A tunnel collapse disaster risk comprehensive assessment method and system

Abstract

This invention discloses a comprehensive assessment method and system for tunnel collapse disaster risk, relating to the field of tunnel engineering technology. The comprehensive assessment method and system for tunnel collapse disaster risk includes the following steps: Step 1: multi-source data acquisition; Step 2: data preprocessing; Step 3: construction of a dynamic indicator system; Step 4: calculation of a triple-fusion assessment model; and Step 5: risk level output and verification. This scheme employs multi-source data acquisition and refined preprocessing to solidify the foundation of assessment data: it adopts a three-dimensional acquisition mode of "static survey + dynamic monitoring + non-contact detection," integrating four types of multi-source data: geological, construction, environmental, and UAV lidar detection. Simultaneously, it completes standardized preprocessing such as outlier removal, data normalization, and missing value completion through the 3σ criterion, Z-score algorithm, and KNN algorithm. The data processing delay is ≤5s, effectively solving the problem of the single data dimension in traditional assessment methods.

Description

Technical Field

[0001] This invention relates to the field of tunnel engineering technology, and in particular to a comprehensive assessment method and system for tunnel collapse disaster risk. Background Technology

[0002] As a core component of infrastructure construction such as transportation and water conservancy, especially mountain water diversion, highway and railway tunnels, tunnel engineering is affected by complex geological conditions, variable construction environment and large differences in rock mass stability. Collapse has become one of the most frequent and harmful geological disasters during construction.

[0003] Currently, approximately 65% ​​of tunnel construction projects in the industry still use a single fuzzy comprehensive evaluation method for collapse risk assessment. This method is also the most widely used traditional technique in the field of tunnel collapse risk assessment. However, this traditional assessment method has many core defects in practical application and is difficult to adapt to the actual risk characteristics of tunnel construction: First, the weight setting lacks dynamism, relying solely on expert subjective scoring to determine the weight of each risk factor, and the weight remains static throughout the entire construction cycle. However, the influence ratio of various risk factors such as geology, construction, and environment varies significantly at different stages of tunnel construction, such as the excavation period and the support completion period. The static weight system cannot match the actual risk changes in construction. Second, the assessment algorithm is simplistic, relying solely on fuzzy comprehensive evaluation to complete risk determination. There is a lack of cross-validation and result optimization using multiple algorithms, and the assessment results are significantly affected by expert subjective judgment, greatly compromising objectivity and accuracy. Thirdly, the data support system is weak. A standardized multi-source data acquisition model has not been established. Only a small amount of basic geological and construction data has been collected. The collection of key data such as environmental factors and rock mass micro-deformation is lacking. At the same time, the preprocessing operations such as outlier removal, missing value completion, and dimension normalization are relatively crude, which further reduces the completeness and accuracy of the assessment data. Fourthly, the assessment and disposal are disconnected. Traditional methods can only complete simple risk level determination without matching targeted risk disposal and monitoring recommendations. This can easily lead to the construction party either failing to take timely prevention and control measures due to missed or misjudged risks, resulting in landslide accidents, or taking unnecessary reinforcement measures due to over-assessment, resulting in a large waste of construction resources.

[0004] Industry data shows that the accuracy rate of traditional single fuzzy comprehensive evaluation method for tunnel collapse risk assessment is only about 76.7%. It is difficult to accurately identify the actual collapse risk level in tunnel construction and cannot provide reliable risk warnings and scientific basis for handling for construction parties. This has become a key technical problem restricting the safe and efficient progress of tunnel construction. Summary of the Invention

[0005] The purpose of this invention is to at least solve one of the technical problems existing in the prior art, and to provide a comprehensive assessment method and system for tunnel collapse disaster risk, which can solve the problem that data collection is difficult to accurately identify the actual collapse risk level in tunnel construction.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a comprehensive assessment method for tunnel collapse disaster risk, comprising step 1: multi-source data acquisition, step 2: data preprocessing, step 3: construction of dynamic indicator system, step 4: calculation of triple fusion assessment model, and step 5: risk level output and verification. Step 1: multi-source data acquisition includes step 101 geological static data, step 102 construction dynamic data, step 103 environmental data, and step 104 UAV three-dimensional data. Step 3: Construction of the dynamic indicator system includes Step 301 Criterion Layer 1, Step 302 Criterion Layer 2, and Step 303 Criterion Layer 3; Step 4: The calculation of the triple fusion evaluation model includes step 401, first-level evaluation, step 402, second-level optimization, and step 403, third-level verification.

