Intelligent tunnel construction system and method based on knowledge enhancement large model

Abstract

The invention provides a tunnel intelligent construction system and method based on a knowledge enhancement large model, and relates to the technical field of artificial intelligence. The system comprises a data layer (a data acquisition module), a knowledge layer (a knowledge base construction module), a decision-making layer (a decision-making module and an evaluation module) and an application layer (a classification module and an intelligent question and answer module), the knowledge layer constructs a knowledge base and a knowledge graph based on professional specification documents and patent documents, and enhances the generalization ability and reasoning ability of a general large model in the tunnel engineering field; the decision-making layer designs a support decision-making method jointly driven by data and knowledge, and constructs a safety risk grading evaluation system to perform safety evaluation on the support scheme; according to the application layer, an external knowledge base is used for enhancing general large model identification event types, working modules corresponding to different events are called respectively, and the functions of integrating excavation support decision making, safety evaluation, tunnel intelligent question answering and the like are achieved. Therefore, according to the embodiment of the invention, by introducing the knowledge enhancement large model, the utilization rate of the multi-source heterogeneous data of the tunnel is improved, and efficient and safe decision support is provided for intelligent construction of the tunnel.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a tunnel intelligent construction system and method based on a knowledge-enhanced large model. Background Art

[0002] With the rapid economic development and continuous advancement of infrastructure construction, tunnel engineering has occupied an important position in high-speed railway and highway construction. Tunnel construction faces challenges of greater burial depth, higher ground stress, and more complex geological conditions. At present, the mechanization and intelligent level of tunnel construction is still relatively low, making it difficult to meet the higher requirements for safety, efficiency, and quality control under complex geological conditions. Therefore, promoting the intelligent transformation of tunnel construction has become an urgent need for the industry's development. Summary of the Invention

[0003] The present invention provides a tunnel intelligent construction system and method based on a knowledge-enhanced large model to at least partially solve the above problems.

[0004] In a first aspect of the present invention, a tunnel intelligent construction system based on a knowledge-enhanced large model is provided. The system includes: A data acquisition module for collecting historical displacement and stress monitoring data of the cross-section of the target tunnel, historical excavation data, expert experience knowledge, and tunnel safety specification documents through data monitoring sensors buried at the tunnel construction site; A knowledge base construction module for constructing a knowledge graph according to the expert experience knowledge, constructing a vector database according to the tunnel safety specification documents, and constructing an external knowledge base according to the knowledge graph and the vector database; A classification module for enhancing the general large model to identify event types through the external knowledge base; in the case where the event type is a support decision request, calling the decision module, in the case where the event type is a support decision request, calling the decision module, in the case where the event type is an adjustment request, calling the intelligent question and answer module, in the case where the event type is an abnormal monitoring request, calling the evaluation module, and in the case where the event type is a geological risk, calling the evaluation module; A decision module for training a teacher model based on expert experience knowledge, training a student model through the teacher model and knowledge distillation technology based on historical excavation and support data to obtain a support decision model; inputting the historical displacement and stress monitoring data of the cross-section of the target tunnel into the support decision model to obtain a support plan; An intelligent question and answer module for processing the text information in the form of natural language sent by the user and the support plan to obtain an adjusted support plan; An evaluation module for evaluating the safety level of the target tunnel based on the historical displacement and stress monitoring data of the cross-section of the target tunnel.

[0005] Optionally, the system further includes: A key parameter prediction module for predicting the surrounding rock deformation parameters, axial force parameters, and bending moment parameters of the target tunnel after applying the support plan based on a key parameter prediction model; An evaluation module is further configured to evaluate the support plan based on the surrounding rock deformation parameters, axial force parameters, and bending moment parameters, the cross-section historical displacement and stress monitoring data of the target tunnel, and the support plan to obtain the safety risk level of the support plan; An optimization module for determining an optimization plan for the support decision model according to the safety risk level of the support plan.

[0006] Optionally, the evaluation module is configured to: Determine the ultimate displacement corresponding to the surrounding rock grade of the target tunnel according to the excavation data of the target tunnel, determine the crown displacement, arch waist displacement, and arch bottom displacement of the target tunnel according to the surrounding rock deformation parameters, and determine the surrounding rock displacement safety factor according to the ultimate displacement, crown displacement, arch waist displacement, and arch bottom displacement; Determine the steel arch area according to the steel arch grade parameters in the support plan, determine the steel arch stress according to the steel arch axial force in the axial force parameters and the steel arch area, and determine the safety factor of the first support structure according to the steel arch stress and the ultimate stress of the steel arch; Determine the compressive strength limit of the shotcrete according to the primary support grade parameters in the support plan, and determine the safety factor of the second support structure according to the minimum principal stress at the crown in the bending moment parameters and the compressive strength limit; Determine the compressive strength limit of the concrete according to the secondary lining grade parameters in the support plan, and determine the safety factor of the third support structure according to the axial forces in five directions of the tunnel in the axial force parameters, the secondary lining thickness parameters in the support plan, and the compressive strength limit of the concrete; Determine the safety index of the support plan according to the surrounding rock displacement safety factor, the safety factor of the first support structure, the safety factor of the second support structure, and the safety factor of the third support structure; Determine the safety risk level of the support plan according to the safety index.

[0007] Optionally, the optimization module is configured to: Determine a full-parameter optimization plan for the support decision model when the safety risk level reaches the first pre-level; Determine a full-parameter optimization plan for the support decision model when the safety index of the support plan continuously decreases by more than the index preset threshold for N times; When the model prediction deviation rate of the support decision-making model is greater than the preset deviation threshold, a full-parameter optimization plan for the support decision-making model is determined.

[0008] Optionally, the data acquisition module is further configured to obtain expert experience knowledge in real time based on the knowledge channel, dynamically update the external knowledge base, and obtain the cross-section historical displacement and stress monitoring data of the target tunnel in real time based on the data channel.

[0009] Optionally, the system further includes: A drawing generation module, configured to generate construction drawings that meet industry standards based on the support plan.

[0010] A second aspect of the present invention provides a tunnel intelligent construction method based on a knowledge-enhanced large model, and the method includes: Collecting the cross-section historical displacement and stress monitoring data of the target tunnel, collecting historical excavation data, collecting expert experience knowledge, and collecting tunnel safety specification documents through data monitoring sensors buried at the tunnel construction site; Constructing a knowledge graph according to the expert experience knowledge, constructing a vector database according to the tunnel safety specification documents, constructing an external knowledge base according to the knowledge graph and the vector database, and obtaining an externally knowledge-base-enhanced general large model; Identifying the event type through the externally knowledge-base-enhanced general large model; When the event type is a support decision request, processing the cross-section historical displacement and stress monitoring data of the target tunnel through the support decision-making model to obtain a support plan, where the support decision-making model is trained for the student model through a teacher model and knowledge distillation technology based on historical excavation support data, and the teacher model is trained based on expert experience knowledge; When the event type is an adjustment request, processing the text information in natural language sent by the user and the support plan to obtain an adjusted support plan; When the event type is an abnormal monitoring request and geological risk, evaluating the safety level of the target tunnel based on the cross-section historical displacement and stress monitoring data of the target tunnel. A third aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes, it implements the tunnel intelligent construction system based on the knowledge-enhanced large model as described in the first aspect of the present invention.

[0011] A fourth aspect of the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the tunnel intelligent construction system based on the knowledge-enhanced large model as described in the first aspect of the present invention.

[0012] In a fifth aspect of the present invention, there is provided a computer program product, including a computer program / instructions, which are implemented by a processor to perform the steps in the tunnel intelligent construction system based on a knowledge-enhanced large model as described in the first aspect of the present invention.

[0013] The tunnel intelligent construction system based on a knowledge-enhanced large model provided by an embodiment of the present invention includes a data layer (including a data acquisition module), a knowledge layer (including a knowledge base construction module), a decision-making layer (including a decision-making module and an evaluation module), and an application layer (including a classification module and an intelligent Q&A module). Among them, the data layer cleans and transforms multi-source heterogeneous data of the tunnel; the knowledge layer constructs a knowledge base and a knowledge graph based on professional specification documents and patent documents to enhance the generalization ability and reasoning ability of the general large model in the field of tunnel engineering; the decision-making layer designs a support decision-making method jointly driven by data and knowledge, and constructs a safety risk grading evaluation system to conduct safety evaluation on the support plan; the application layer enhances the general large model by an external knowledge base to identify event types, and respectively calls working modules corresponding to different events to implement functions such as integrated excavation support decision-making, safety evaluation, and tunnel intelligent Q&A. Thus, by introducing a knowledge-enhanced large model, the embodiment of the present invention helps to improve the utilization rate of multi-source heterogeneous data of the tunnel and provides efficient and safe decision-making support for tunnel intelligent construction. Description of the Drawings

[0014] In order to more clearly illustrate the technical solutions of the present invention, the drawings required for the description of the present invention will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings without creative efforts based on these drawings.

[0015] Figure 1 is a structural block diagram of the tunnel intelligent construction system based on a knowledge-enhanced large model provided by the present invention; Figure 2 is a system hierarchy diagram of the tunnel intelligent construction system based on a knowledge-enhanced large model provided by the present invention; Figure 3 is a step flow chart of the tunnel intelligent construction method based on a knowledge-enhanced large model provided by the present invention. Detailed Embodiments

[0016] To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below with reference to the drawings and specific embodiments.

[0017] The core challenges in intelligent tunnel construction lie in how to effectively integrate and analyze multi-source heterogeneous data to achieve accurate geological environment exploration, surrounding rock classification, support design, and blasting parameter optimization. At the same time, it is also necessary to solve the problems of real-time perception and dynamic regulation of the construction state of intelligent equipment, and build a collaborative management and control and visual remote control system of "ground-tunnel-machine-communication-human". These technical difficulties need to be solved through advanced information processing, artificial intelligence methods, and multi-professional collaboration. In recent years, artificial intelligence technology has been gradually applied in the field of tunnel engineering, such as the surrounding rock classification method based on deep learning and the intelligent support design algorithm, which provide new ideas for solving the above problems. However, the current methods are mostly based on single data-driven, and their performance highly depends on the training data set, making it difficult to cope with the complexity and diversity of geological conditions, resulting in insufficient generalization ability.

[0018] The rapid development of large language models (LLMs) has brought new opportunities to intelligent tunnel construction. Since the Transformer architecture was proposed, large models represented by GPT have made breakthrough progress in the field of natural language processing. Especially the emergence of technologies such as Retrieval-Augmented Generation (RAG) and LangChain has further improved the ability of large models to process complex queries and multi-modal data. These technologies make it possible to efficiently utilize multi-source heterogeneous data in tunnel engineering, such as realizing functions like intelligent design and optimization, risk warning, and intelligent question answering through the semantic understanding ability of large models. However, the current application of large models in tunnel engineering is still in the exploration stage, especially in the aspect of intelligent decision-making for the whole process and all elements of tunnel construction combined with empirical knowledge, which still needs further in-depth research.

[0019] Although certain progress has been made in the field of intelligent tunnel construction, there are still significant gaps in intelligent decision-making combined with empirical knowledge and the application of large models. The existing methods have limitations in generalization ability and adaptability, making it difficult to meet the requirements of tunnel construction under complex geological conditions. Therefore, inventing a tunnel intelligent construction decision-making system and method based on knowledge-enhanced large models has important theoretical value and practical significance, and can provide a new solution for the intelligent transformation of tunnel engineering.

[0020] Specifically, as Figure 1 shown, it shows the structural block diagram of a tunnel intelligent construction system based on a knowledge-enhanced large model provided by an embodiment of the present invention. The system 100 includes: A data acquisition module 101, configured to collect the cross-section historical displacement and stress monitoring data of the target tunnel, collect historical excavation data, collect expert experience knowledge, and collect tunnel safety specification documents through data monitoring sensors buried at the tunnel construction site.

[0021] In the embodiments of the present invention, data monitoring sensors can be buried at the tunnel construction site to collect the cross-section historical displacement and stress monitoring data of the target tunnel in real time. Specifically, the data monitoring sensors include XYJ-4 pressure cells, XJH-2 strain gauges, and CL-XZ-B steel bar gauges, and data monitoring is carried out in 8 directions including the crown, left arch shoulder, right arch shoulder, left arch waist, right arch waist, left arch foot, right arch foot, and invert.

[0022] In the embodiments of the present invention, the cross-section historical displacement and stress monitoring data include: the stress data and displacement data in 8 directions of the crown, left arch shoulder, right arch shoulder, left arch waist, right arch waist, left arch foot, right arch foot, and invert, as well as the surrounding rock geological environment data of the target tunnel.

[0023] In the embodiments of the present invention, the historical excavation data includes the surrounding rock geological environment of the historically excavated tunnel (hardness, integrity, rock layer thickness, underground water volume, burial depth level, geological structure, inner contour form + burial depth, density, elastic modulus, Poisson's ratio, cohesion, internal friction angle, tensile strength, vertical stress, horizontal stress, longitudinal stress) and the support scheme parameters of the historically excavated tunnel (anchor bolts, steel arch frames, secondary lining, advanced support, shotcrete, steel mesh).

[0024] In the embodiments of the present invention, the expert experience knowledge records the support parameters adopted in different construction environments, and the construction environments include hardness, integrity, rock layer thickness, underground water volume, burial depth level, geological structure, and inner contour form; the support parameters include shotcrete parameters, steel mesh parameters, anchor bolt parameters, steel arch frame parameters, secondary lining parameters, and advanced support parameters.

[0025] In the embodiments of the present invention, the tunnel safety specification documents include professional specification documents, patent documents and other files.

[0026] The tunnel intelligent construction system based on the knowledge-enhanced large model provided by the embodiments of the present invention can be divided into a data layer, a knowledge layer, a decision-making layer, and an application layer according to the hierarchy. Specifically, Figure 2 The system hierarchy diagram of the tunnel intelligent construction system based on the knowledge-enhanced large model provided by the embodiments of the present invention is shown. The data layer is used to collect multi-source heterogeneous data of the tunnel (including the cross-section historical displacement and stress monitoring data, historical excavation data, expert experience knowledge, and tunnel safety specification documents) of the target tunnel, and preprocess, manage, and update the multi-source heterogeneous data of the tunnel.

[0027] Specifically, the data collection module is used to collect data, and in addition, the preprocessing module can preprocess the data, including: data cleaning, data conversion, etc., specifically including processing methods such as missing value filling, outlier removal, data type conversion, and structured storage.

[0028] The knowledge base construction module 102 is used to construct a knowledge graph according to the expert experience knowledge, construct a vector database according to the tunnel safety specification document, and construct an external knowledge base based on the knowledge graph and the vector database, so as to obtain an external knowledge base enhanced general large model.

[0029] In the embodiment of the present invention, entity recognition and relationship extraction can be performed on the expert experience knowledge collected in the data layer through a large language model to obtain a knowledge graph. The tunnel safety specification document collected in the data layer is segmented through an embedding model to realize vectorized storage and obtain a vector database.

[0030] In the embodiment of the present invention, the knowledge base construction module is located in the knowledge layer and is used to organize the data collected in the data layer to form an external knowledge base.

[0031] In the embodiment of the present invention, an external knowledge base enhanced general large model is constructed based on the above external knowledge base to improve the generalization ability and reasoning ability of the general large model in the field of tunnel engineering. Subsequently, event classification and intelligent question answering can be realized based on the external knowledge base enhanced general large model in the application layer.

[0032] The classification module 103 is used to identify the event type through the external knowledge base enhanced general large model; in the case where the event type is a support decision request, the decision module is called, in the case where the event type is an adjustment request, the intelligent question answering module is called, in the case where the event type is an abnormal monitoring request, the evaluation module is called, and in the case where the event type is a geological risk, the evaluation module is called.

[0033] In the embodiment of the present invention, as Figure 2 shown, in the application layer, through the classification module, based on the large model agent (the above external knowledge base enhanced general large model), the event type is identified through the event classifier, and the response process is generated based on the workflow engine, the corresponding response module is scheduled, and the scheduling priority is dynamically allocated according to the event urgency.

[0034] In the embodiment of the present invention, in the application layer, based on a multi-agent collaborative architecture, multiple downstream applications are integrated, including excavation support decision-making (outputting a support plan through the decision module, generating specific construction parameters and controlling construction machinery), safety assessment (performing dynamic risk assessment through the evaluation module), standardized drawing generation (automatically generating construction drawings that meet industry standards based on the support plan parameters through the drawing generation module), tunnel intelligent question answering (providing construction specification query and abnormal working condition diagnosis services through the intelligent question answering module based on the large model enhanced by the knowledge graph and the vector database), and data knowledge management (realizing the incremental update of monitoring data and the knowledge base through the data acquisition module), etc.

[0035] In the embodiment of the present invention, event classification includes the following types: (1)Support decision request: This event is triggered when the surrounding rock grading and surrounding rock environment parameters are measured at the excavation face of the target tunnel, and the support decision module is called to generate support parameters and construction parameters.

[0036] (2)Adjustment request: This event is triggered when construction personnel submit a requirement for modifying the plan through the interactive terminal. The intelligent Q&A module is called to provide specification support, and at the same time, the evaluation module can be started to verify the feasibility of the method plan submitted by the user.

[0037] (3)Abnormal monitoring request: This event is triggered when the data monitoring sensor alarms when detecting a sudden stress change or the surrounding rock deformation exceeding the limit. Based on this event, the data acquisition module is triggered to update the abnormal data, and the evaluation module is called to start a secondary evaluation based on the cross-sectional historical displacement and stress monitoring data of the target tunnel collected in real time.

[0038] (4)Geological risk: This event is activated when the advanced geological prediction system identifies geological risks such as hidden structures or abnormal water inrush. The evaluation module is preferentially scheduled to conduct a safety level assessment based on the cross-sectional historical displacement and stress monitoring data of the target tunnel collected in real time, predict the risk, and simultaneously start the support decision module to generate an emergency plan based on the cross-sectional historical displacement and stress monitoring data of the target tunnel collected in real time.

[0039] In the embodiments of the present invention, all event processing results are transmitted back to the knowledge layer through the data acquisition module, driving the large model to update the decision logic through incremental learning, forming a closed-loop control mechanism of "event trigger - collaborative response - knowledge iteration".

[0040] The decision module 104 is used to train a teacher model based on expert experience knowledge, train a student model based on historical excavation support data through the teacher model and knowledge distillation technology to obtain a support decision model; process the cross-sectional historical displacement and stress monitoring data of the target tunnel through the support decision model to obtain a support plan.

[0041] The embodiments of the present invention propose a data-knowledge jointly driven support decision method, which conducts knowledge distillation based on expert experience knowledge, combines historical excavation support data to train a support decision model, and improves the dynamic decision-making ability of the model in complex geological environments.

[0042] In the embodiments of the present invention, the decision module is located in the decision-making layer. As Figure 2 shown, the data-knowledge jointly driven support decision method proposed by the decision-making layer is a method of training a teacher model based on expert experience knowledge and guiding the student model to update parameters through knowledge distillation technology. The student model is trained by historical excavation support data and learns the expert experience knowledge distilled by the teacher model.

[0043] The intelligent Q&A module 105 is used to process the text information in the form of natural language sent by the user and the support plan to obtain an adjusted support plan.

[0044] In the embodiment of the present invention, the intelligent Q&A module processes the text information sent by the user based on an external knowledge base enhanced general large model, and obtains an adjusted support plan according to safety specification knowledge, and can also provide users with construction specification query and abnormal working condition diagnosis services.

[0045] The evaluation module 106 is used to evaluate the safety level of the target tunnel based on the cross-section historical displacement and stress monitoring data of the target tunnel.

[0046] In the embodiment of the present invention, the evaluation module can evaluate the safety level of the target tunnel based on the cross-section historical displacement and stress monitoring data of the target tunnel monitored by sensors and the tunnel excavation data of the target tunnel.

[0047] Specifically, the system further includes: The key parameter prediction module is used to predict the surrounding rock deformation parameters, axial force parameters and bending moment parameters of the target tunnel after applying the support plan based on the key parameter prediction model.

[0048] In the embodiment of the present invention, the key parameter prediction model adopts a multi-model integrated learning architecture to predict the convergence values of the surrounding rock deformation, axial force and bending moment generated by applying this plan.

[0049] In the embodiment of the present invention, as Figure 2 shown, the key parameter prediction module and the evaluation module are also located in the decision-making layer, and are used to perform safety evaluation on the current support plan (including: the to-be-applied support plan output by the support decision model, or the actually applied support plan, or the support plan adjusted by the user).

[0050] In the embodiment of the present invention, the multi-model integrated learning architecture adopts machine learning models such as KRR (Kernel Ridge Regression), Enet (Elastic Net), XGBoost and GBoost (Gradient Boosting) as base models, and further introduces LASSO (Least Absolute Shrinkage and Selection Operator) as a meta-model to integrate the features output by the base models, re-evaluate the relationship between the prediction and the target variable at a higher level, and perform final parameter prediction.

[0051] An evaluation module is further configured to evaluate the support plan based on the surrounding rock deformation parameters, axial force parameters, bending moment parameters, historical displacement and stress monitoring data of the cross-section of the target tunnel, and the support plan, so as to obtain the safety risk level of the support plan.

[0052] In the embodiment of the present invention, the evaluation module can obtain the safety risk level of the support plan based on the actual data of the current tunnel and the predicted surrounding rock deformation parameters, axial force parameters, and bending moment parameters (or the actual surrounding rock deformation parameters, axial force parameters, and bending moment parameters after applying the current support plan).

[0053] Specifically, the evaluation module is configured to: Determine the ultimate displacement corresponding to the surrounding rock grade of the target tunnel according to the excavation data of the target tunnel, determine the crown displacement, waist displacement, and invert displacement of the target tunnel according to the surrounding rock deformation parameters, and determine the surrounding rock displacement safety factor according to the ultimate displacement, crown displacement, waist displacement, and invert displacement; Determine the area of the steel arch according to the steel arch grade parameters in the support plan, determine the stress of the steel arch according to the axial force of the steel arch in the axial force parameters and the area of the steel arch, and determine the safety factor of the first support structure according to the stress of the steel arch and the ultimate stress of the steel arch; Determine the ultimate compressive strength of the shotcrete according to the primary support grade parameters in the support plan, and determine the safety factor of the second support structure according to the minimum principal stress at the crown in the bending moment parameters and the ultimate compressive strength; Determine the ultimate compressive strength of the concrete according to the secondary lining grade parameters in the support plan, and determine the safety factor of the third support structure according to the axial forces in five directions of the tunnel in the axial force parameters, the secondary lining thickness parameters in the support plan, and the ultimate compressive strength of the concrete; Determine the safety index of the support plan according to the surrounding rock displacement safety factor, the safety factor of the first support structure, the safety factor of the second support structure, and the safety factor of the third support structure; Determine the safety risk level of the support plan according to the safety index.

[0054] In the embodiment of the present invention, four safety factors are calculated, and the safety performance of the plan is analyzed through an evaluation function. The safety factors include the surrounding rock displacement safety factor , the safety factor of the first support structure , the safety factor of the second support structure , the safety factor of the third support structure , and their calculation methods are as follows: ; ; ; ; In the formula: is the ultimate displacement; is the average value of the convergence values of the vault displacement, the waist displacement, and the invert displacement; is the ultimate stress of the steel arch frame; is the stress of the steel arch frame; is the ultimate compressive strength of the shotcrete; is the convergence value of the minimum principal stress at the vault; is the ultimate compressive strength of the concrete; is the thickness of the secondary lining in the support plan; - are the convergence values of the axial forces in the five directions of the tunnel vault, arch shoulder, arch waist, arch foot, and invert, respectively.

[0055] Specifically, in the calculation process of the surrounding rock displacement safety factor , the calculation method of the ultimate displacement is as follows: First, the surrounding rock grade can be deduced from the elastic modulus of the target tunnel, and the corresponding relationship is shown in Table 1.

[0056] Table 1 Corresponding relationship between surrounding rock grades and elastic modulus

[0057] For the surrounding rock with an elastic modulus between 1.3 and 6, it is Class IV surrounding rock; for that between 1 and 1.3, it is Class V surrounding rock. For other grades of surrounding rock, K1 does not need to be calculated. To obtain the ultimate displacement, the buried depth also needs to be calculated, and its calculation formula is: . Through the buried depth and the surrounding rock grade, the ultimate displacements of the surrounding rock at different grades can be obtained, as shown in Table 2.

[0058] Table 2 Ultimate displacements of the surrounding rock at different grades

[0059] In the calculation process of the first support structure safety factor , the parameter of the steel arch frame grade can be taken out from the support plan. If it is empty or 0, it means that there is no steel arch frame in this plan, that is, K2 does not need to be calculated. If the steel arch frame grade has a definite value, the area of the steel arch frame needs to be deduced, and their corresponding relationship is shown in Table 3.

[0060] Table 3 Areas of steel arch frames of different grades

[0061] After that, the predicted stress of the steel arch frame is obtained by dividing the axial force of the steel arch frame predicted by the key parameter model by the area of the steel arch frame. The ultimate stress of the steel arch frame is a fixed value of 300e6, denoted as Calculate K2 from the predicted stress of the steel arch and the ultimate stress of the steel arch.

[0062] In the calculation process of the safety factor of the second support structure Take a parameter of the primary support grade from the support plan. Different primary support grades correspond to different compressive strength limits of the shotcrete, denoted as , and the corresponding relationship is shown in Table 4.

[0063] Table 4 Tensile strength limits of different primary support grades

[0064] After that, obtain K3 from the minimum principal stress at the crown (denoted as ) predicted by the key parameter prediction model and the tensile strength limit.

[0065] In the calculation process of the safety factor of the third support structure First, take out the secondary lining grade and the secondary lining thickness from the support plan. The secondary lining grade is used to back-calculate the compressive strength limit of the concrete (denoted as ), and the corresponding relationship is shown in Table 5.

