CN112765809B

CN112765809B - Railway line comparing and selecting method and device based on typical disasters of high-ground stress tunnel

Info

Publication number: CN112765809B
Application number: CN202110058285.4A
Authority: CN
Inventors: 李天斌; 张广泽; 马俊杰; 张雨露; 曾鹏; 丁浩江; 冯君; 张小林
Original assignee: Chengdu Univeristy of Technology
Current assignee: Chengdu Univeristy of Technology
Priority date: 2021-01-14
Filing date: 2021-01-14
Publication date: 2022-11-11
Anticipated expiration: 2041-01-14
Also published as: CN112765809A

Abstract

The embodiment of the invention provides a railway line comparison and selection method and device based on typical disasters of a high ground stress tunnel, which are applied to the technical field of geological route selection research.

Description

Railway line comparison and selection method and device based on typical disasters of high-ground-stress tunnel

Technical Field

The invention relates to the technical field of geological route selection research, in particular to a railway route comparison and selection method and device based on high ground stress tunnel typical disasters.

Background

In recent years, with the more and more planning and construction of plateau complex mountain railway engineering, such as Sichuan-Tibet railway, dian-Tibet railway, newly-Tibet railway, gannan-Tibet railway and the like, the traditional technical and economic route selection method cannot meet the requirements of actual route selection work, so that geological route selection plays an increasingly important role.

In the prior art, some schemes carry out major earthquake risk analysis on a Chuanzang railway boil-Ganzui section based on the minimum As reasonability practical (ALARP) criterion, and carry out railway line scheme comparison and selection from macroscopic and microscopic aspects by combining an improved scheme evaluation method, individual engineering arrangement measures and the like. Some schemes also establish a risk quantitative analysis and evaluation model of the landslide hazard by adopting a contribution rate method, so that the acceptable risk level of the landslide hazard along the Chuanghai-Tibet railway Kangding-Linzhi section is quantitatively analyzed, and reference is provided for railway route selection of a research section. In some schemes, a three-stage underground river identification method is provided by researching a tunnel case which passes through an underground river development area and generates sudden water gushing, and a tunnel sudden water gushing risk evaluation model is established by adopting a hierarchical analysis-expert grading method, so that reference is provided for tunnel route selection of the underground river development area.

In conclusion, the geological route selection research content is rich at present, and reference is provided for geological route selection of railways and highways. However, at present, for typical tunnel geological disasters in a high ground stress environment, particularly railway route selection research on rock burst and large deformation is less, the existing geological route selection method still relies on more artificial evaluation, and the artificial evaluation relies on more massive historical data, so that how to carry out quantitative risk prediction on the rock burst and the large deformation in the high ground stress environment based on limited data in the early route selection stage of railway engineering is a problem to be solved urgently.

Disclosure of Invention

The embodiment of the invention provides a railway line comparison and selection method and device based on typical disasters of a high-ground stress tunnel, which adopts a naive Bayes machine learning method to predict the probability of the high-ground stress typical tunnel disasters of the to-be-compared selected railway tunnel in a high-ground stress environment according to geological factors of a tunnel engineering section in the to-be-compared selected railway line in the high-ground stress area, and then adopts a product measurement mode to quantify the risk value of the railway line disasters according to the probability of the high-ground stress typical tunnel disasters of the to-be-compared selected railway tunnel in the high-ground stress area.

The first aspect of the embodiments of the present application provides a railway line comparison and selection method based on a high ground stress tunnel typical disaster, where the method includes:

counting specific tunnels in the to-be-compared route selection way to obtain tunnel engineering sections, wherein the tunnel engineering sections are in one-to-one correspondence with the to-be-compared route selection way; the specific tunnel is a tunnel located in a high ground stress environment;

acquiring first rock strength, first ground stress, a first geological structure and a first surrounding rock level of the tunnel engineering section;

inputting the first rock strength, the first ground stress, the first geological structure and the first surrounding rock grade into a preset rock burst probability prediction model to obtain a first posterior probability of rock burst of the tunnel engineering section;

inputting the first rock strength, the first ground stress, the first geological structure and the first surrounding rock level into a preset large deformation probability prediction model to obtain a second posterior probability of large deformation of the tunnel engineering section;

calculating a comprehensive risk length value of the tunnel engineering section according to the first posterior probability and the second posterior probability;

and determining the tunnel engineering section with the minimum comprehensive risk length value as a target tunnel engineering section, and taking the candidate railway route corresponding to the target tunnel engineering section as an optimal railway route.

Optionally, the method further comprises:

obtaining a plurality of sampling positions, and collecting a first feature vector of each sampling position; the first feature vector includes: a second rock strength, a second geostress, a second geological formation, and a second country rock class;

according to the actual rock burst grade of the current sampling site, sequentially marking first marking information on the first eigenvector of each sampling site to obtain second eigenvectors carrying the first marking information, wherein the second eigenvectors correspond to the sampling sites one by one;

according to the actual large deformation level of the current sampling site, sequentially marking second marking information on the first eigenvector of each sampling site to obtain third eigenvectors carrying the second marking information, wherein the third eigenvectors correspond to the sampling sites one by one;

aiming at a plurality of sampling places, respectively training a naive Bayesian classification model by using the second feature vectors which are in one-to-one correspondence with the sampling places to obtain the preset rockburst probability prediction model;

and aiming at a plurality of sampling places, respectively training a naive Bayes classification model by using the third feature vectors which are in one-to-one correspondence with the sampling places to obtain the preset large-deformation probability prediction model.

Optionally, the first labeling information and the second labeling information each include a first level, a second level, a third level, and a fourth level; the method further comprises the following steps:

respectively counting the number of the first category, the number of the second category, the number of the third category and the number of the fourth category; wherein the first category number is the number of the second feature vectors of the first level of which the first label information is, or the number of the third feature vectors of the first level of which the second label information is; the second category number is the number of the second feature vectors of which the first label information is the second level, or the number of the third feature vectors of which the second label information is the second level; the third category number is the number of the second feature vectors of which the first label information is the third level, or the number of the third feature vectors of which the second label information is the third level; the fourth category number is the number of the second feature vectors of which the first label information is the fourth level, or the number of the third feature vectors of which the second label information is the fourth level;

obtaining prior probabilities of rock burst occurrence of a plurality of sampling sites according to the first category quantity, the second category quantity, the third category quantity, the fourth category quantity and the total quantity of the second characteristic vectors;

obtaining the prior probability of large deformation of the plurality of sampling sites according to the first category number, the second category number, the third category number, the fourth category number and the total number of the third feature vectors;

respectively training a naive Bayesian classification model by using the second feature vectors in one-to-one correspondence with the sampling places to obtain the preset rockburst probability prediction model, including:

predicting a third posterior probability of the rockburst occurring at the sampling site to which any second eigenvector belongs by using the naive Bayesian classification model based on the prior probabilities of the rockburst occurring at the multiple sampling sites;

calculating the actual rockburst level of the sampling site to which any second eigenvector belongs and a first loss value of the third posterior probability;

according to the first loss value, adjusting the prior probability of the rock burst at the sampling site and the parameters of the naive Bayes classification model;

when the first loss value is smaller than a preset threshold value, determining the naive Bayes classification model after multiple training as the preset rockburst probability prediction model;

respectively training a naive Bayes classification model by using the third feature vectors in one-to-one correspondence with the sampling places to obtain the preset large deformation probability prediction model, comprising:

predicting a fourth posterior probability of large deformation of a sampling place to which any third feature vector belongs by utilizing the naive Bayes classification model based on the prior probabilities of large deformation of a plurality of sampling places;

calculating the actual large deformation level of the sampling site where any third feature vector belongs and a second loss value of the fourth posterior probability;

according to the second loss value, adjusting the prior probability of large deformation of the sampling place and the parameters of the naive Bayes classification model;

and when the second loss value is smaller than a preset threshold value, determining the naive Bayesian classification model after multiple times of training as the preset large deformation probability prediction model.

Optionally, the method further comprises:

obtaining the length of a risk assessment segment of the tunnel engineering segment;

calculating a comprehensive risk length value of the tunnel engineering section according to the first posterior probability and the second posterior probability, wherein the calculation comprises the following steps:

calculating to obtain the length of the rock burst risk segment according to the first posterior probability and the length of the risk assessment segment;

calculating to obtain the length of the large deformation risk segment according to the second posterior probability and the length of the risk assessment segment;

and accumulating the length of the rockburst risk segment and the length of the large deformation risk segment to obtain a comprehensive risk length value of the tunnel engineering segment.

