in the absorption probability direction are further fed into the scale parameter and translation parameter generation stages to form affine coupling transformation parameters, which are used to update the tensor to be updated, resulting in a candidate tensor field. Through the above steps, this invention embeds the structural plane orientation, transmission relationship, and absorption probability into the parameter update process, enhancing the consistency between the generated results and the actual seepage connectivity.

[0137] S7. Calculate the water level and groundwater inflow response at the excavation face based on the candidate permeability tensor field, and match and determine the water level and groundwater inflow response at the excavation face based on the water level monitoring data and the groundwater inflow data at the excavation face, and output the three-dimensional anisotropic permeability tensor field of the rock mass and the directed connectivity state from the recharge boundary to the excavation face.

[0138] Reading the candidate penetration tensor field Supply boundary marker field Excavation face marking field 3D unit center coordinates Relationship between adjacent three-dimensional units Water level monitoring data Groundwater inflow data at the excavation face and the absorption probability from the supply boundary to the excavation face .in, The layer number is The six-dimensional anisotropic permeation tensor parameters output by the affine coupling layer; Represents three-dimensional cell index The supply boundary marker, with a value of or ; Represents three-dimensional cell index The excavation face marker, with a value of or ; Represents three-dimensional cell index The set of adjacent three-dimensional unit indices; This indicates water level monitoring data. This indicates the water level monitoring value. Indicates a valid marker for water level monitoring; This represents the groundwater inflow data at the excavation face. Indicates the inflow value. Indicates a valid inflow marker; The layer number is The absorption probability from the supply boundary to the excavation face is calculated from the three-dimensional element transmission relationship diagram within the affine coupling layer.

[0139] According to the supply boundary marker field Determine the set of 3D element indices corresponding to the supply boundary. , will satisfy 3D cell index writing According to the excavation face marking field Determine the set of three-dimensional element indices corresponding to the excavation face. , will satisfy 3D cell index writing According to the effective markers of water level monitoring Determine the set of three-dimensional unit indexes for water level monitoring to participate in matching. , will satisfy 3D cell index writing According to the inflow valid marker Determine the set of three-dimensional element indices for matching groundwater inflow at the excavation face. , will satisfy 3D cell index writing .

[0140] Based on the candidate penetration tensor field , set of three-dimensional element indices corresponding to the supply boundary , set of three-dimensional element indices corresponding to the excavation face 3D unit center coordinates Relationship with adjacent three-dimensional units The calculation boundaries for the groundwater inflow response at the excavation face and the water level are determined. The hydraulic response calculation conditions include the candidate permeability tensor field, the set of 3D element indices corresponding to the recharge boundary, the set of 3D element indices corresponding to the excavation face, the coordinates of the 3D element centers, the relationships between adjacent 3D elements, and the recharge boundary head. Drainage head at the excavation face Seepage solution residual threshold Maximum number of iterations for solving seepage Replenishment boundary head and the drainage head of the excavation face To pre-determine the engineering boundary head, and Seepage solution residual threshold For non-negative numbers, the maximum number of iterations for percolation solution is [value missing]. It is a positive integer.

[0141] When calculating the seepage response of the candidate permeability tensor field, each three-dimensional element is indexed. The parameters of the six-dimensional three-dimensional anisotropic permeability tensor are reduced to the three-dimensional symmetric permeability tensor matrix. Three-dimensional symmetric permeation tensor matrix The main diagonal elements are arranged in sequence using , and Non-diagonal elements are written in symmetrical positions. , and For each , the center coordinates of the three-dimensional unit Pointing to the coordinates of the three-dimensional unit center The unit vector is used as the adjacent connection direction. ,Will and The distance between them is used as the center distance. Connect adjacent directions Projected to and The above yields the three-dimensional unit index. One-side and adjacent 3D cell index The permeability value along the adjacent connection direction on one side. The two permeability values ​​are harmonicly averaged to obtain the equivalent permeability value of the adjacent interface. This is based on the area of ​​the common surface of adjacent 3D units and the center distance. Convert the equivalent permeability value into the interfacial conductivity coefficient. The area of ​​the common surface is obtained by multiplying the scales of two adjacent, perpendicularly connected three-dimensional elements; the interface conductivity coefficient... Used to represent three-dimensional cell index index of adjacent 3D cells The head transfer capacity between them.

