A calculation method for determining tunnel excavation footage based on broken rock mass joint model

Abstract

The invention discloses a calculation method for determining tunnel excavation footage based on a broken rock mass joint model, including obtaining the joint information of the broken rock mass, and constructing an orthogonal test table according to the joint information and the excavation footage; calculating the vault settlement and tunnel perimeter convergence corresponding to the orthogonal test table by a three-dimensional discrete element method, and establishing a machine learning sample; training the learning sample by a machine learning algorithm, and establishing a regression model; optimizing the regression model by a cuckoo intelligent optimization algorithm; obtaining the joint information of the stratum area to be excavated, and establishing a broken rock mass engineering joint distribution calculation table and a distribution model; calculating the corresponding vault settlement and tunnel perimeter convergence values ​​by a regression model according to the distribution calculation table and the excavation footage calculation table; and outputting a cyclic excavation footage value when the result is less than the limit value of the vault settlement and tunnel perimeter convergence in the project. The tunnel excavation footage value can be determined quickly, and the accuracy of the calculation result is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of machine learning and image recognition, and in particular to a calculation method for determining tunnel excavation footage based on a broken rock mass joint model. Background Art

[0002] During tunnel construction, the stability of the project must be ensured first. The excavation progress of the tunnel is a key factor in determining the stability of the project. The mechanical characteristics of the tunnel surrounding rock, especially the joint information of the broken rock mass, are important factors affecting the excavation progress.

[0003] Due to the influence of geological movement, blasting, freeze-thaw cycles and other factors, joints and fissures are developed and widely exist in tunnel rocks, and these joints are often discontinuous. The variability of their parameters such as trace length, density, and inclination angle leads to a high degree of uncertainty in the stability state of the tunnel surrounding rock system. Although the traditional pervasive joint model takes into account the influence of joint factors, its joints are through and linear and cannot reflect the overall state of the surrounding rock. When using traditional numerical calculation software to calculate the tunnel excavation advance under the pervasive joint model, it is necessary to first establish a numerical model of the tunnel, then input the mechanical parameters of the surrounding rock, and finally perform numerical calculations manually. The time period is long and the result error of a single calculation is large, which has certain limitations. Summary of the invention

[0004] The present invention provides a calculation method for determining tunnel excavation footage based on a broken rock mass joint model to overcome the above technical problems.

[0005] A method for calculating tunnel excavation footage based on a broken rock mass joint model comprises the following steps:

[0006] S1. Determine the joint information of the broken rock mass using an image recognition method and advanced geological prediction, and construct an orthogonal test table according to the joint information and excavation footage, wherein the joint information includes the maximum edge of the broken rock mass, the minimum edge of the broken rock mass, the joint cohesion, and the joint friction angle;

[0007] S2. A broken rock mass model is established according to the joint information of the broken rock mass. The established broken rock mass model is calculated by a three-dimensional discrete element method in combination with different excavation footages to obtain the corresponding final vault settlement and tunnel perimeter convergence, and a set of machine learning samples is established. The attributes of the learning samples include joint information, excavation footage, vault settlement, and tunnel perimeter convergence.

[0008] S3. Use machine learning algorithms to train the learning sample set and establish a regression model of the nonlinear mapping relationship between the learning samples;

[0009] S4, optimizing the regression model by using the cuckoo intelligent optimization algorithm;

[0010] S5. Sampling the stratum area to be excavated to obtain the broken rock mass joint information, and establishing a calculation table of the broken rock mass engineering joint distribution of the corresponding project;

[0011] S6. Establish a cyclic excavation footage calculation table based on engineering experience, randomly extract excavation footage values ​​from the cyclic excavation footage calculation table based on the fractured rock mass engineering joint distribution calculation table, and calculate the corresponding vault settlement and tunnel perimeter convergence values ​​through the regression model;

[0012] S7, compare the arch settlement and hole perimeter convergence values ​​with the limit values ​​of the arch settlement and hole perimeter convergence in the project, if they are less than the limit values, go to S8, if they are greater than the limit values, go to S6;

[0013] S8. Output the excavation footage value as the current cycle excavation footage.