Optionally, the tunnel engineering section comprises a plurality of tunnel regions; the first posterior probability comprises a plurality of first posterior sub-probabilities; the tunnel regions correspond to the first posterior sub-probabilities one to one; the first posterior sub-probability includes: a first and a second rockburst probability; the second posterior sub-probability includes: a first large deformation probability and a second large deformation probability;

calculating to obtain the length of the rock burst risk segment according to the first posterior probability and the length of the risk assessment segment, wherein the method comprises the following steps:

sequentially obtaining a first rock burst probability and a second rock burst probability from each first posterior sub-probability;

calculating the length of a first risk subsection of each tunnel region with medium rock burst according to the first rock burst probability of each tunnel region and the length of the risk assessment section to obtain a plurality of lengths of the first risk subsections;

calculating the length of a second risk subsection of each tunnel region with strong rock burst according to the second rock burst probability of each tunnel region and the length of the risk evaluation subsection to obtain a plurality of lengths of the second risk subsections;

accumulating the lengths of the first risk subsegments and the lengths of the second risk subsegments to obtain the length of the rockburst risk segment;

calculating to obtain the length of the large deformation risk segment according to the second posterior probability and the length of the risk assessment segment, wherein the calculation comprises the following steps:

sequentially obtaining a first large deformation probability and a second large deformation probability from each second posterior sub probability;

according to the first large deformation probability of each tunnel region and the length of the risk assessment segment, calculating the length of a third risk subsection of each tunnel region with medium and large deformation to obtain a plurality of lengths of the third risk subsections;

according to the second large deformation probability of each tunnel region and the length of the risk assessment segment, calculating the length of a fourth risk subsection of each tunnel region with strong large deformation to obtain a plurality of lengths of the fourth risk subsections;

and accumulating the lengths of the third risk subsections and the lengths of the fourth risk subsections to obtain the length of the large deformation risk subsection.

In a second aspect of the embodiments of the present application, there is provided a railway line comparison and selection device based on a typical disaster of a high-ground-stress tunnel, where the device includes:

the first obtaining module is used for counting a specific tunnel in a to-be-compared route selection path to obtain a tunnel engineering section, and the tunnel engineering section is in one-to-one correspondence with the to-be-compared route selection path; the specific tunnel is a tunnel in a high ground stress environment;

the first acquisition module is used for acquiring first rock strength, first ground stress, a first geological structure and a first surrounding rock grade of the tunnel engineering section;

the first input module is used for inputting the first rock strength, the first ground stress, the first geological structure and the first surrounding rock level into a preset rock burst probability prediction model to obtain a first posterior probability of rock burst of the tunnel engineering section;

the second input module is used for inputting the first rock strength, the first ground stress, the first geological structure and the first surrounding rock grade into a preset large deformation probability prediction model to obtain a second posterior probability of large deformation of the tunnel engineering section;

the calculation module is used for calculating a comprehensive risk length value of the tunnel engineering section according to the first posterior probability and the second posterior probability;

and the determining module is used for determining the tunnel engineering section with the minimum comprehensive risk length value as a target tunnel engineering section and taking the to-be-compared selected railway line corresponding to the target tunnel engineering section as an optimal railway line.

Optionally, the apparatus further comprises:

the second acquisition module is used for acquiring a plurality of sampling places and acquiring a first feature vector of each sampling place; the first feature vector includes: a second rock strength, a second geostress, a second geological formation, and a second country rock class;

the first marking module is used for marking first marking information on the first characteristic vector of each sampling site in sequence according to the actual rock burst grade of the current sampling site to obtain a second characteristic vector carrying the first marking information, and the second characteristic vectors correspond to the sampling sites one by one;

the second marking module is used for marking second marking information on the first eigenvector of each sampling site in sequence according to the actual large deformation grade of the current sampling site to obtain third eigenvectors carrying the second marking information, wherein the third eigenvectors correspond to the sampling sites one by one;

the first training module is used for training a naive Bayesian classification model by respectively utilizing the second feature vectors which are in one-to-one correspondence with the sampling places aiming at the sampling places to obtain the preset rockburst probability prediction model;

and the second training module is used for training a naive Bayes classification model by respectively utilizing the third feature vectors which are in one-to-one correspondence with the sampling positions aiming at the sampling positions to obtain the preset large-deformation probability prediction model.

Optionally, the first annotation information and the second annotation information each include a first level, a second level, a third level, and a fourth level; the device further comprises:

the statistical module is used for respectively counting the number of the first category, the number of the second category, the number of the third category and the number of the fourth category; wherein the first category number is the number of the second feature vectors of the first level of which the first label information is, or the number of the third feature vectors of the first level of which the second label information is; the second category number is the number of the second feature vectors of which the first label information is the second level, or the number of the third feature vectors of which the second label information is the second level; the third category number is the number of the second feature vectors of which the first label information is the third level, or the number of the third feature vectors of which the second label information is the third level; the fourth category number is the number of the second feature vectors of which the first label information is the fourth level, or the number of the third feature vectors of which the second label information is the fourth level;

a first obtaining module, configured to obtain prior probabilities of rock bursts occurring at the multiple sampling sites according to the first category number, the second category number, the third category number, the fourth category number, and the total number of the second feature vectors;

a second obtaining module, configured to obtain prior probabilities of large deformation of the multiple sampling locations according to the first category number, the second category number, the third category number, the fourth category number, and the total number of the third feature vectors;

the first training module comprises:

the first prediction sub-module is used for predicting a third posterior probability of rock burst occurrence of a sampling site to which any second feature vector belongs based on the prior probability of rock burst occurrence of a plurality of sampling sites by using the naive Bayes classification model;

the first calculation submodule is used for calculating the actual rockburst level of the sampling site to which any second eigenvector belongs and a first loss value of the third posterior probability;

the first adjusting submodule is used for adjusting the prior probability of the rock burst occurring at the sampling site and the parameters of the naive Bayes classification model according to the first loss value;

the first determining submodule is used for determining the naive Bayesian classification model after multiple times of training as the preset rockburst probability prediction model when the first loss value is smaller than a preset threshold value;

the second training module comprises:

the second prediction sub-module is used for predicting a fourth posterior probability of large deformation of a sampling place to which any third feature vector belongs based on the prior probability of large deformation of a plurality of sampling places by using the naive Bayes classification model;

the second calculation submodule is used for calculating the actual large deformation level of the sampling site where any third feature vector belongs and a second loss value of the fourth posterior probability;

a second adjusting sub-module, configured to adjust, according to the second loss value, a prior probability of a large deformation occurring at the sampling location and a parameter of the naive bayes classification model;

and the second determining submodule is used for determining the naive Bayes classification model after multiple times of training as the preset large deformation probability prediction model when the second loss value is smaller than a preset threshold value.

Optionally, the apparatus further comprises:

the second obtaining module is used for obtaining the length of the risk assessment segment of the tunnel engineering segment;

the calculation module comprises:

the third calculation sub-module is used for calculating the length of the rock burst risk segment according to the first posterior probability and the length of the risk assessment segment;

the fourth calculation submodule is used for calculating the length of the large deformation risk segment according to the second posterior probability and the length of the risk evaluation segment;

and the accumulation submodule is used for accumulating the length of the rockburst risk segment and the length of the large deformation risk segment to obtain a comprehensive risk length value of the tunnel engineering segment.

Optionally, the tunnel engineering section comprises a plurality of tunnel regions; the first posterior probability comprises a plurality of first posterior sub-probabilities; the tunnel regions correspond to the first posterior sub-probabilities one to one; the first posterior sub-probability includes: a first and second rockburst probability; the second posterior sub-probability includes: a first large deformation probability and a second large deformation probability; the third calculation submodule comprises:

the first obtaining subunit is used for obtaining a first rock burst probability and a second rock burst probability from each first posterior sub probability in sequence;

the first calculating subunit is used for calculating the length of a first risk subsection of each tunnel region, wherein the length of the first risk subsection is used for generating medium rock burst, and a plurality of lengths of the first risk subsections are obtained;

the second calculation subunit is used for calculating the length of a second risk subsection of each tunnel region, in which strong rock burst occurs, according to the second rock burst probability of each tunnel region and the length of the risk assessment section, so as to obtain a plurality of lengths of the second risk subsections;

the first accumulation subunit is configured to accumulate the lengths of the first risk subsections and the lengths of the second risk subsections to obtain the length of the rockburst risk section;

the fourth calculation submodule includes:

the second obtaining subunit is used for sequentially obtaining a first large deformation probability and a second large deformation probability from each second posterior sub-probability;

the third calculation subunit is used for calculating the length of a third risk subsection of each tunnel region with medium and large deformation according to the first large deformation probability of each tunnel region and the length of the risk assessment subsection to obtain a plurality of lengths of the third risk subsections;

the fourth calculating subunit is configured to calculate, according to the second large deformation probability of each tunnel region and the risk assessment segment length, a fourth risk sub-segment length of each tunnel region with a strong large deformation, so as to obtain a plurality of fourth risk sub-segment lengths;

and the second accumulation subunit accumulates the lengths of the third risk subsections and the lengths of the fourth risk subsections to obtain the length of the large deformation risk subsection.