[0007] Preferably, the geological static data in step 101 includes the following indicators: surrounding rock grade, rock mass integrity coefficient, and fault density, which are collected by combining ground-penetrating radar and borehole sampling. The measured data of a certain tunnel K10+230 section are surrounding rock grade IV, rock mass integrity coefficient 0.62, and fault density 0.3 faults / m.

[0008] Preferably, the construction dynamic data in step 102 includes the following indicators: the arch settlement rate, the surrounding convergence value, and the support stress, which are monitored in real time by fiber optic grating sensors and convergence meters; the monitoring data for the tunnel K10+230 section are: arch settlement 1.8mm / d, convergence value 3.2mm, and support stress 18MPa.

[0009] Preferably, the environmental data in step 103 includes: daily rainfall, groundwater level, and air humidity, which are collected using rain gauges and water level sensors; the daily rainfall in this monitoring section is 25 mm and the groundwater level is 6 m deep, and the on-site monitoring specifications are followed.

[0010] Preferably, in step 104, the UAV three-dimensional data focuses on the width of rock mass fissures and the deformation of the tunnel face, using a point cloud density of 50 points / cm². 2 The laser radar drone was used for detection; the measured width of the crack in this section was 0.8 mm and the local protrusion on the face of the tunnel was 2 cm, in accordance with the non-contact measurement standard.

[0011] Preferably, step 2 involves data preprocessing. Outlier removal: The 3σ criterion is used to filter the data. For example, if the threshold for the arch subsidence rate is set to 5 mm / d, and a monitoring point initially collects data of 6.2 mm / d, it is determined to be an outlier and removed. Normalization: The Z-score algorithm is used to standardize data of different dimensions. The algorithm formula is: Z=(x-μ) / Ï; (Where x is the original data, μ is the mean, and σ is the standard deviation), taking a support stress of 18MPa as an example, the standardized value is calculated to be 0.72; Data completion: KNN algorithm is used to interpolate for missing data (neighborhood parameter k=5). For a certain period of time, the groundwater level data is missing. By calling the synchronous data of 5 surrounding monitoring points, the interpolation is completed to 6.1m.

[0012] Preferably, step 3 involves constructing a dynamic indicator system. The indicator system uses "collapse risk level" as the target layer, with three criterion layers and corresponding indicator layers. The weights and indicator thresholds for each layer are as follows: Step 301 Criterion Layer 1: Geological Factors, basic weight 0.45, includes three indicator layers: Surrounding rock grade: dynamic weight 0.18, threshold range is II-V; Rock mass integrity coefficient: dynamic weight 0.12, threshold range 0.3-0.8; Fault density: dynamic weight 0.08, threshold range 0-0.5 faults / m; Step 302, Criterion Layer Two: Construction Factors, with a basic weight of 0.35, includes three indicator layers: The rate of subsidence of the arch is: dynamic weight 0.15, threshold range 0-5mm / d (where >2mm / d triggers an early warning). Support stress: dynamic weight 0.10, threshold range 0-30MPa (where >25MPa triggers an early warning); Vibration frequency of construction equipment: dynamic weight 0.06, threshold range 0-10Hz (where >8Hz triggers an early warning); Step 303, Criterion Layer 3: Environmental Factors, with a base weight of 0.20, includes three indicator layers: Daily rainfall: dynamic weight 0.09, threshold range 0-50mm (where >20mm triggers an alert); Groundwater level depth: dynamic weight 0.07, threshold range 0-10m (where <5m increases the risk level); Rock mass fracture width: dynamic weight 0.04, threshold range 0-2mm (where >1mm triggers an early warning).