[0014] Furthermore, the broken rock mass model established according to the joint information of the broken rock mass in step 2 includes the following steps:

[0015] S21, setting a suitable discrete point spatial density according to the maximum edge of the broken rock mass and the minimum edge of the broken rock mass, and randomly forming discrete points;

[0016] S22, using the discrete points as vertices, the maximum edge and minimum edge of the broken rock mass as the maximum edge and minimum edge of the triangle as the upper and lower limits of the side length, constructing a Delaunay triangle while ensuring that each triangle does not contain an independent discrete point pair, and forming a spatial triangulation network, starting from the origin of the broken rock mass model, numbering each discrete point and triangle in a clockwise order from near to far, and from bottom to top;

[0017] S23, finding all triangles with the discrete point as a common vertex in the triangular network in sequence according to the discrete point number, and calculating the center of the circumscribed circle of each triangle;

[0018] S24, connecting the centers of the circumscribed circles of adjacent triangles with the discrete points as common vertices in sequence according to the discrete point numbering sequence to obtain a spatial Thiessen polygon;

[0019] S25, delete the spatial triangulation network and delete all spatial discrete points;

[0020] S26. Set the constitutive model of the spatial Thiessen polygons to the Mohr-Coulomb constitutive model, import the obtained joint cohesion, joint friction angle and related basic mechanical parameters of the rock into the model, and form a broken rock model that can be calculated using the three-dimensional discrete element method.

[0021] Furthermore, in step 3, establishing a regression model for the nonlinear mapping relationship between learning samples includes the following steps:

[0022] S31, performing normalization processing on all samples in the learning sample set, specifically, selecting any sample in the learning sample set, and performing normalization processing on all attributes of the sample by formula (1) to obtain a sample set;

[0023]

[0024] Among them, X j_max represents the maximum value of the mth attribute of the sample in the learning sample set; X j_min Then it represents the minimum value of the mth attribute of the sample in the learning sample set, X ij represents the value of the jth attribute of the i-th sample, 0<i<n, n is the number of samples in the sample set, 0<j<m, m is the number of sample attributes, X ij =(X i1 , X i2 , …X im );

[0025] S32, by M = n / (h × ω) the sample set X ij The sample X obtained after convolution pooling is X=(x1, x2, ..., x M ), where h×ω is the convolution kernel size, and the convolution kernel contains m feature convolution pooling operations of n samples, where the pooling method is g(x)=max(x k ), 1≤k≤h×ω;

[0026] S33, sample X is convolved and transformed through formula (2) to obtain a regression model with nonlinear mapping relationship Obtain a regression model with nonlinear mapping relationship The formula is;

[0027]

[0028] in, ω ji and B l They represent the convolution kernel and bias corresponding to the neuron i in layer l and the jth input, respectively. The predicted output of the kth node is y k , the actual output is d k , the error term is δ k , then the error term is δ k =(d k -y k )y k (1-y k ), then the error function of all nodes trained using the machine learning algorithm is

[0029] Further, in S4, the regression model is optimized by the cuckoo intelligent optimization algorithm, specifically, the node error function in the machine learning algorithm is optimized, and the steps of optimizing the node error function in the machine learning algorithm are:

[0030] S41. Use As the error function in machine learning, the parameters γ and σ representing the bird's nest in the cuckoo algorithm are determined empirically. 2 The range of values, where γ is the regularization parameter indicating the degree of interval fitting, σ 2 is the square bandwidth, where k is the bird's nest parameter;

[0031] S42, setting the initial exclusion probability parameter P a , P a = [0, 1], the location of the bird's nest is randomly generated and used express, For the location of the bird's nest, They are respectively composed of the maximum edge of the broken rock mass, the minimum edge of the broken rock mass, the cohesion of the joints, the friction angle of the joints, the excavation footage, the settlement of the arch, and the convergence of the hole. Each bird's nest corresponds to a parameter k. The fitting degree of each group of bird's nest positions corresponding to the training set is calculated to find the current optimal bird's nest position. And the optimal fit F max , the optimal fit F max The optimal nest location is the one with the largest fitting value. The best fit F max The corresponding nest location;

[0032] S43, retain the best bird's nest position of the previous generation And according to formula (3) and formula (4), calculate the Levy flight step length;

