Deep learning grading prediction method and system based on large deformation of surrounding rock of hard rock tunnel

Abstract

The invention provides a deep learning grading prediction method and system based on hard rock tunnel surrounding rock large deformation. The method comprises the steps of collecting characteristic data of a target tunnel project in a preset time period, performing processing to obtain multiple groups of same test data sets, performing processing through multiple different preset prediction models according to the test data sets, and obtaining multiple groups of corresponding grading prediction results, performing comparison processing on the multiple groups of grading prediction results and an actual grading prediction result to obtain multiple groups of corresponding prediction deviation rate data, screening out a preset prediction model set meeting requirements, obtaining real-time characteristic data of the target tunnel engineering, performing processing through the preset prediction model set to obtain a corresponding grading prediction result set, and performing classification prediction on the real-time characteristic data of the target tunnel engineering. And processing to obtain a final grading prediction result, obtaining a corresponding risk grade according to the final grading prediction result, and responding to a corresponding plan, thereby realizing a deep learning grading prediction technology based on the large deformation of the surrounding rock of the hard rock tunnel.

Description

Technical Field

[0001] The present application relates to the technical field of large deformation prediction of surrounding rock, and specifically to a deep learning-based hierarchical prediction method and system for large deformation of surrounding rock in hard rock tunnels. Background Art

[0002] In the construction of hard rock tunnels, large deformation of surrounding rock is a key issue affecting construction safety and project quality. Traditional methods for predicting large deformation of surrounding rock mainly rely on empirical formulas, numerical simulations, and simple analysis of field monitoring data. The empirical formulas are based on summaries of past engineering cases and are difficult to adapt to complex working conditions under different geological conditions. Although numerical simulations can theoretically calculate surrounding rock deformation, the selection of model parameters is subjective, and the calculated results deviate from the actual situation. Simple analysis of field monitoring data lacks in-depth data mining and cannot effectively predict the development trend and degree of large deformation of surrounding rock. With the deepening application of deep learning technology in the engineering field, although some studies have attempted to use it for tunnel surrounding rock deformation prediction, existing methods mostly use a single model, which makes it difficult to accurately capture the surrounding rock deformation characteristics in complex and changeable hard rock tunnel projects, resulting in inaccurate prediction and classification results, which cannot provide a reliable basis for engineering decision-making and seriously threaten the safety and progress of tunnel construction.

[0003] In response to the above problems, effective technical solutions are urgently needed. Summary of the Invention

[0004] The purpose of this application is to provide a deep learning hierarchical prediction method and system based on large deformation of the surrounding rock of hard rock tunnels. Multiple groups of identical test data sets can be obtained by processing the characteristic data of the target tunnel project within a preset time period, and processed through multiple different preset prediction models to obtain corresponding multiple groups of hierarchical prediction results. These results are compared with the actual hierarchical prediction results to obtain corresponding multiple groups of prediction deviation rate data, and the preset prediction model sets that meet the requirements are screened out. The preset prediction model sets are processed according to the real-time characteristic data to obtain the corresponding hierarchical prediction result sets, and the final hierarchical prediction results are obtained by processing, the corresponding risk level is obtained, and the corresponding plan is responded to, thereby realizing the technology of deep learning hierarchical prediction based on large deformation of the surrounding rock of hard rock tunnels.

[0005] This application also provides a deep learning-based hierarchical prediction method for large deformation of surrounding rock in hard rock tunnels, comprising the following steps: Collect characteristic data of the target tunnel project within a preset time period and process it to obtain multiple sets of identical test data sets; Processing the test data set through multiple different preset prediction models to obtain corresponding multi-group hierarchical prediction results; Comparing the multiple groups of graded prediction results with the actual graded prediction results, respectively, to obtain corresponding multiple groups of prediction deviation rate data, and screening out a preset prediction model set that meets the requirements; Acquire real-time characteristic data of the target tunnel project, process it through the preset prediction model set, obtain a corresponding hierarchical prediction result set, and process it to obtain a final hierarchical prediction result; Obtain the corresponding risk level based on the final graded prediction result and respond to the corresponding plan.

[0006] Optionally, in the deep learning hierarchical prediction method for large deformation of surrounding rock of hard rock tunnels described in the present application, the step of collecting characteristic data of the target tunnel project within a preset time period and processing to obtain multiple sets of identical test data sets includes: Collect characteristic data of the target tunnel project within a preset time period, including geological survey data, tunnel design data, construction data, and on-site monitoring data, and process them to obtain multiple sets of identical test data sets; The geological survey data include rock type, strength, elastic modulus and ground stress; The tunnel design data includes burial depth, cross-sectional dimensions and shape; The construction data includes excavation method, support time and support strength; The on-site monitoring data includes displacement and strain.

[0007] Optionally, in the deep learning hierarchical prediction method based on large deformation of surrounding rock of hard rock tunnels described in the present application, the processing of the test data set by multiple different preset prediction models to obtain corresponding multiple groups of hierarchical prediction results includes: Processing the test data set through multiple different preset prediction models to obtain corresponding multi-group hierarchical prediction results; The preset prediction models include a convolutional neural network prediction model, a recurrent neural network prediction model, a long short-term memory network prediction model, a gated recurrent unit prediction model and a random forest prediction model.