A third aspect of embodiments of the present application provides a readable storage medium, on which a computer program is stored, which, when executed by a processor, implements the steps in the method according to the first aspect of the present application.

A fourth aspect of the embodiments of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor executes the computer program to implement the steps of the method according to the first aspect of the present application.

According to geological factors influencing the occurrence of rock burst and large deformation, the geological factors causing typical tunnel disasters, namely the rock burst and the large deformation are selected: rock strength, ground stress, geological formation, and surrounding rock grade. Counting the specific tunnels of each to-be-compared railway route to obtain tunnel engineering sections, collecting the measured values of the geological factors for each tunnel engineering section, and inputting the collected rock strength, the ground stress, the geological structure and the surrounding rock level into two independent models respectively: presetting a rock burst probability prediction model and a large deformation probability prediction model, and calculating to obtain the rock burst probability of the tunnel engineering section of each to-be-compared iron selection route and the large deformation probability of the tunnel engineering section of each to-be-compared iron selection route based on the rock strength, the ground stress, the geological structure and the surrounding rock level of the tunnel engineering section of each to-be-compared iron selection route by using a machine learning method through the preset rock burst probability prediction model and the large deformation probability prediction model; selecting a risk evaluation segment length according to the length of a tunnel region in the tunnel engineering segment, obtaining the product of the risk evaluation segment length and the probability of the rock burst of the tunnel engineering segment of each to-be-compared iron selection route by adopting a product measurement method, obtaining the rock burst risk segment length, and quantifying the probability of the rock burst of the tunnel engineering segment of each to-be-compared iron selection route; and obtaining the product of the risk evaluation segment length and the probability of the large deformation of the tunnel engineering segment of each to-be-compared iron selection route to obtain a large deformation risk segment length, quantizing the probability of the large deformation of the tunnel engineering segment of each to-be-compared iron selection route, and accumulating the rock burst risk segment length and the large deformation risk segment length for each to-be-compared iron selection route to obtain a comprehensive risk length value of each to-be-compared iron selection route. And finally, selecting the railway line with the minimum comprehensive risk length value as the optimal railway line according to the comprehensive risk length value of each railway line to be compared and selected, reducing potential rock burst and large deformation disasters as much as possible in the line comparison and selection stage, reducing the risk in the tunnel construction period, not depending on artificial evaluation and relying on a large amount of historical data to finish the comparison and selection of the railway line, quantifying the probability of rock burst and large deformation of the tunnel of the railway line, and performing comparison and selection on the railway line according to the objective quantified risk value, thereby ensuring that the necessary selection result of the railway line is more objective and accurate.

Drawings

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings used in the embodiments or the prior art descriptions will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative efforts.

FIG. 1 is a flowchart of the steps of a railway line selection method based on a typical disaster of a high-geostress tunnel according to an embodiment of the present application;

FIG. 2 is a system diagram of a prediction index for a risk of rock burst occurring in a tunnel in a high ground stress environment according to an embodiment of the present application;

FIG. 3 is an index system diagram for predicting the risk of large deformation of a tunnel in a high ground stress environment according to an embodiment of the present application;

FIG. 4 is a flowchart illustrating the steps of training a naive Bayesian classification model according to an embodiment of the application;

fig. 5 is a functional module schematic diagram of a railway line comparison and selection device based on a typical disaster of a high-ground stress tunnel according to an embodiment of the present application.

Detailed Description

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

The high-ground stress areas are high in the terrain stress of high-altitude complex mountain areas such as Sichuan-Tibet plateau, yunobi plateau and the like. The geological structure of the high ground stress area is complex, the new structure moves actively, the fold is broken and developed, and the high ground stress area has the necessary conditions for rock burst and large deformation disasters.

The railway lines of Sichuan-Tibet railway, dian-Tibet railway, new-Tibet railway, ganjin railway and the like all need to pass through the high ground stress area, and because the topography of the plateau mountain area is complex, the tunnel ratio of the railway lines of Sichuan-Tibet railway, dian-Tibet railway, new-Tibet railway, ganjin railway and the like is high, the buried depth of the tunnel is large, and the soft and hard edges along the tunnel are distributed. The tunnel construction can explode the mountain land and the like under external force action, and under the condition that tunnel rock burst and large deformation disasters easily occur in a high ground stress area, the risk of typical tunnel disasters is increased. Typical tunnel disasters are rockburst disasters and large deformation disasters.

In construction, tunnel disasters can directly threaten the safety of constructors and equipment, seriously slow down the construction period and increase the investment of engineering cost. Therefore, how to conduct quantitative risk prediction on rock burst and large deformation based on limited data in the early route selection stage of railway engineering so as to guide the railway route selection is an urgent problem to be solved.

In view of the above problems, the present application provides a railway line comparing and selecting method based on typical disasters of a high ground stress tunnel, which quantifies the risks of potential rock burst and large deformation disasters of a plurality of high ground stress railway lines, selects a railway line with the minimum risk of rock burst and large deformation disasters according to the quantification result of rock burst and large deformation disasters, and reduces the risk in the tunnel construction period in the early stage of line selection of railway engineering.

Fig. 1 is a flowchart of steps of a railway line selection method based on a typical disaster of a high-geostress tunnel in an embodiment of the present application, and refer to fig. 1:

typical tunnel disasters are rockburst disasters and large deformation disasters.

Step S11: counting specific tunnels in the to-be-compared iron route selection road to obtain tunnel engineering sections, wherein the tunnel engineering sections correspond to the to-be-compared iron route selection road one by one; the specific tunnel is a tunnel in a high ground stress environment.

The candidate selection of the railway line means that: before a railway in a high ground stress area is built, namely before an early route selection stage, a plurality of railway routes to be selected are planned, and in the early route selection stage, a railway route with the minimum disaster occurrence risk is determined from the plurality of railway routes to be selected according to the quantified disaster occurrence risk. Specifically, an economic route planning method or the like may be used to design a plurality of routes sequentially passing through the departure place, the waypoints and the destination.

The specific tunnel refers to a tunnel in a railway line in a high ground stress area. Illustratively, 20 tunnels are planned in a railway line, wherein 5 tunnels (tunnel 1, tunnel 2, tunnel 3, tunnel 4 and tunnel 5) are located in a high ground stress area, then the 5 tunnels are all special tunnels, and the 5 tunnels form a tunnel engineering section; the tunnel engineering section comprises the following tunnel regions: the area where tunnel 1 is located, the area where tunnel 2 is located, the area where tunnel 3 is located, the area where tunnel 4 is located, and the area where tunnel 5 is located.

The tunnel engineering section refers to a tunnel section planned in a railway line to be compared. One candidate selection railway line comprises a corresponding tunnel engineering section, and one tunnel engineering section comprises a plurality of tunnel areas. Taking the chuanheng railway line as an example, before the line selection stage, A, B, C three candidate comparison railway line selection paths are planned, wherein the railway line a needs to pass through four high mountains: a tunnel constructed in order to pass the railway line a through the Kunlun mountain, the aeolian volcano, the Tanggulan mountain or the Grignard snow mountain is a tunnel region; the total of the tunnels constructed in order to make the railway line A pass through four high mountains in sequence is a tunnel engineering section.

Step S12: collecting a first rock strength, a first ground stress, a first geological structure and a first surrounding rock level of the tunnel engineering section;

the first rock strength, the first ground stress, the first geological structure and the first surrounding rock grade respectively refer to the strength, the ground stress, the geological structure and the surrounding rock grade of the rock of the tunnel engineering section which are measured or collected in order to predict the probability of the typical tunnel disaster occurring in the tunnel engineering section.