[0142] The head of the three-dimensional unit corresponding to the recharge boundary is fixed as the recharge boundary head. The water head of the three-dimensional unit corresponding to the excavation face is fixed as the drainage water head of the excavation face. For those that do not belong to And does not belong to The three-dimensional element is used, and the water balance at all adjacent interfaces is taken as the discrete seepage equation. The unknowns in the discrete seepage equation are the head response of the three-dimensional element. The interfacial conductivity coefficient is used. Assemble the sparse coefficient matrix, write the fixed head boundary into the right-hand side, and use Gauss-Seidel iteration to solve for the head response of the 3D element. In each iteration, update the head response of one non-boundary 3D element, making the 3D element index... The difference between the sum of the inflow and outflow at adjacent interfaces decreases. After all non-boundary 3D elements have undergone one update, the residuals of the discrete seepage equation are calculated; when the residuals of the discrete seepage equation are not greater than the seepage solution residual threshold... The solution is stopped when the number of iterations reaches the maximum number of iterations for percolation solution. The solution is stopped when the time is reached. The solution yields the 3D element index. Water level response For the three-dimensional element corresponding to the excavation face, read the three-dimensional elements of the non-excavation face adjacent to the three-dimensional element corresponding to the excavation face. Calculate the interfacial flow rate into the corresponding three-dimensional element of the excavation face based on the interfacial conductivity coefficient and head difference. Then, sum all the inflow interfacial flows to form the groundwater inflow response at the excavation face. .

[0143] Matching error was calculated based on water level monitoring data and groundwater inflow data at the excavation face. For each three-dimensional element index, calculate the water level response. Water level monitoring value The absolute difference, and a preset water level normalization scale. Normalization is performed to obtain the water level error; It is a positive number, and its dimensions are consistent with those of the water level monitoring values. For each 3D element index, calculate the groundwater inflow response at the excavation face. With inflow value The absolute difference, and using a preset inflow normalization scale. Normalization is performed to obtain the inflow error; It is a positive number, and its dimensions are consistent with those of the inflow value. The average value of all water level errors is denoted as... The average value of all inflow errors is denoted as .when When empty, Write And write the water level matching weight into ;when When empty, Write And write the inflow matching weights to Match water level with weights Matching weights with inflow Normalized to a dimensionless nonnegative number, and According to the water level matching weight and the inflow matching weight, and Weighting is applied to generate matching error. Matching error It is a dimensionless numerical value.

[0144] Matching error Matching threshold The comparison is performed to form a matching result. Preset matching threshold It is a dimensionless, non-negative number. At that time, the matching result will be determined. Write This indicates that the matching result meets the comparison condition, and the candidate penetration tensor field is... It was determined to be the three-dimensional anisotropic permeability tensor field of the rock mass. When At that time, the matching result will be determined. Write This indicates that the matching result does not meet the comparison conditions, and the candidate permeability tensor field is not written into the three-dimensional anisotropic permeability tensor field of the rock mass.

[0145] In the matching result In this case, based on the absorption probability from the supply boundary to the excavation face... This forms a directed connected state from the supply boundary to the excavation face. Let this directed connected state be denoted as... , Represents three-dimensional cell index Whether it points towards the excavation face along the transmission relationship. The absorption probability from the replenishment boundary to the excavation face. With preset directed connectivity threshold Comparison; preset directed connectivity threshold It is a dimensionless numerical value, and its range is [value range missing]. .when At that time, Write ;when At that time, Write . Represents three-dimensional cell index This represents a directed connectivity state from the supply boundary to the excavation face. Represents three-dimensional cell index It does not belong to the directed connectivity state from the supply boundary to the excavation face.

[0146] like Figure 4The diagram illustrates the response calculation and result determination process for the candidate permeability tensor field. The candidate tensor field, along with water level monitoring and inflow data, enters the response calculation stage to obtain the water level response and inflow response, respectively. These responses are then matched with the corresponding monitoring results. When the matching conditions are met, a three-dimensional anisotropic permeability tensor field and the directed connectivity state from the recharge boundary to the excavation face are output. The inversion results are directly correlated with the two types of engineering responses: water level and groundwater inflow at the excavation face. The output results include not only the tensor field but also the connectivity state, making the results more closely aligned with engineering applications.

[0147] Actual testing revealed that without this invention, significant water accumulation, dampness, and wetness were observed at the inlet of the drainage ditch and the junction with the sidewalls, indicating a strong connection between the local recharge and the excavation face. However, with this invention, the area of ​​water accumulation was significantly reduced, the wetness was lessened, and the surface became drier. Engineering results demonstrate that the three-dimensional anisotropic permeability tensor field and directed connectivity output by this invention have practical guiding value, providing a more effective basis for identifying and controlling groundwater inflow at the excavation face, and significantly improving the engineering relevance of the results.