[0008] Optionally, in the deep learning hierarchical prediction method based on large deformation of surrounding rock of hard rock tunnels described in the present application, the multiple groups of hierarchical prediction results are compared with the actual hierarchical prediction results respectively to obtain corresponding multiple groups of prediction deviation rate data, and a preset prediction model set that meets the requirements is screened out, including: Obtaining actual graded prediction results of the target tunnel project within the preset time period, and comparing them with the multiple groups of graded prediction results to obtain corresponding multiple groups of prediction deviation rate data; Comparing the plurality of sets of predicted deviation rate data with a first preset threshold and a second preset threshold respectively; If the prediction deviation rate data is greater than the first preset threshold and less than the second preset threshold, the corresponding graded prediction result meets the requirements; A preset prediction model set that meets the requirements is obtained based on all comparison results.

[0009] Optionally, in the deep learning hierarchical prediction method based on large deformation of surrounding rock of hard rock tunnels described in the present application, acquiring real-time characteristic data of the target tunnel project, processing the data using the preset prediction model set, obtaining a corresponding hierarchical prediction result set, and processing the data to obtain a final hierarchical prediction result, includes: monitoring the target tunnel project and acquiring real-time characteristic data, including real-time geological survey data, real-time tunnel design data, real-time construction data, and real-time field monitoring data; Processing the real-time characteristic data through the preset prediction model set to obtain a corresponding hierarchical prediction result set; A weighted process is performed on the hierarchical prediction result set to obtain a final hierarchical prediction result.

[0010] Optionally, in the deep learning hierarchical prediction method based on large deformation of surrounding rock of hard rock tunnels described in the present application, obtaining a corresponding risk level according to the final hierarchical prediction result and responding to a corresponding plan include: Compare the final classification prediction result with the preset classification prediction threshold, obtain the corresponding risk level according to the comparison result, and respond to the corresponding plan; If the comparison result is less than the first preset classification prediction threshold, the corresponding risk level is Level I, and the corresponding emergency plan is a Level I emergency plan, which includes enhanced monitoring and regular inspections; If the comparison result is greater than or equal to the first preset classification prediction threshold and less than the second preset classification prediction threshold, the corresponding risk level is Level II, and the corresponding emergency plan is a Level II emergency plan, which includes suspension of construction, reinforcement and support, and cause analysis; If the comparison result is greater than or equal to the second preset classification prediction threshold, the corresponding risk level is Level III, and the corresponding plan is a Level III plan, which includes immediate evacuation, emergency reinforcement, and the formulation of special treatment plans.

[0011] In a second aspect, the present application provides a deep learning hierarchical prediction system for large deformation of surrounding rock of hard rock tunnels, the system comprising: a memory and a processor, the memory comprising a program for a deep learning hierarchical prediction method for large deformation of surrounding rock of hard rock tunnels, the program for a deep learning hierarchical prediction method for large deformation of surrounding rock of hard rock tunnels being executed by the processor to implement the following steps: Collect characteristic data of the target tunnel project within a preset time period and process it to obtain multiple sets of identical test data sets; Processing the test data set through multiple different preset prediction models to obtain corresponding multi-group hierarchical prediction results; Comparing the multiple groups of graded prediction results with the actual graded prediction results, respectively, to obtain corresponding multiple groups of prediction deviation rate data, and screening out a preset prediction model set that meets the requirements; Acquire real-time characteristic data of the target tunnel project, process it through the preset prediction model set, obtain a corresponding hierarchical prediction result set, and process it to obtain a final hierarchical prediction result; Obtain the corresponding risk level based on the final graded prediction result and respond to the corresponding plan.

[0012] Optionally, in the deep learning hierarchical prediction system based on large deformation of surrounding rock of hard rock tunnels described in the present application, the acquisition of characteristic data of the target tunnel project within a preset time period and processing to obtain multiple sets of identical test data sets include: Collect characteristic data of the target tunnel project within a preset time period, including geological survey data, tunnel design data, construction data, and on-site monitoring data, and process them to obtain multiple sets of identical test data sets; The geological survey data include rock type, strength, elastic modulus and ground stress; The tunnel design data includes burial depth, cross-sectional dimensions and shape; The construction data includes excavation method, support time and support strength; The on-site monitoring data includes displacement and strain.

[0013] Optionally, in the deep learning hierarchical prediction system based on large deformation of surrounding rock of hard rock tunnels described in the present application, the processing of the test data set by multiple different preset prediction models to obtain corresponding multiple groups of hierarchical prediction results includes: Processing the test data set through multiple different preset prediction models to obtain corresponding multi-group hierarchical prediction results; The preset prediction models include a convolutional neural network prediction model, a recurrent neural network prediction model, a long short-term memory network prediction model, a gated recurrent unit prediction model and a random forest prediction model.

[0014] Optionally, in the deep learning hierarchical prediction system based on large deformation of surrounding rock of hard rock tunnels described in the present application, the multiple groups of hierarchical prediction results are compared with the actual hierarchical prediction results respectively to obtain corresponding multiple groups of prediction deviation rate data, and a preset prediction model set that meets the requirements is screened out, including: Obtaining actual graded prediction results of the target tunnel project within the preset time period, and comparing them with the multiple groups of graded prediction results to obtain corresponding multiple groups of prediction deviation rate data; Comparing the plurality of sets of predicted deviation rate data with a first preset threshold and a second preset threshold respectively; If the prediction deviation rate data is greater than the first preset threshold and less than the second preset threshold, the corresponding graded prediction result meets the requirements; A preset prediction model set that meets the requirements is obtained based on all comparison results.