[0013] Preferably, step 4 involves calculating the triple fusion evaluation model; Step 401: Level 1 Assessment: Fuzzy Comprehensive Evaluation; Establish a rating scale: low risk (0-0.3), medium risk (0.3-0.6), high risk (0.6-1.0); Constructing a fuzzy matrix: Taking geological factors in segment K10+230 as an example; Level 1 output result: Overall evaluation score 0.58, corresponding to medium risk level; Step 402 Secondary Optimization: Backpropagation Neural Network; Training data support: The model was trained using 1200 historical landslide examples, including 320 small landslides, 580 medium landslides, and 300 large landslides. Network structure design: The input layer contains 12 indicators, the hidden layer has 8 neurons, and the output layer corresponds to 3 risk levels; Optimization results: The fuzzy comprehensive evaluation result of 0.58 was corrected to 0.63, corresponding to the high-risk level, and the model test accuracy reached 89.7%. Step 403: Level 3 Validation: Monte Carlo Simulation; Uncertainty parameters selected: surrounding rock grade (value range IV-V), crown settlement rate (value range 1.8-2.5mm / d); Simulation parameter settings: After 1000 simulations, the risk probability distribution was found to be 78.3% high risk and 21.7% medium risk. Final assessment result: High risk level.

[0014] Preferably, step 5: risk level output and verification; Low risk: The comprehensive evaluation value ranges from 0 to 0.3. The recommended treatment is routine monitoring (monitoring frequency once every 2 hours). The verification case is a tunnel section at K8+450, with an evaluation value of 0.27. No collapse occurred during construction. Medium risk: The comprehensive evaluation value ranges from 0.3 to 0.6. The recommended treatment is to increase the frequency of monitoring (once every 30 minutes) and strengthen the support. The verification case is a local area of ​​section K8+450 of a tunnel, where the evaluation value was 0.45. After the support was reinforced, the evaluation value dropped to 0.29. High risk: The comprehensive evaluation value ranges from 0.6 to 1.0. The recommended treatment is to stop work and evacuate + special reinforcement (advanced small pipe support); the verification case is a tunnel section K10+230, with an evaluation value of 0.63. After special reinforcement, the evaluation value dropped to 0.25. Verification data source: 30 monitoring sections of the Xinjiang Kashuang Water Diversion Tunnel. The accuracy rate of this method reached 92.3%, which is 15.6% higher than the traditional fuzzy evaluation method.

[0015] Preferably, a comprehensive risk assessment system for tunnel collapse disasters includes the following system modules; Multi-source data acquisition module: Its core function is to receive four types of data: geological, construction, environmental, and UAV data, and it supports the transmission of three protocols: RS485, 5G, and IPsec. Hardware and software support includes fiber optic grating sensors, lidar UAVs, and data acquisition gateways. Data preprocessing module: Core functions include outlier removal, normalization, and data completion, with a processing latency of ≤5s; hardware and software support is provided by edge computing nodes; Indicator system management module: Core functions include custom indicator configuration, threshold parameter setting, and dynamic weight algorithm integration; hardware and software support is a visual configuration platform; Dynamic weight calculation module: Its core function is to realize the collaborative calculation of AHP algorithm and entropy weight algorithm; the hardware and software support is the Python programming environment; Triple fusion evaluation module: The core function is to integrate three algorithms: fuzzy evaluation, BP neural network and Monte Carlo simulation; hardware and software support includes GPU server, TensorFlow framework and MATLAB tool. Risk output and early warning module: The core functions are risk level visualization display, risk zoning map generation, audible and visual alarms and SMS notifications, with an early warning delay of ≤10s; hardware and software support includes industrial tablets, SMS modems, and LED early warning screens; Database module: Its core functions are to store historical data, real-time monitoring data, and evaluation result data, and it supports MySQL database and Redis caching technology; the hardware and software support is cloud server.

[0016] Compared with the prior art, the beneficial effects of the present invention are: 1. The comprehensive assessment method and system for tunnel collapse disaster risk: This scheme uses multi-source data acquisition and refined preprocessing to solidify the foundation of assessment data. It adopts a three-dimensional acquisition mode of "static survey + dynamic monitoring + non-contact detection" to integrate four types of multi-source data: geology, construction, environment, and UAV lidar detection. At the same time, it completes standardized preprocessing such as outlier removal, data normalization, and missing value completion through the 3σ criterion, Z-score algorithm, and KNN algorithm. The data processing delay is ≤5s, which effectively solves the problems of single data dimension and rough processing in traditional assessment methods, and greatly improves the completeness and accuracy of assessment data.