Fig. 2 is a system diagram of prediction indexes of risk of rock burst occurring in a tunnel in a high ground stress environment according to an embodiment of the present application, and fig. 3 is a system diagram of prediction indexes of risk of large deformation occurring in a tunnel in a high ground stress environment according to an embodiment of the present application, with reference to fig. 2 and 3;

in order to quantify the risk of rock burst and large deformation disaster in high ground stress areas, the applicant finds that: the rock burst is related to lithology and rock mass structure, and the rock burst usually occurs in rock mass with hard rock quality, good structural integrity, low fresh or weathering degree, low development degree of weak structural surfaces such as fracture joints and the like and good brittleness. Rock burst is also related to the buried depth and the ground stress, the buried depth is generally in direct proportion to the magnitude of the ground stress, and rock mass is more prone to rock burst or higher in rock burst damage degree in a high ground stress area. The rock burst is related to underground water, the rock burst usually occurs in dry rock mass, the underground water can soften the rock mass, the strength of the rock mass is reduced, and the reserved elastic energy is not enough to generate the rock burst; rock burst is related to structure, and rock burst is more likely to occur or the damage degree of rock burst is higher in stress concentration areas such as a fold core part and an extrusion belt; rock burst is also related to section shape, excavation mode, support time and the like.

The applicant has also found that: the main influencing factors of the large deformation of the tunnel surrounding rock can be summarized as the surrounding rock conditions, the lithological conditions and the ground stress conditions. In terms of surrounding rock conditions, the more broken the rock mass structure, the more underground water, the higher the surrounding rock level, and the worse the surrounding rock quality, the higher the probability of large deformation; in terms of lithological conditions, the lower the uniaxial compressive strength, the lower the elastic modulus and the higher the expansibility of the rock, the greater the probability of large deformation; in terms of the ground stress condition, the larger the initial ground stress value of the rock mass is, the greater the probability of large deformation of the rock mass is.

Based on the findings, the applicant obtains concrete measured values of the rock strength, the ground stress, the geological structure and the surrounding rock grade of rocks along the tunnel engineering section by measurement, and predicts the risk of rock burst and large deformation disaster of the tunnel engineering section by taking the rock strength, the ground stress, the geological structure and the surrounding rock grade as the basis.

Step S13: and inputting the first rock strength, the first ground stress, the first geological structure and the first surrounding rock grade into a preset rock burst probability prediction model to obtain a first posterior probability of the tunnel engineering section for rock burst.

Step S14: and inputting the first rock strength, the first ground stress, the first geological structure and the first surrounding rock grade into a preset large deformation probability prediction model to obtain a second posterior probability of the tunnel engineering section with large deformation.

The applicant further finds that the rock strength of a certain high ground stress area is more than 60MPa, the ground stress is more than 40MPa, the geological structure is a fold core part, and when the surrounding rock level is I level, strong rock burst is easy to occur; the rock strength is more than 30MPa and less than 60MPa, the ground stress is more than 30MPa and less than 40MPa, the geological structure is a special geological structure, and when the grade of surrounding rocks is between I grade and II grade, medium rock burst is easy to occur; the rock strength is more than 15MPa and less than 30MPa, the ground stress is more than 20MPa and less than 30MPa, the geological structure is near the fault, and when the surrounding rock level is between II level and III level, slight rock burst is easy to occur; the rock strength is less than 15MPa, the ground stress is less than 10MPa, the geological structure is a fold two-wing fault fracture zone, and rock burst basically cannot occur when the surrounding rock level is greater than the III level. The rock strength of a certain high ground stress area is less than 5MPa, the ground stress is more than 25MPa, the geological structure is an extrusion belt and a fold core part, and when the surrounding rock grades are V-grade and IV-grade, strong and large deformation is easy to occur; the rock strength is more than 5MPa and less than 15MPa, the ground stress is more than 15MPa and less than 25MPa, the geological structure is near a fault and a fault fracture zone, and when the surrounding rock level is V level, medium and large deformation is easy to occur; the rock strength is more than 15MPa and less than 30MPa, the ground stress is more than 8MPa and less than 15MPa, the geological structure is two folded wings, and when the surrounding rock grade is between IV grade and V grade, slight large deformation is easy to occur; the rock strength is more than 30MPa, the ground stress is less than 8MPa, the geological structure is wrinkle-free and special, and when the surrounding rock level is less than IV level, large deformation basically does not occur. In summary, the impact of the rock strength, the ground stress, the geological structure and the surrounding rock grade on the rock burst severity is different from the impact of the rock strength, the ground stress, the geological structure and the surrounding rock grade on the large deformation severity, and has no correlation, so that the applicant inputs the rock strength, the ground stress, the geological structure and the surrounding rock grade into the preset rock burst probability prediction model and the preset large deformation probability prediction model respectively, calculates the probability of the tunnel engineering section for the rock burst by using the preset rock burst probability prediction model, and calculates the probability of the tunnel engineering section for the large deformation by using the preset large deformation probability prediction model.

Wherein, the grade of the surrounding rock is I grade, which means that the rock is fresh and complete, the structural influence is slight, the joint crack does not develop or develops slightly, the rock is closed and does not extend long, no or little weak structural plane and the fault bandwidth is less than 0.1 meter, and the rock is nearly orthogonal to the hole and is in an integral or blocky masonry structure. The grade II of the surrounding rock means that the rock is fresh or slightly weathered, and the joint crack develops or develops slightly; has a small number of weak structural surfaces and poor interlayer bonding. The fault breaking bandwidth is less than 0.5 m, the fault breaking bandwidth is oblique or orthogonal to the tunnel direction, and the rock mass is of a blocky masonry or layered masonry structure. The surrounding rock grade III means that the rock is slightly weathered or weakly weathered, and the fracture develops and is partially opened to fill mud under the influence of a geological structure; the soft structural surface is distributed more, the fault fracture zone is less than 1 meter, the fault fracture zone is oblique or parallel to the tunnel line, and the rock is in a stone-broken mosaic structure. The surrounding rock grade IV is the same as the class III; the fracture and weak structural surface is more, the fault fracture zone is less than 2 m and is parallel to the hole, the rock body is in a rubble-shaped embedded structure, and the local part is in a rubble-shaped crushing structure. A surrounding rock grade v refers to a dispersion: the rock mass is mostly sand layer landslide accumulation, broken, pebbled and gravelly soil. The surrounding rock grade VI means that most rock bodies are broken and scattered soft soil bodies.

The concrete steps of collecting the rock strength, the ground stress, the geological structure and the surrounding rock grade of the tunnel engineering section can comprise: at least one measuring point is selected from the tunnel engineering section, and geological factors of the measuring point are detected to obtain first rock strength, first ground stress, a first geological structure and a first surrounding rock level.

The first posterior probability refers to a probability vector output by a preset rockburst probability prediction model, and the probability vector comprises the probability of occurrence of a strong rockburst, the probability of occurrence of a medium rockburst, the probability of occurrence of a slight rockburst and the probability of occurrence of no rockburst. And the sum of the four probabilities in the probability vector is one.

The second posterior probability refers to a probability vector output by a preset large deformation probability prediction model, and the probability vector comprises the probability of generating strong large deformation, the probability of generating medium large deformation, the probability of generating slight large deformation and the probability of not generating large deformation. And the sum of the four probabilities in the probability vector is one.

In an example of the present application, after obtaining the first rock strength, the first ground stress, the first geological structure and the first surrounding rock level, the vector P (x) may be further generated according to the measured value of the first rock strength, the measured value of the first ground stress, the measured value of the first geological structure and the measured value of the first surrounding rock level ₁ 、x ₂ 、x ₃ 、x ₄ ). Will vector P (x) ₁ 、x ₂ 、x ₃ 、x ₄ ) Respectively inputting a preset rockburst probability prediction model and a preset large deformation probability prediction model, and extracting P (x) from the preset rockburst probability prediction model ₁ 、x ₂ 、x ₃ 、x ₄ ) Calculating to obtain a first posterior probability, and presetting a large deformation probability prediction model to extract P (x) ₁ 、x ₂ 、x ₃ 、x ₄ ) And calculating to obtain a second posterior probability.

Step S15: calculating a comprehensive risk length value of the tunnel engineering section according to the first posterior probability and the second posterior probability;

the method for calculating the comprehensive risk length value comprises the following specific steps: step S15-1: obtaining the length of a risk assessment segment of the tunnel engineering segment; the length of the risk assessment segment can be selected according to the site selection and the total length of the tunnel engineering segment, the lengths of the tunnel engineering segments of all the railway routes to be compared can be scaled in an equal proportion, and the minimum length is selected as the length of the risk assessment segment. Illustratively, the total lengths of the four candidate railroads are 4 × 10 respectively ⁴ Kilometer, 8X 10 ⁴ Kilometer, 16X 10 ⁴ Kilometer and 1X 10 ⁴ Kilometers, after the total length is reduced in equal proportion, 100 meters is taken as the length of the risk assessment segment.