[0015] From the above, it can be seen that the deep learning hierarchical prediction method and system based on large deformation of surrounding rock of hard rock tunnel provided by the present application collects characteristic data of the target tunnel project within a preset time period, and processes it to obtain multiple groups of identical test data sets, processes the test data sets through multiple different preset prediction models, obtains corresponding multiple groups of hierarchical prediction results, compares the multiple groups of hierarchical prediction results with the actual hierarchical prediction results, obtains corresponding multiple groups of prediction deviation rate data, and screens out the preset prediction model set that meets the requirements, obtains real-time characteristic data of the target tunnel project, processes it through the preset prediction model set, obtains the corresponding hierarchical prediction result set, and processes it to obtain the final hierarchical prediction result, obtains the corresponding risk level according to the final hierarchical prediction result, and responds to the corresponding plan, thereby realizing the technology of deep learning hierarchical prediction based on large deformation of surrounding rock of hard rock tunnel.

[0016] Other features and advantages of the present application will be described in the following description, and in part will become apparent from the description, or understood by practicing the embodiments of the present application. The objectives and other advantages of the present application can be achieved and obtained through the structures particularly pointed out in the written description and the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments of the present application. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.

[0018] Figure 1 A flowchart of a deep learning hierarchical prediction method for large deformation of surrounding rock in hard rock tunnels provided in an embodiment of the present application; Figure 2 A flowchart of obtaining multiple sets of identical test data sets for the deep learning hierarchical prediction method for large deformation of surrounding rock of hard rock tunnels provided in an embodiment of the present application; Figure 3 A flowchart of screening out a set of preset prediction models that meet the requirements of the deep learning hierarchical prediction method for large deformation of surrounding rock of hard rock tunnels provided in an embodiment of the present application; Figure 4 A flowchart for obtaining the final graded prediction results corresponding to the deep learning graded prediction method for large deformation of surrounding rock of hard rock tunnels provided in an embodiment of the present application. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work fall within the scope of protection of the present application.

[0020] It should be noted that similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and should not be understood as indicating or implying relative importance.

[0021] Please refer to Figure 1 , Figure 1 This is a flowchart of a deep learning hierarchical prediction method for large deformation of surrounding rock in hard rock tunnels in some embodiments of the present application. This deep learning hierarchical prediction method for large deformation of surrounding rock in hard rock tunnels is used in terminal devices such as computers and mobile phones. This deep learning hierarchical prediction method for large deformation of surrounding rock in hard rock tunnels includes the following steps: S11, collecting characteristic data of the target tunnel project within a preset time period, and processing to obtain multiple sets of identical test data sets; S12, processing the test data set using a plurality of different preset prediction models to obtain corresponding multi-group hierarchical prediction results; S13, comparing the multiple groups of hierarchical prediction results with the actual hierarchical prediction results respectively, obtaining corresponding multiple groups of prediction deviation rate data, and screening out a preset prediction model set that meets the requirements; S14, acquiring real-time characteristic data of the target tunnel project, processing the data using the preset prediction model set to obtain a corresponding hierarchical prediction result set, and processing the data to obtain a final hierarchical prediction result; S15. Obtain the corresponding risk level according to the final classification prediction result and respond to the corresponding plan.

[0022] It should be noted that during the construction of the target tunnel project, a high-precision sensor network is used to continuously collect characteristic data such as geological survey data, tunnel design data, construction data, and on-site monitoring data within a preset time period, and multiple sets of identical test data sets are generated through cloning technology; then, these test data sets are respectively imported into multiple different preset prediction models such as neural networks, random forests, and support vector machines; each model conducts in-depth data mining based on its own algorithm and outputs corresponding multi-group hierarchical prediction results, and then, the multi-group hierarchical prediction results are compared one by one with the actual hierarchical prediction results, and the corresponding multi-group prediction deviation rate data are calculated; based on Based on the preset error threshold, the preset prediction models with prediction deviation rates within an acceptable range are screened out to form a preset prediction model set that meets the requirements; after obtaining the real-time characteristic data of the target tunnel project, it is input into the preset prediction model set, and each model independently calculates to generate a corresponding graded prediction result set; these results are integrated through weighted averaging to obtain the final graded prediction result, accurately locking the risk level; finally, based on the final graded prediction result, the system quickly matches the corresponding risk level plan, which includes level I, II, and III plans; if it is judged to be a level III plan, immediate evacuation, emergency reinforcement, and the formulation of a special treatment plan will be initiated.

[0023] Please refer to Figure 2 , Figure 2 This is a flowchart for obtaining multiple sets of identical test data sets based on a deep learning hierarchical prediction method for large deformation of surrounding rock in hard rock tunnels in some embodiments of the present application. According to embodiments of the present invention, collecting characteristic data of a target tunnel project within a preset time period and processing it to obtain multiple sets of identical test data sets includes: S21. Collect characteristic data of the target tunnel project within a preset time period, including geological survey data, tunnel design data, construction data, and on-site monitoring data, and process the data to obtain multiple sets of identical test data sets; S22. The geological survey data includes rock type, strength, elastic modulus, and ground stress; S23, the tunnel design data includes burial depth, cross-sectional dimensions and shape; S24, the construction data includes excavation method, support time and support strength; S25. The on-site monitoring data includes displacement and strain.