Claims

1. A three-dimensional seepage field inversion analysis method applicable to anisotropic rock masses, characterized in that, include: S1. Obtain borehole lithology, structural plane occurrence, segmented water pressure data, water level monitoring data, groundwater inflow data at the excavation face, recharge boundary and excavation face of the underground engineering rock mass to form basic data on rock mass seepage. S2. Based on the rock mass seepage basic data, determine the lithological state, structural plane direction, local seepage constraint, and boundary relationship from the recharge boundary to the excavation face to form the conditions for the generation of the seepage tensor; S3. Construct an affine coupling layer for conditionally normalized flow based on the permeation tensor generation conditions. Limit the generated object to three-dimensional anisotropic permeation tensor parameters constrained by the structural surface direction. Divide the permeation tensor parameters retained in this layer and the permeation tensor parameters to be updated in this layer within the affine coupling layer to form the intermediate state of the permeation tensor in this layer. S4. Based on the retained permeability tensor parameters of this layer, the structural plane direction and the relationship between adjacent three-dimensional elements, determine the conduction relationship between the three-dimensional elements and form a three-dimensional element conduction relationship diagram. S5. Calculate the random walk absorption probability based on the three-dimensional unit transmission relationship diagram and the boundary relationship from the supply boundary to the excavation face, form the absorption probability from the supply boundary to the excavation face, and form the change in the direction of absorption probability based on the difference in absorption probability between adjacent three-dimensional units. S6. Input the absorption probability and the change in the direction of absorption probability from the supply boundary to the excavation face into the scale parameter and translation parameter generation stage of the affine coupling layer to form affine coupling transformation parameters, and update the permeability tensor parameters to be updated in this layer according to the affine coupling transformation parameters to obtain the candidate permeability tensor field. S7. Calculate the water level and groundwater inflow response at the excavation face based on the candidate permeability tensor field, and match and determine the water level and groundwater inflow response at the excavation face based on the water level monitoring data and the groundwater inflow data at the excavation face, and output the three-dimensional anisotropic permeability tensor field of the rock mass and the directed connectivity state from the recharge boundary to the excavation face.

2. The three-dimensional seepage field inversion analysis method applicable to anisotropic rock masses according to claim 1, characterized in that, S1 specifically refers to: The spatial extent of the underground engineering rock mass is discretized into a set of three-dimensional units according to a preset three-dimensional unit scale; The borehole lithology and structural plane occurrence of the underground engineering rock mass are obtained, and the borehole lithology and structural plane occurrence are mapped to three-dimensional units according to the borehole location and the location of the structural plane measuring points to form lithological and structural spatial data. Acquire segmented water pressure data and water level monitoring data, and map the segmented water pressure data and water level monitoring data to three-dimensional units according to the location of the water pressure test section and the location of the water level monitoring, to form hydraulic response spatial data; Acquire groundwater inflow data, recharge boundary and excavation face at the excavation face, and form boundary inflow spatial data based on the inflow location, recharge boundary location and excavation face location; Based on lithological structure spatial data, hydraulic response spatial data, and boundary inflow spatial data, rock mass seepage basic data are formed according to a three-dimensional unit index. The three-dimensional unit index is used to identify the corresponding unit positions of the lithological structure spatial data, hydraulic response spatial data, and boundary inflow spatial data of the same three-dimensional unit.

3. The three-dimensional seepage field inversion analysis method applicable to anisotropic rock masses according to claim 2, characterized in that, S2 specifically refers to: Based on the borehole lithology in the rock mass seepage basic data, determine the lithology category and lithology interface location corresponding to the three-dimensional unit, and form a lithology state to limit the range of values ​​of the three-dimensional anisotropic permeability tensor parameter; Based on the orientation of the structural planes in the rock mass seepage basic data, determine the dip direction and dip angle of the structural planes corresponding to the three-dimensional units, and convert the dip direction and dip angle of the structural planes into the structural plane directions used to define the principal conduction directions of the three-dimensional anisotropic permeability tensor parameters; Based on the segmented water pressure data and water level monitoring data in the rock mass seepage basic data, the local seepage value range corresponding to the three-dimensional unit is determined, forming a local seepage constraint to constrain the magnitude of the three-dimensional anisotropic seepage tensor parameter; The water level monitoring characteristics are determined based on the water level monitoring data in the rock mass seepage basic data. The water level monitoring characteristics include the reference water level and the water level change range. The groundwater inflow data characterization of the excavation face is determined based on the groundwater inflow data of the rock mass seepage basic data. The groundwater inflow data characterization of the excavation face includes the baseline inflow value and the inflow variation range. The boundary relationship between the recharge boundary and the excavation face is determined based on the groundwater inflow data at the excavation face, the recharge boundary, and the excavation face in the rock mass seepage basic data. The conditions for generating the permeability tensor are formed by combining lithological state, structural plane orientation, local permeability constraint, water level monitoring characterization, groundwater inflow data characterization at the excavation face, boundary relationship from the recharge boundary to the excavation face, and the coordinates of the three-dimensional unit center.