Step S15-2: calculating to obtain the length of the rock burst risk segment according to the first posterior probability and the length of the risk assessment segment; since the tunnel engineering section comprises a plurality of tunnel regions, the first posterior probability comprises a plurality of first posterior sub-probabilities. The tunnel regions correspond to the first posterior sub-probabilities one by one; the first posterior sub-probability includes: the first rockburst probability, the second rockburst probability, the third rockburst probability and the fourth rockburst probability; the first rockburst probability refers to the probability of strong rockburst in the tunnel, the second rockburst probability refers to the probability of medium rockburst in the tunnel, the third rockburst probability refers to the probability of slight rockburst in the tunnel, and the fourth rockburst probability refers to the probability of no rockburst in the tunnel.

Step S15-3: calculating to obtain the length of the large deformation risk segment according to the second posterior probability and the length of the risk assessment segment; the second posterior sub-probability includes: a first large deformation probability, a second large deformation probability, a third large deformation probability, and a fourth large deformation probability. The first large deformation probability refers to the probability that the tunnel has strong large deformation, the second large deformation probability refers to the probability that the tunnel has medium large deformation, the third large deformation probability refers to the probability that the tunnel has slight large deformation, and the fourth large deformation probability refers to the probability that the tunnel does not have large deformation.

Step S15-4: and accumulating the length of the rockburst risk segment and the length of the large deformation risk segment to obtain a comprehensive risk length value of the tunnel engineering segment.

Specifically, the comprehensive risk length value is calculated by adopting the formula (1):

the method comprises the following steps of obtaining a comprehensive risk length value S, obtaining a large deformation risk length D, obtaining a large deformation risk length I, obtaining a large deformation risk length N, and obtaining a large deformation risk length N, wherein S is a comprehensive risk length value, R is a rockburst risk length, D is a large deformation risk length, i is each tunnel area in a railway line, and n is the number of tunnel areas in the railway line. R is equal to the product of the first posterior probability and the risk assessment segment length, and D is equal to the product of the second posterior probability and the risk assessment segment length.

Continuing with the example of the tibetan railway line, railway line a needs to cross four mountains: the tunnel comprises four tunnel areas, namely a tunnel penetrating through the Kunlun mountain, a tunnel penetrating through the wind volcano, a tunnel penetrating through the Tanggulan mountain and a tunnel penetrating through the Grignard. The four first posterior sub-probabilities are respectively a first posterior probability of passing through a tunnel of Kunlun mountain, a first posterior probability of passing through a tunnel of wind volcano, a first posterior probability of passing through a tunnel of Tanggula mountain, and a first posterior probability of passing through a tunnel of Grignard. The four second posterior sub-probabilities are respectively a second posterior probability of traversing a tunnel of Kunlun mountain, a second posterior probability of traversing a tunnel of wind volcano, a second posterior probability of traversing a tunnel of Tanggula mountain, and a second posterior probability of traversing a tunnel of Grignard.

The specific method for executing step S15-2 is as follows: sequentially obtaining a first rock burst probability and a second rock burst probability from each first posterior sub-probability; because the loss caused by the slight rock burst and the slight large deformation to the tunnel engineering construction is small, the risks caused by medium and strong rock burst and large deformation disasters to the tunnel construction are mainly considered.

Calculating the length of a first risk subsection of each tunnel region, in which medium rock burst occurs, according to the first rock burst probability and the length of the risk evaluation section of each tunnel region to obtain a plurality of lengths of the first risk subsections; calculating the length of a second risk subsection of each tunnel region with strong rock burst according to the second rock burst probability and the length of the risk evaluation section of each tunnel region to obtain a plurality of lengths of the second risk subsections; and accumulating the lengths of the first risk subsegments and the lengths of the second risk subsegments to obtain the length of the rockburst risk segment.

And (3) calculating the length of the rock burst risk segment by adopting a formula (2).

Wherein, B is the length of the risk assessment segment, R is the length of the rockburst risk segment, i is the tunnel region, and R is the length of the tunnel region ₁₁ 、R ₁₂ 、R _1i Probability of occurrence of strong rock burst for different tunnel regions in tunnel engineering section, R ₂₁ 、R ₂₂ 、R _2i The probability of medium rock burst for different tunnel regions in the tunnel engineering section. The first posterior sub-probabilities of two different tunnel regions in a tunnel project section can be respectively expressed as (R) ₁₁ ，R ₂₁ ，R ₃₁ ，R ₄₁ ) And (R) ₁₂ ，R ₂₂ ，R ₃₂ ，R ₄₂ ) Wherein R is ₁₁ And R ₁₂ Is the first rock burst probability, respectively represents the probability of strong rock burst in two different tunnel regions, R ₂₁ And R ₂₂ Is a second rock burst probability, which respectively represents the probability of two different tunnel regions generating medium rock bursts, R ₃₁ And R ₃₂ Is the third rock burst probability, which respectively represents the probability of slight rock burst in two different tunnel regions, R ₄₁ And R ₄₂ And the fourth rock burst probability respectively represents the probability that two different tunnel regions do not generate rock bursts.

The first risk sub-segment length is R _2i X B; the second risk sub-segment length is R _1i ×B。

The specific method for performing step S15-3 is as follows: sequentially obtaining a first large deformation probability and a second large deformation probability from each second posterior sub probability; calculating the length of a third risk subsegment of each tunnel region with medium and large deformation according to the first large deformation probability of each tunnel region and the length of the risk assessment segment to obtain a plurality of lengths of the third risk subsegments; calculating the length of a fourth risk subsegment of each tunnel region with strong large deformation according to the second large deformation probability of each tunnel region and the length of the risk assessment segment to obtain a plurality of lengths of the fourth risk subsegments; and accumulating the lengths of the third risk subsections and the lengths of the fourth risk subsections to obtain the length of the large deformation risk subsection.

And (3) calculating the length of the large deformation risk segment.

Wherein, B is the length of the risk assessment segment, D is the length of the large deformation risk segment, i is the tunnel region, D ₁₁ 、D ₁₂ 、D _1j Probability of strong deformation of different tunnel regions in the tunnel engineering section, D ₂₁ 、D ₂₂ 、D _2i The probability of medium-sized deformation of different tunnel regions in the tunnel engineering section is determined. The second posterior sub-probabilities of two different tunnel regions in the tunnel engineering section can be respectively expressed as (D) ₁₁ ，D ₂₁ ，D ₃₁ ，D ₄₁ ) And (D) ₁₂ ，D ₂₂ ，D ₃₂ ，D ₄₂ ) Wherein D is ₁₁ And D ₁₂ Is the first large deformation probability, respectively representing the probability of the two different tunnel regions having strong large deformation, D ₂₁ And D ₂₂ Is the second largest deformation probability, respectively representing the probability of the two different tunnel regions having medium and large deformation, D ₃₁ And D ₃₂ Is the third largest deformation probability, respectively represents the probability of slightly large deformation of two different tunnel regions, D ₄₁ And D ₄₂ And the fourth maximum deformation probability respectively represents the probability that two different tunnel regions do not have large deformation.

Third risk sub-segment length D _2i X B; fourth Risk subsection Length D _1i ×B。

In one example of the present application, the tunnel engineering section of the railway line a comprises four tunnel zones: a Kunlun mountain tunnel region, a wind-fire mountain tunnel region, a Tanggul mountain tunnel region and a Gray snow mountain tunnel region, wherein the risk evaluation section length B is 100m, and the calculation formula of the rock burst risk section length R is

The first risk sub-segment length comprises: r ₂₁ ×B、R ₂₂ ×B、R ₂₃ X B and R ₂₄ And (B) is multiplied by. Wherein R is ₂₁ The meaning of x B is: selecting a 100-meter risk evaluation segment length from the Kunlun mountain tunnel region, multiplying the segment length by the probability of medium rock burst in the Kunlun mountain tunnel region, and calculating to obtain a risk segment length for representing the medium rock burst in the Kunlun mountain tunnel region; r is ₂₂ The meaning of the multiplied by B is that the length of a risk evaluation segment of 100 meters is selected from the tunnel region of the wind-fire mountain, and is multiplied by the probability of the medium rock burst in the tunnel region of the wind-fire mountain, so that the length of the risk segment for expressing the medium rock burst in the tunnel region of the wind-fire mountain is calculated. R ₂₃ X B and R ₂₄ Meaning of xB expression and R ₂₁ ×B、R ₂₂ The meaning of XB expression is similar and will not be repeated in this application. And finally, accumulating the lengths of the first risk subsections of the strong rock burst occurring segment lengths of the Kunlun mountain tunnel region, the Fenghuoshan tunnel region, the Tanggulshan tunnel region and the Gray snow mountain tunnel with the lengths of the second risk subsections of the medium rock burst occurring segments of the four tunnels to obtain the lengths of the rock burst risk segments. And similarly, accumulating the lengths of the third risk subsegments and the lengths of the fourth risk subsegments of the four tunnels by using a similar method to obtain the length of the large deformation risk subsegment.