[0024] It should be noted that the characteristic data of the target tunnel project within a preset time period, including geological survey data, tunnel design data, construction data and field monitoring data, are collected and processed through cloning or replication technology to obtain multiple sets of identical test data sets. Among them, the geological survey data include rock type, strength, elastic modulus and ground stress; the tunnel design data include burial depth, cross-sectional size and shape; the construction data include excavation method, support time and support strength; the field monitoring data include displacement and strain.

[0025] According to an embodiment of the present invention, the processing of the test data set through a plurality of different preset prediction models to obtain corresponding multi-group hierarchical prediction results includes: Processing the test data set through multiple different preset prediction models to obtain corresponding multi-group hierarchical prediction results; The preset prediction models include a convolutional neural network prediction model, a recurrent neural network prediction model, a long short-term memory network prediction model, a gated recurrent unit prediction model and a random forest prediction model.

[0026] It should be noted that multiple groups of identical test data sets are input into different preset prediction models for processing. These preset prediction models include convolutional neural network prediction models, recurrent neural network prediction models, long short-term memory network prediction models, gated recurrent unit prediction models and random forest prediction models, and each preset prediction model outputs a set of graded prediction results of surrounding rock deformation after processing the data.

[0027] Please refer to Figure 3 , Figure 3 This is a flowchart for selecting a preset prediction model set that meets the requirements of a deep learning hierarchical prediction method for large deformation of surrounding rock of hard rock tunnels in some embodiments of the present application. According to an embodiment of the present invention, the multiple sets of hierarchical prediction results are compared with the actual hierarchical prediction results to obtain corresponding multiple sets of prediction deviation rate data, and the preset prediction model set that meets the requirements is selected, including: S31, obtaining the actual graded prediction results of the target tunnel project within the preset time period, and comparing them with the multiple groups of graded prediction results to obtain corresponding multiple groups of prediction deviation rate data; S32, comparing the multiple sets of prediction deviation rate data with a first preset threshold and a second preset threshold respectively; S33. If the prediction deviation rate data is greater than the first preset threshold and less than the second preset threshold, the corresponding graded prediction result meets the requirements; S34. Obtain a preset prediction model set that meets the requirements based on all comparison results.

[0028] It should be noted that the actual graded prediction results within the preset time period are accurately obtained. The results are based on comprehensive judgment of multiple sources of information such as geological radar detection, displacement monitoring data, and expert on-site evaluation, and serve as the gold standard for subsequent comparisons. This actual result is carefully compared with the multi-group graded prediction results output by each preset prediction model using quantitative methods such as mean square error and mean absolute error, and the corresponding multiple groups of prediction deviation rate data are calculated to intuitively present the prediction accuracy of each model in digital form. Subsequently, the multiple groups of prediction deviation rate data are compared with the pre-set first preset threshold and second preset threshold respectively; among which the first preset threshold is set to be relatively loose. The basic standard is used to eliminate models that deviate significantly from reality; the second preset threshold serves as a high-precision screening threshold, allowing only models with extremely small errors to pass; during the screening process, if a certain prediction deviation rate data is greater than the first preset threshold and less than the second preset threshold, it means that the corresponding graded prediction result has a certain error, but is still within the acceptable range of the project, and the result is judged to meet the requirements. This standard not only avoids missing potential models due to excessive strictness, but also prevents low-precision models from mixing in; finally, the system will screen out the preset prediction models that output graded prediction results that meet the requirements based on all comparison results, and integrate them to form a set of preset prediction models that meet the requirements.

[0029] Please refer to Figure 4 , Figure 4 This is a flowchart of obtaining the final graded prediction results of a deep learning graded prediction method for large deformation of surrounding rock of hard rock tunnels in some embodiments of the present application. According to embodiments of the present invention, obtaining real-time characteristic data of the target tunnel project, processing it through the preset prediction model set, obtaining a corresponding graded prediction result set, and processing to obtain the final graded prediction results includes: S41, monitoring the target tunnel project and obtaining real-time characteristic data, including real-time geological survey data, real-time tunnel design data, real-time construction data, and real-time on-site monitoring data; S42: Process the real-time characteristic data using the preset prediction model set to obtain a corresponding hierarchical prediction result set; S43: Perform weighted processing according to the hierarchical prediction result set to obtain a final hierarchical prediction result.

[0030] It should be noted that the target tunnel project is monitored and real-time characteristic data is obtained, including real-time geological survey data, real-time tunnel design data, real-time construction data and real-time field monitoring data; among which, the real-time geological survey data includes real-time rock type, real-time strength, real-time elastic modulus and real-time ground stress; the real-time tunnel design data includes real-time burial depth, real-time cross-sectional size and real-time shape; the real-time construction data includes real-time excavation method, real-time support time and real-time support strength; the real-time field monitoring data includes real-time displacement and real-time strain; based on these data, each prediction model in the preset prediction model set screened out previously is processed separately, and each model outputs a graded prediction result, and finally a corresponding graded prediction result set is obtained; the graded prediction results in the graded prediction result set are weighted to obtain the final graded prediction result.