[0017] 2. The comprehensive assessment method and system for tunnel collapse disaster risk: This scheme constructs a dynamic weighting system to adapt to the dynamic risk characteristics of tunnel construction. It adopts a combination mode of subjective weighting by AHP and objective weighting by entropy weighting, and can dynamically adjust the weight of each criterion layer according to the tunnel construction stage. During the excavation period, the weight of construction factors is increased to 0.42, and after the support is completed, the weight of geological factors is increased to 0.51. This solves the problem of static weighting and mismatch with actual risk characteristics in traditional assessment methods, making the risk factor weighting more in line with the actual construction and improving the rationality of the assessment.

[0018] 3. The comprehensive risk assessment method and system for tunnel collapse disasters, with its triple-fusion assessment model, significantly improves the accuracy of risk assessment: integrating fuzzy comprehensive evaluation, BP neural network, and Monte Carlo simulation algorithms to complete the entire process of risk calculation from "first-level assessment to second-level optimization to third-level verification." Verified by actual measurements on 30 monitoring sections of the Xinjiang Kashuang Water Diversion Tunnel, the accuracy of this method in assessing collapse risk reached 92.3%, an improvement of 15.6% compared to the traditional single fuzzy comprehensive evaluation method. This effectively solves the problems of traditional methods being greatly affected by subjective factors and lacking multi-algorithm verification, achieving accurate determination of collapse risk levels.

[0019] 4. The comprehensive assessment method and system for tunnel collapse disaster risk: This scheme features a low-latency assessment and early warning system to achieve rapid risk response. Relying on GPU servers and 5G slicing network hardware support, an efficient system architecture has been constructed. The triple fusion assessment module is accelerated by NVIDIA A100 GPU servers. The early warning delay of the risk output and early warning module is ≤10s. The risk level can be pushed in real time and accompanied by audible and visual alarms through construction terminals and LED early warning screens. This solves the problem of delayed early warning in traditional assessment methods and buys time for the construction party to take timely measures.

[0020] 5. The comprehensive risk assessment method and system for tunnel collapse disasters, along with the tiered treatment recommendations and practical verification, provide a scientific basis for construction decisions: Based on the assessment results, three risk levels—low, medium, and high—are defined, and targeted treatment recommendations are matched accordingly. High-risk sections are addressed with work stoppage and evacuation, plus advanced small-diameter pipe support. All treatment recommendations have been verified through actual tunnel construction cases. After treatment, the risk assessment value of the affected sections can be significantly reduced to the low-risk range. This solves the problem of traditional assessment methods that only determine risk without providing scientific treatment guidance, achieving closed-loop management of "risk assessment - precise treatment." This effectively reduces the incidence of tunnel collapse accidents while avoiding the waste of construction resources caused by over-assessment, ensuring the safe and efficient progress of tunnel construction. Attached Figure Description

[0021] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a schematic diagram of the system modules of the present invention; Figure 2 This is a schematic diagram of the fuzzy matrix of the present invention. Detailed Implementation

[0022] This section will describe in detail specific embodiments of the present invention. Preferred embodiments of the present invention are shown in the accompanying drawings. The purpose of the drawings is to supplement the textual description with graphics, so that people can intuitively and vividly understand each technical feature and overall technical solution of the present invention, but they should not be construed as limiting the scope of protection of the present invention.

[0023] In the description of this invention, it should be understood that the orientation descriptions, such as up, down, front, back, left, right, etc., are based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention.

[0024] In the description of this invention, terms such as greater than, less than, and exceeding are understood to exclude the stated number, while terms such as above, below, and within are understood to include the stated number. The use of "first" and "second" in the description is merely for distinguishing technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the order of the indicated technical features.

[0025] In the description of this invention, unless otherwise explicitly defined, terms such as "set up," "install," and "connect" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this invention in conjunction with the specific content of the technical solution.