Step S16: and determining the tunnel engineering section with the minimum comprehensive risk length value as a target tunnel engineering section, and taking the candidate railway route corresponding to the target tunnel engineering section as an optimal railway route.

In another embodiment of the application, when the lengths of medium and strong rock bursts and large deformation risk segments between the selection lines are the same, the third rock burst probability is continuously selected from the first posterior sub-probabilities, the third large deformation probability is selected from the second posterior sub-probabilities, and the rock burst risk segment length and the large deformation risk segment length are calculated by adopting the formulas (4) and (5).

Wherein R is ₃₁ 、R ₃₂ 、R _3i Probability of slight rock burst occurring for different tunnel zones in a tunnel engineering section, D ₃₁ 、D ₃₂ 、D _3i The probability of slightly large deformations occurring for different tunnel regions in the tunnel engineering section.

According to geological factors influencing the occurrence of rock burst and large deformation, the geological factors causing typical tunnel disasters, namely the rock burst and the large deformation are selected: rock strength, ground stress, geological formations, and grade of surrounding rock. And acquiring the measured values of the geological factors aiming at the tunnel engineering section of each candidate railway line, and respectively inputting the acquired rock strength, ground stress, geological structure and surrounding rock level into two independent models: presetting a rock burst probability prediction model and a large deformation probability prediction model, and calculating the probability of rock burst occurrence of the tunnel engineering section of each to-be-compared iron selection route and the probability of large deformation of the tunnel engineering section of each to-be-compared iron selection route by using the preset rock burst probability prediction model and the large deformation probability prediction model and based on the rock strength, the ground stress, the geological structure and the surrounding rock level of the tunnel engineering section of each to-be-compared iron selection route; selecting a risk evaluation segment length according to the length of a tunnel region in the tunnel engineering segment, obtaining the product of the risk evaluation segment length and the probability of rock burst generation of the tunnel engineering segment of each to-be-compared iron selection route by adopting a product measurement method to obtain the rock burst risk segment length, and quantifying the probability of rock burst generation of the tunnel engineering segment of each to-be-compared iron selection route; and (3) taking the product of the risk evaluation segment length and the probability of large deformation of the tunnel engineering segment of each to-be-compared iron selection route to obtain a large-deformation risk segment length, quantizing the probability of large deformation of the tunnel engineering segment of each to-be-compared iron selection route, and accumulating the rock burst risk segment length and the large-deformation risk segment length for each to-be-compared iron selection route to obtain a comprehensive risk length value of each to-be-compared iron selection route. And finally, selecting the railway line with the minimum comprehensive risk length value as the optimal railway line according to the comprehensive risk length value of each railway line to be compared and selected, reducing potential rock burst and large deformation disasters as much as possible in the line comparison and selection stage, reducing the risk in the tunnel construction period, not depending on artificial evaluation and relying on a large amount of historical data to finish the comparison and selection of the railway line, quantifying the probability of rock burst and large deformation of the tunnel of the railway line, and performing comparison and selection on the railway line according to the objective quantified risk value, thereby ensuring that the necessary selection result of the railway line is more objective and accurate.

In order to more intelligently implement the method proposed by the applicant and enable the application range of the method to be wider, the applicant also constructs two naive Bayes classification models to be trained, trains the two naive Bayes classification models based on the obtained samples to obtain a preset rockburst probability prediction model and a preset large deformation probability prediction model, and executes part or all of the steps in the method by using the preset rockburst probability prediction model and the preset large deformation probability prediction model.

Fig. 4 is a flowchart of steps of training a naive bayes classification model according to an embodiment of the application, referring to fig. 4.

Step S41: obtaining a plurality of sampling positions, and collecting a first feature vector of each sampling position; the first feature vector comprises: a second rock strength, a second geostress, a second geological formation, and a second country rock class.

The sampling site can be any region of western plateau mountain areas, the region with rockburst or large deformation can be used as a first sampling site, the region without rockburst or large deformation is used as a second sampling site, the rock strength, the ground stress, the geological structure and the surrounding rock level of a plurality of first sampling sites are collected, and the rock strength, the ground stress, the geological structure and the surrounding rock level of a plurality of second sampling sites are collected as a comparison group.

The second rock strength, the second ground stress, the second geological structure and the second surrounding rock level respectively refer to the rock strength, the ground stress, the geological structure and the surrounding rock level of the sampling site which are measured or collected for collecting geological factors of different sites of the plateau mountain area.

Step S42: according to the actual rock burst grade of the current sampling site, sequentially marking first marking information on the first eigenvector of each sampling site to obtain second eigenvectors carrying the first marking information, wherein the second eigenvectors correspond to the sampling sites one by one;

step S43: according to the actual large deformation level of the current sampling site, sequentially marking second marking information on the first eigenvector of each sampling site to obtain third eigenvectors carrying the second marking information, wherein the third eigenvectors correspond to the sampling sites one by one;

the first marking information and the second marking information respectively comprise a first grade, a second grade, a third grade and a fourth grade, if a sampling site has an over-strong rockburst, the first characteristic vector corresponding to the sampling site is marked with the first grade, if the sampling site has an over-medium rockburst, the first characteristic vector corresponding to the sampling site is marked with the second grade, if the sampling site has an over-light rockburst, the first characteristic vector corresponding to the sampling site is marked with the third grade, and if the sampling site has no rockburst, the first characteristic vector corresponding to the sampling site is marked with the fourth grade to obtain a second characteristic vector. If the sampling location is deformed too strongly, a first grade is marked on the first eigenvector corresponding to the sampling location, if the sampling location is deformed moderately, a second grade is marked on the first eigenvector corresponding to the sampling location, if the sampling location is deformed slightly greatly, a third grade is marked on the first eigenvector corresponding to the sampling location, and if the sampling location is not deformed too greatly, a fourth grade is marked on the first eigenvector corresponding to the sampling location, so that a third eigenvector is obtained.

The samples for training the naive Bayesian classification model are a second feature vector carrying the first labeling information and a third feature vector carrying the second labeling information.

Step S44: aiming at a plurality of sampling places, respectively training a naive Bayesian classification model by using the second feature vectors which are in one-to-one correspondence with the sampling places to obtain the preset rockburst probability prediction model;

step S45: and aiming at a plurality of sampling places, respectively training a naive Bayes classification model by using the third feature vectors which are in one-to-one correspondence with the sampling places to obtain the preset large-deformation probability prediction model.

The method comprises the steps that a naive Bayes classification model to be trained is constructed according to the condition that whether rock burst or large deformation can occur in a certain area of the nature in the future is unknown objective fact, through multiple times of training, geological factors in a learning sample of the naive Bayes classification model are associated with the rock burst and the association of the geological factors with the large deformation, and accurate experience values are obtained, namely for any place, the prior probability of whether the rock burst occurs is obtained according to historical rock burst data based on the rock strength, the ground stress, the geological structure and the surrounding rock level of the place; or for any place, obtaining the prior probability of whether large deformation occurs according to historical large deformation data based on the rock strength, the ground stress, the geological structure and the surrounding rock level of the place.

Firstly, the applicant obtains an initial prior probability according to a sample before training a naive Bayes classification model. The specific method comprises the following steps: respectively counting the number of the first category, the number of the second category, the number of the third category and the number of the fourth category; wherein the first category number is the number of the second feature vectors of the first level of which the first label information is, or the number of the third feature vectors of the first level of which the second label information is; the second category number is the number of the second feature vectors of which the first label information is the second level, or the number of the third feature vectors of which the second label information is the second level; the third category number is the number of the second feature vectors of which the first label information is the third level, or the number of the third feature vectors of which the second label information is the third level; the fourth category number is the number of the second feature vectors of which the first label information is the fourth level, or the number of the third feature vectors of which the second label information is the fourth level;

obtaining the prior probability of the rock burst of the plurality of sampling sites according to the first category quantity, the second category quantity, the third category quantity, the fourth category quantity and the total quantity of the second characteristic vectors; and obtaining the prior probability of large deformation of the plurality of sampling sites according to the first category number, the second category number, the third category number, the fourth category number and the total number of the third feature vectors.