[0031] According to an embodiment of the present invention, obtaining a corresponding risk level according to the final graded prediction result and responding to a corresponding plan includes: Compare the final classification prediction result with the preset classification prediction threshold, obtain the corresponding risk level according to the comparison result, and respond to the corresponding plan; If the comparison result is less than the first preset classification prediction threshold, the corresponding risk level is Level I, and the corresponding emergency plan is a Level I emergency plan, which includes enhanced monitoring and regular inspections; If the comparison result is greater than or equal to the first preset classification prediction threshold and less than the second preset classification prediction threshold, the corresponding risk level is Level II, and the corresponding emergency plan is a Level II emergency plan, which includes suspension of construction, reinforcement and support, and cause analysis; If the comparison result is greater than or equal to the second preset classification prediction threshold, the corresponding risk level is Level III, and the corresponding plan is a Level III plan, which includes immediate evacuation, emergency reinforcement, and the formulation of special treatment plans.

[0032] It should be noted that the final graded prediction result is compared with the preset graded prediction threshold, wherein the preset graded prediction threshold includes a first preset graded prediction threshold and a second preset graded prediction threshold, and the first preset graded prediction threshold is less than the second preset graded prediction threshold; then the corresponding risk level is obtained according to the comparison result, and the corresponding plan is responded to; if the comparison result is less than the first preset graded prediction threshold, the corresponding risk level is level I, that is, low risk, and the corresponding plan is a level I plan, and the plan content includes strengthening monitoring and regular inspections; if the comparison result is greater than or equal to the first preset graded prediction threshold and less than the second preset graded prediction threshold, the corresponding risk level is level II, that is, medium risk, and the corresponding plan is a level II plan, and the plan content includes suspending construction, strengthening support and cause analysis; if the comparison result is greater than or equal to the second preset graded prediction threshold, the corresponding risk level is level III, that is, high risk, and the corresponding plan is a level III plan, and the plan content includes immediate evacuation, emergency reinforcement and formulation of special treatment plans.

[0033] In a second aspect, the present invention further discloses a deep learning hierarchical prediction system based on large deformation of surrounding rock of hard rock tunnels, comprising a memory and a processor, wherein the memory includes a deep learning hierarchical prediction method program based on large deformation of surrounding rock of hard rock tunnels, and when the deep learning hierarchical prediction method program based on large deformation of surrounding rock of hard rock tunnels is executed by the processor, the following steps are implemented: Collect characteristic data of the target tunnel project within a preset time period and process it to obtain multiple sets of identical test data sets; Processing the test data set through multiple different preset prediction models to obtain corresponding multi-group hierarchical prediction results; Comparing the multiple groups of graded prediction results with the actual graded prediction results, respectively, to obtain corresponding multiple groups of prediction deviation rate data, and screening out a preset prediction model set that meets the requirements; Acquire real-time characteristic data of the target tunnel project, process it through the preset prediction model set, obtain a corresponding hierarchical prediction result set, and process it to obtain a final hierarchical prediction result; Obtain the corresponding risk level based on the final graded prediction result and respond to the corresponding plan.

[0034] It should be noted that during the construction of the target tunnel project, a high-precision sensor network is used to continuously collect characteristic data such as geological survey data, tunnel design data, construction data, and on-site monitoring data within a preset time period, and multiple sets of identical test data sets are generated through cloning technology; then, these test data sets are respectively imported into multiple different preset prediction models such as neural networks, random forests, and support vector machines; each model conducts in-depth data mining based on its own algorithm and outputs corresponding multi-group hierarchical prediction results, and then, the multi-group hierarchical prediction results are compared one by one with the actual hierarchical prediction results, and the corresponding multi-group prediction deviation rate data are calculated; based on Based on the preset error threshold, the preset prediction models with prediction deviation rates within an acceptable range are screened out to form a preset prediction model set that meets the requirements; after obtaining the real-time characteristic data of the target tunnel project, it is input into the preset prediction model set, and each model independently calculates to generate a corresponding graded prediction result set; these results are integrated through weighted averaging to obtain the final graded prediction result, accurately locking the risk level; finally, based on the final graded prediction result, the system quickly matches the corresponding risk level plan, which includes level I, II, and III plans; if it is judged to be a level III plan, immediate evacuation, emergency reinforcement, and the formulation of a special treatment plan will be initiated.

[0035] According to an embodiment of the present invention, collecting characteristic data of a target tunnel project within a preset time period and processing to obtain multiple sets of identical test data sets includes: Collect characteristic data of the target tunnel project within a preset time period, including geological survey data, tunnel design data, construction data, and on-site monitoring data, and process them to obtain multiple sets of identical test data sets; The geological survey data include rock type, strength, elastic modulus and ground stress; The tunnel design data includes burial depth, cross-sectional dimensions and shape; The construction data includes excavation method, support time and support strength; The on-site monitoring data includes displacement and strain.

[0036] It should be noted that the characteristic data of the target tunnel project within a preset time period, including geological survey data, tunnel design data, construction data and field monitoring data, are collected and processed through cloning or replication technology to obtain multiple sets of identical test data sets. Among them, the geological survey data include rock type, strength, elastic modulus and ground stress; the tunnel design data include burial depth, cross-sectional size and shape; the construction data include excavation method, support time and support strength; the field monitoring data include displacement and strain.