[0026] Please see Figure 1-2 This invention provides a technical solution: a comprehensive assessment method for tunnel collapse disaster risk, comprising the following steps: Step 1: Multi-source data acquisition; (corresponding to the system data acquisition module) The three-dimensional data acquisition mode, which combines static surveying, dynamic monitoring, and non-contact detection, is adopted. Specific data is as follows: Step 101: Geological static data: The collected indicators include the surrounding rock grade, rock mass integrity coefficient, and fault density. The data are collected using a combination of ground-penetrating radar and borehole sampling. The measured data of section K10+230 of a certain tunnel are: surrounding rock grade IV, rock mass integrity coefficient 0.62, and fault density 0.3 faults / m. The data collection is based on the industry standard JTG / T3660—2020. Step 102 Construction Dynamic Data: The collected indicators include the crown settlement rate, the surrounding convergence value, and the support stress, which are monitored in real time through fiber optic grating sensors and convergence meters; the monitoring data for the K10+230 section of the tunnel are a crown settlement of 1.8 mm / d, a convergence value of 3.2 mm, and a support stress of 18 MPa, which meet the requirements of JTG / T3660—2020 specification. Step 103 Environmental Data: Focus on collecting daily rainfall, groundwater level, and air humidity using rain gauges and water level sensors; the daily rainfall in this monitoring section is 25mm and the groundwater level is 6m deep, and on-site monitoring specifications are followed. Step 104: UAV 3D Data: Focusing on rock mass fracture width and tunnel face deformation, using a point cloud density of 50 points / cm². 2 The laser radar drone was used for detection; the measured width of the crack in this section was 0.8 mm, and the local protrusion on the working face was 2 cm, following the non-contact measurement standard; Step 2: Data preprocessing (corresponding to the system's data preprocessing module) Outlier removal: The 3σ criterion is used to filter the data. For example, if the threshold for the arch subsidence rate is set to 5 mm / d, and a monitoring point initially collects data of 6.2 mm / d, it is determined to be an outlier and removed. Normalization: The Z-score algorithm is used to standardize data of different dimensions. The algorithm formula is: Z=(x-μ) / Ï; (Where x is the original data, μ is the mean, and σ is the standard deviation), taking a support stress of 18MPa as an example, the standardized value is calculated to be 0.72; Data completion: KNN algorithm was used to interpolate for missing data (neighborhood parameter k=5). For a certain period of time, the groundwater level data was missing. By calling the synchronous data of 5 surrounding monitoring points, the interpolation was completed to 6.1m. Step 3: Construction of a dynamic indicator system (corresponding to the system indicator system management module) The indicator system uses "collapse risk level" as the target layer, with three criterion layers and corresponding indicator layers. The weights and indicator thresholds for each layer are as follows: Step 301 Criterion Layer 1: Geological Factors, basic weight 0.45, includes three indicator layers: Surrounding rock level: dynamic weight 0.18, threshold range is II-V (where V corresponds to the highest risk); Rock mass integrity coefficient: dynamic weight 0.12, threshold range 0.3-0.8 (the lower the value, the higher the risk). Fault density: dynamic weight 0.08, threshold range 0-0.5 faults / m; Step 302, Criterion Layer Two: Construction Factors, with a basic weight of 0.35, includes three indicator layers: The rate of subsidence of the arch is: dynamic weight 0.15, threshold range 0-5mm / d (where >2mm / d triggers an early warning). Support stress: dynamic weight 0.10, threshold range 0-30MPa (where >25MPa triggers an early warning); Vibration frequency of construction equipment: dynamic weight 0.06, threshold range 0-10Hz (where >8Hz triggers an early warning); Step 303, Criterion Layer 3: Environmental Factors, with a base weight of 0.20, includes three indicator layers: Daily rainfall: dynamic weight 0.09, threshold range 0-50mm (where >20mm triggers an alert); Groundwater level depth: dynamic weight 0.07, threshold range 0-10m (where <5m increases the risk level); Rock mass fracture width: dynamic weight 0.04, threshold range 0-2mm (where >1mm triggers an early warning); Note: The weight calculation adopts a combination of AHP subjective weight (based on expert scoring) and entropy weight method objective weight (based on data dispersion), and is dynamically adjusted according to the construction stage: the weight of construction factors during the excavation period is increased to 0.42, and the weight of geological factors after the support is completed is increased to 0.51. Step 4: Calculation of the triple fusion evaluation model (corresponding to the system evaluation module) Step 