[0033] step i =step min +(step max -step min )d i (3)

[0034]

[0035] Among them, d i is the flight step length of the i-th generation, n i is the nest position of the i-th generation, n best is the optimal nest position before the i-th generation, d max is the maximum step length, step min is the minimum step size, step maxis the maximum step length, N is the number of iterations, and the positions of other bird nests are updated by the step length to obtain a new set of bird nest positions And calculate their fitness F, compare the position of the new bird's nest with the fitness F of the position of the previous generation of bird's nest, if the fitness of the new bird's nest is higher than the fitness of the previous generation of bird's nest, then replace the position of the previous generation of bird's nest with the new bird's nest;

[0036] S44, find the optimal bird's nest position in step S43 The corresponding fitting degree F is used to determine whether the fitting degree of the current optimal bird's nest position is greater than the set required value. If it is greater, the k value is output; if it does not meet the requirements, return to S42 to continue optimizing.

[0037] Furthermore, the specific steps of establishing the broken rock mass engineering joint distribution model of the corresponding project according to the broken rock mass engineering joint distribution calculation table in S5 are as follows:

[0038] S51, establish an excavation footage calculation table with 0.20m as the minimum value, 2.0m as the maximum value, and 0.20m as the incremental value;

[0039] S52, starting from the entrance of the engineering tunnel, based on the advanced geological forecast and the address exploration map data, setting a section every 0.20m, constructing a set of broken rock and soil joint data according to the sections, and establishing a calculation table for the distribution of broken rock engineering joints of the corresponding project;

[0040] S53. Using the joint data of the current section as the joint data between the current section and the adjacent section, a fractured rock mass engineering joint distribution model of the corresponding project is constructed.

[0041] The present invention provides a calculation method for determining the tunnel excavation footage based on a broken rock mass joint model, so as to realize the rapid determination of the tunnel cyclic excavation footage of the broken rock mass joint parameters. By using a machine learning algorithm and optimizing it, the accuracy of the calculation result is improved, and the errors caused by previous manual subjective evaluation or numerical simulation calculation are avoided. An accurate and safe evaluation method is provided for determining the tunnel cyclic excavation footage. At the same time, spatial Thiessen polygons are used to transform the joints into a broken line type, which can better reflect the irregular fragmentation form of the broken rock mass. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0043] Figure 1is a flow chart of the method of the present invention;

[0044] Figure 2 It is a flow logic diagram of the present invention;

[0045] Figure 3 is a logic flow chart of the present invention using an algorithm to optimize a machine learning model;

[0046] Figure 4 It is the principle diagram of generating Thiessen polygon unit of the present invention;

[0047] Figure 5 is a numerical calculation model of an embodiment of the present invention;

[0048] Figure 6 It is a vertical displacement cloud diagram numerically calculated in part of the embodiment of the present invention;

[0049] Figure 7 It is a vertical displacement diagram of a part of a tunnel in an embodiment of the present invention;

[0050] Figure 8 This is a tunnel displacement measurement point diagram according to an embodiment of the present invention. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0052] Figure 1 The method of the present invention is a flow chart. The method of this embodiment may include:

[0053] A calculation method for determining tunnel excavation footage based on a fractured rock mass joint model, such as Figure 1-3 As shown, the following steps are included:

[0054] S1. Determine the joint information of the broken rock mass using an image recognition method and advanced geological prediction, and construct an orthogonal test table according to the joint information and excavation footage, wherein the joint information includes the maximum edge of the broken rock mass, the minimum edge of the broken rock mass, the joint cohesion, and the joint friction angle;

[0055] S2. A broken rock mass model is established according to the joint information of the broken rock mass. The established broken rock mass model is calculated by a three-dimensional discrete element method in combination with different excavation footages to obtain the corresponding final vault settlement and tunnel perimeter convergence, and a set of machine learning samples is established. The attributes of the learning samples include joint information, excavation footage, vault settlement, and tunnel perimeter convergence.