[0037] According to an embodiment of the present invention, the processing of the test data set through a plurality of different preset prediction models to obtain corresponding multi-group hierarchical prediction results includes: Processing the test data set through multiple different preset prediction models to obtain corresponding multi-group hierarchical prediction results; The preset prediction models include a convolutional neural network prediction model, a recurrent neural network prediction model, a long short-term memory network prediction model, a gated recurrent unit prediction model and a random forest prediction model.

[0038] It should be noted that multiple groups of identical test data sets are input into different preset prediction models for processing. These preset prediction models include convolutional neural network prediction models, recurrent neural network prediction models, long short-term memory network prediction models, gated recurrent unit prediction models and random forest prediction models, and each preset prediction model outputs a set of graded prediction results of surrounding rock deformation after processing the data.

[0039] According to an embodiment of the present invention, the multiple groups of hierarchical prediction results are compared with the actual hierarchical prediction results to obtain corresponding multiple groups of prediction deviation rate data, and a preset prediction model set that meets the requirements is screened out, including: Obtaining actual graded prediction results of the target tunnel project within the preset time period, and comparing them with the multiple groups of graded prediction results to obtain corresponding multiple groups of prediction deviation rate data; Comparing the plurality of sets of predicted deviation rate data with a first preset threshold and a second preset threshold respectively; If the prediction deviation rate data is greater than the first preset threshold and less than the second preset threshold, the corresponding graded prediction result meets the requirements; A preset prediction model set that meets the requirements is obtained based on all comparison results.

[0040] It should be noted that the actual graded prediction results within the preset time period are accurately obtained. The results are based on comprehensive judgment of multiple sources of information such as geological radar detection, displacement monitoring data, and expert on-site evaluation, and serve as the gold standard for subsequent comparisons. This actual result is carefully compared with the multi-group graded prediction results output by each preset prediction model using quantitative methods such as mean square error and mean absolute error, and the corresponding multiple groups of prediction deviation rate data are calculated to intuitively present the prediction accuracy of each model in digital form. Subsequently, the multiple groups of prediction deviation rate data are compared with the pre-set first preset threshold and second preset threshold respectively; among which the first preset threshold is set to be relatively loose. The basic standard is used to eliminate models that deviate significantly from reality; the second preset threshold serves as a high-precision screening threshold, allowing only models with extremely small errors to pass; during the screening process, if a certain prediction deviation rate data is greater than the first preset threshold and less than the second preset threshold, it means that the corresponding graded prediction result has a certain error, but is still within the acceptable range of the project, and the result is judged to meet the requirements. This standard not only avoids missing potential models due to excessive strictness, but also prevents low-precision models from mixing in; finally, the system will screen out the preset prediction models that output graded prediction results that meet the requirements based on all comparison results, and integrate them to form a set of preset prediction models that meet the requirements.

[0041] According to an embodiment of the present invention, acquiring the real-time characteristic data of the target tunnel project, processing the data using the preset prediction model set to obtain a corresponding hierarchical prediction result set, and processing the data to obtain a final hierarchical prediction result includes: monitoring the target tunnel project and acquiring real-time characteristic data, including real-time geological survey data, real-time tunnel design data, real-time construction data, and real-time field monitoring data; Processing the real-time characteristic data through the preset prediction model set to obtain a corresponding hierarchical prediction result set; A weighted process is performed on the hierarchical prediction result set to obtain a final hierarchical prediction result.

[0042] It should be noted that the target tunnel project is monitored and real-time characteristic data is obtained, including real-time geological survey data, real-time tunnel design data, real-time construction data and real-time field monitoring data; among which, the real-time geological survey data includes real-time rock type, real-time strength, real-time elastic modulus and real-time ground stress; the real-time tunnel design data includes real-time burial depth, real-time cross-sectional size and real-time shape; the real-time construction data includes real-time excavation method, real-time support time and real-time support strength; the real-time field monitoring data includes real-time displacement and real-time strain; based on these data, each prediction model in the preset prediction model set screened out previously is processed separately, and each model outputs a graded prediction result, and finally a corresponding graded prediction result set is obtained; the graded prediction results in the graded prediction result set are weighted to obtain the final graded prediction result.

[0043] According to an embodiment of the present invention, obtaining a corresponding risk level according to the final graded prediction result and responding to a corresponding plan includes: Compare the final classification prediction result with the preset classification prediction threshold, obtain the corresponding risk level according to the comparison result, and respond to the corresponding plan; If the comparison result is less than the first preset classification prediction threshold, the corresponding risk level is Level I, and the corresponding emergency plan is a Level I emergency plan, which includes enhanced monitoring and regular inspections; If the comparison result is greater than or equal to the first preset classification prediction threshold and less than the second preset classification prediction threshold, the corresponding risk level is Level II, and the corresponding emergency plan is a Level II emergency plan, which includes suspension of construction, reinforcement and support, and cause analysis; If the comparison result is greater than or equal to the second preset classification prediction threshold, the corresponding risk level is Level III, and the corresponding plan is a Level III plan, which includes immediate evacuation, emergency reinforcement, and the formulation of special treatment plans.