401: Level 1 Assessment: Fuzzy Comprehensive Evaluation; Establish a rating scale: low risk (0-0.3), medium risk (0.3-0.6), high risk (0.6-1.0); Constructing a fuzzy matrix: Taking geological factors in segment K10+230 as an example (please refer to...) Figure 2 ); Level 1 output result: Overall evaluation score 0.58, corresponding to medium risk level; Step 402 Secondary Optimization: Backpropagation Neural Network; Training data support: The model was trained using 1200 historical landslide examples, including 320 small landslides, 580 medium landslides, and 300 large landslides. Network structure design: The input layer contains 12 indicators, the hidden layer has 8 neurons, and the output layer corresponds to 3 risk levels; Optimization results: The fuzzy comprehensive evaluation result of 0.58 was corrected to 0.63, corresponding to the high-risk level, and the model test accuracy reached 89.7%. Step 403 Level 3 Verification: Monte Carlo Simulation Uncertainty parameters selected: surrounding rock grade (value range IV-V), crown settlement rate (value range 1.8-2.5mm / d); Simulation parameter settings: After 1000 simulations, the risk probability distribution was found to be 78.3% high risk and 21.7% medium risk. Final assessment result: High risk level (95% confidence level); Step 5: Risk level output and verification; Low risk: The comprehensive evaluation value ranges from 0 to 0.3. The recommended treatment is routine monitoring (monitoring frequency once every 2 hours). The verification case is a tunnel section at K8+450, with an evaluation value of 0.27. No collapse occurred during construction. Medium risk: The comprehensive evaluation value ranges from 0.3 to 0.6. The recommended treatment is to increase the frequency of monitoring (once every 30 minutes) and strengthen the support. The verification case is a local area of ​​section K8+450 of a tunnel, where the evaluation value was 0.45. After the support was reinforced, the evaluation value dropped to 0.29. High risk: The comprehensive evaluation value ranges from 0.6 to 1.0. The recommended treatment is to stop work and evacuate + special reinforcement (advanced small pipe support); the verification case is a tunnel section K10+230, with an evaluation value of 0.63. After special reinforcement, the evaluation value dropped to 0.25. Verification data source: 30 monitoring sections of the Xinjiang Kashuang Water Diversion Tunnel. The accuracy rate of this method reached 92.3%, which is 15.6% higher than the traditional fuzzy evaluation method. A comprehensive risk assessment system for tunnel collapse disasters, comprising the following modules; Multi-source data acquisition module: Its core function is to receive four types of data: geological, construction, environmental, and UAV data, and it supports the transmission of three protocols: RS485, 5G, and IPsec. Hardware and software support includes fiber optic grating sensors, lidar UAVs, and data acquisition gateways. Data preprocessing module: Core functions include outlier removal, normalization, and data completion, with a processing latency of ≤5s; hardware and software support is provided by edge computing nodes (CPU model: Intel i7-12700). Indicator system management module: Core functions include custom indicator configuration, threshold parameter setting, and dynamic weight algorithm integration; hardware and software support is a visual configuration platform (technical framework: Vue3 + ElementPlus). Dynamic weight calculation module: The core function is to realize the collaborative calculation of AHP algorithm (with built-in expert scoring interface) and entropy weight algorithm (data dispersion analysis); the hardware and software support is the Python programming environment (dependent on numpy and scipy libraries). The triple fusion evaluation module integrates three algorithms: fuzzy evaluation, BP neural network, and Monte Carlo simulation. Hardware and software support includes a GPU server (model: NVIDIA A100), TensorFlow framework (BP neural network), and MATLAB tools (Monte Carlo simulation). Risk output and early warning module: The core functions are risk level visualization display, risk zoning map generation, audible and visual alarms and SMS notifications, with an early warning delay of ≤10s; hardware and software support includes industrial tablets, SMS modems, and LED early warning screens; Database module: The core functions are to store historical data (storage capacity ≥ 100,000 records), real-time monitoring data, and evaluation result data, and support MySQL database and Redis caching technology; hardware and software support is cloud server (configuration: 8 cores 16G, storage capacity ≥ 1TB). Furthermore, the solution employs multi-source data acquisition and refined preprocessing to solidify the foundation of assessment data: It adopts a three-dimensional acquisition mode of "static survey + dynamic monitoring + non-contact detection" to integrate four types of multi-source data: geological, construction, environmental, and UAV lidar