[0056] S3. Use machine learning algorithms to train the learning sample set and establish a regression model of the nonlinear mapping relationship between the learning samples;

[0057] S4, optimizing the regression model by using the cuckoo intelligent optimization algorithm;

[0058] S5. Sampling the stratum area to be excavated to obtain the broken rock mass joint information, and establishing a calculation table of the broken rock mass engineering joint distribution of the corresponding project;

[0059] S6. Establish a cyclic excavation footage calculation table based on engineering experience, randomly extract excavation footage values ​​from the cyclic excavation footage calculation table based on the fractured rock mass engineering joint distribution calculation table, and calculate the corresponding vault settlement and tunnel perimeter convergence values ​​through the regression model;

[0060] S7, compare the arch settlement and hole perimeter convergence values ​​with the limit values ​​of the arch settlement and hole perimeter convergence in the project, if they are less than the limit values, go to S8, if they are greater than the limit values, go to S6;

[0061] S8. Output the excavation footage value as the current cycle excavation footage.

[0062] For S1, the joint information of the broken rock mass is determined by using image recognition methods and advanced geological prediction, and the joint information includes the maximum edge of the broken rock mass, the minimum edge of the broken rock mass, the joint cohesion and the joint friction angle, and an orthogonal experimental table or orthogonal test table is constructed according to the joint information in combination with different excavation footages. Specifically, an orthogonal scheme is designed for the joint information of the broken rock mass to form a basic group of calculation data. The basis for determining the range of parameter values ​​during the orthogonal test is the engineering survey results in the actual application case. In order to further clearly demonstrate the implementation and application process of the method of the present invention, the detailed application process of the method of the present invention is described with the Daya section in Changchun City, Jilin Province as the engineering support.

[0063] The joint parameters of the broken rock mass in the project were sampled and counted, and the distribution range of the joint parameters in this section was obtained as shown in Table 1:

[0064]

[0065]

[0066] Table 1 Distribution range of joint parameters

[0067] According to the survey results, an orthogonal experimental table of geometric parameters of broken rock joints was constructed, and a five-level design was carried out on the five parameters of the maximum edge of the broken rock mass, the minimum edge of the broken rock mass, the joint cohesion, the joint friction angle and the excavation footage, forming a total of 25 groups of orthogonal samples. The orthogonal experimental table of joint geometric parameters is shown in Table 2.

[0068]

[0069]

[0070] Table 2 Orthogonal test table of joint geometric parameters

[0071] S2. Calculate the joint information of the broken rock mass and different excavation footages by three-dimensional discrete element method software to obtain the corresponding final tunnel vault settlement and tunnel perimeter convergence, and establish a machine learning sample set, wherein the attributes of the learning sample include joint information, excavation footage, vault settlement and tunnel perimeter convergence;

[0072] Figure 4 A schematic diagram is generated for the Thiessen polygon unit of the present invention, such as Figure 4 Shown

[0073] Step 1: According to the maximum edge and the minimum edge of the broken rock mass, a certain spatial density is set, and the system randomly forms discrete points (serial number ①).

[0074] Step 2: Using discrete points as vertices, construct Delaunay triangles while ensuring that each triangle does not contain independent discrete point pairs, and form a spatial triangulation network (serial number ②).

[0075] Step 3: Connect the centers of the circumscribed circles of adjacent triangles with common vertices to obtain the spatial Thiessen polygon (serial number ③).

[0076] Step 4: Delete the spatial triangulation network (serial number ④).

[0077] Step 5: Remove all spatial discretization (serial number ⑤).

[0078] The simulation of rock joints is done by using a fragmented rock mass unit, which is a set of continuous polygons consisting of perpendicular bisectors connecting two adjacent line segments. The characteristics of a fragmented rock mass are:

[0079] 1. Each fractured rock mass contains only one discrete point data.

[0080] 2. The distance between the point in the fractured rock mass and the corresponding discrete point is the shortest.

[0081] 3. The distances from a point on the edge of the fractured rock mass to the discrete points on both sides are equal.

[0082] By establishing a relevant model through jointed and fractured rock mass division in the three-dimensional discrete element method software, the parameters of each orthogonal scheme are input into the three-dimensional discrete element method to calculate the corresponding arch settlement and cave circumference convergence of each group, and the orthogonal schemes and results are sorted out to establish a learning sample.