[0044] It should be noted that the final graded prediction result is compared with the preset graded prediction threshold, wherein the preset graded prediction threshold includes a first preset graded prediction threshold and a second preset graded prediction threshold, and the first preset graded prediction threshold is less than the second preset graded prediction threshold; then the corresponding risk level is obtained according to the comparison result, and the corresponding plan is responded to; if the comparison result is less than the first preset graded prediction threshold, the corresponding risk level is level I, that is, low risk, and the corresponding plan is a level I plan, and the plan content includes strengthening monitoring and regular inspections; if the comparison result is greater than or equal to the first preset graded prediction threshold and less than the second preset graded prediction threshold, the corresponding risk level is level II, that is, medium risk, and the corresponding plan is a level II plan, and the plan content includes suspending construction, strengthening support and cause analysis; if the comparison result is greater than or equal to the second preset graded prediction threshold, the corresponding risk level is level III, that is, high risk, and the corresponding plan is a level III plan, and the plan content includes immediate evacuation, emergency reinforcement and formulation of special treatment plans.

[0045] The present invention discloses a deep learning hierarchical prediction method and system based on large deformation of surrounding rock of hard rock tunnels. The method collects characteristic data of the target tunnel project within a preset time period and processes it to obtain multiple groups of identical test data sets. The test data sets are processed through multiple different preset prediction models to obtain corresponding multiple groups of hierarchical prediction results. The multiple groups of hierarchical prediction results are compared with the actual hierarchical prediction results to obtain corresponding multiple groups of prediction deviation rate data, and the preset prediction model sets that meet the requirements are screened out. The real-time characteristic data of the target tunnel project are obtained, and the data are processed through the preset prediction model set to obtain the corresponding hierarchical prediction result set. The final hierarchical prediction result is obtained, and the corresponding risk level is obtained according to the final hierarchical prediction result. The corresponding plan is responded to, thereby realizing the technology of deep learning hierarchical prediction based on large deformation of surrounding rock of hard rock tunnels.

[0046] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0047] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed across multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0048] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0049] Those skilled in the art will appreciate that all or part of the steps of the above-mentioned method embodiments may be implemented by hardware related to program instructions, and the aforementioned program may be stored in a readable storage medium. When the program is executed, the program executes the steps of the above-mentioned method embodiments. The aforementioned storage medium includes various media that can store program codes, such as mobile storage devices, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0050] Alternatively, if the integrated units described above are implemented as software functional modules and sold or used as standalone products, they can also be stored on a readable storage medium. Based on this understanding, the technical solutions of the embodiments of the present invention, or the portion that contributes to the prior art, can be embodied in the form of a software product. This software product, stored on a storage medium, includes instructions for enabling a computer device (such as a personal computer, server, or network device) to execute all or part of the methods described in the various embodiments of the present invention. The aforementioned storage media include various media capable of storing program code, such as removable storage devices, ROM, RAM, magnetic disks, or optical disks.

Claims

1. A deep learning hierarchical prediction method for large deformation of surrounding rock in hard rock tunnels, characterized by: The following steps are involved: Collect characteristic data of the target tunnel project within a preset time period and process it to obtain multiple sets of identical test data sets; Processing the test data set through multiple different preset prediction models to obtain corresponding multi-group hierarchical prediction results; Comparing the multiple groups of graded prediction results with the actual graded prediction results, respectively, to obtain corresponding multiple groups of prediction deviation rate data, and screening out a preset prediction model set that meets the requirements; Acquire real-time characteristic data of the target tunnel project, process it through the preset prediction model set, obtain a corresponding hierarchical prediction result set, and process it to obtain a final hierarchical prediction result; Obtain the corresponding risk level based on the final graded prediction result and respond to the corresponding plan.

2. The deep learning hierarchical prediction method based on large deformation of surrounding rock of hard rock tunnel according to claim 1 is characterized in that: The collecting characteristic data of the target tunnel project within a preset time period and processing to obtain multiple sets of identical test data sets includes: Collect characteristic data of the target tunnel project within a preset time period, including geological survey data, tunnel design data, construction data, and on-site monitoring data, and process them to obtain multiple sets of identical test data sets; The geological survey data include rock type, strength, elastic modulus and ground stress; The tunnel design data includes burial depth, cross-sectional dimensions and shape; The construction data includes excavation method, support time and support strength; The on-site monitoring data includes displacement and strain.

3. The deep learning hierarchical prediction method based on large deformation of surrounding rock of hard rock tunnel according to claim 2 is characterized in that: The processing of the test data set through a plurality of different preset prediction models to obtain corresponding multi-group hierarchical prediction results includes: Processing the test data set through multiple different preset prediction models to obtain corresponding multi-group hierarchical prediction results; The preset prediction models include a convolutional neural network prediction model, a recurrent neural network prediction model, a long short-term memory network prediction model, a gated recurrent unit prediction model and a random forest prediction model.