In one example of the application, 80 sampling locations are obtained, wherein 20 sampling locations have had too strong rock bursts, 10 sampling locations have had too medium rock bursts, 15 sampling locations have had light rock bursts, and 35 sampling locations have not had rock bursts. A first category number of 20, a second category number of 10, a third category number of 15, and a fourth category number of 35 are obtained. The prior probability of the occurrence of the rock burst of the samples obtained based on the 80 sampling sites, namely the second characteristic vector carrying the first marking information is

The specific method for executing step S44 is as follows: predicting a third posterior probability of rockburst occurrence of a sampling site to which any second feature vector belongs by using the naive Bayes classification model based on prior probabilities of rockburst occurrence of a plurality of the sampling sites; calculating the actual rockburst level of the sampling site to which any second eigenvector belongs and a first loss value of the third posterior probability; according to the first loss value, adjusting the prior probability of the rock burst at the sampling site and the parameters of the naive Bayes classification model; and when the first loss value is smaller than a preset threshold value, determining the naive Bayes classification model after multiple times of training as the preset rockburst probability prediction model. And when the first loss value is smaller than the preset threshold value, the naive Bayes classification model can output the posterior probability of the rock burst in the tunnel engineering section in preparation.

Before inputting the second feature vector into the naive Bayes classification model, the likelihood of a plurality of second feature vectors is trained, so that the distribution of the second feature vectors meets the distribution of the naive Bayes classification model, the parameters of the distribution are estimated through training set data, the commonly used Gaussian naive Bayes, polynomial naive Bayes and Bernoulli naive Bayes are provided, and the distribution rule of the Gaussian naive Bayes distribution closest to the sample data feature is obtained by analyzing the application ranges of the three types of distribution models and respectively carrying out trial training. Taking the case that the classification model based on the naive Bayes is established based on the Gaussian naive Bayes distribution, the likelihood of the second feature vector is calculated by adopting the formula (6).

Wherein, mu _y 、σ _y ² Is the expected value and variance of the characteristic of the model parameter to be learned, P (x) _i Y) is the likelihood of the second feature vector, y is the probability that the second feature vector satisfies the Gaussian distribution, and the value of i is 4,P (x) _i |y)＝P(x ₁ ,x ₂ ,…,x ₄ |y)，x ₁ Is a measure of rock strength, x ₂ Is a measure of the earth stress, x ₃ Is a measure of the geological formation, x ₄ Is a measurement of the grade of the surrounding rock.

Specifically, based on the formula (7), the naive bayesian classification model is used for predicting the third posterior probability of the rockburst of the sampling site to which any second feature vector belongs:

wherein the value of n is 4,P (y), the prior probability of occurrence of rockburst of a second characteristic vector, P (x) ₁ ,x ₂ ,…,x _n Y) is the likelihood of the second feature vector, P (x) ₁ ,x ₂ ,…,x _n ) Is the second feature vector.

In the initial stage of training, the predicted third posterior probability inevitably has a larger difference with the first labeled information, the actual rockburst grade of the sampling site is obtained according to the first labeled information, the first loss value of the third posterior probability compared with the actual rockburst grade is calculated, the parameter of the preset rockburst probability prediction model is adjusted according to the first loss value, the preset rockburst probability prediction model can more accurately extract the rock strength, the ground stress, the geological structure and the correlation between the surrounding rock grade and the rockburst in the second characteristic vector, the prior probability is adjusted according to the first loss value, so that the naive Bayes classification model obtains a more accurate experience value under the objective fact that whether the unknown natural world can have the rockburst or the large deformation in the future, and the probability of the rockburst or the large deformation in the arbitrary site can be obtained.

The specific method for executing step S45 is as follows: predicting a fourth posterior probability of large deformation of a sampling place to which any third feature vector belongs by using the naive Bayes classification model based on the prior probabilities of large deformation of a plurality of the sampling places; calculating the actual large deformation level of the sampling site where any third feature vector belongs and a second loss value of the fourth posterior probability; according to the second loss value, adjusting the prior probability of large deformation of the sampling place and the parameters of the naive Bayes classification model; and when the second loss value is smaller than a preset threshold value, determining the naive Bayes classification model after multiple times of training as the preset large deformation probability prediction model. And when the second loss value is smaller than the preset threshold value, the naive Bayes classification model can output the posterior probability of large deformation of the tunnel engineering section in preparation.

The formula and principle used for executing step S45 are the same as those used for executing step S45, and are not described again in this embodiment of the present application.

The method comprises the steps of selecting a plurality of sampling places, collecting rock strength, ground stress, geological structure and surrounding rock level of the sampling places, and marking geological factors according to actual rock burst conditions of the sampling places and actual large deformation conditions of the sampling places to obtain training samples; meanwhile, a naive Bayes classification model is constructed according to a naive Bayes theory, different naive Bayes classification models are trained by training samples, the probability of occurrence of rock burst or large deformation of a sampling place predicted by the naive Bayes classification model and the loss value between the rock burst or large deformation actually occurring at the sampling place are adopted, and the parameter of the naive Bayes classification model and the prior probability of use of the probability of occurrence of rock burst or large deformation predicted by the naive Bayes classification model are adjusted, so that the naive Bayes classification model is learned to obtain an empirical value of whether rock burst occurs at any place in the objective fact that whether rock burst occurs at the future in an unknown natural world or not, or whether large deformation occurs at any place in the objective fact that whether large deformation occurs at the future in the unknown natural world; meanwhile, the naive Bayes classification model can predict the probability of strong rock burst, medium rock burst and slight rock burst or strong deformation, medium deformation and slight deformation of any place more accurately based on the rock strength, ground stress, geological structure and surrounding rock level of the place.

Based on the same inventive concept, the embodiment of the application provides a railway line comparing and selecting device based on typical disasters of a high ground stress tunnel. Referring to fig. 5, fig. 5 is a functional module schematic diagram of a railway line comparison and selection device based on a typical disaster of a high-ground stress tunnel according to an embodiment of the present application. The device includes:

a first obtaining module 51, configured to count a specific tunnel in a to-be-compared route selection route to obtain a tunnel engineering section, where the tunnel engineering section corresponds to the to-be-compared route selection route one to one; the specific tunnel is a tunnel in a high ground stress environment;

a first collecting module 52, configured to collect a first rock strength, a first ground stress, a first geological structure, and a first surrounding rock level of the tunnel engineering section;

a first input module 53, configured to input the first rock strength, the first ground stress, the first geological structure, and the first surrounding rock level into a preset rock burst probability prediction model, so as to obtain a first posterior probability of occurrence of rock burst in the tunnel engineering section;

a second input module 54, configured to input the first rock strength, the first ground stress, the first geological structure, and the first surrounding rock level into a preset large deformation probability prediction model, so as to obtain a second posterior probability of large deformation of the tunnel engineering section;

a calculating module 55, configured to calculate a comprehensive risk length value of the tunnel engineering segment according to the first posterior probability and the second posterior probability;

and the ground determining module 56 is configured to determine the tunnel engineering section with the smallest comprehensive risk length value as a target tunnel engineering section, and use the candidate railway route corresponding to the target tunnel engineering section as an optimal railway route.