[0083] The jointed and fractured rock mass division refers to replacing the existing joints in the rock with the broken lines of the fractured rock mass. The development degree of the joints can be solved by setting the maximum edge of the fractured rock mass, the minimum edge of the fractured rock mass, the joint cohesion, and the joint friction angle. A three-dimensional discrete element method numerical calculation model is established based on the on-site survey drawings and tunnel design drawings.

[0084] exist Figure 4 On the basis of the above, the soil around the tunnel with an area of ​​4.5 times the diameter is retained, the bottom of the model is fixed, and the boundaries on both sides are normal constrained. The positions and sizes of each structure are as follows: Figure 5 As shown, the tunnel section is established according to the tunnel design drawings, and the surrounding rock and soil are established according to the on-site investigation drawings.

[0085] The three-dimensional model is divided into fractured rock mass units. In order to statistically analyze the deformation law, three displacement monitoring points are arranged in the tunnel monitoring section. The position and coordinate information of each point are as follows: Figure 8 shown. Figure 8 A, B, and C are displacement measuring points, A is the displacement measuring point of the arch crown, and BC is the displacement measuring point of the arch waist. Figure 7 The curve is the data recorded at the arch displacement measuring point.

[0086] The mechanical parameters of rock mass and joints were obtained through laboratory tests on specimens made on site. The results are shown in Table 3: Table of numerical mechanical parameters:

[0087]

[0088] Table 3 Numerical calculation mechanical parameters

[0089] The processing flow of step S3 and step S4 is as follows: Figure 6 shown.

[0090] S3. Using a machine learning algorithm to train the learning samples, and establishing a regression model of the nonlinear mapping relationship between the learning samples, wherein the regression model is a machine learning neural network, including:

[0091] S31: performing normalization processing on all samples in the learning sample set, specifically, selecting any sample in the learning sample set, and performing normalization processing on all attributes of the sample by formula (1) to obtain a sample set;

[0092]

[0093] Among them, X j_max represents the maximum value of the mth attribute of the sample in the learning sample set; X j_min Then it represents the minimum value of the mth attribute of the sample in the learning sample set, X ijDenotes the value of the j-th attribute of the i-th sample, where 0 < i < n and n is the number of samples in the sample set, and 0 < j < m and m is the number of sample attributes, X ij =(X i1 , X i2 , … X im ); The schematic situation of the specific input sample working conditions is shown in Table 4;

[0094] Input Sample Condition 1 Condition 2 Condition 3 ...... Working condition <![CDATA[The maximum edge X1 of the fractured rock mass]]> <![CDATA[X 11 ]]> <![CDATA[X 12 ]]> <![CDATA[X 13 ]]> ...... <![CDATA[X 1n ]]> <![CDATA[Minimum edge X2 of fractured rock mass]]> <![CDATA[X 21 ]]> <![CDATA[X 22 ]]> <![CDATA[X 23 ]]> ...... <![CDATA[X 2n ]]> <![CDATA[Joint cohesion X3]]> <![CDATA[X 31 ]]> <![CDATA[X 32 ]]> <![CDATA[X 33 ]]> ...... <![CDATA[X 3n ]]> <![CDATA[Joint friction angle X4]]> <![CDATA[X 41 ]]> <![CDATA[X 42 ]]> <![CDATA[X 43 ]]> ...... <![CDATA[X 4n ]]> Excavation footage X5 <![CDATA[X 51 ]]> <![CDATA[X 52 ]]> <![CDATA[X 53 ]]> ...... <![CDATA[X 5n ]]> Vault Settlement X6 <![CDATA[X 61 ]]> <![CDATA[X 62 ]]> <![CDATA[X 63 ]]> ...... <![CDATA[X 6n ]]> Convergence around the hole X7 <![CDATA[X 71 ]]> <![CDATA[X 72 ]]> <![CDATA[X 73 ]]> ...... <![CDATA[X 7n ]]>

[0095] Table 4 Schematic Table of Input Sample Working Conditions

[0096] S32: Through M = n / (h × ω), the sample set X ij After convolution and pooling, the newly obtained sample X, X = (x1, x2, …, x M ), where h × ω is the convolution kernel size, and the convolution kernel contains m feature convolution pooling operations of n samples. The pooling method is g(x) = max(x k ), 1 ≤ k ≤ h × ω;