4. The three-dimensional seepage field inversion analysis method applicable to anisotropic rock masses according to claim 3, characterized in that, S3 specifically refers to: Based on the lithological state and local permeability constraints in the permeability tensor generation conditions, the value boundaries of the three-dimensional anisotropic permeability tensor parameters in each three-dimensional unit are determined, forming tensor value constraints. Based on the tensor value constraints and the structural surface orientation, the transmission correspondence between the three-dimensional anisotropic permeability tensor parameters and the structural surface orientation is determined, thus forming the structural surface orientation constraint relationship; Based on the structural plane orientation constraint relationship, the generated object of the conditional normalized flow is set as the three-dimensional anisotropic permeation tensor parameter constrained by the structural plane orientation, thus forming a tensor generated object. Based on the tensor-generated object, an affine coupling layer with conditionally normalized flow is constructed, and the generation target of the affine coupling layer is limited to the three-dimensional anisotropic permeation tensor parameter, thus forming the affine coupling layer generation structure. Based on the affine coupling layer generation structure, the three-dimensional anisotropic permeation tensor parameters within the affine coupling layer are set as the tensor state to be partitioned in this layer, forming a tensor state that can participate in the partitioning within the layer. Based on the tensor states that can participate in the intra-layer partitioning and the boundary relationship from the supply boundary to the excavation face, the parameter part used to preserve the transmission relationship between three-dimensional elements and the parameter part used to receive intra-layer updates are determined, forming the basis for tensor parameter partitioning. Based on the tensor parameter partitioning criteria, the affine coupling layer is divided into the retained permeation tensor parameters and the permeation tensor parameters to be updated in this layer. The retained permeation tensor parameters and the permeation tensor parameters to be updated in this layer are then combined to form the intermediate state of the permeation tensor in this layer.

5. The three-dimensional seepage field inversion analysis method applicable to anisotropic rock masses according to claim 4, characterized in that, When dividing the permeation tensor parameters to be retained and the permeation tensor parameters to be updated within the affine coupling layer, the propagation direction constraint of the three-dimensional anisotropic permeation tensor parameters entering the current affine coupling layer is applied according to the structural plane direction, and the three-dimensional anisotropic permeation tensor parameters participating in the division are determined in combination with the tensor value constraint. The parameter part corresponding to the structural plane direction and used to form the propagation relationship between three-dimensional units is divided into the permeation tensor parameters to be retained in the current layer, and the parameter part other than the permeation tensor parameters to be retained in the current layer and used to receive the update of the affine coupling transformation parameters is divided into the permeation tensor parameters to be updated in the current layer.

6. The three-dimensional seepage field inversion analysis method applicable to anisotropic rock masses according to claim 4, characterized in that, S4 specifically refers to: Based on the intermediate state of the permeability tensor of this layer, the parameters of the retained permeability tensor of this layer are extracted, and the main conduction direction of the tensor corresponding to the parameters of the retained permeability tensor of this layer is defined according to the direction of the structural surface, thus forming the conduction capacity of the unit structural surface; Based on the relationship between adjacent three-dimensional elements, determine the adjacent three-dimensional elements and adjacent connection directions corresponding to each three-dimensional element, and match the adjacent connection directions with the conduction capacity of the element structure surface to form the adjacent direction matching result. The conduction strength between adjacent three-dimensional elements along the adjacent connection direction is calculated based on the adjacent direction matching results, and the conduction strength on both sides of the adjacent three-dimensional elements is constrained to form the inter-element conduction strength. The inter-element conduction strength is compared with a preset conduction threshold, and a conduction relationship is determined between adjacent three-dimensional elements if the comparison conditions are met. Based on the conduction relationship and adjacent connection direction, conduction edges connecting three-dimensional units are constructed, and the conduction strength between units is used as the conduction weight of the conduction edges; thus forming a three-dimensional unit conduction relationship graph for calculating the random walk absorption probability.

7. The three-dimensional seepage field inversion analysis method applicable to anisotropic rock masses according to claim 6, characterized in that, In S4, the specific steps for forming the inter-unit conduction strength are as follows: calculate the conduction strength from any three-dimensional unit to an adjacent three-dimensional unit according to the adjacent connection direction, and calculate the conduction strength from an adjacent three-dimensional unit to any three-dimensional unit according to the opposite adjacent connection direction; match each conduction strength with the tensor principal conduction direction defined by the structural plane direction, and compare it with the preset conduction threshold respectively. If the conduction intensity in any direction satisfies the comparison condition, a conduction relationship between three-dimensional elements is formed along the direction that satisfies the comparison condition, and the conduction intensity that satisfies the comparison condition is used as the conduction weight of the corresponding conduction edge.

8. The three-dimensional seepage field inversion analysis method applicable to anisotropic rock masses according to claim 6, characterized in that, S5 specifically refers to: Based on the boundary relationship between the supply boundary and the excavation face, the three-dimensional elements corresponding to the supply boundary and the three-dimensional elements corresponding to the excavation face are determined in the three-dimensional element transmission relationship diagram, forming the boundary calculation object of the random walk absorption probability; Based on the transmission relationships and transmission weights in the three-dimensional unit transmission relationship diagram, the random walk transfer relationship between adjacent three-dimensional units is determined, and the random walk transfer relationship is restricted to adjacent three-dimensional units with transmission relationships, thus forming a random walk transfer constraint; Based on the random walk transition constraints and boundary calculation objects, the three-dimensional element corresponding to the excavation face is set as the absorption calculation element, and the three-dimensional element corresponding to the supply boundary is set as the random walk initial constraint element, thus forming the absorption calculation conditions from the supply boundary to the excavation face. Based on the absorption calculation conditions, the random walk absorption probability of each three-dimensional element reaching the absorption calculation element is calculated in the three-dimensional element transmission relationship diagram, forming the absorption probability from the supply boundary to the excavation face. Based on the absorption probability from the supply boundary to the excavation face, the difference in absorption probability between adjacent three-dimensional elements along the direction of the conduction relationship is calculated, thus forming the difference in absorption probability between adjacent three-dimensional elements. Based on the difference in absorption probability between adjacent three-dimensional units and the direction of conduction in the three-dimensional unit conduction diagram, the difference in absorption probability corresponding to each three-dimensional unit is assigned to the adjacent conduction direction, forming the change in absorption probability direction.

9. The three-dimensional seepage field inversion analysis method applicable to anisotropic rock masses according to claim 8, characterized in that, S6 specifically refers to: Based on the permeability tensor parameters and permeability tensor generation conditions retained in this layer, the basic generation inputs used to generate scale parameters and translation parameters in the generation process are determined. The absorption probability from the supply boundary to the excavation face is embedded into the basic generation input according to the correspondence of three-dimensional elements, which is the basis for the generation of the permeability tensor parameter to be updated in this layer, thus forming the absorption probability constraint input; The transmission direction corresponding to the permeability tensor parameter to be updated in this layer is determined based on the structural plane direction, and the change in the absorption probability direction is allocated to the transmission direction to form the direction change constraint input. Based on the absorption probability constraint input and the direction change constraint input, the scale parameters and translation parameters corresponding to the permeability tensor parameters to be updated in this layer are calculated in the scale parameter and translation parameter generation stage to form affine coupling transformation parameters. The permeation tensor parameters to be updated in this layer are updated by performing an affine update based on the affine coupling transformation parameters, thus forming the updated permeation tensor parameters to be updated in this layer. The updated permeability tensor parameters to be updated in this layer are combined with the permeability tensor parameters retained in this layer to form the updated three-dimensional anisotropic permeability tensor parameters; The updated three-dimensional anisotropic permeation tensor parameters are used as the output of the affine coupling layer, and the candidate permeation tensor field is obtained based on the output of the affine coupling layer.

10. The three-dimensional seepage field inversion analysis method applicable to anisotropic rock masses according to claim 8, characterized in that, S7 specifically refers to: Based on the candidate permeability tensor field and the boundary relationship between the recharge boundary and the excavation face, the calculation boundary of the water level and the groundwater inflow response at the excavation face is determined, thus forming the hydraulic response calculation conditions. Based on the hydraulic response calculation conditions, seepage response calculations are performed on the candidate permeability tensor field to generate the water level and groundwater inflow response at the excavation face. Based on water level monitoring data and groundwater inflow data at the excavation face, the matching error of water level and groundwater inflow response at the excavation face is calculated, and the matching error is compared with a preset matching threshold to form a matching judgment result. If the matching judgment result meets the comparison conditions, the corresponding candidate permeability tensor field is determined as the three-dimensional anisotropic permeability tensor field of the rock mass, and a directed connectivity state from the supply boundary to the excavation face is formed according to the absorption probability from the supply boundary to the excavation face.

Images