Tunnel three-dimensional line selection design method based on geological intelligent evaluation

Abstract

The invention relates to a tunnel three-dimensional line selection design method based on geological intelligent evaluation, and the method comprises the steps: obtaining the geological exploration data of a tunnel line selection section, and constructing a drilling information database; based on the drilling information database, a pre-trained single stratum excavation benefit evaluation model is adopted for benefit evaluation, and the excavation benefit score of each single stratum is obtained; based on the drilling information database, virtual drilling data are introduced, and a three-dimensional geologic model is constructed; based on the three-dimensional geologic model, performing benefit evaluation on the adjacent GTP voxels by adopting a pre-trained adjacent stratum excavation benefit evaluation model to obtain excavation benefit scores of the adjacent stratums; and constructing an underground existing structure model, superposing the underground existing structure model to the three-dimensional geologic model, and performing comprehensive evaluation according to the excavation benefit score of each single stratum and the excavation benefit score of the adjacent stratum to obtain a final three-dimensional line selection scheme. Compared with the prior art, the method has the advantages that various factors are considered, and the path is optimal.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel line selection, and in particular to a three-dimensional tunnel line selection design method based on geological intelligent assessment. Background Art

[0002] Tunnel route selection is an integral part of highway and railway route design. It is influenced by multiple objective factors, including topography, geology, hydrology, and technology, as well as supervisory factors such as expert experience, local transportation needs, and government opinions. Designers must qualitatively and quantitatively design schemes based on multiple perspectives, including route capacity, economic rationality, technical rationality, economic benefits, and coordination with other projects. Currently, tunnel route design technology is still in the 2D design era. This involves manually interpreting the topography, geology, and socioeconomic information of the design project through 2D CAD drawings. Experts then compare and select, based on various requirements, to determine technical standards and route controls. This type of tunnel route selection has limitations and can lead to efficiency losses and economic waste during the construction and operation phases of tunnel projects.

[0003] At present, there are some intelligent three-dimensional line selection methods in the existing technology, such as: the intelligent line selection method for transmission lines based on adaptive node optimization ant colony algorithm disclosed in patent application CN115186929A, the intelligent line selection method for transmission lines based on GIS multi-objective dynamic programming technology disclosed in patent application CN113256011A, the rapid modeling method for railway intelligent line selection direction scheme based on BIM disclosed in patent application CN112231799A, an improved three-dimensional geological modeling method based on Kriging interpolation disclosed in patent application CN117671182A, a method based on multi-source data disclosed in patent CN112070890B. Patent application CN117371098A discloses a method for rapid and refined three-dimensional geological modeling, patent application CN117371098A discloses a method for highway route selection based on AI intelligence, and patent CN112560215B discloses a method for power line selection based on deep reinforcement learning. Among them, the intelligent line selection patents mostly adopt the form of grid diagrams, and the optimal line plan is obtained by calculating the unit cost in the two-dimensional grid diagram, and the cost calculation is mostly carried out in a comprehensive evaluation manner; the three-dimensional line selection patents mostly adopt a method of combining manual and system design, mainly by setting control points, and then according to the linear elements, combined with cost analysis and other methods to obtain the final line selection plan; currently there are few related technologies for shield tunnel line selection.

[0004] A route selection method combining GIS and genetic algorithms has also been proposed in the prior art. Its basic design process is as follows: 1) Build a GIS terrain model based on geological survey and aerial photography data; 2) Use a genetic algorithm to randomly generate a route plan between the starting and ending points, obtaining a series of points; 3) Input the point coordinates into the geographic information system to obtain a route plan; 4) Based on the digital elevation model in the GIS, cut the digital elevation model along the route plan line to obtain a longitudinal section; 5) Cut the digital elevation model at equal intervals and perpendicular to the route direction to obtain a cross-section; 6) Calculate the route construction cost by comprehensively considering the plan, longitudinal and cross-section diagrams, and various factors affecting route selection; 7) Output the results to the genetic algorithm program for optimization decision-making, find the optimal solution, and calculate the change in the objective function; 8) Input the more optimal solution found in step 7 into the geographic information system, and repeat steps 3 to 7 until the objective function value reaches a certain change and ends the cycle. The disadvantage of this method is that it requires calculating the construction cost of the entire route, which requires the development of very detailed and complex calculation formulas for each influencing factor, which is very labor-intensive. Secondly, genetic algorithm programming is also very difficult, and because the iterative process is in a black box state, when problems occur in the result line, it is difficult to make reasonable modifications to the algorithm itself based on the results.

[0005] A comprehensive analysis of the reasons reveals the following: 1) Route selection schemes designed using two-dimensional plan drawings are difficult to interpret; 2) Traditional manual route selection often focuses on the influence of a single factor; 3) Some route selection decisions consider the influence of multiple factors, but the route selection process is highly subjective. Summary of the Invention

[0006] The purpose of the present invention is to provide a three-dimensional tunnel line selection design method based on geological intelligent evaluation with a comprehensive optimal line scheme.

[0007] The purpose of the present invention can be achieved by the following technical solutions:

[0008] A three-dimensional tunnel line selection design method based on geological intelligent assessment includes the following steps:

[0009] Obtain geological exploration data for the tunnel alignment section and construct a drilling information database to record stratum information;

[0010] Based on the drilling information database, a pre-trained single stratum excavation benefit evaluation model is used to perform benefit evaluation, obtain an excavation benefit score for each single stratum, and record the score in the drilling information database;

[0011] Based on the borehole information database, virtual borehole data is introduced, and a three-dimensional geological model is constructed with GTP voxels as units, wherein a single GTP voxel represents single stratum information;

[0012] Based on the three-dimensional geological model, a pre-trained adjacent stratum excavation benefit evaluation model is used to evaluate the benefits of adjacent GTP voxels to obtain an adjacent stratum excavation benefit score;

[0013] A model of existing underground structures is constructed and superimposed on the three-dimensional geological model. A comprehensive evaluation is performed based on the excavation benefit scores of each single stratum and the excavation benefit scores of adjacent strata to obtain a final three-dimensional line selection plan.

[0014] Furthermore, the geological exploration data includes spatial location data, drill holes, geological profiles, structural maps, and stratigraphic attributes.

[0015] Furthermore, the borehole information database includes multiple sets, including a stratum set, a borehole set, a borehole-stratum set, and an attribute set, wherein the stratum set includes a stratum number and a stratum name.

[0016] The drill hole set includes the drill hole number, plane coordinates, surface height, number of layers and overall geological characteristics of the region.

[0017] The borehole-stratum set includes a borehole-stratum number, a borehole number, a stratum number, a bottom coordinate, an attribute number, and an excavation benefit score;

[0018] The attribute set includes borehole-stratum number, gravity, water content, saturation, cohesion, internal friction angle, elastic modulus, Poisson's ratio, porosity, earth pressure coefficient, compression coefficient, compression modulus, standard penetration test hits, cone tip resistance, side wall friction resistance, permeability coefficient, surrounding hydrological type, water area orientation and water area clear distance.

[0019] Furthermore, the training steps of the single stratum excavation benefit evaluation model include:

[0020] Obtain each expert's initial excavation benefit score for different single stratum samples, and use the weighted average method to calculate the final excavation benefit score for different single stratum samples;

[0021] Build a single-layer neural network model;

[0022] The attribute set in the drilling information database is used as input, and the final excavation benefit scores of different single stratum samples are used as output. The single-layer neural network model is trained to establish the correlation between the stratum samples and the excavation benefit scores of the single stratum samples, thereby forming a single stratum excavation benefit evaluation model. The single stratum excavation benefit evaluation model uses formula (1) to calculate the excavation benefit score of each single stratum sample, and uses formula (2) to map the excavation benefit score of each single stratum sample to the (0, 1) interval to obtain the corresponding occurrence probability:

[0023]

[0024] in:

[0025] P1+P2+P3+P4+P5=1

[0026] 0≤P1,P2,P3,P4,P5≤1

[0027] Where, and is the model output, i.e. the five stratum excavation benefit ratings, is the stratum attribute in the attribute set, w is the weight scalar, b is the bias, P i is the probability of occurrence, and e is a natural constant.