[0097] S33: The sample X undergoes convolution and conversion operations through formula (2) to obtain a regression model of the non-linear mapping relationship

[0098]

[0099] Among them, ω ji and B l respectively represent the convolution kernel and bias corresponding to the i-th neuron of the l-th layer and the j-th input. The predicted output of the k-th node is y k , the actual output is d k , and the error term is δ k , then the error term is δ k =(d k -y k )y k (1 - y k ), then the error function of all nodes trained using the machine learning algorithm is

[0100] S4. Optimize the regression model through the cuckoo intelligent optimization algorithm, specifically referring to using the fitness F function to replace the original machine learning algorithm to form an optimized machine learning algorithm, and perform parameter setting according to the optimal parameter k corresponding to the optimal fitness F max , including:

[0101] S41. Use As the error function in machine learning, the parameters γ and σ representing the bird's nest in the cuckoo algorithm are determined empirically. 2 The range of values, where γ is the regularization parameter indicating the degree of interval fitting, σ 2 is the square bandwidth;

[0102] S42, setting the initial exclusion probability parameter P a , P a = [0,1], the location of the bird's nest is randomly generated and used express, For the location of the bird's nest, They are respectively composed of the maximum edge of the broken rock mass, the minimum edge of the broken rock mass, the cohesion of the joints, the friction angle of the joints, the excavation footage, the settlement of the arch, and the convergence of the hole. Each bird's nest corresponds to a parameter k. The fitting degree of each group of bird's nest positions corresponding to the training set is calculated to find the current optimal bird's nest position. And the optimal fit F max , the optimal fit F max The optimal nest location is the one with the largest fitting value. The best fit F max The corresponding nest location;

[0103] Keep the best bird's nest location of the previous generation And according to formula (3) and formula (4), calculate the Levy flight step length;

[0104] step i =step min +(step max -step min )d i (3)

[0105]

[0106] Among them, d i is the flight step length of the i-th generation, n i is the nest position of the i-th generation, n best is the optimal nest position before the i-th generation, d max is the maximum step length, step min is the minimum step size, step max is the maximum step length, N is the number of iterations, and the positions of other bird nests are updated by the step length to obtain a new set of bird nest positions And calculate their fitness F, compare the position of the new bird's nest with the fitness F of the position of the previous generation of bird's nest, if the fitness of the new bird's nest is higher than the fitness of the previous generation of bird's nest, then replace the position of the previous generation of bird's nest with the new bird's nest;

[0107] S44, find the optimal bird's nest position in step S43 The corresponding fitting degree F is used to determine whether the current optimal nest position fitting degree is greater than the set required value. If it is greater, the k value is output; if it does not meet the requirements, it returns to S42 to continue to optimize. The Thiessen polygon unit generated by the optimization in S4 is as follows Figure 4 As shown;

[0108] S5. Sampling the stratum area to be excavated to obtain broken rock mass joint information, establishing a broken rock mass engineering joint distribution calculation table for the corresponding project, and establishing a broken rock mass engineering joint distribution model for the corresponding project based on the broken rock mass engineering joint distribution calculation table, including:

[0109] S51: With 0.20m as the minimum value, 2.0m as the maximum value, and 0.20m as the increment value, a calculation table for excavation footage is established. There are 10 excavation footages in the table;

[0110] S52: starting from the entrance of the engineering tunnel, based on the advanced geological forecast and the address exploration map data, a section is set every 0.20m, a set of broken rock and soil joint data is constructed according to the section, and a calculation table of broken rock engineering joint distribution of the corresponding engineering is established;

[0111] S53: Using the joint data of the current section as the joint data between the current section and the adjacent section, a fractured rock mass engineering joint distribution model of the corresponding project is constructed.