Optionally, the apparatus further comprises:

the first training module is used for training a naive Bayesian classification model by respectively using the second feature vectors which are in one-to-one correspondence with the sampling places aiming at the plurality of sampling places to obtain the preset rockburst probability prediction model;

and the second training module is used for training a naive Bayesian classification model by respectively using the third feature vectors in one-to-one correspondence with the sampling places aiming at the plurality of sampling places to obtain the preset large-deformation probability prediction model.

the statistical module is used for respectively counting the first category quantity, the second category quantity, the third category quantity and the fourth category quantity; wherein the first category number is the number of the second feature vectors of the first level of which the first label information is, or the number of the third feature vectors of the first level of which the second label information is; the second category number is the number of the second feature vectors of which the first label information is the second level, or the number of the third feature vectors of which the second label information is the second level; the third category number is the number of the second feature vectors of which the first label information is the third level, or the number of the third feature vectors of which the second label information is the third level; the fourth category number is the number of the second feature vectors of which the first label information is the fourth level, or the number of the third feature vectors of which the second label information is the fourth level;

a first obtaining module, configured to obtain prior probabilities of rock bursts occurring at multiple sampling sites according to the first category number, the second category number, the third category number, the fourth category number, and the total number of the second feature vectors;

the first training module comprises:

the first calculation submodule is used for calculating the actual rock burst level of the sampling site to which any second eigenvector belongs and the first loss value of the third posterior probability;

the second training module comprises:

the second adjusting submodule is used for adjusting the prior probability of large deformation of the sampling place and the parameters of the naive Bayes classification model according to the second loss value;

Optionally, the apparatus further comprises:

the calculation module comprises:

Alternatively,

the tunnel engineering section comprises a plurality of tunnel regions; the first posterior probability comprises a plurality of first posterior sub-probabilities; the tunnel regions correspond to the first posterior sub-probabilities one to one; the first posterior sub-probability includes: a first and second rockburst probability; the second posterior sub-probability includes: a first large deformation probability and a second large deformation probability; the third computation submodule comprises:

the fourth calculation submodule includes:

the second obtaining subunit is used for obtaining a first large deformation probability and a second large deformation probability from each second posterior sub probability in sequence;

Based on the same inventive concept, another embodiment of the present application provides a readable storage medium, on which a computer program is stored, where the program, when executed by a processor, implements the steps in the method for selecting a railway line based on a typical disaster of a high-ground-stress tunnel according to any of the above embodiments of the present application.

Based on the same inventive concept, another embodiment of the present application provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the steps of the method for selecting a railway line based on a high-geostress tunnel typical disaster according to any of the above embodiments of the present application.

It should be noted that, in the present specification, the embodiments are all described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments may be referred to each other.

It should also be noted that, in this document, the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings, and are only for convenience in describing the present invention and simplifying the description, but do not indicate or imply that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and thus, should not be construed as limiting the present invention. Moreover, relational terms such as "first" and "second" are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions or should not be construed as indicating or implying relative importance. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or terminal. Without further limitation, an element defined by the phrase "comprising a … …" does not exclude the presence of another identical element in a process, method, article, or terminal that comprises the element.

The technical solutions provided by the present application are described in detail above, and the principles and embodiments of the present application are described herein by using specific examples, which are only used to help understanding the present application, and the content of the present description should not be construed as limiting the present application. While various modifications of the illustrative embodiments and applications will be apparent to those skilled in the art based upon this disclosure, it is not necessary or necessary to exhaustively enumerate all embodiments, and all obvious variations and modifications can be resorted to, falling within the scope of the disclosure.

Claims

1. A railway line comparison and selection method based on typical disasters of a high-ground-stress tunnel is characterized by comprising the following steps of:

counting specific tunnels in the to-be-compared iron route selection road to obtain tunnel engineering sections, wherein the tunnel engineering sections correspond to the to-be-compared iron route selection road one by one; the specific tunnel is a tunnel in a high ground stress environment;

inputting the first rock strength, the first ground stress, the first geological structure and the first surrounding rock level into a preset rock burst probability prediction model to obtain a first posterior probability of rock burst of the tunnel engineering section;

the first posterior probability comprises a plurality of first posterior sub-probabilities; the tunnel regions correspond to the first posterior sub-probabilities one to one; the first posterior sub-probability comprises: a first and a second rockburst probability; the second posterior sub-probability includes: a first large deformation probability and a second large deformation probability;

sequentially obtaining a first rockburst probability and a second rockburst probability from each first posterior sub-probability;

according to the second rockburst probability of each tunnel region and the length of the risk assessment segment, calculating the length of a second risk subsection of each tunnel region, wherein the strong rockburst happens to the tunnel region, and obtaining a plurality of lengths of the second risk subsections;

sequentially obtaining a first large deformation probability and a second large deformation probability from each second posterior sub-probability;

calculating the length of a fourth risk subsegment of each tunnel region with strong large deformation according to the second large deformation probability of each tunnel region and the length of the risk assessment segment to obtain a plurality of lengths of the fourth risk subsegments;

accumulating the lengths of the third risk subsegments and the lengths of the fourth risk subsegments to obtain the length of the large deformation risk subsegment;

2. The method of claim 1, further comprising:

obtaining a plurality of sampling places, and collecting a first feature vector of each sampling place; the first feature vector comprises: a second rock strength, a second geostress, a second geological formation, and a second country rock class;

3. The method of claim 2, wherein the first annotation information and the second annotation information each comprise a first level, a second level, a third level, and a fourth level; the method further comprises the following steps:

obtaining the prior probability of the rock burst of the plurality of sampling sites according to the first category quantity, the second category quantity, the third category quantity, the fourth category quantity and the total quantity of the second characteristic vectors;

respectively training a naive Bayesian classification model by using the second feature vectors in one-to-one correspondence with the sampling places to obtain the preset rockburst probability prediction model, comprising:

predicting a third posterior probability of rockburst occurrence of a sampling site to which any second feature vector belongs by using the naive Bayes classification model based on prior probabilities of rockburst occurrence of a plurality of the sampling sites;

when the first loss value is smaller than a preset threshold value, determining the naive Bayesian classification model after multiple times of training as the preset rockburst probability prediction model;

predicting a fourth posterior probability of large deformation of a sampling place to which any third feature vector belongs by using the naive Bayes classification model based on the prior probabilities of large deformation of a plurality of the sampling places;

4. The method of claim 1, further comprising:

5. A railway line comparing and selecting device based on typical disasters of a high ground stress tunnel is characterized by comprising:

the first obtaining module is used for counting a specific tunnel in a to-be-compared route selection path to obtain a tunnel engineering section, and the tunnel engineering section is in one-to-one correspondence with the to-be-compared route selection path; the specific tunnel is a tunnel in a high-ground-stress environment;

the first acquisition module is used for acquiring first rock strength, first ground stress, a first geological structure and a first surrounding rock level of the tunnel engineering section;

the second input module is used for inputting the first rock strength, the first ground stress, the first geological structure and the first surrounding rock level into a preset large deformation probability prediction model to obtain a second posterior probability of large deformation of the tunnel engineering section;

the first posterior probability comprises a plurality of first posterior sub-probabilities; the tunnel regions correspond to the first posterior sub-probabilities one by one; the first posterior sub-probability includes: a first and a second rockburst probability; the second posterior sub-probability includes: a first large deformation probability and a second large deformation probability;

a third computation submodule comprising:

the first obtaining subunit is used for sequentially obtaining a first rockburst probability and a second rockburst probability from each first posterior sub-probability;

the first calculation subunit is used for calculating the length of a first risk subsection of each tunnel region, wherein the first risk subsection generates medium rock burst, and the length of the first risk subsection is obtained according to the first rock burst probability and the risk assessment section length of each tunnel region;

the first accumulation subunit is used for accumulating the lengths of the first risk subsections and the lengths of the second risk subsections to obtain the length of the rockburst risk section;

a fourth computation submodule, comprising:

the second accumulation subunit accumulates the lengths of the third risk subsections and the lengths of the fourth risk subsections to obtain the length of a large deformation risk subsection;

6. The apparatus of claim 5, further comprising:

the second acquisition module is used for acquiring a plurality of sampling places and acquiring a first feature vector of each sampling place; the first feature vector comprises: a second rock strength, a second geostress, a second geological formation, and a second country rock class;

7. The apparatus of claim 6, wherein the first annotation information and the second annotation information each comprise a first level, a second level, a third level, and a fourth level; the device further comprises:

the statistical module is used for respectively counting the number of the first category, the number of the second category, the number of the third category and the number of the fourth category; wherein the first category number is the number of the second feature vectors of the first level of which the first label information is, or the number of the third feature vectors of the first level of which the second label information is; the second category number is the number of the second feature vectors of the second level of which the first labeling information is, or the number of the third feature vectors of the second level of which the second labeling information is; the third category number is the number of the second feature vectors of which the first label information is the third level, or the number of the third feature vectors of which the second label information is the third level; the fourth category number is the number of the second feature vectors of which the first label information is the fourth level, or the number of the third feature vectors of which the second label information is the fourth level;

the first training module comprises:

the first determining submodule is used for determining the naive Bayes classification model after multiple times of training as the preset rockburst probability prediction model when the first loss value is smaller than a preset threshold value;

the second training module comprises:

and the second determining submodule is used for determining the naive Bayesian classification model after multiple times of training as the preset large deformation probability prediction model when the second loss value is smaller than a preset threshold value.

8. The apparatus of claim 5, further comprising: