A risk intelligent assessment and early warning method and system for underground engineering construction

Abstract

This invention belongs to the field of tunnel and underground engineering construction safety technology, and provides a method and system for intelligent risk assessment and early warning in underground engineering construction. The method includes: acquiring engineering and monitoring data and binding applicability metadata; establishing an indicator system and configuring constraints and weights; using a large language model to parse standard provisions into structured scoring rules, and generating an applicable rule set through metadata matching; extracting actual values ​​and calculating risk scores and risk levels based on weights; outputting an early warning when the risk level or monitoring data exceeds a preset acceptable threshold; generating disposal suggestions based on the large language model combined with trigger indicators and applicability metadata, and performing matching and standard compliance checks on the suggestions. This invention, through multi-dimensional metadata adaptation and large-model structured parsing, effectively avoids rule misuse under complex working conditions, and achieves objective dynamic early warning and closed-loop self-evolving intelligent risk management.

Description

Technical Field

[0001] This invention relates to the field of tunnel and underground engineering construction safety technology, specifically to a method and system for intelligent risk assessment and early warning in underground engineering construction. Background Technology

[0002] Underground engineering projects typically include tunnel engineering, underground stations and structures, urban integrated utility tunnels, underground roads and underground space development, and deep foundation pits. They are characterized by limited working space, high degree of concealment, uncertain geological and hydrological conditions, and the coupling of construction disturbances and environmental effects. During underground engineering construction, risks often manifest as instability of surrounding rock or soil, water and mud inrushes, surface subsidence and deformation of surrounding buildings and structures, abnormal stress on the support system, and impacts on the safety of existing lines and pipelines. These risks evolve continuously as different construction methods, geological conditions, and engineering stages (such as excavation, support, structural construction, dewatering, and reinforcement) progress.

[0003] Currently, construction risk assessment requires the comprehensive use of multi-source heterogeneous information, including design specifications, construction organization and special plans, geological survey reports, monitoring and measurement data, construction logs, and surrounding environmental surveys. Because this information has inconsistent definitions and significant timeliness differences, traditional assessment methods relying primarily on manual experience and static tables struggle to maintain consistency and timeliness, and are even less capable of consistently adapting to different construction methods and geological conditions.

[0004] The existing intelligent risk assessment methods mainly suffer from the following technical shortcomings: (1) Automatic evaluation can be achieved by solidifying the rule base or expert scoring model, but rule extraction and maintenance are highly dependent on manual labor. When faced with different versions of specifications, different construction methods and different geological conditions, it is easy to encounter problems such as missing thresholds and applicable conditions, conflicts or incomplete coverage; (2) Learning risk relationships from historical data through machine learning, but due to the scarcity of underground engineering accident samples and significant differences between projects, the generalization and interpretability of the model are seriously insufficient, and it is difficult to meet the strict compliance requirements of the construction industry to "follow the standards". (3) Reference to standard provisions based on general text understanding or simple retrieval, but standard provisions often contain complex combination constraints (such as "applicable conditions - indicators - thresholds - classification - exceptions"). If the system cannot reliably identify and accurately match the "construction method - stratum - stage" of a specific project, it is very easy to cause misuse of rules, drift of scoring caliber and inconsistency of acceptable thresholds, which in turn leads to evaluation deviation and distortion of disposal recommendations.

[0005] Furthermore, due to the continuous updating of monitoring data, the risk status changes dynamically. Existing assessment models generally lack the ability to version rules, trace differences, and dynamically calibrate, making it difficult to form an auditable closed-loop management system. Summary of the Invention

[0006] To address the technical flaw that underground engineering standards typically include multi-dimensional applicable constraints such as construction methods, geological conditions, and engineering stages, which can easily lead to misapplication of rules in practical applications, this invention provides a method and system for intelligent risk assessment and early warning in underground engineering construction. The specific technical solution is as follows: A method for intelligent risk assessment and early warning in underground engineering construction, specifically including: S1. Obtain engineering data, risk event information, and monitoring and measurement data of underground engineering projects, and bind applicability metadata to the engineering data and risk event information to obtain the bound applicability metadata; wherein, the applicability metadata includes at least construction method data, geological condition data, and engineering stage data; S2. Establish a risk assessment index system and quantitative standards for underground engineering projects. The quantitative standards are used to define the scoring benchmark and risk classification criteria of the risk assessment index system. The risk assessment index system includes at least an occurrence probability assessment index and a consequence severity assessment index. Corresponding applicability constraints and weights are configured for the occurrence probability assessment index and the consequence severity assessment index, respectively. The applicability constraints include construction method data, geological condition data, and engineering stage data. S3. A pre-trained large language model is used to parse and structure the normative clauses in the risk assessment standard library, converting the normative clauses in natural language form into structured scoring rules that include the rules' applicable conditions, indicator items, and scoring thresholds, thereby generating a set of scoring rules. S4. Perform an applicability check on the set of scoring rules. The applicability check includes matching the bound applicability metadata with the rule applicability conditions in the structured scoring rules to filter out inapplicable rules and generate an applicable rule set. S5. Based on the applicable rule set and the scoring thresholds contained in each scoring rule in the applicable rule set, extract the actual values ​​of the corresponding indicator items from the engineering data and the monitoring measurement data, combine the weights, quantify the probability score and the severity score of the consequences of the risk event information, and determine the risk level according to the preset risk matrix. S6. When the risk level or the monitoring and measurement data and its changing trend exceed the preset acceptable threshold, the indicator item that causes the limit to exceed the limit will be determined as the trigger indicator, and a warning message will be triggered and output. S7. Based on the pre-trained large language model, take the risk event information, construction method data, geological condition data, engineering stage data in the applicability metadata, and the triggering indicators as input to generate disposal suggestions, and perform constraint verification on the disposal suggestions, including suggestion matching verification and standard compliance verification.

[0007] In a preferred implementation, the risk assessment index system includes detailed quantitative indicators characterizing the characteristics of underground engineering projects, wherein the detailed quantitative indicators include at least two or more of the following groups: Detailed indicators of geological and surrounding rock conditions, including surrounding rock grade or soil type, and adverse geological phenomena; Detailed indicators of groundwater and seepage conditions include groundwater level, inflow rate, permeability coefficient, and risk of sudden inflow; The detailed indicators of excavation and support parameters include excavation step distance, support type and parameters, and initial support closure time; The detailed indicators of construction technology and equipment parameters include shield propulsion parameters, cutterhead torque or thrust, synchronous grouting pressure and volume, and soil improvement parameters; The monitoring and measurement indicators are detailed in terms of the surrounding environment, including surface subsidence, horizontal displacement of the retaining structure, axial force of supports or anchor cables, pit bottom heave indicators, deformation of surrounding buildings and structures, geometric state of existing track, as well as monitoring alarm thresholds, monitoring frequency and trend indicators.

[0008] In a preferred implementation, step S3 involves converting the normative text in natural language form into structured scoring rules that include rule application conditions, indicator items, and scoring thresholds, including: The normative provisions in natural language form are parsed into structured scoring rules that include applicable conditions, scoring thresholds, grading intervals and exception clauses. The construction method data, geological condition data and engineering stage data fields are explicitly extracted from the applicable conditions of the rules and structured. The qualitative descriptions in the normative provisions are mapped to measurable indicators or graded intervals for the corresponding indicator items.

[0009] In a preferred implementation, step S4, before performing the applicability check, further includes: Perform a consistency check on the set of scoring rules, specifically: perform consistency checks on the structured scoring rules that conflict with or overlap with each other, and output conflict descriptions.

[0010] In a preferred implementation, in step S6, the triggering condition for the warning information includes at least one of the following: Triggered based on the risk level exceeding the static acceptable threshold; Triggered based on monitoring and measurement indicators exceeding the corresponding static acceptable threshold; Triggered based on monitoring measurement trends, acceleration, or abnormal scores exceeding dynamic acceptable thresholds; Triggered by a multi-indicator joint rule.

[0011] In a preferred implementation, in step S7, the step of generating a disposal suggestion based on the pre-trained large language model, using the risk event information, construction method data, geological condition data, engineering stage data in the applicability metadata, and the triggering indicators as input, and performing constraint verification on the disposal suggestion, including suggestion matching verification and standard compliance verification, includes: A set of candidate measures is formed based on the risk event information, the construction method data and geological condition data in the applicability metadata, and the project stage data; The pre-trained large language model is used to combine and sort the candidate measure set to generate disposal suggestions that include measure items, implementation order, resource requirements and target indicators; The proposed actions are subject to a matching check with the triggering indicators, a compliance check, and an executability check that takes into account on-site resources and process constraints.

[0012] In a preferred implementation, after step S7, a feedback and dynamic calibration step is further included, specifically: Based on historical statistical data, monitoring trend predictions, or engineering mechanism models, predict the risk changes after implementing the proposed measures, and obtain the response feedback data after implementing the proposed measures. The monitoring data collected after the treatment and the on-site feedback are used as the treatment feedback data. The treatment feedback data, which reflects the actual risk changes, is compared with the prediction results to generate a deviation record. When the deviation record exceeds the preset deviation threshold, dynamic calibration of the scoring rule set or grading interval is triggered, generating a new rule version and supporting rollback.

[0013] A risk intelligent assessment and early warning system for underground engineering construction, specifically including: The data acquisition and preprocessing module is used to acquire engineering data, risk event information, and monitoring and measurement data of underground engineering projects, and bind applicability metadata to the engineering data and risk event information to obtain the bound applicability metadata; wherein, the applicability metadata includes at least construction method data, geological condition data, and engineering stage data; The indicator system and standard management module is used to establish a risk assessment indicator system and quantitative standards for underground engineering projects. The quantitative standards are used to define the scoring benchmark and risk classification criteria of the risk assessment indicator system. The risk assessment indicator system includes at least an occurrence probability assessment indicator and a consequence severity assessment indicator, and corresponding applicability constraints and weights are configured for the occurrence probability assessment indicator and the consequence severity assessment indicator, respectively. The applicability constraints include construction method data, geological condition data, and engineering stage data. The standard clause structuring module is used to parse and structure the normative clauses in the risk assessment standard library using a pre-trained large language model. It converts the normative clauses in natural language form into structured scoring rules containing rule application conditions, indicator items, and scoring thresholds, generating a set of scoring rules. At the same time, it performs applicability verification on the set of scoring rules. The applicability verification includes matching the bound applicability metadata with the rule application conditions in the structured scoring rules to filter out inapplicable rules and generate an applicable rule set. The risk scoring and grading module is used to extract the actual values ​​of corresponding indicator items from the engineering data and the monitoring measurement data based on the applicable rule set and the scoring thresholds contained in each scoring rule in the applicable rule set, combine the weights, quantify the probability score and the severity score of the consequences of the risk event information, and determine the risk level according to the preset risk matrix. The early warning module is used to determine the indicator item that causes the limit to exceed the preset acceptable threshold when the risk level or the monitoring measurement data and its changing trend exceed the preset acceptable threshold, and to trigger and output early warning information. The disposal suggestion and constraint verification module is used to generate disposal suggestions based on the pre-trained large language model, taking the risk event information, construction method data, geological condition data, engineering stage data in the applicability metadata, and the trigger indicators as input, and to perform constraint verification on the disposal suggestions, including suggestion matching verification and standard compliance verification.

[0014] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the intelligent risk assessment and early warning method for underground engineering construction as described in any one of the claims.

[0015] A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, it implements the intelligent risk assessment and early warning method for underground engineering construction as described in any one of the claims.

[0016] Compared with the prior art, the beneficial effects of the present invention are: (1) This invention utilizes a pre-trained large language model to parse complex natural language standard provisions into structured scoring rules, and innovatively binds applicable metadata such as "construction method data, geological condition data, and engineering stage data" to engineering data and standard provisions in a two-way manner. By strictly aligning and matching to filter out inapplicable rules, it solves the evaluation bias caused by the generalization of existing technologies when facing variable working conditions, and greatly improves the consistency of standard implementation and the traceability of scoring results.

[0017] (2) This invention breaks through the limitations of traditional static ledgers. It extracts actual values ​​from engineering data and real-time monitoring and measurement data for quantitative scoring, and introduces monitoring and measurement data and their changing trends as triggering conditions for dynamic early warning. This enables the system to keenly capture abnormal acceleration and trend changes during the construction process, transforming delayed post-event discovery into objective, real-time data-driven early warning.

[0018] (3) This invention not only generates highly relevant on-site handling suggestions based on a large language model combined with specific engineering constraints, but also introduces a prediction process based on historical data and mechanistic models. By collecting actual handling feedback data to generate deviation records, dynamic calibration of scoring rules or weights is achieved, thereby enabling the system to have a closed-loop self-evolutionary capability that self-corrects and continuously optimizes as the project progresses. Attached Figure Description

[0019] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0020] Figure 1 This is a flowchart illustrating the intelligent risk assessment and early warning method for underground engineering construction in this invention.

[0021] Figure 2 This is a schematic diagram of the structured rule set formation process in an embodiment of the present invention; Figure 3 This is a schematic diagram of the internal structure of the structured rule set in an embodiment of the present invention; Figure 4 This is a schematic diagram of the closed loop of generating and calling the applicable rule set for the project in this embodiment of the invention. Detailed Implementation

[0022] The present invention will be further described below through specific embodiments, but this is not a limitation of the present invention. Those skilled in the art can make various modifications or improvements based on the basic idea of ​​the present invention, but as long as they do not depart from the basic idea of ​​the present invention, they are all within the protection scope of the present invention.

[0023] Example 1: See Figure 1 This invention provides a method for intelligent risk assessment and early warning in underground engineering construction, specifically including: Step S1: Obtain engineering data, risk event information, and monitoring measurement data for the underground engineering project, and bind applicability metadata to the engineering data and risk event information to obtain the bound applicability metadata; wherein, the applicability metadata includes at least construction method data, geological condition data, and engineering stage data; In practical implementation, the data acquisition and preprocessing process encompasses the access and cleaning of multi-source heterogeneous data. First, engineering data is accessed through file uploads, direct database connections, or API interfaces. This data includes not only core design specifications, construction organization plans, and geological survey reports, but also specialized construction plans, construction logs, construction monitoring data, and survey data on surrounding buildings, structures, and pipelines. Second, the system binds structured applicability metadata to the aforementioned data and risk event information. In addition to construction methods (such as tunnel boring, cut-and-cover, open-cut, and foundation pits), geological conditions (such as soft soil, karst, and gravel), and engineering stages (such as excavation, support, structural construction, and dewatering), it further binds the formation time, work site location, and spatial segment identifiers (mileage / section / zone).

[0024] Furthermore, for risk event information (including risk event lists, text descriptions, or structured records), the system performs event standardization processing, extracting and completing the event category, occurrence time, involved components, and triggering indicators. For monitoring and measurement data, the system not only collects raw values ​​but also performs time alignment, outlier detection, and missing value processing. It also automatically calculates derived features such as short-term slope, acceleration, and anomaly scores to generate standardized monitoring summaries, laying the data foundation for subsequent dynamic early warnings.

[0025] S2. Establish a risk assessment index system and quantitative standards for underground engineering projects. The quantitative standards are used to define the scoring benchmark and risk classification criteria of the risk assessment index system. The risk assessment index system includes at least an occurrence probability assessment index and a consequence severity assessment index. Corresponding applicability constraints and weights are configured for the occurrence probability assessment index and the consequence severity assessment index, respectively. The applicability constraints include construction method data, geological condition data, and engineering stage data. In this embodiment, the risk assessment index system established by the present invention not only distinguishes between probability assessment indicators and consequence severity assessment indicators, but also deeply integrates detailed quantitative indicators characterizing the characteristics of underground engineering to reflect the characteristics of underground engineering. Specifically, the detailed quantitative indicators include at least two or more of the following groups: Detailed indicators of geological and surrounding rock conditions, including surrounding rock grade or soil type, and adverse geological phenomena; Detailed indicators of groundwater and seepage conditions include groundwater level, inflow rate, permeability coefficient, and risk of sudden inflow; The detailed indicators of excavation and support parameters include excavation step distance, support type and parameters, and initial support closure time; The detailed indicators of construction technology and equipment parameters include shield propulsion parameters, cutterhead torque or thrust, synchronous grouting pressure and volume, and soil improvement parameters; The monitoring and measurement indicators are detailed in terms of the surrounding environment, including surface subsidence, horizontal displacement of the retaining structure, axial force of supports or anchor cables, pit bottom heave indicators, deformation of surrounding buildings and structures, geometric state of existing track, as well as monitoring alarm thresholds, monitoring frequency and trend indicators.

[0026] After establishing the indicator system, a data mapping relationship of "data field / monitoring summary field → indicator item" is further established. For example, the stratigraphic type in the geological exploration report is directly mapped to the geological indicator. At the same time, applicable constraints are configured for each indicator item, and weights are configured by expert experience or historical data statistics. The source and version of the weights are recorded to form a quantitative standard underlying table of "indicator item – grade interval – score – applicable conditions – threshold source".

[0027] S3. A pre-trained large language model is used to parse and structure the normative clauses in the risk assessment standard library, converting the normative clauses in natural language form into structured scoring rules that include the rules' applicable conditions, indicator items, and scoring thresholds, thereby generating a set of scoring rules. To overcome the problems of incomplete coverage and high maintenance costs caused by traditional systems relying on manual rule extraction, this invention deeply integrates a pre-trained Large Language Model (LLM). In specific implementation, such as... Figure 2 As shown, firstly, a risk assessment standard library containing national regulations, industry standards, and enterprise procedures is maintained, and the standard texts are parsed and segmented into clauses. Then, utilizing the semantic understanding capabilities of a pre-trained Large Language Model (LLM), complex and lengthy natural language standard clauses are accurately parsed into structured scoring rules containing "applicable conditions, scoring thresholds, grading intervals, and exception clauses." During this process, the Large Language Model explicitly extracts and structures the construction method data, geological condition data, and engineering stage data fields from the condition descriptions; simultaneously, qualitative expressions in the clauses (such as "significant settlement" and "fractured rock mass") are directly mapped to measurable indicators or specific quantitative grading intervals for the corresponding indicator items.

[0028] Further, please refer to Figure 3This demonstrates the internal structure of the structured rule set generated by this invention. This rule set not only includes macro-level rule set header information (such as rule set ID, source standard version, etc.), but each rule entry is also strongly correlated with applicable construction methods, applicable geological conditions, and applicable engineering stages. It also encompasses elements such as indicator items, triggering conditions, threshold / grading intervals, and weights. This highly structured data paradigm provides precise call output interfaces for subsequent scoring, early warning, and handling modules.

[0029] S4. Perform an applicability check on the set of scoring rules. The applicability check includes matching the bound applicability metadata with the rule applicability conditions in the structured scoring rules to filter out inapplicable rules and generate an applicable rule set. Furthermore, as a preferred embodiment, before performing the suitability check, the following steps are also included: Perform a consistency check on the set of scoring rules, specifically: perform consistency checks on the structured scoring rules that conflict with or overlap with each other, and output conflict descriptions.

[0030] Consistency verification refers to the system automatically detecting and identifying conflicting (such as contradictory judgment criteria for the same indicator) or overlapping structured scoring rules, and outputting conflict explanations and correction suggestions for expert review.

[0031] Specifically, applicability verification is the core of the entire rule adaptation process. This invention strictly aligns and matches the "bound applicability metadata" extracted in step S1 at the project level with the "rule application conditions" in the structured scoring rules. For example, if the project metadata is "shield tunneling method", then the standard rules with "cut-and-cover tunneling method" application conditions are automatically filtered out, and an applicable rule set that is highly suitable for the current working conditions is generated.

[0032] Furthermore, the applicable rule sets that pass verification will be solidified, and the system will strictly manage their versions, recording the generation time, review status, and version number, supporting comparison of rule version differences and historical rollback. For example... Figure 4 As shown downstream, once the applicable rule set for the project is generated, the system can smoothly enter the complete closed-loop process of subsequent risk scoring calls (calculating P / C values), level determination and early warning triggering, and finally trigger handling and dynamic calibration deviation analysis.

[0033] S5. Based on the applicable rule set and the scoring thresholds contained in each scoring rule in the applicable rule set, extract the actual values ​​of the corresponding indicator items from the engineering data and the monitoring measurement data, combine the weights, quantify the probability score and the severity score of the consequences of the risk event information, and determine the risk level according to the preset risk matrix. In its implementation, this invention, based on the established mapping relationship, automatically retrieves the current actual value of each indicator from the summary features of engineering data and monitoring measurement data. It then uses an applicable rule set to determine the scoring threshold or grading interval into which the value falls, thus deriving the basic score for each indicator. Subsequently, combining the corresponding weights configured in step S2, a linear weighted summation or combined aggregation algorithm is used to calculate the probability score P and the severity score C of the risk event. The probability score P includes aggregated geological factors, groundwater, process deviations, monitoring anomalies, etc., while the severity score C includes aggregated impacts on personnel, economy, construction period, and environmentally sensitive targets.

[0034] Finally, the calculated P and C values ​​are substituted into a preset 5×5 two-dimensional risk matrix for mapping calculation to determine the current comprehensive risk level (e.g., Level I to Level IV), and a risk assessment record containing details of triggering indicators and rule tracing evidence is generated. The risk record includes risk event information, P value, C value, risk level, set of triggering indicators, evidence citation path, and the currently effective rule version number.

[0035] S6. When the risk level or the monitoring and measurement data and its changing trend exceed the preset acceptable threshold, the indicator item that causes the limit to exceed the limit will be determined as the trigger indicator, and a warning message will be triggered and output. In practical implementation, this invention configures multi-level thresholds and notification strategies such as attention, early warning, and alarm, presenting a multi-dimensional and comprehensive set of conditions for triggering early warning information. The triggering conditions for early warning information include at least one of the following: Triggered based on the risk level exceeding the static acceptable threshold; Triggered based on monitoring and measurement indicators exceeding the corresponding static acceptable threshold; Triggered based on monitoring measurement trends, acceleration, or abnormal scores exceeding dynamic acceptable thresholds; Triggered by a multi-indicator joint rule.

[0036] When any of the above conditions are met, the indicator item that causes the limit to exceed the limit will be determined as a "trigger indicator", and an early warning message containing the warning level, trigger indicator, evidence reference and suggested handling direction will be generated and pushed to the relevant responsible persons through preset channels.

[0037] S7. Based on the pre-trained large language model, take the risk event information, construction method data, geological condition data, engineering stage data in the applicability metadata, and the triggering indicators as input to generate disposal suggestions, and perform constraint verification on the disposal suggestions, including suggestion matching verification and standard compliance verification.

[0038] In practice, the process of generating disposal recommendations is subject to strict engineering logic constraints. First, based on risk event information, construction method data, geological condition data and engineering stage data in the applicability metadata, a set of candidate measures is extracted from the built-in measure library (including grouting adjustment, parameter control, support reinforcement, drainage and dewatering, monitoring encryption, speed and load limits and emergency sealing, etc.).

[0039] Subsequently, a pre-trained large language model is used to combine and logically sort the candidate measure set, generating disposal suggestions that include specific measure items, implementation order, resource requirements, and target indicators. To prevent the large model from outputting "illusions," the system enforces constraint verification on the disposal suggestions. This constraint verification includes at least a matching verification of the suggestions with the trigger indicators and a compliance verification of the normative provisions that do not violate the prohibitions of the standard clauses. If necessary, an executability verification can also be performed based on the actual resources on site.

[0040] Furthermore, in a preferred embodiment, the system also possesses closed-loop calibration capabilities. Following S7, a response feedback and dynamic calibration step may be included: the system predicts risk changes in advance based on historical statistical data or engineering mechanism models. After actual construction and response, on-site monitoring data is collected as response feedback data and compared with the prediction results to generate deviation records. When the absolute value of the deviation record exceeds a preset deviation threshold, dynamic calibration of the scoring rule set or grading interval is automatically triggered, generating a new rule version and supporting rollback. This embodiment has removed weight calibration, focusing on the objective evolution of the rules.

[0041] Based on this, the system uses historical statistical data, monitoring trend algorithms, or underground engineering physical mechanism models to predict the risk changes of P and C after implementing the proposed treatment. After actual construction and treatment, the system collects post-treatment monitoring data and on-site assessment feedback as treatment feedback data, and compares the treatment feedback data reflecting actual risk changes with the previous prediction results to generate deviation records. When the absolute value of the deviation record exceeds a preset deviation threshold (indicating that the original assessment system has blind spots or distortions), the system will automatically trigger dynamic calibration and fine-tuning of the weights, scoring thresholds, or grading intervals mentioned in steps S2 and S3, generating new rule and weight versions and supporting rollback, thereby giving the assessment system a closed-loop capability of continuous self-correction and autonomous evolution as the project progresses.

[0042] Example 2: Based on the same inventive concept as Embodiment 1 above, Embodiment 2 also provides a risk intelligent assessment and early warning system for underground engineering construction. The system has multiple functional modules internally, and the internal processing logic of each module completely corresponds to steps S1-S7 of the above method, specifically including: The data acquisition and preprocessing module is used to acquire engineering data, risk event information, and monitoring and measurement data of underground engineering projects, and bind applicability metadata to the engineering data and risk event information to obtain the bound applicability metadata; wherein, the applicability metadata includes at least construction method data, geological condition data, and engineering stage data; The indicator system and standard management module is used to establish a risk assessment indicator system and quantitative standards for underground engineering projects. The quantitative standards are used to define the scoring benchmark and risk classification criteria of the risk assessment indicator system. The risk assessment indicator system includes at least an occurrence probability assessment indicator and a consequence severity assessment indicator, and corresponding applicability constraints and weights are configured for the occurrence probability assessment indicator and the consequence severity assessment indicator, respectively. The applicability constraints include construction method data, geological condition data, and engineering stage data. The standard clause structuring module is used to parse and structure the normative clauses in the risk assessment standard library using a pre-trained large language model. It converts the normative clauses in natural language form into structured scoring rules containing rule application conditions, indicator items, and scoring thresholds, generating a set of scoring rules. At the same time, it performs applicability verification on the set of scoring rules. The applicability verification includes matching the bound applicability metadata with the rule application conditions in the structured scoring rules to filter out inapplicable rules and generate an applicable rule set. The risk scoring and grading module is used to extract the actual values ​​of the corresponding indicator items from the engineering data and the monitoring measurement data based on the applicable rule set and the scoring thresholds contained in each scoring rule in the applicable rule set, combine the weights, quantify the probability score and the severity score of the consequences of the risk event information, and determine the risk level according to the preset risk matrix. The early warning module is used to determine the indicator item that causes the limit to exceed the preset acceptable threshold when the risk level or the monitoring measurement data and its changing trend exceed the preset acceptable threshold, and to trigger and output early warning information. The disposal suggestion and constraint verification module is used to generate disposal suggestions based on the pre-trained large language model, taking the risk event information, construction method data, geological condition data, engineering stage data in the applicability metadata, and the trigger indicators as input, and to perform constraint verification on the disposal suggestions, including suggestion matching verification and standard compliance verification.

[0043] Example 3: This embodiment also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements all the steps of the intelligent risk assessment and early warning method for underground engineering construction described in Embodiment 1. This computer device can be a cloud server cluster, an edge computing node, or a local industrial control computer, possessing the computing power to process massive amounts of monitoring data and run large-scale language models.

[0044] This embodiment also provides a computer-readable storage medium, such as ROM, RAM, disk, flash memory, or optical disk, on which a computer program is stored. When the computer program is executed by a processor, it also implements all the steps of the methods in Embodiment 1 above.

[0045] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the present invention. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in the present invention should still be covered by the claims of the present invention.

Claims

1. A risk intelligent assessment and early warning method for underground engineering construction, characterized in that, Specifically, it includes: S1. Obtain engineering data, risk event information, and monitoring and measurement data of underground engineering projects, and bind applicability metadata to the engineering data and risk event information to obtain the bound applicability metadata; wherein the applicability metadata includes at least construction method data, geological condition data, and engineering stage data; S2. Establish a risk assessment index system and quantitative standards for underground engineering projects. The quantitative standards are used to define the scoring benchmark and risk classification criteria of the risk assessment index system. The risk assessment index system includes at least an occurrence probability assessment index and a consequence severity assessment index. Corresponding applicability constraints and weights are configured for the occurrence probability assessment index and the consequence severity assessment index, respectively. The applicability constraints include construction method data, geological condition data, and engineering stage data. S3. A pre-trained large language model is used to parse and structure the normative clauses in the risk assessment standard library, converting the normative clauses in natural language form into structured scoring rules that include the rules' applicable conditions, indicator items, and scoring thresholds, thereby generating a set of scoring rules. S4. Perform an applicability check on the set of scoring rules. The applicability check includes matching the bound applicability metadata with the rule applicability conditions in the structured scoring rules to filter out inapplicable rules and generate an applicable rule set. S5. Based on the applicable rule set and the scoring thresholds contained in each scoring rule in the applicable rule set, extract the actual values ​​of the corresponding indicator items from the engineering data and the monitoring measurement data, combine the weights, quantify the probability score and the severity score of the consequences of the risk event information, and determine the risk level according to the preset risk matrix. S6. When the risk level or the monitoring and measurement data and its changing trend exceed the preset acceptable threshold, the indicator item that causes the limit to exceed the limit will be determined as the trigger indicator, and a warning message will be triggered and output. S7. Based on the pre-trained large language model, take the risk event information, construction method data, geological condition data, engineering stage data in the applicability metadata, and the triggering indicators as input to generate disposal suggestions, and perform constraint verification on the disposal suggestions, including suggestion matching verification and standard compliance verification.

2. The intelligent risk assessment and early warning method for underground engineering construction according to claim 1, characterized in that, The risk assessment indicator system includes detailed quantitative indicators characterizing the characteristics of underground engineering projects, and these detailed quantitative indicators include at least two or more of the following groups: Detailed indicators of geological and surrounding rock conditions, including surrounding rock grade or soil type, and adverse geological phenomena; Detailed indicators of groundwater and seepage conditions include groundwater level, inflow rate, permeability coefficient, and risk of sudden inflow; The detailed indicators of excavation and support parameters include excavation step distance, support type and parameters, and initial support closure time; The detailed indicators of construction technology and equipment parameters include shield propulsion parameters, cutterhead torque or thrust, synchronous grouting pressure and volume, and soil improvement parameters; The monitoring and measurement indicators are detailed in terms of the surrounding environment, including surface subsidence, horizontal displacement of the retaining structure, axial force of supports or anchor cables, pit bottom heave indicators, deformation of surrounding buildings and structures, geometric state of existing track, as well as monitoring alarm thresholds, monitoring frequency and trend indicators.

3. The intelligent risk assessment and early warning method for underground engineering construction according to claim 1, characterized in that, In step S3, the normative provisions in natural language form are converted into structured scoring rules that include rule application conditions, indicator items, and scoring thresholds, including: The normative provisions in natural language form are parsed into structured scoring rules that include applicable conditions, scoring thresholds, grading intervals and exception clauses. The construction method data, geological condition data and engineering stage data fields are explicitly extracted from the applicable conditions of the rules and structured. The qualitative descriptions in the normative provisions are mapped to measurable indicators or graded intervals for the corresponding indicator items.

4. The intelligent risk assessment and early warning method for underground engineering construction according to claim 3, characterized in that, In step S4, before performing the suitability check, the following is also included: Perform a consistency check on the set of scoring rules, specifically: perform consistency checks on the structured scoring rules that conflict with or overlap with each other, and output conflict descriptions.

5. The intelligent risk assessment and early warning method for underground engineering construction according to claim 1, characterized in that, In step S6, the triggering condition for the early warning information includes at least one of the following: Triggered based on the risk level exceeding the static acceptable threshold; Triggered based on monitoring and measurement indicators exceeding the corresponding static acceptable threshold; Triggered based on monitoring measurement trends, acceleration, or abnormal scores exceeding dynamic acceptable thresholds; Triggered by a multi-indicator joint rule.

6. The intelligent risk assessment and early warning method for underground engineering construction according to claim 1, characterized in that, In step S7, based on the pre-trained large language model, a disposal suggestion is generated using the risk event information, construction method data, geological condition data, engineering stage data in the applicability metadata, and the triggering indicators as input. Constraint checks are then performed on the disposal suggestion, including suggestion matching verification and regulatory compliance verification. A set of candidate measures is formed based on the risk event information, the construction method data, the geological condition data, and the project stage data in the applicability metadata; The pre-trained large language model is used to combine and sort the candidate measure set to generate disposal suggestions that include measure items, implementation order, resource requirements and target indicators; The proposed actions are subject to a matching check with the triggering indicators, a compliance check, and an executability check that takes into account on-site resources and process constraints.

7. The intelligent risk assessment and early warning method for underground engineering construction according to claim 3, characterized in that, Following S7, a feedback and dynamic calibration step is also included, specifically: Based on historical statistical data, monitoring trend predictions, or engineering mechanism models, predict the risk changes after implementing the proposed measures, and obtain the response feedback data after implementing the proposed measures. The monitoring data collected after the treatment and the on-site feedback are used as the treatment feedback data. The treatment feedback data, which reflects the actual risk changes, is compared with the prediction results to generate a deviation record. When the deviation record exceeds the preset deviation threshold, dynamic calibration of the scoring rule set or grading interval is triggered, generating a new rule version and supporting rollback.

8. A risk intelligent assessment and early warning system for underground engineering construction, characterized in that, Specifically, it includes: The data acquisition and preprocessing module is used to acquire engineering data, risk event information, and monitoring and measurement data of underground engineering projects, and bind applicability metadata to the engineering data and risk event information to obtain the bound applicability metadata; wherein, the applicability metadata includes at least construction method data, geological condition data, and engineering stage data; The indicator system and standard management module is used to establish a risk assessment indicator system and quantitative standards for underground engineering projects. The quantitative standards are used to define the scoring benchmark and risk classification criteria of the risk assessment indicator system. The risk assessment indicator system includes at least an occurrence probability assessment indicator and a consequence severity assessment indicator, and corresponding applicability constraints and weights are configured for the occurrence probability assessment indicator and the consequence severity assessment indicator, respectively. The applicability constraints include construction method data, geological condition data, and engineering stage data. The standard clause structuring module is used to parse and structure the normative clauses in the risk assessment standard library using a pre-trained large language model. It converts the normative clauses in natural language form into structured scoring rules containing rule application conditions, indicator items, and scoring thresholds, generating a set of scoring rules. At the same time, it performs applicability verification on the set of scoring rules. The applicability verification includes matching the bound applicability metadata with the rule application conditions in the structured scoring rules to filter out inapplicable rules and generate an applicable rule set. The risk scoring and grading module is used to extract the actual values ​​of corresponding indicator items from the engineering data and the monitoring measurement data based on the applicable rule set and the scoring thresholds contained in each scoring rule in the applicable rule set, combine the weights, quantify the probability score and the severity score of the consequences of the risk event information, and determine the risk level according to the preset risk matrix. The early warning module is used to determine the indicator item that causes the limit to exceed the preset acceptable threshold when the risk level or the monitoring measurement data and its changing trend exceed the preset acceptable threshold, and to trigger and output early warning information. The disposal suggestion and constraint verification module is used to generate disposal suggestions based on the pre-trained large language model, taking the risk event information, construction method data, geological condition data, engineering stage data in the applicability metadata, and the trigger indicators as input, and to perform constraint verification on the disposal suggestions, including suggestion matching verification and standard compliance verification.

9. A computer device, characterized in that, The system includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the intelligent risk assessment and early warning method for underground engineering construction as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the method for intelligent risk assessment and early warning of underground engineering construction as described in any one of claims 1-7.

Images