[0112] S6. Establish a cyclic excavation footage calculation table based on engineering experience, and calculate corresponding vault settlement and tunnel perimeter convergence values ​​through a regression model based on the fractured rock mass engineering joint distribution calculation table and randomly extracting excavation footage values ​​from the cyclic excavation footage calculation table;

[0113] S7, compare the arch settlement and hole perimeter convergence values ​​with the limit values ​​of the arch settlement and hole perimeter convergence in the project, if they are less than the limit values, go to S8, if they are greater than the limit values, go to S6;

[0114] S8. Output the excavation footage value as the current cycle excavation footage.

[0115] The overall beneficial effect is that the present invention realizes the rapid determination of tunnel excavation footage of broken rock mass joint parameters as a whole. Through the machine learning algorithm, the accuracy of the calculation results is improved, the errors in previous manual subjective evaluation or numerical simulation calculation are avoided, and an accurate and safe evaluation method is provided for the determination of tunnel cycle excavation footage.

[0116] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A calculation method for determining tunnel excavation footage based on a broken rock mass joint model. It is characterized in that The following steps are involved: S1. Determine the joint information of the broken rock mass using an image recognition method and advanced geological prediction, and construct an orthogonal test table according to the joint information and excavation footage, wherein the joint information includes the maximum edge of the broken rock mass, the minimum edge of the broken rock mass, the joint cohesion, and the joint friction angle; S2. A broken rock mass model is established according to the joint information of the broken rock mass. The established broken rock mass model is calculated by a three-dimensional discrete element method in combination with different excavation footages to obtain the corresponding final vault settlement and tunnel perimeter convergence, and a set of machine learning samples is established. The attributes of the learning samples include joint information, excavation footage, vault settlement, and tunnel perimeter convergence. S3. Use machine learning algorithms to train the learning sample set and establish a regression model of the nonlinear mapping relationship between the learning samples; S4, optimizing the regression model by using the cuckoo intelligent optimization algorithm; S5. Sampling the stratum area to be excavated to obtain the broken rock mass joint information, and establishing a calculation table of the broken rock mass engineering joint distribution of the corresponding project; S6. Establish a cyclic excavation footage calculation table based on engineering experience, randomly extract excavation footage values ​​from the cyclic excavation footage calculation table based on the fractured rock mass engineering joint distribution calculation table, and calculate the corresponding vault settlement and tunnel perimeter convergence values ​​through the regression model; S7, compare the arch settlement and hole perimeter convergence values ​​with the limit values ​​of the arch settlement and hole perimeter convergence in the project, if they are less than the limit values, go to S8, if they are greater than the limit values, go to S6; S8, outputting the excavation footage value as the current cycle excavation footage; The broken rock mass model established according to the joint information of the broken rock mass in step S2 comprises the following steps: S21, setting a suitable discrete point spatial density according to the maximum edge of the broken rock mass and the minimum edge of the broken rock mass, and randomly forming discrete points; S22, using the discrete points as vertices, the maximum edge and minimum edge of the broken rock mass as the maximum edge and minimum edge of the triangle as the upper and lower limits of the side length, constructing a Delaunay triangle while ensuring that each triangle does not contain an independent discrete point pair, and forming a spatial triangulation network, starting from the origin of the broken rock mass model, numbering each discrete point and triangle in a clockwise order from near to far, and from bottom to top; S23, finding all triangles with the discrete point as a common vertex in the triangular network in sequence according to the discrete point number, and calculating the center of the circumscribed circle of each triangle; S24, connecting the centers of the circumscribed circles of adjacent triangles with the discrete points as common vertices in sequence according to the discrete point numbering sequence to obtain a spatial Thiessen polygon; S25, delete the spatial triangulation network and delete all spatial discrete points; S26. Set the constitutive model of the spatial Thiessen polygons to the Mohr-Coulomb constitutive model, import the obtained joint cohesion, joint friction angle and related basic mechanical parameters of the rock into the model, and form a broken rock model that can be calculated using the three-dimensional discrete element method.