[0066] Table 5 Compressive strength limits of concrete for different secondary lining grades

[0067] After that, calculate K4 based on the axial forces in five directions (crown, shoulder, waist, foot, and bottom of the arch, denoted as ), the secondary lining thickness (denoted as ), and the compressive strength limit of the concrete predicted by the model.

[0068] In the embodiments of the present invention, the evaluation function is as follows:

[0069] In the formula: is the magnitude of the i-th safety factor; is the value of the evaluation function corresponding to the safety factor ; when the safety factor of the support structure approaches 1.5, the evaluation function takes the maximum value. On the left side of the maximum value is equal to 1.0, the evaluation function is 0, indicating that the support structure is about to undergo instability failure at this time. A too large safety factor will cause material waste and increase the construction cost. As the safety factor continues to increase, the tunnel design becomes more conservative, and the evaluation function decreases accordingly.

[0070] is the sum of the evaluation function values of all safety factors and represents the construction safety index. The larger this value is, the safer the support structure is. The calculation formula is as follows:

[0071] Based on the value of the construction safety index, the safety risk of the support structure of the plan is graded, and the grading standard is shown in Table 6 below: Table 6. Safety grading standard

[0072] In the embodiment of the present invention, the system further includes: an optimization module, configured to determine an optimization plan for the support decision model according to the safety risk level of the support plan.

[0073] In the embodiment of the present invention, aiming at the safety problem of the support plan, a safety risk grading evaluation system is constructed to evaluate the support plan, and aiming at the uncertainty in tunnel construction, based on the incremental learning technology, a dynamic feedback tuning mechanism is constructed to continuously iteratively optimize the decision-making effect.

[0074] Specifically, the optimization module is configured to: In the case where the safety risk level reaches the first pre-level, determine a full-parameter optimization plan for the support decision model.

[0075] In the case where the safety index of the support plan continuously decreases by more than the index preset threshold for N times, determine a full-parameter optimization plan for the support decision model.

[0076] When the model prediction deviation rate of the support decision model is greater than the deviation preset threshold, determine a full-parameter optimization plan for the support decision model.

[0077] Specifically, in the embodiment of the present invention, three-level optimization trigger conditions are set: First-level trigger (emergency optimization): When the safety risk level reaches level IV, forcefully start the full-parameter optimization of the support decision model.

[0078] Second-level trigger (conventional optimization): Start the full-parameter optimization of the support decision model when the safety index of the support plan continuously decreases by more than 10% for 3 consecutive times.

[0079] Third-level trigger (preventive optimization): Activate the full-parameter optimization of the support decision model when the model prediction deviation rate ε≥15%.

[0080] In the embodiment of the present invention, the full-parameter optimization plan refers to: optimizing the parameters of the decision model (including the teacher model and the student model), specifically based on the incremental data, performing secondary training on the model, and optimizing the model parameters through neural network backpropagation.

[0081] In the embodiments of the present invention, a model version management module may also be constructed to retain the optimal model in the most recent 7 days as the baseline version, and automatically switch to the optimal historical version when the performance of the latest model decays.

[0082] In the embodiments of the present invention, short-term verification may also be carried out. Compare the predicted values of the optimized model with the measured data of the next construction cycle, allowing a deviation ≤ 10%, and carry out long-term verification to analyze the economic indicators of 5 consecutive construction cycles (material loss rate ≤ 5%). Implement intelligent fusing rules. When the safety index does not improve after 2 consecutive optimizations, roll back to the baseline version; if a single optimization causes the safety index to drop by 10%, freeze the model parameters and initiate an artificial review process.

[0083] Specifically, in the embodiments of the present invention, the data acquisition module is further configured to obtain expert experience knowledge in real time based on the knowledge channel, dynamically update the external knowledge base, and obtain the historical displacement and stress monitoring data of the cross-section of the target tunnel in real time based on the data channel.

[0084] In the embodiments of the present invention, the knowledge channel monitors expert experience knowledge and dynamically updates the knowledge base; the data channel accesses the monitoring data stream of the data layer in real time and extracts the latest construction samples based on a sliding time window. When optimization is triggered, full-parameter optimization of the support decision-making model is performed based on the latest data obtained from the knowledge channel and the data channel.

[0085] Specifically, the system further includes: A drawing generation module, configured to generate construction drawings that meet industry standards based on the support plan. In the embodiments of the present invention, for the support plan that passes the safety assessment, based on the standard drawing generation algorithm, excavation construction method drawings, support design drawings are generated, and the parameters required for the construction of the bolt trolley, wet shotcrete trolley, and arch trolley are generated.

[0086] By introducing a large language model in the embodiments of the present invention, a large amount of unstructured data generated during the tunnel construction process, such as geological reports, construction logs, etc., can be efficiently processed. This not only improves the utilization rate of data, but also enhances the comprehensive data analysis ability, providing more accurate and comprehensive information support for tunnel construction.

[0087] Through the real-time inference ability of the large language model in the embodiments of the present invention, event classification is realized, enabling corresponding instant decision support to be provided for different events during the tunnel construction process. By real-time analyzing multi-modal data, the system can quickly identify potential safety hazards and construction problems, and provide optimization suggestions, thus significantly improving the efficiency and safety of tunnel construction.

[0088] Embodiments of the present invention integrate professional knowledge in the field of tunnel construction into large language models, enhancing the intelligence level and adaptability of the system. Through continuous learning and updating, the system can better handle complex and changing geological conditions and construction environments, provide more accurate and personalized solutions, and provide strong support for the high-quality development of tunnel construction.

[0089] Based on the same inventive concept, the present invention also provides a tunnel intelligent construction method based on a knowledge-enhanced large model, as Figure 3 shown, which shows a step flow chart of the tunnel intelligent construction method based on a knowledge-enhanced large model. The method includes the following steps: S301, collecting historical displacement and stress monitoring data of the cross-section of the target tunnel, collecting historical excavation data, collecting expert experience knowledge, and collecting tunnel safety specification documents through data monitoring sensors buried at the tunnel construction site; S302, constructing a knowledge graph according to the expert experience knowledge, constructing a vector database according to the tunnel safety specification documents, and constructing an external knowledge base according to the knowledge graph and the vector database to obtain an external knowledge base-enhanced general large model; S303, identifying the event type through the external knowledge base-enhanced general large model; S304, in the case where the event type is a support decision request, processing the historical displacement and stress monitoring data of the cross-section of the target tunnel through the support decision model to obtain a support plan. The support decision model is trained for the student model through a teacher model and knowledge distillation technology based on historical excavation support data, and the teacher model is trained based on expert experience knowledge; S305, in the case where the event type is an adjustment request, processing the text information in natural language form sent by the user and the support plan to obtain an adjusted support plan; S306, in the case where the event type is an abnormal monitoring request and geological risk, evaluating the safety level of the target tunnel based on the historical displacement and stress monitoring data of the cross-section of the target tunnel.

[0090] In the embodiments of the present invention, the surrounding rock parameters of the construction environment can be collected, missing values can be filled, outliers can be removed, and normalization processing can be performed, and then converted into a structured data format and stored in the data layer. Based on the preprocessed surrounding rock parameters, the knowledge base and knowledge graph in the knowledge layer are called, and combined with the support decision-making model in the decision-making layer, a preliminary support plan is generated, including an excavation plan, an advanced support plan, an initial support plan, and a secondary lining plan. Based on the key parameter prediction model, the convergence values of the surrounding rock deformation, axial force, and bending moment generated by the support plan are predicted, and the safety factor and safety index are calculated. The safety risk grading evaluation of the support plan is carried out based on the safety index; if the safety risk level exceeds the limit, the support plan is regenerated; for the support plan that passes the safety assessment, based on the standard drawing generation algorithm, the excavation construction method drawing and the support design drawing are generated, and the parameters required for the construction of the bolt trolley, wet shotcrete trolley, and arch trolley are generated. During the construction process, data such as surrounding rock deformation and support structure stress are collected in real time and compared and analyzed with the model prediction results; if the deviation exceeds the limit, all parameters of the support decision-making model are updated, the support plan is regenerated, and the model parameters are optimized based on the dynamic feedback optimization mechanism described in the decision-making layer.

[0091] Based on the same inventive concept, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes, it implements the steps in the tunnel intelligent construction method based on the knowledge-enhanced large model as described in any one of the above embodiments.

[0092] Based on the same inventive concept, the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the steps in the tunnel intelligent construction method based on the knowledge-enhanced large model as described in any one of the above embodiments.

[0093] Based on the same inventive concept, the present invention provides a computer program product, including a computer program / instructions. When the computer program / instructions are executed by a processor, it implements the steps in the tunnel intelligent construction method based on the knowledge-enhanced large model as described in any one of the above embodiments.

[0094] Each embodiment in this specification is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other.

[0095] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, apparatus, or computer program product. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) that contain computer-usable program code.

[0096] The present invention is described with reference to the flowcharts and / or block diagrams of methods, terminal devices (apparatuses), and computer program products according to the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable terminal devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable terminal devices generate means for implementing the functions specified in Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.

[0097] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable terminal devices to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.

[0098] These computer program instructions can also be loaded onto a computer or other programmable terminal devices, so that a series of operation steps are executed on the computer or other programmable terminal devices to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable terminal devices provide steps for implementing the functions specified in Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.

[0099] Although the preferred embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications to these embodiments once they know the basic creative concepts. Therefore, the appended claims are intended to be construed as including the preferred embodiments and all changes and modifications falling within the scope of the present invention.

[0100] Finally, it should also be noted that in the present invention, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or terminal device comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or elements inherent to such process, method, article or terminal device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or terminal device comprising the said element.

[0101] The above has introduced in detail a tunnel intelligent construction method based on a knowledge-enhanced large model provided by the present invention. Specific examples are used in the present invention to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A tunnel intelligent construction system based on a knowledge-enhanced large model, characterized in that The system includes: A data acquisition module, which is used to collect the cross-section historical displacement and stress monitoring data of the target tunnel through the data monitoring sensors buried at the tunnel construction site, collect the historical excavation data, collect the expert experience knowledge, and collect the tunnel safety specification documents; A knowledge base construction module, which is used to construct a knowledge graph according to the expert experience knowledge, construct a vector database according to the tunnel safety specification documents, construct an external knowledge base according to the knowledge graph and the vector database, and obtain an externally knowledge base enhanced general large model; A classification module, which is used to identify the event type through the externally knowledge base enhanced general large model; in the case that the event type is a support decision request, call the decision module, in the case that the event type is an adjustment request, call the intelligent question answering module, in the case that the event type is an abnormal monitoring request, call the evaluation module, and in the case that the event type is a geological risk, call the evaluation module; A decision module, which is used to train a teacher model based on the expert experience knowledge, train a student model through the teacher model and the knowledge distillation technology based on the historical excavation support data, and obtain a support decision model; process the cross-section historical displacement and stress monitoring data of the target tunnel through the support decision model to obtain a support plan; An intelligent question answering module, which is used to process the text information in the form of natural language sent by the user and the support plan to obtain an adjusted support plan; An evaluation module, which is used to evaluate the safety level of the target tunnel based on the cross-section historical displacement and stress monitoring data of the target tunnel.

2. The tunnel intelligent construction system based on the knowledge-enhanced large model according to claim 1, wherein The system further includes: A key parameter prediction module, which is used to predict the surrounding rock deformation parameters, axial force parameters and bending moment parameters of the target tunnel after applying the support plan based on the key parameter prediction model; The evaluation module is further used to evaluate the support plan based on the surrounding rock deformation parameters, axial force parameters and bending moment parameters, the cross-section historical displacement and stress monitoring data of the target tunnel and the support plan to obtain the safety risk level of the support plan; An optimization module, which is used to determine the optimization plan of the support decision model according to the safety risk level of the support plan.

3. The tunnel intelligent construction system based on the knowledge-enhanced large model according to claim 2, wherein The evaluation module is used for: Determine the ultimate displacement corresponding to the surrounding rock grade of the target tunnel according to the excavation data of the target tunnel, determine the crown displacement, waist displacement and invert displacement of the target tunnel according to the surrounding rock deformation parameters, and determine the surrounding rock displacement safety factor according to the ultimate displacement, crown displacement, waist displacement and invert displacement; Determine the steel arch area according to the steel arch grade parameter in the support plan, determine the steel arch stress according to the steel arch axial force in the axial force parameter and the steel arch area, and determine the safety factor of the first support structure according to the steel arch stress and the ultimate stress of the steel arch; Determine the compressive strength limit of the shotcrete according to the primary support grade parameter in the support plan, and determine the safety factor of the second support structure according to the minimum principal stress at the crown in the bending moment parameter and the compressive strength limit; Determine the ultimate compressive strength of the concrete according to the secondary lining grade parameters in the support plan, and determine the safety factor of the third support structure according to the axial forces in five directions of the tunnel in the axial force parameters, the secondary lining thickness parameters in the support plan, and the ultimate compressive strength of the concrete; Determine the safety index of the support plan according to the surrounding rock displacement safety factor, the safety factor of the first support structure, the safety factor of the second support structure, and the safety factor of the third support structure; Determine the safety risk level of the support plan according to the safety index; 4. The tunnel intelligent construction system based on the knowledge-enhanced large model according to claim 3, wherein, The optimization module is used for: When the safety risk level reaches the first pre-level, determine the full-parameter optimization plan for the support decision model; When the safety index of the support plan continuously decreases by more than the index preset threshold for N times, determine the full-parameter optimization plan for the support decision model; When the model prediction deviation rate of the support decision model is greater than the deviation preset threshold, determine the full-parameter optimization plan for the support decision model; 5. The tunnel intelligent construction system based on the knowledge-enhanced large model according to claim 1, characterized in that, The data acquisition module is further used to obtain expert experience knowledge in real time based on the knowledge channel, dynamically update the external knowledge base, and obtain the historical displacement and stress monitoring data of the cross-section of the target tunnel in real time based on the data channel; 6. The tunnel intelligent construction system based on the knowledge-enhanced large model according to claim 1, characterized in that, The system further includes: The drawing generation module is used to generate construction drawings that meet industry standards based on the support plan; 7. A tunnel intelligent construction method based on a knowledge-enhanced large model, characterized in that, The method includes: Collect the historical displacement and stress monitoring data of the cross-section of the target tunnel, the historical excavation data, the expert experience knowledge, and the tunnel safety specification documents through the data monitoring sensors buried at the tunnel construction site; Construct a knowledge graph according to the expert experience knowledge, construct a vector database according to the tunnel safety specification documents, construct an external knowledge base according to the knowledge graph and the vector database, and obtain an external knowledge base enhanced general large model; Identify the event type through the external knowledge base enhanced general large model; When the event type is a support decision request, process the historical displacement and stress monitoring data of the cross-section of the target tunnel through the support decision model to obtain a support plan. The support decision model is trained for the student model through the teacher model and knowledge distillation technology based on the historical excavation support data, and the teacher model is trained based on the expert experience knowledge; When the event type is an adjustment request, process the text information in natural language form sent by the user and the support plan to obtain an adjusted support plan; When the event types are abnormal monitoring requests and geological risks, evaluate the safety level of the target tunnel based on the historical displacement and stress monitoring data of the cross-section of the target tunnel; 8. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the tunnel intelligent construction system based on the knowledge-enhanced large model as claimed in claim 7; 9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the tunnel intelligent construction system based on the knowledge-enhanced large model as claimed in claim 7; 10. A computer program product, comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, it implements the steps in the tunnel intelligent construction system based on the knowledge-enhanced large model as claimed in claim 7;

Images