4. The deep learning hierarchical prediction method based on large deformation of surrounding rock of hard rock tunnel according to claim 3 is characterized in that: The step of comparing the multiple groups of hierarchical prediction results with the actual hierarchical prediction results to obtain corresponding multiple groups of prediction deviation rate data and screening out a preset prediction model set that meets the requirements includes: Obtaining actual graded prediction results of the target tunnel project within the preset time period, and comparing them with the multiple groups of graded prediction results to obtain corresponding multiple groups of prediction deviation rate data; Comparing the plurality of sets of predicted deviation rate data with a first preset threshold and a second preset threshold respectively; If the prediction deviation rate data is greater than the first preset threshold and less than the second preset threshold, the corresponding graded prediction result meets the requirements; A preset prediction model set that meets the requirements is obtained based on all comparison results.

5. The deep learning hierarchical prediction method based on large deformation of surrounding rock of hard rock tunnel according to claim 4 is characterized in that: The acquiring of real-time characteristic data of the target tunnel project, processing the data through the preset prediction model set to obtain a corresponding hierarchical prediction result set, and processing the data to obtain a final hierarchical prediction result includes: monitoring the target tunnel project and acquiring real-time characteristic data, including real-time geological survey data, real-time tunnel design data, real-time construction data, and real-time field monitoring data; Processing the real-time characteristic data through the preset prediction model set to obtain a corresponding hierarchical prediction result set; A weighted process is performed on the hierarchical prediction result set to obtain a final hierarchical prediction result.

6. The deep learning hierarchical prediction method based on large deformation of surrounding rock of hard rock tunnel according to claim 5 is characterized in that: Obtaining the corresponding risk level according to the final classification prediction result and responding to the corresponding plan includes: Compare the final classification prediction result with the preset classification prediction threshold, obtain the corresponding risk level according to the comparison result, and respond to the corresponding plan; If the comparison result is less than the first preset classification prediction threshold, the corresponding risk level is Level I, and the corresponding emergency plan is a Level I emergency plan, which includes enhanced monitoring and regular inspections; If the comparison result is greater than or equal to the first preset classification prediction threshold and less than the second preset classification prediction threshold, the corresponding risk level is Level II, and the corresponding emergency plan is a Level II emergency plan, which includes suspension of construction, reinforcement and support, and cause analysis; If the comparison result is greater than or equal to the second preset classification prediction threshold, the corresponding risk level is Level III, and the corresponding plan is a Level III plan, which includes immediate evacuation, emergency reinforcement, and the formulation of special treatment plans.

7. A deep learning hierarchical prediction system based on large deformation of surrounding rock in hard rock tunnels is characterized by: The system includes: a memory and a processor, wherein the memory includes a program of a deep learning hierarchical prediction method based on large deformation of surrounding rock of hard rock tunnels, and when the program of the deep learning hierarchical prediction method based on large deformation of surrounding rock of hard rock tunnels is executed by the processor, the following steps are implemented: Collect characteristic data of the target tunnel project within a preset time period and process it to obtain multiple sets of identical test data sets; Processing the test data set through multiple different preset prediction models to obtain corresponding multi-group hierarchical prediction results; Comparing the multiple groups of graded prediction results with the actual graded prediction results, respectively, to obtain corresponding multiple groups of prediction deviation rate data, and screening out a preset prediction model set that meets the requirements; Acquire real-time characteristic data of the target tunnel project, process it through the preset prediction model set, obtain a corresponding hierarchical prediction result set, and process it to obtain a final hierarchical prediction result; Obtain the corresponding risk level based on the final graded prediction result and respond to the corresponding plan.

8. The deep learning hierarchical prediction system based on large deformation of surrounding rock of hard rock tunnel according to claim 7 is characterized in that: The collecting characteristic data of the target tunnel project within a preset time period and processing to obtain multiple sets of identical test data sets includes: Collect characteristic data of the target tunnel project within a preset time period, including geological survey data, tunnel design data, construction data, and on-site monitoring data, and process them to obtain multiple sets of identical test data sets; The geological survey data include rock type, strength, elastic modulus and ground stress; The tunnel design data includes burial depth, cross-sectional dimensions and shape; The construction data includes excavation method, support time and support strength; The on-site monitoring data includes displacement and strain.

9. The deep learning hierarchical prediction system based on large deformation of surrounding rock of hard rock tunnel according to claim 8 is characterized in that: The processing of the test data set through a plurality of different preset prediction models to obtain corresponding multi-group hierarchical prediction results includes: Processing the test data set through multiple different preset prediction models to obtain corresponding multi-group hierarchical prediction results; The preset prediction models include a convolutional neural network prediction model, a recurrent neural network prediction model, a long short-term memory network prediction model, a gated recurrent unit prediction model and a random forest prediction model.

10. The deep learning hierarchical prediction system based on large deformation of surrounding rock of hard rock tunnel according to claim 9 is characterized in that: The step of comparing the multiple groups of hierarchical prediction results with the actual hierarchical prediction results to obtain corresponding multiple groups of prediction deviation rate data and screening out a preset prediction model set that meets the requirements includes: Obtaining actual graded prediction results of the target tunnel project within the preset time period, and comparing them with the multiple groups of graded prediction results to obtain corresponding multiple groups of prediction deviation rate data; Comparing the plurality of sets of predicted deviation rate data with a first preset threshold and a second preset threshold respectively; If the prediction deviation rate data is greater than the first preset threshold and less than the second preset threshold, the corresponding graded prediction result meets the requirements; A preset prediction model set that meets the requirements is obtained based on all comparison results.

Images