[0028] Furthermore, the step of obtaining the final single stratum sample excavation benefit score includes:

[0029] Obtain the expert's ratings and corresponding benefit weights for time benefit, cost benefit, and risk benefit of different single stratum samples, and calculate the initial excavation benefit scores of each expert for different single stratum samples. The initial excavation benefit score of a single stratum sample = time benefit weight * time benefit grade + cost benefit weight * cost benefit grade + risk benefit weight * risk benefit grade, and the sum of all benefit weights is 1.

[0030] Considering the attributes of each expert, the weight of each expert suggestion for each single stratum sample is calculated, where the calculation expression of the suggestion weight is:

[0031]

[0032] Where Ω is the weight of expert recommendation, α1 is the weight of the expert's unit, α2 is the weight of familiarity, α3 is the weight of professional title, and α4 is the weight of years of work experience;

[0033] Based on the initial excavation benefit scores of the different single stratum samples and the weight values ​​recommended by the experts, a weighted average is performed to obtain the final excavation benefit score of the single stratum sample.

[0034] Furthermore, the steps of constructing the three-dimensional geological model include:

[0035] Based on the drilling information database, the Kriging interpolation method is used to interpolate missing points to establish virtual drilling holes, thereby obtaining a drilling information database including virtual drilling holes, wherein the expression of the Kriging interpolation method is:

[0036]

[0037] where λ i Satisfies the equation:

[0038]

[0039] Where K * (x0) is the estimated value of any block, K(x i ) is the value of n valid samples within the influence range of the block segment, λ i is related to the variable K(x i ) is used to represent the contribution of each variable value to the valuation, γ(x i ,x j ) is the semivariogram of the two points, μ is the Lagrange multiplier;

[0040] Based on the drilling information database including the virtual drilling holes, the surface of each layer is constructed by TIN grid division to form a multi-layer DEM model;

[0041] Based on the multi-layer DEM model, GTP voxels corresponding to a single stratum are constructed according to the drilling points and the Delaunay triangulation to form a three-dimensional geological model.

[0042] Furthermore, the step of forming a multi-layer DEM model includes:

[0043] 1) Projecting the discrete point data of each single stratum in the drilling information database including the virtual borehole onto a horizontal plane to obtain a point set S1;

[0044] 2) Let the minimum coordinate point in point set S1 be p1, set p1 as the coordinate origin through axis-shift transformation, and obtain the local coordinate system xy and point set S2;

[0045] 3) Based on the point set S2, calculate p1p n Sort the angles with the positive direction of the x-axis to get p1,…,p n , where p1 is the starting point of the convex hull boundary, p n is the vertex of the convex hull, and the calculation expression of the angle is:

[0046] cotα i =x i / y i

[0047] The sorting operation is performed using the characteristics of the cotangent function, and the expression is:

[0048]

[0049] In the formula, (x i ,y i ) is point p i Coordinate, α i For p1p n Angle with the positive direction of x-axis;

[0050] 4) Let p n+1 =p1;

[0051] 5) Connect p i p j , where initially i=1, j=2;

[0052] 6) Determine point p k With directed straight line The position relationship, where the initial k = 3, the judgment expression is:

[0053]

[0054] Where S is Δp i p j The area of ​​p, S is positive, negative and zero, respectively, representing the point p k On a directed straight line On the left, right or straight line, (x i ,y i )、(x j ,y j ) are points p i 、p j The coordinates of , (x, y) are the coordinates of the new point p;

[0055] 7) If point p k On a directed straight line On the left side, then i←j, j←k, k←k+1, return to step 5), if point p k On a directed straight line , or on the left side of the line, delete the line segment. j←k, k←k+1, return to step 5);

[0056] 8) Until j=n+1, connect p i p j , forming a convex hull;

[0057] 9) Sort the points in the convex hull to obtain the sequence p1, p2, ..., p m ;

[0058] 10) Take point p1 from the sequence, connect point p1 with each point on the convex hull boundary to obtain an initial triangulated network, and set initial i=1;

[0059] 11) Let i = i + 1, and take out point p i, perform empty circumscribed circle detection on all initial triangulated networks, obtain the initial triangulated networks that do not meet the empty circumscribed circle criteria, and mark them. In the process of empty circumscribed circle detection, the circumscribed circle of the triangle is used to generate triangles along the circumscribed circle line under the condition of fixing one side of the triangle. The property that the inner angle of the triangle corresponding to the free moving point remains unchanged is used to judge the point p i Whether it is inside the circle, thus determining whether it meets the empty circumcircle criterion;

[0060] 12) Delete the marked internal edges of the initial triangulated network to form a Delaunay hole C;

[0061] 13) Move point p i Connect the vertices of the boundary of Delaunay hole C to form a new triangulated DEM;

[0062] 14) Repeat steps 11)-13) until all points inside the convex hull are processed, and add the z coordinate of each point to finally form a multi-layer DEM model.

[0063] Furthermore, the steps of constructing the GTP voxel include:

[0064] The triangulated network DEM is vertically extended downward along the borehole one by one to generate GTP voxels. During the extension generation process, if the stratum codes of the three vertices of the triangulated network DEM are the same, the drill point chain of the vertex is searched downward until the three nearest points with the same stratum code are found. The three points form a TIN surface as the lower TIN surface of the current GTP voxel and the upper TIN surface of the next GTP voxel in the GTP voxel chain. This process is repeated to complete the construction process of all GTP voxels.

[0065] The GTP voxels generated by expansion include the following situations:

[0066] (1) If the difference between the stratum numbers of the upper and lower TIN surfaces is 1, it is a homogeneous GTP voxel;

[0067] (2) If the difference between the stratum numbers of the upper and lower TIN surfaces is not 1, it is a heterogeneous GTP voxel and needs to be further decomposed to form a homogeneous GTP voxel or a degenerate tetrahedron, where the vertices of the heterogeneous GTP voxel are set to A, B, C, A1, B1, and C1. The decomposition step includes:

[0068] a) The heterogeneous GTP volume has only one edge with a stratum number greater than 2:

[0069] Set midpoints G, H, G1, and H1 on both sides of the upper and lower TIN corner points AB, BC, A1B1, and B1C1 respectively;

[0070] Connect HH1 and GG1, set points E and F on HH1 and GG1, where the Z coordinates of points E and F are the same as the Z coordinate of point D, which is located on BB1;

[0071] The quadrilateral AGHC is meshed according to the Delaunay criterion to decompose the heterogeneous GTP voxel into multiple homogeneous GTP voxels;

[0072] b) The number of strata contained in two edges of the heterogeneous GTP voxel is greater than 2, and the two edges have the same number of strata:

[0073] Set midpoints G, F, G1, and F1 on both sides of the upper and lower TIN corner points AC, BC, A1C1, and B1C1 of the edges with a small stratum proportion;

[0074] Connect F1F and GG1, set points E and E1 on F1F and GG1, where the Z coordinate of point E is the same as the Z coordinate of point D, the Z coordinate of point E1 is the same as the Z coordinate of point D1, and point D is on BB1;

[0075] The quadrilateral ABFG is meshed according to the Delaunay criterion to decompose the inhomogeneous GTP voxel into multiple homogeneous GTP voxels;

[0076] c) The number of strata contained in two edges of the heterogeneous GTP voxel is greater than 2, and the two edges contain different numbers of strata:

[0077] Find the midpoints of the three edges of the upper and lower TIN surfaces G, H, I, G1, H1, and I1 respectively;

[0078] Connect GG1, HH1, and II1, and set points D1, F1, and E1 on GG1, HH1, and II1. The Z coordinate of point D1 is the same as that of point D, the Z coordinate of point E1 is equal to the average of the Z coordinates of points E and F, and the Z coordinate of point F1 is the average of the Z coordinates of points D, E, and F. Point D is located on BB1, and points E and F are located on CC1.

[0079] The quadrilateral AGHC is meshed according to the Delaunay criterion to decompose the heterogeneous GTP voxel into multiple homogeneous GTP voxels.

[0080] Furthermore, the training steps of the adjacent strata excavation benefit evaluation model include:

[0081] Obtain each expert's initial adjacent stratum excavation benefit score for the adjacent stratum sample, and consider the expert's attributes to calculate the final adjacent stratum sample excavation benefit score using the weighted average method, where the initial adjacent stratum excavation benefit score = time benefit weight * time benefit level + cost benefit weight * cost benefit level + risk benefit weight * risk benefit level, and the sum of each benefit weight is equal to 1;

[0082] A neural network model is constructed and trained to obtain a trained adjacent stratum excavation benefit evaluation model, wherein the input of the neural network model includes the attribute set in the drilling information database and the adjacent stratum sample grouping, and the output is the final adjacent stratum sample excavation benefit score.

[0083] Furthermore, the step of obtaining the final three-dimensional line selection solution includes:

[0084] For a 3D geological model that includes an existing underground structure model, select tunnel entrances and exits, set the total number of path curve midpoints, and set the safety control distance of the existing structure;

[0085] Connecting tunnel entrances and exits;

[0086] Generate random arrangement of curve midpoints in the path according to the total number of curve midpoints set;

[0087] A curve path is generated using the midpoints of each curve as control points, wherein the curve path includes a horizontal curve formed by projection on the plane XY and a longitudinal curve formed by projection on the longitudinal plane XZ. The process of generating the horizontal curve includes:

[0088] Set the control elements of the horizontal curve, including the intersection point coordinates JD, the straight-slow point coordinates ZH, the smooth-straight point coordinates HZ, the radius R, the turning angle value, the transition curve length Ls, the tangent length T, and the outer distance;

[0089] Discrete the control elements of the flat curve into a group of line elements connected end to end to form a flat curve;

[0090] The generation process of the longitudinal curve includes:

[0091] Set the control elements of the longitudinal curve, including the intersection coordinate JD, the foresight slope i1, the backsight slope i2, the curvature radius R, the tangent length T, and the outer distance E;

[0092] Discrete the control elements of the longitudinal curve into a group of line elements connected end to end to form a longitudinal curve;

[0093] Determine whether the curved path satisfies the following conditions in addition to the preset building spacing:

[0094] ① Longitudinal slope inspection: Calculate the longitudinal slope of the straight line segment. If the slope of any segment does not meet the preset range, it is considered that it does not meet the line design requirements. Otherwise, it is considered to meet the requirements.

[0095] ② Curvature radius check: Calculate the plane curvature radius of the line curve segment. If the curvature radius of any segment is less than the preset value, it is considered that the line design requirements are not met. Otherwise, it is met.

[0096] Among them, the line straight segment and the line curve segment are connected to each other, and a curve path is generally composed of two straight segments with a curve segment in the middle.

[0097] For the line selection paths that meet the conditions, a comprehensive evaluation is performed based on the excavation benefit score of the single stratum corresponding to the GTP volume and the excavation benefit score of the adjacent strata to obtain the final three-dimensional line selection scheme, where the expression for the comprehensive evaluation is:

[0098]

[0099] Where C is the total benefit evaluation value, α is the benefit evaluation weight of single stratum excavation, X i is the excavation benefit evaluation score of a single stratum of GTP volume i, β is the excavation benefit score of the adjacent stratum, and Y i is the crossing benefit evaluation level of adjacent stratum relationship j.

[0100] Compared with the prior art, the present invention has the following beneficial effects:

[0101] (1) With the help of the grid evaluation concept, the present invention uses the characteristics of GTP elements carrying stratum benefits and stratum properties to construct a three-dimensional geological model, transforming the two-dimensional grid network into a three-dimensional grid network. It not only targets the unit body, but also cooperates with the interaction characteristics between strata unique to tunnel engineering, and considers the benefit impact of crossing between adjacent strata. The tunnel line specifications and the safety range of surrounding existing building structures are used as limiting constraints. Finally, the comprehensive optimal shield tunnel intelligent line selection in the three-dimensional stratum is achieved through benefit comparison.

[0102] (2) The present invention is designed to realize an intelligent design method for three-dimensional tunnel line selection and obtain the optimal path. The optimal tunnel line selection path is derived by comparing the benefit scores of all options. Based on the attribute information of each stratum obtained by drilling exploration, a stratum benefit evaluation model is established by combining expert scoring with a neural network algorithm, thereby obtaining the construction benefit evaluation score of each stratum. The stratum information and the benefit evaluation score are assigned to the three-dimensional geological model together through the DEM-GTP modeling method to obtain a three-dimensional geological model with transparent unit benefits. At the same time, since there are different degrees of construction risks when the tunnel passes through adjacent strata, the expert scoring is also combined with the neural network method to obtain an adjacent crossing construction risk benefit decision model that takes into account the spatial position between each stratum. During route selection, after importing the existing underground structure control points, it is only necessary to preset the tunnel entrances and exits, the number of path curve midpoints, the structure control distance, the tunnel cross-sectional area, and the weight ratio of the two benefit assessments. Based on the relevant specifications for tunnel line types, while maintaining a preset distance from existing structures, the total benefit assessment of all units crossed by the route and the total construction risk benefit assessment of all crossings of adjacent strata can be combined through intelligent calculation and comparison to obtain the route plan with the least benefit, that is, the optimal tunnel path decision.

[0103] (3) The present invention regards stratum elements as nodes and establishes lines between stratum interfaces, dividing the entire line selection section into a point-line-point structure. This model fixes the evaluation objects, i.e., points and lines, in advance. In the early work, by collecting expert evaluation opinions, a benefit decision model is constructed using a machine learning algorithm, which realizes the use of models to replace expert evaluation. Compared with traditional line selection work, it can achieve rapid line benefit evaluation. At the same time, the evaluation results of the decision model obtained by training a large number of samples with the weighted matching evaluation of expert personal attributes, on the one hand, avoid the problem of focusing on a single influencing factor when evaluating line selection by a single person; on the other hand, it avoids the large subjective shortcomings in line selection work. In addition, the benefit evaluation model with points and lines as evaluation targets is constructed, and the large number of characteristic values ​​that need to be considered in line benefit evaluation are standardized into two categories, which reduces the modeling complexity and shortens the decision evaluation time. The transparent unit decision benefit value can provide later verification work. Combined with the three-dimensional geological model modeling technology, the line selection basis is assigned to the three-dimensional geological model, which can intuitively display the line selection results, solving the problem of difficulty in interpreting the two-dimensional plane design line selection scheme in traditional line selection work. BRIEF DESCRIPTION OF THE DRAWINGS

[0104] Figure 1 Schematic diagram of the method flow of the present invention;

[0105] Figure 2 The single stratum sample excavation benefit evaluation process of the present invention;

[0106] Figure 3 A diagram of a single-layer neural network computing model of the present invention;

[0107] Figure 4 The three-dimensional data hybrid model of the present invention;

[0108] Figure 5 This is a data relationship diagram of the drilling information database of the present invention;

[0109] Figure 6 This is a schematic diagram of sorting point sets by viewing angle according to the present invention;

[0110] Figure 7 A diagram showing the process of generating the convex hull of a point set according to the present invention;

[0111] Figure 8 A diagram of the point set triangulation generation process of the present invention;

[0112] Figure 9 This is a schematic diagram of empty circumscribed circle detection according to the present invention;

[0113] Figure 10 Expanding the GTP voxel graph for the drilling point chain of the present invention;

[0114] Figure 11 Schematic diagram of the decomposition of heterogeneous GTP of the present invention, wherein Figure a is the first heterogeneous GTP case, Figure b is the second heterogeneous GTP case, and Figure c is the third heterogeneous GTP case;

[0115] Figure 12 It is a schematic diagram of the flat curve elements of the present invention;

[0116] Figure 13 It is a schematic diagram of the vertical curve element of the present invention. DETAILED DESCRIPTION

[0117] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0118] This embodiment provides a three-dimensional tunnel line selection design method based on geological intelligent assessment, such as Figure 1 As shown, the method includes the following steps:

[0119] S1. Design an expert scoring table with stratum properties as the evaluation target, collect a large amount of stratum property information and provide it to experts for excavation benefit evaluation, and obtain a unique scoring result for all stratum samples by weighted average.

[0120] S2. Combining the stratum sample attributes with the excavation benefit scoring results, a stratum excavation benefit evaluation decision model is obtained through training with a machine learning algorithm, and the stratum excavation benefit can be evaluated through the attribute centralized data information.

[0121] The result of the stratum excavation benefit evaluation is one of the important factors in determining the final line selection decision. It will evaluate and assign a value to the excavation benefit of each unit body of the 3D geological model. This section explains steps S1 to S2 in detail. The detailed method is as follows:

[0122] (1) Prepare expert opinion form for ground excavation benefit evaluation

[0123] The evaluation of stratum excavation benefit is based on expert evaluation opinions. In order to eliminate the subjective judgment factors of experts as much as possible and realize stratum benefit scoring from multiple angles, the evaluation opinion form sent to experts is designed to integrate the three factors of time, cost and risk. Experts can allocate benefit weights according to their personal ideas and combine geological information to finally obtain the stratum benefit score, as shown in Table 1.

[0124] in,

[0125] Time efficiency: This mainly evaluates the safe excavation speed of the excavated soil sample. The faster the excavation speed, the higher the efficiency level.

[0126] Cost-effectiveness: The excavation cost of the sample soil is estimated from the perspectives of “manpower, machinery, materials, methods, and environment”. The lower the cost, the higher the benefit evaluation.

[0127] Risk-benefit: mainly refers to a series of risk problems that may be faced when excavating sample soil, such as the floating problem of shield tunnel excavation segments in soft soil strata, rock burst problem in highway tunnel excavation in rocky strata, etc. By judging the degree of excavation risk, the benefit is scored. The lower the risk, the higher the benefit.

[0128] At the same time, the reference value of the expert's personal opinion is weighted according to the expert's personal resume, and the personal attributes of the evaluation expert are investigated, as shown in Table 2.

[0129] Table 1 Expert opinions on formation benefit assessment

[0130]

[0131] Note: The factor benefit level is positively correlated with the stratum excavation benefit score. The higher the factor benefit level, the higher the stratum excavation benefit score. The final stratum excavation benefit score = time benefit weight * time benefit level + cost benefit weight * cost benefit level + risk benefit weight * risk benefit level, and the sum of all benefit weights is equal to 1.

[0132] Table 2 Expert attribute questionnaire

[0133]

[0134] (2) Create a sample attribute record form

[0135] As a statistical stratigraphic information sample, the stratigraphic sample information table provided to experts as a reference for evaluation and decision-making must simultaneously meet the characteristics of being comprehensive enough for benefit evaluation and being easily accessible.

[0136] a. Overall regional geological characteristics. The overall regional geological conditions are primarily determined by the stratigraphic attributes within the 10-meter radius above and below the crossing area. These attributes are categorized into four types: soft soil, hard soil, clay soil, and sandy soil. Multiple selections are permitted. After a preliminary determination of the construction method, the benefit assessment level should be reduced for stratigraphic characteristics that differ significantly from the overall regional geological characteristics.

[0137] b. Surrounding hydrological characteristics. The information table provides assessment experts with information on surrounding hydrological characteristics in two forms: type and clearance. Clearance is recorded in two types: lateral and overhead, combined with four range levels. Lateral describes riverside and coastal areas, while overhead describes river crossings and sea crossings.

[0138] c. Basic properties of strata, refer to general geological survey data.

[0139] The final table design is shown in the table:

[0140] Table 3 Stratum sample information

[0141]

[0142] (3) Collect and organize expert opinions

[0143] Collect and organize the expert opinion forms and extract the data from the expert opinion forms.

[0144] (4) Formulate standards for weighting expert opinions

[0145] To reduce the influence of expert subjectivity, we further processed the expert opinions using a weighted average. The final evaluation score for each sample number was determined using the sample number as the unit. The weighted recommendations for each expert are recorded in the table below, with the weights accurate to one decimal place.

[0146] Table 4 Expert weight recommendation value record table

[0147]

[0148] The stratum excavation benefit scores given by each expert in the stratum benefit evaluation expert opinion form are calculated based on the expert weight recommendation value to obtain the weighted average benefit score of each stratum sample as its final evaluation score, and the result is rounded to an integer.

[0149] The expert opinion evaluation process for a single formation sample is as follows: Figure 2 As shown in the figure, first, the experts receive the evaluation information, evaluate the sample information in terms of time benefit, cost benefit and risk benefit, and assign benefit weights to sum them up. Then, they assign individual weights to the experts, and take the weighted average of the expert decision-making levels to obtain the single stratum sample excavation benefit evaluation score (level).

[0150] (5) Constructing a model for evaluating the benefits of ground excavation

[0151] The model uses stratum sample information provided to experts for evaluation as input, and the final stratum excavation benefit score, extracted based on the weighted expert decisions, as the model output. The characteristic weights of different stratum attribute parameters in the benefit score decision-making process are trained to construct a stratum excavation benefit evaluation model. In practical applications, using stratum sample information from geological survey data as model input, the evaluation model can determine the excavation benefit score for that stratum.

[0152] The specific implementation method is as follows: a single-layer neural network is used to establish the correlation between stratum sample information and stratum excavation benefit screen. The output layer is a fully connected layer. The calculation model structure is as follows: Figure 3 shown.

[0153] The 19 formation attributes in the formation sample information table are used as the input of the model (x in ), since there are 5 stratum excavation benefit ratings as output (z out ), the weight contains 30 scalars (w), and the bias contains 5 scalars (b). The decision scores corresponding to the excavation benefit rating levels of each stratum can be calculated according to formula (1).

[0154]

[0155] The output result is mapped to the interval (0, 1) using formula (2) to obtain the corresponding decision probability (P).

[0156]

[0157] Where, e is a natural constant, which is 2.718.

[0158] The result satisfies the following two conditions:

[0159] P1+P2+P3+P4+P5=1 (3)

[0160] 0≤P1,P2,P3,P4,P5≤1 (4)

[0161] Where P and z out One-to-one correspondence, P i The bigger the description The greater the possibility that the corresponding stratum excavation benefit score level will be used as the decision result, the corresponding category with the highest probability of decision occurrence will be taken as the excavation benefit evaluation level.

[0162] S3. The survey unit shall provide geological exploration data for the selected route section, including spatial location data, drill holes, geological profiles, structural maps, stratigraphic properties, etc.

[0163] S4. Preprocess the data and establish a borehole information database, which is divided into stratum set, borehole set, borehole-stratum set, and attribute set, record stratum information, and perform benefit evaluation on each borehole-stratum set object through the stratum excavation benefit evaluation model.

[0164] S5. Arrange the drilling sampling point data on different stratigraphic interfaces according to stratigraphic sequence, including number, elevation, coordinates, attributes, benefit evaluation, etc., and supplement the point density through interpolation method, construct the surface of each stratigraphic layer in the form of TIN, and form a multi-layer DEM model.

[0165] S6. Based on the Delaunay triangulation determined and generated according to the drilling points on each layer of DEM, the corresponding vertices of the upper and lower layers are connected to obtain a three-dimensional geological model composed of a plurality of generalized triangular prisms.

[0166] use Figure 4 The 3D data hybrid model shown here performs dimensionality reduction and subdivision of the 3D geological model from top to bottom, including geometric topological relationships and stratigraphic attribute information, where (X, Y, Z) refers to the global coordinates of the drill point. This section will mainly explain S3 to S6 in detail. The detailed method is as follows:

[0167] (1) Drilling data processing

[0168] The data sources for 3D geological modeling include three aspects: first, direct measurement data, such as field fill, on-site surveys, satellite remote sensing, and GPS data; second, borehole data and well data obtained through drilling; and third, seismic data. This data, which corresponds to the data provided to experts for benefit evaluation, is organized into geological distribution data in the form of borehole data records to form borehole data.

[0169] Borehole data consists of multiple sets of discrete points. Each borehole point includes plane coordinates (X, Y) in global coordinates and elevation information (surface elevation and elevation of each stratum boundary node). Since the original borehole data are of many types, highly discrete, and have different reliability levels, the borehole data are first classified and preprocessed, and a borehole information database is constructed. Figure 5 The recorded information includes all the characteristic values ​​required by the stratum excavation benefit evaluation model. The evaluation model can be connected through the data interface to perform batch benefit evaluation on the stratum samples of the selected line address and record them.

[0170] (2) Introduction of virtual drilling

[0171] In order to supplement the data missing caused by insufficient borehole arrangement density in actual work, the Kriging interpolation method is used to interpolate missing points and establish virtual boreholes. The existing borehole data are used as samples, and a certain weight is assigned to each sample value. The weighted average method is used to interpolate the unknown attributes of the estimated area.

[0172] The interpolation calculation of the Kriging method is as shown in formula (5):

[0173]

[0174] Among them, K * (x0) is the estimated value of any block, K(x i ) is the value of n valid samples within the influence range of the block segment, λ i is related to the variable K(x i ) is used to represent the contribution of each variable value to the valuation.

[0175] According to the linear unbiased and minimum estimation variance properties, the weighted coefficient satisfies the equation:

[0176]

[0177] Among them, γ(x i ,x j ) is the semivariogram of the two points, μ is the Lagrange multiplier, and solving this system of equations yields the weighting coefficient λ i , and then we can find the value K * (x0).

[0178] (3) TIN grid division

[0179] First, the convex hull generation algorithm of spatial scattered points is used to generate the convex hull based on the borehole data point set. Taking the surface point set S0 as an example, n discrete points are projected onto the horizontal plane to obtain a new point set S1. Let the point with the minimum y coordinate in S1 be p1. By using the axis-shift transformation, p1 is set as the coordinate origin to obtain the local coordinate system xy and the point set S2. Each point in S2 is transformed according to its perspective, that is, p1p n Sort by the angles with the positive direction of the x-axis to get p1, ..., p n ; p1 is the starting point of the convex hull boundary, p n It must also be a vertex of the convex hull, such as Figure 6 As shown, the sorting judgment method is as follows:

[0180] a. Set point p i The coordinates are (x i ,y i ), calculate p i The angle α between p1 and the positive half axis of the x-axis i Cotangent value cotα i =x i / y i ;

[0181] b. Using the characteristics of the cotangent function (when x∈(0,π), y=cot x is monotonically decreasing), we have formula (7):

[0182]

[0183] Let p n+1 = p1, follow the steps below and Figure 7 Generate the convex hull:

[0184] a. Connect p i p j (Initially i=1, j=2);

[0185] b. Determine p k (Initial k = 3) and directed straight line Positional relationship;

[0186] It can be judged by formula (8):

[0187]

[0188] Where (x i ,y i )、(x j ,y j ) are points p i 、p j The coordinates of the new point p are (x, y), and S is Δp i p j The area of ​​p. S is positive, negative and zero, respectively, and determines the area of ​​point p on the directed line. to the left, right or in a straight line.

[0189] c. If it is on the left, then i←j, j←k, k←k+1, and go to a;

[0190] If on the right or above, delete the line segment j←k, k←k+1, turn to a;

[0191] d. Until j = n + 1, then connect p i p j , a convex hull is formed.

[0192] Then, to complete the TIN segmentation, we insert it point by point, combined with Figure 8 , the specific steps are as follows:

[0193] a. Sort the points inside the convex hull to get the sequence p1, p2, ..., p m ;

[0194] b. Take point p1 and connect it with the points on the convex hull boundary to obtain the initial triangulation, and set the initial i=1;

[0195] c. Let i = i + 1, and take out point p i , perform empty circumcircle detection on all triangles, find and mark the triangles that do not meet the empty circumcircle criteria;

[0196] d. Delete the internal mesh edges of the triangle edges marked above to form Delaunay holes;

[0197] e.P i Connect with the vertices on the boundary of hole C to form a new triangulated network;

[0198] f. Repeat steps c to e until all points inside the convex hull have been processed, and finally add the z coordinate of each point.

[0199] Among them, such as Figure 9The empty circumcircle detection method shown here uses the circumcircle of a triangle to generate a triangle along the circumcircle line while keeping one side of the triangle fixed. The property that the interior angle of the triangle corresponding to a freely moving point remains unchanged allows for determining whether the test point is within the circle. Given a known side length, the interior angle of the triangle can be calculated using the cosine law.

[0200] in:

[0201] At this time, Δ does not meet the empty circumcircle criterion;

[0202] At this time, Δ meets the empty circumcircle criterion.

[0203] (4) GTP body construction

[0204] The modeling principle of GTP voxels is to use the TIN surface composed of the triangles of the upper and lower bases of the GTP to express different ground layers. Then, the spatial quadrilateral surfaces of the GTP side are used to describe the spatial relationship between layers, and the GTP cylinders are used to express the internal entities between layers. The implementation steps are as follows:

[0205] a. The triangular network DEM can be obtained by steps (1) to (3). The triangles are vertically connected and extended one by one along the borehole to generate GTP, and the overall description model of the geological body is established.

[0206] b. To reflect the order of strata appearance and pinch-out, the stratum numbers of the three borehole data points in the upper and lower TIN surfaces of the GTP in the borehole data are the same. This means that, assuming the three vertices of the current triangle have the same stratum number, a downward search is performed along the chain of borehole points at each vertex until the closest three points with the same stratum number are found. The TIN surface formed by these three points serves as the lower TIN surface of the current GTP and also as the upper TIN surface of the next GTP in the chain.

[0207] c. Repeat step b until all GTP constructions are completed

[0208] Since the strata in the borehole are uniformly numbered and arranged in order, there are two situations in the GTP generated by the above method: one is that the strata numbers of the upper and lower TIN surfaces differ by 1, which is a homogeneous GTP; the other is that the strata numbers of the upper and lower TIN surfaces differ by not 1, which is a heterogeneous GTP. Figure 10 shown.

[0209] Due to the complexity of the geological situation, including faults, pinchouts, lenses, and other features, the appearance of heterogeneous GTPs reflects these characteristics and requires further decomposition into smaller homogeneous GTPs or degenerate tetrahedrons. Here are three cases where this decomposition is necessary: ​​the first is when only one edge contains more than two strata; the second is when two edges contain more than two strata, and the number of strata on these two edges is the same; the third is when two edges contain more than two strata, and the number of strata on these two edges is different. The decomposition methods for these three cases are as follows:

[0210] 1) For the first case ( Figure 11 (Figure a in the figure), taking the three edges with a stratum ratio of 3:2:2 as an example:

[0211] a. Set the midpoints G, H, G1, and H1 on both sides of the corner points of the upper and lower TIN surfaces where the stratum accounts for a large proportion, that is, sides AB, BC, A1B1, and B1C1.

[0212] b. Connect HH1 and GG1, set E and F to be located at HH1 and GG1, and set their Z coordinate to be the same as D.

[0213] c. According to the Delaunay criterion, the quadrilateral AGHC can be divided into four homogeneous GTPs.

[0214] When the number of strata containing the largest number of edges continues to increase, the partitioned GTP model can be degenerated, as shown by the tetrahedron DI'FE, and when the number of strata contained in DB is greater than 2, a recursive degenerate form should be formed until the number of strata contained is equal to 2.

[0215] 2) For the second case ( Figure 11 (Figure b) Take the three edges with a stratum ratio of 3:3:2 as an example:

[0216] a. Set the midpoints G, F, G1, and F1 on both sides of the corner points of the upper and lower TIN surfaces, i.e., sides AC, BC, A1C1, and B1C1, respectively, where the stratum accounts for a small proportion.

[0217] b. Connect F1F and GG1, set E and E1 to be located at F1F and GG1, with the Z coordinates of E and D being the same, and the Z coordinates of E1 and D1 being the same.

[0218] c. According to the Delaunay criterion, the quadrilateral ABFG can be divided into 5 homogeneous GTPs.

[0219] When the number of strata on two edges with more than 2 layers continues to increase, the tetrahedral degeneration of the partitioned GTP model is required as above.

[0220] 3) For the third case ( Figure 11(Figure c in the figure), taking the three edges with a stratum ratio of 4:3:2 as an example:

[0221] a. Find the midpoints of the three edges of the upper and lower TIN surfaces, namely points G, H, I, G1, H1, and I1.

[0222] b. Connect GG1, HH1, and II1, and set D1, F1, and E1 to be located at GG1, HH1, and II1. The Z coordinates of D1 and D are the same, the Z coordinate of E1 is equal to the average of the Z coordinates of E and F, and the Z coordinate of F1 is the average of the Z coordinates of D, E, and F.

[0223] c. The quadrilateral AGHC is decomposed according to the Delaunay criterion, and the original GTP can be divided into 6 mean GTPs + 1 degenerate tetrahedron.

[0224] When the strata on the three edges differ more from each other, the redundant strata need to be tetrahedron degenerated in a recursive manner.

[0225] S7. Using a method similar to S1 and S2, an expert scoring table is designed with adjacent stratum crossing conditions as the evaluation target. A large amount of stratum information is paired with each other based on positional relationships to obtain a combination of adjacent stratum excavation conditions, which is provided to experts for benefit evaluation. Combined with a machine learning algorithm, an adjacent stratum excavation benefit evaluation model is obtained.

[0226] The preceding content successfully segments the geological body into GTP voxels and assigns attributes. Each GTP voxel now fully represents the stratigraphic information for its location, and a single GTP voxel only represents information for one type of stratigraphic element. Steps S1 and S2 describe methods for evaluating the excavation benefits of each stratum. This section, aimed at evaluating the benefits of crossing adjacent strata, employs a similar approach to steps S1 and S2 to conduct S7, evaluating the benefits of crossing adjacent strata. This section primarily describes the design of the expert survey form. The remaining content follows the same principles as steps S1 and S2.

[0227] (1) Existing form introduction

[0228] The expert attribute survey form is shown in Table 2, the stratum attribute table is shown in Table 3, and the expert weight recommended values ​​are shown in Table 4.

[0229] (2) Design of adjacent stratum crossing benefit evaluation table

[0230] The form format provided to the experts is shown in Table 5. Experts can select stratum samples from the stratum attribute table based on their personal experience, combine them in pairs, and evaluate them based on their spatial arrangement. Similarly, they can start from the three aspects of cost-effectiveness, time-effectiveness, and risk-effectiveness to finally obtain the benefit rating of crossing adjacent strata.

[0231] Table 1 Benefit evaluation table of crossing adjacent strata

[0232]

[0233] Note: The factor benefit level is positively correlated with the adjacent stratum crossing benefit score. The higher the factor benefit level, the higher the adjacent stratum crossing benefit score. The final adjacent stratum crossing benefit score = time benefit weight * time benefit level + cost benefit weight * cost benefit level + risk benefit weight * risk benefit level, and the sum of all benefit weights is equal to 1.

[0234] (3) Decision-making model for benefit evaluation of crossing adjacent strata

[0235] Using a similar approach to steps S1 and S2, the input is based on the stratum sample attributes, supplemented by adjacent stratum sample grouping. The output is also labeled with the expert benefit assessment results, resulting in a decision model for the benefit assessment of adjacent stratum crossing. This decision model yields a benefit score for adjacent stratum crossing. Because this decision model and the excavation benefit assessment decision model produce two different benefit assessment results, the weighting of the two model assessment results should be adjusted as needed during benefit calculation to meet design requirements.

[0236] S8. Import the model of existing underground structures into the three-dimensional geological model.

[0237] Before starting to design the tunnel route, S8 arranges the existing underground structures to maintain a safe distance when selecting the tunnel route. Here, the control points are directly imported to generate a rough model of the existing structures.

[0238] S9. Designers select tunnel entry and exit control points on the 3D geological model, preset the total number of path curve midpoints, set the safety control distance from existing structures, set the tunnel structure cross-sectional dimensions, and set the weight ratios for the two benefit assessments. This automatically generates multiple virtual tunnel path segments that meet tunnel line specifications and maintain control distances from existing underground structures. The optimal route is determined by comprehensively considering the sum of the benefits of each virtual tunnel structure passing through the strata and the sum of the excavation benefits of passing through adjacent strata. By selecting a path, a tunnel model formed by expanding the path segment as the central axis is generated. A semi-transparent geological model is used to visualize the tunnel model and stratum distribution.

[0239] S10. Automatically output 2D horizontal and vertical section drawings, benefit evaluation reports, and 3D BIM models, and record information such as the path through the stratum coding sequence, attributes, and benefit evaluation.

[0240] In S9, after the designer selects the tunnel entrance and exit, sets the total number of path curve points, the safety control distance of existing structures, and the tunnel structure cross-sectional dimensions, multiple paths to be evaluated can be automatically generated for subsequent benefit evaluation. The specific steps for path generation are as follows:

[0241] (1) Connecting tunnel entrances and exits;

[0242] (2) generating random arrangement of midpoints in the path according to the total number of pre-set midpoints;

[0243] (3) Generate a curve path using the midpoint of each curve as a control point. The projection of each curve path on the plane XY coordinate is a horizontal curve, and the projection on the longitudinal plane XZ is a longitudinal curve. The control elements of horizontal and longitudinal curves are as follows:

[0244] a. The control elements of a flat curve include the intersection coordinates JD, the coordinates of the straight-slope point ZH, the coordinates of the straight-slope point HZ, the radius R, the angle value, the length of the transition curve Ls, the length of the tangent line T, the outer distance, etc. These parameters are discretized into a set of line elements connected end to end to form a flat curve, such as Figure 12 shown.

[0245] b. The control factors of the vertical curve include the intersection coordinates JD, the foresight slope i1, the backsight slope i2, the curvature radius R, the tangent length T, the outer distance E, etc. These parameters are discretized into a set of end-to-end line elements to form a vertical curve, such as Figure 13 shown.

[0246] For path control, in addition to meeting the preset building spacing, it also meets the relevant design requirements of the code, such as:

[0247] (1) Longitudinal slope check: Calculate the longitudinal slope of the straight line segment. If the slope of any segment is less than 0.3% or greater than 5%, the line design is rejected.

[0248] (2) Curvature radius check: Calculate the plane curvature radius of the line curve segment. If the plane curvature radius of any segment is less than 40D (D is the outer diameter of the tunnel), the line design is rejected.

[0249] After obtaining multiple alternative routes, the excavation benefit evaluation level in the GTP unit is combined with the crossing benefit evaluation between the strata through which the route passes, and the final route benefit evaluation score is calculated using formula (9). The final route selection scheme can be obtained by comparison.

[0250]

[0251] Where: C is the total benefit evaluation value, and the number of decimal places is the same as the number of weights given by the two benefit evaluation models; α is the excavation benefit evaluation weight; X i is the excavation benefit evaluation level of element i; β is the benefit evaluation weight of adjacent strata crossing; Y i is the crossing benefit evaluation level of adjacent stratum relationship j.

[0252] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0253] It will be understood by those skilled in the art that the embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention may be implemented in various computer languages, for example, the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0254] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0255] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0256] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0257] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0258] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A three-dimensional tunnel line selection design method based on geological intelligent assessment, characterized in that: The following steps are involved: Obtain geological exploration data for the tunnel alignment section and construct a drilling information database to record stratum information; Based on the drilling information database, a pre-trained single stratum excavation benefit evaluation model is used to perform benefit evaluation, obtain an excavation benefit score for each single stratum, and record the score in the drilling information database; Based on the borehole information database, virtual borehole data is introduced, and a three-dimensional geological model is constructed with GTP voxels as units, wherein a single GTP voxel represents single stratum information; Based on the three-dimensional geological model, a pre-trained adjacent stratum excavation benefit evaluation model is used to evaluate the benefits of adjacent GTP voxels to obtain an adjacent stratum excavation benefit score; A model of existing underground structures is constructed and superimposed on the three-dimensional geological model. A comprehensive evaluation is performed based on the excavation benefit scores of each single stratum and the excavation benefit scores of adjacent strata to obtain a final three-dimensional line selection plan.

2. A three-dimensional tunnel line selection design method based on geological intelligent assessment according to claim 1, characterized in that: The geological exploration data includes spatial location data, drill holes, geological profiles, structural maps, and stratum attributes.

3. The method for three-dimensional tunnel line selection and design based on geological intelligent assessment according to claim 1, characterized in that: The borehole information database includes multiple sets, including a stratum set, a borehole set, a borehole-stratum set, and an attribute set, wherein the stratum set includes a stratum number and a stratum name. The drill hole set includes the drill hole number, plane coordinates, surface height, number of layers and overall geological characteristics of the region. The borehole-stratum set includes a borehole-stratum number, a borehole number, a stratum number, a bottom coordinate, an attribute number, and an excavation benefit score; The attribute set includes borehole-stratum number, gravity, water content, saturation, cohesion, internal friction angle, elastic modulus, Poisson's ratio, porosity, earth pressure coefficient, compression coefficient, compression modulus, standard penetration test hits, cone tip resistance, side wall friction resistance, permeability coefficient, surrounding hydrological type, water area orientation and water area clear distance.

4. The method for three-dimensional tunnel line selection and design based on geological intelligent assessment according to claim 1, characterized in that: The training steps of the single stratum excavation benefit evaluation model include: Obtain each expert's initial excavation benefit score for different single stratum samples, and use the weighted average method to calculate the final excavation benefit score for different single stratum samples; Build a single-layer neural network model; The attribute set in the drilling information database is used as input, and the final excavation benefit scores of different single stratum samples are used as output. The single-layer neural network model is trained to establish the correlation between the stratum samples and the excavation benefit scores of the single stratum samples, thereby forming a single stratum excavation benefit evaluation model. The single stratum excavation benefit evaluation model uses formula (1) to calculate the excavation benefit score of each single stratum sample, and uses formula (2) to map the excavation benefit score of each single stratum sample to the (0, 1) interval to obtain the corresponding occurrence probability: in: P1+P2+P3+P4+P5=1 0≤P1,P2,P3,P4,P5≤1 Where, and is the model output, i.e. the five stratum excavation benefit ratings, is the stratum attribute in the attribute set, w is the weight scalar, b is the bias, P i is the probability of occurrence, and e is a natural constant.

5. The method for three-dimensional tunnel line selection and design based on geological intelligent assessment according to claim 4, characterized in that: The steps of obtaining the final single stratum sample excavation benefit score include: Obtain the expert's ratings and corresponding benefit weights for time benefit, cost benefit, and risk benefit of different single stratum samples, and calculate the initial excavation benefit scores of each expert for different single stratum samples. The initial excavation benefit score of a single stratum sample = time benefit weight * time benefit grade + cost benefit weight * cost benefit grade + risk benefit weight * risk benefit grade, and the sum of all benefit weights is 1. Considering the attributes of each expert, the weight of each expert suggestion for each single stratum sample is calculated, where the calculation expression of the suggestion weight is: Where Ω is the weight of expert recommendation, α1 is the weight of the expert's unit, α2 is the weight of familiarity, α3 is the weight of professional title, and α4 is the weight of years of work experience; Based on the initial excavation benefit scores of the different single stratum samples and the weight values ​​recommended by the experts, a weighted average is performed to obtain the final excavation benefit score of the single stratum sample.

6. The method for three-dimensional tunnel line selection and design based on geological intelligent assessment according to claim 1, characterized in that: The steps of constructing the three-dimensional geological model include: Based on the drilling information database, the Kriging interpolation method is used to interpolate missing points to establish virtual drilling holes, thereby obtaining a drilling information database including virtual drilling holes, wherein the expression of the Kriging interpolation method is: where λ i Satisfies the equation: Where K * (x0) is the estimated value of any block, K(x i ) is the value of n valid samples within the influence range of the block segment, λ i is related to the variable K(x i ) is used to represent the contribution of each variable value to the valuation, γ(x i ,x j ) is the semivariogram of the two points, μ is the Lagrange multiplier; Based on the drilling information database including the virtual drilling holes, the surface of each layer is constructed by TIN grid division to form a multi-layer DEM model; Based on the multi-layer DEM model, GTP voxels corresponding to a single stratum are constructed according to the drilling points and the Delaunay triangulation to form a three-dimensional geological model.

7. The method for three-dimensional tunnel line selection and design based on geological intelligent assessment according to claim 6, characterized in that: The steps of forming a multi-layer DEM model include: 1) Projecting the discrete point data of each single stratum in the drilling information database including the virtual borehole onto a horizontal plane to obtain a point set S1; 2) Let the minimum coordinate point in point set S1 be p1, set p1 as the coordinate origin through axis-shift transformation, and obtain the local coordinate system xy and point set S2; 3) Based on the point set S2, calculate p1p n Sort the angles with the positive direction of the x-axis to get p1,…,p n , where p1 is the starting point of the convex hull boundary, p n is the vertex of the convex hull, and the calculation expression of the angle is: cotα i =x i y i The sorting operation is performed using the characteristics of the cotangent function, and the expression is: In the formula, (x i ,y i ) is point p i Coordinate, α i For p1p n Angle with the positive direction of x-axis; 4) Let p n+1 =p1; 5) Connect p i p j , where initially i=1, j=2; 6) Determine point p k With directed straight line The position relationship, where the initial k = 3, the judgment expression is: Where S is Δp i p j The area of ​​p, S is positive, negative and zero, respectively, representing the point p k On a directed straight line On the left, right or straight line, (x i ,y i )、(x j ,y j ) are points p i 、p j The coordinates of , (x, y) are the coordinates of the new point p; 7) If point p k On a directed straight line On the left side, then i←j, j←k, k←k+1, return to step 5), if point p k On a directed straight line , or on the left side of the line, delete the line segment. j←k, k←k+1, return to step 5); 8) Until j=n+1, connect p i p j , forming a convex hull; 9) Sort the points in the convex hull to obtain the sequence p1, p2, ..., p m ; 10) Take point p1 from the sequence, connect point p1 with each point on the convex hull boundary to obtain an initial triangulated network, and set initial i=1; 11) Let i = i + 1, and take out point p i , perform empty circumscribed circle detection on all initial triangulated networks, obtain the initial triangulated networks that do not meet the empty circumscribed circle criteria, and mark them. In the process of empty circumscribed circle detection, the circumscribed circle of the triangle is used to generate triangles along the circumscribed circle line under the condition of fixing one side of the triangle. The property that the inner angle of the triangle corresponding to the free moving point remains unchanged is used to judge the point p i Whether it is inside the circle, thus determining whether it meets the empty circumcircle criterion; 12) Delete the marked internal edges of the initial triangulated network to form a Delaunay hole C; 13) Move point p i Connect the vertices of the boundary of Delaunay hole C to form a new triangulated DEM; 14) Repeat steps 11)-13) until all points inside the convex hull are processed, and add the z coordinate of each point to finally form a multi-layer DEM model.

8. The method for three-dimensional tunnel line selection and design based on geological intelligent assessment according to claim 7, characterized in that: The steps of constructing the GTP body element include: The triangulated network DEM is vertically extended downward along the borehole one by one to generate GTP voxels. During the extension generation process, if the stratum codes of the three vertices of the triangulated network DEM are the same, the drill point chain of the vertex is searched downward until the three nearest points with the same stratum code are found. The three points form a TIN surface as the lower TIN surface of the current GTP voxel and the upper TIN surface of the next GTP voxel in the GTP voxel chain. This process is repeated to complete the construction process of all GTP voxels. The GTP voxels generated by expansion include the following situations: (1) If the difference between the stratum numbers of the upper and lower TIN surfaces is 1, it is a homogeneous GTP voxel; (2) If the difference between the stratum numbers of the upper and lower TIN surfaces is not 1, it is a heterogeneous GTP voxel and needs to be further decomposed to form a homogeneous GTP voxel or a degenerate tetrahedron, where the vertices of the heterogeneous GTP voxel are set to A, B, C, A1, B1, and C1. The decomposition step includes: a) The heterogeneous GTP volume has only one edge with a stratum number greater than 2: Set midpoints G, H, G1, and H1 on both sides of the upper and lower TIN corner points AB, BC, A1B1, and B1C1 respectively; Connect HH1 and GG1, set points E and F on HH1 and GG1, where the Z coordinates of points E and F are the same as the Z coordinate of point D, which is located on BB1; The quadrilateral AGHC is meshed according to the Delaunay criterion to decompose the heterogeneous GTP voxel into multiple homogeneous GTP voxels; b) The number of strata contained in two edges of the heterogeneous GTP voxel is greater than 2, and the two edges have the same number of strata: Set midpoints G, F, G1, and F1 on both sides of the upper and lower TIN corner points AC, BC, A1C1, and B1C1 of the edges with a small stratum proportion; Connect F1F and GG1, set points E and E1 on F1F and GG1, where the Z coordinate of point E is the same as the Z coordinate of point D, the Z coordinate of point E1 is the same as the Z coordinate of point D1, and point D is on BB1; The quadrilateral ABFG is meshed according to the Delaunay criterion to decompose the inhomogeneous GTP voxel into multiple homogeneous GTP voxels; c) The number of strata contained in two edges of the heterogeneous GTP voxel is greater than 2, and the two edges contain different numbers of strata: Find the midpoints of the three edges of the upper and lower TIN surfaces G, H, I, G1, H1, and I1 respectively; Connect GG1, HH1, and II1, and set points D1, F1, and E1 on GG1, HH1, and II1. The Z coordinate of point D1 is the same as that of point D, the Z coordinate of point E1 is equal to the average of the Z coordinates of points E and F, and the Z coordinate of point F1 is the average of the Z coordinates of points D, E, and F. Point D is located on BB1, and points E and F are located on CC1. The quadrilateral AGHC is meshed according to the Delaunay criterion to decompose the heterogeneous GTP voxel into multiple homogeneous GTP voxels.

9. The method for three-dimensional tunnel line selection and design based on geological intelligent assessment according to claim 1, characterized in that: The training steps of the adjacent stratum excavation benefit evaluation model include: Obtain each expert's initial adjacent stratum excavation benefit score for the adjacent stratum sample, and consider the expert's attributes to calculate the final adjacent stratum sample excavation benefit score using the weighted average method, where the initial adjacent stratum excavation benefit score = time benefit weight * time benefit level + cost benefit weight * cost benefit level + risk benefit weight * risk benefit level, and the sum of each benefit weight is equal to 1; A neural network model is constructed and trained to obtain a trained adjacent stratum excavation benefit evaluation model, wherein the input of the neural network model includes the attribute set in the drilling information database and the adjacent stratum sample grouping, and the output is the final adjacent stratum sample excavation benefit score.

10. The method for three-dimensional tunnel line selection and design based on geological intelligent assessment according to claim 1, characterized in that: The steps of obtaining the final three-dimensional line selection solution include: For a 3D geological model that includes an existing underground structure model, select tunnel entrances and exits, set the total number of path curve midpoints, and set the safety control distance of the existing structure; Connecting tunnel entrances and exits; Generate random arrangement of curve midpoints in the path according to the total number of curve midpoints set; A curve path is generated using the midpoints of each curve as control points, wherein the curve path includes a flat curve formed by projection on the plane XY and a longitudinal curve formed by projection on the longitudinal plane XZ. The generation process of the flat curve includes: Set the control elements of the horizontal curve, including the intersection point coordinates JD, the straight-slow point coordinates ZH, the smooth-straight point coordinates HZ, the radius R, the turning angle value, the transition curve length Ls, the tangent length T, and the outer distance; Discrete the control elements of the flat curve into a group of line elements connected end to end to form a flat curve; The generation process of the longitudinal curve includes: Set the control elements of the longitudinal curve, including the intersection coordinate JD, the foresight slope i1, the backsight slope i2, the curvature radius R, the tangent length T, and the outer distance E; Discrete the control elements of the longitudinal curve into a group of line elements connected end to end to form a longitudinal curve; Determine whether the curved path satisfies the following conditions in addition to the preset building spacing: ① Longitudinal slope inspection: Calculate the longitudinal slope of the straight line segment. If the slope of any segment does not meet the preset range, it is considered that it does not meet the line design requirements. Otherwise, it is considered to meet the requirements. ② Curvature radius check: Calculate the plane curvature radius of the line curve segment. If the curvature radius of any segment is less than the preset value, it is considered that the line design requirements are not met. Otherwise, it is met. For the line selection paths that meet the conditions, a comprehensive evaluation is performed based on the excavation benefit score of the single stratum corresponding to the GTP volume and the excavation benefit score of the adjacent strata to obtain the final three-dimensional line selection scheme, where the expression for the comprehensive evaluation is: Where C is the total benefit evaluation value, α is the benefit evaluation weight of single stratum excavation, X i is the excavation benefit evaluation score of a single stratum of GTP volume i, β is the excavation benefit score of the adjacent stratum, and Y i is the crossing benefit evaluation level of adjacent stratum relationship j.

Images