2. A method for calculating tunnel excavation footage based on a broken rock mass joint model according to claim 1, It is characterized in that The step S3 of establishing a regression model for the nonlinear mapping relationship between learning samples includes the following steps: S31, performing normalization processing on all samples in the learning sample set, specifically, selecting any sample in the learning sample set, and performing normalization processing on all attributes of the sample by formula (1) to obtain a sample set; Among them, X j_max represents the maximum value of the m-th attribute of the sample in the learning sample set; X j_min represents the minimum value of the m-th attribute of the sample in the learning sample set, X ij represents the value of the j-th attribute of the i-th sample, where 0 < i < n and n is the number of samples in the sample set, and 0 < j < m and m is the number of sample attributes, X ij =(X i1 , X i2 ,... X im ); S32, by M = n / (h × ω) the sample set X ij The sample X obtained after convolution pooling, X = (x 1 ,x 2 ,…,x M ), where h×ω is the convolution kernel size, and the convolution kernel contains m feature convolution pooling operations of n samples, where the pooling method is g(x)=max(x k ), 1≤k≤h×ω; S33, sample X is convolved and transformed by formula (2) to obtain the nonlinear mapping relationship. Normalized Model x j , and obtain the regression model x of the nonlinear mapping relationship j The formula is; in, ω ji and B l They represent the convolution kernel and bias corresponding to the neuron i in layer l and the jth input, respectively. The predicted output of the kth node is y k , the actual output is d k , the error term is δ k , then the error term is δ k =(d k -y k )y k (1-y k ), then the error function of all nodes trained using the machine learning algorithm is 3. A method for calculating tunnel excavation footage based on a broken rock mass joint model according to claim 2, It is characterized in that In S4, the regression model is optimized by the cuckoo intelligent optimization algorithm, specifically, the node error function in the machine learning algorithm is optimized. The steps of optimizing the node error function in the machine learning algorithm are: S41. Use As the error function in machine learning, the parameters γ and σ representing the bird's nest in the cuckoo algorithm are determined empirically. 2 The range of values, where γ is the regularization parameter indicating the degree of interval fitting, σ 2 is the square bandwidth, where k is the bird's nest parameter; S42, setting the initial exclusion probability parameter P a , P a = [0,1]), the location of the bird's nest is randomly generated and express, For the location of the bird's nest, They are respectively composed of the maximum edge of the broken rock mass, the minimum edge of the broken rock mass, the cohesion of the joints, the friction angle of the joints, the excavation footage, the settlement of the arch, and the convergence of the hole. Each bird's nest corresponds to a parameter k. The fitting degree of each group of bird's nest positions corresponding to the training set is calculated to find the current optimal bird's nest position. And the optimal fit F max , the optimal fit F max The optimal nest location is the one with the largest fitting value. The best fit F max The corresponding nest location; S43, retain the best bird's nest position of the previous generation And according to formula (3) and formula (4), calculate the Levy flight step length; step i =step min +(step max -step min )d i (3) Among them, d i is the flight step length of the i-th generation, n i is the nest position of the i-th generation, n best is the optimal nest position before the i-th generation, d max is the maximum step length, step min is the minimum step size, step max is the maximum step length, N is the number of iterations, and the positions of other bird nests are updated by the step length to obtain a new set of bird nest positions And calculate their fitness F, compare the position of the new bird's nest with the fitness F of the position of the previous generation of bird's nest, if the fitness of the new bird's nest is higher than the fitness of the previous generation of bird's nest, then replace the position of the previous generation of bird's nest with the new bird's nest; S44, find the optimal bird's nest position in step S43 The corresponding fitting degree F is used to determine whether the fitting degree of the current optimal bird's nest position is greater than the set required value. If it is greater, the k value is output; if it does not meet the requirements, return to S42 to continue optimizing.

4. A method for calculating tunnel excavation footage based on a broken rock mass joint model according to claim 3, It is characterized in that The specific steps of establishing the broken rock mass engineering joint distribution model of the corresponding project according to the broken rock mass engineering joint distribution calculation table described in S5 are: S51, establish an excavation footage calculation table with 0.20m as the minimum value, 2.0m as the maximum value, and 0.20m as the incremental value; S52, starting from the entrance of the engineering tunnel, based on the advanced geological forecast and the address exploration map data, setting a section every 0.20m, constructing a set of broken rock and soil joint data according to the sections, and establishing a calculation table for the distribution of broken rock engineering joints of the corresponding project; S53. Using the joint data of the current section as the joint data between the current section and the adjacent section, a fractured rock mass engineering joint distribution model of the corresponding project is constructed.

Images