detection. At the same time, it completes standardized preprocessing such as outlier removal, data normalization, and missing value completion through the 3σ criterion, Z-score algorithm, and KNN algorithm. The data processing delay is ≤5s, which effectively solves the problems of single data dimension and rough processing in traditional assessment methods, and greatly improves the completeness and accuracy of assessment data. Furthermore, the scheme constructs a dynamic weighting system to adapt to the dynamic risk characteristics of tunnel construction: it adopts a combination mode of subjective weighting by AHP and objective weighting by entropy weighting, and can dynamically adjust the weight of each criterion layer according to the tunnel construction stage. During the excavation period, the weight of construction factors is increased to 0.42, and after the support is completed, the weight of geological factors is increased to 0.51. This solves the problem that the weights of traditional assessment methods are static and do not match the actual risk characteristics, making the risk factor weight settings more in line with the actual construction and improving the rationality of the assessment. Furthermore, the proposed triple-fusion evaluation model significantly improves the accuracy of risk assessment: it integrates three algorithms—fuzzy comprehensive evaluation, BP neural network, and Monte Carlo simulation—to complete the entire risk calculation process of "first-level assessment - second-level optimization - third-level verification." Verified by actual measurements in 30 monitoring sections of the Xinjiang Kashuang Water Diversion Tunnel, the accuracy of this method in assessing landslide risk reached 92.3%, a 15.6% improvement over the traditional single fuzzy comprehensive evaluation method. This effectively solves the problems of traditional methods being greatly affected by subjective factors and lacking multi-algorithm verification, achieving accurate determination of landslide risk levels. Furthermore, the low-latency assessment and early warning system of this solution enables rapid risk response: relying on GPU servers and 5G slicing network hardware support, an efficient system architecture has been built. The triple fusion assessment module is accelerated by NVIDIA A100 GPU server, and the early warning latency of the risk output and early warning module is ≤10s. The risk level can be pushed in real time and sound and light alarms can be achieved through construction terminals and LED early warning screens, which solves the problem of delayed early warning in traditional assessment methods and buys time for construction parties to take timely measures. Furthermore, the proposed tiered risk assessment scheme, validated through practical application, provides a scientific basis for construction decisions: Based on the assessment results, risks are categorized into low, medium, and high risk levels, with corresponding targeted mitigation recommendations. High-risk sections are addressed through work stoppage and evacuation, coupled with advanced small-diameter pipe support. All mitigation recommendations have been validated through actual tunnel construction cases. After mitigation, the risk assessment value of the affected sections significantly decreases to the low-risk range. This solves the problem of traditional assessment methods that only determine risk without providing scientific mitigation guidance, achieving a closed-loop management system of "risk assessment - precise mitigation." This effectively reduces the incidence of tunnel collapses while avoiding the waste of construction resources caused by over-assessment, ensuring the safe and efficient progress of tunnel construction.

[0027] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

Claims

1. A comprehensive risk assessment method for tunnel collapse disasters, comprising step 1: multi-source data acquisition, step 2: data preprocessing, step 3: construction of a dynamic indicator system, step 4: calculation of a triple fusion assessment model, and step 5: risk level output and verification, characterized in that: Step 1: Multi-source data acquisition includes step 101 geological static data, step 102 construction dynamic data, step 103 environmental data, and step 104 UAV 3D data; Step 3: Construction of the dynamic indicator system includes Step 301 Criterion Layer 1, Step 302 Criterion Layer 2, and Step 303 Criterion Layer 3; Step 4: The calculation of the triple fusion evaluation model includes step 401, first-level evaluation, step 402, second-level optimization, and step 403, third-level verification.

2. The method and system for comprehensive risk assessment of tunnel collapse disasters according to claim 1, characterized in that: Step 101: Geological static data collection indicators include surrounding rock grade, rock mass integrity coefficient, and fault density. The data are collected using a combination of ground-penetrating radar and borehole sampling. The measured data for section K10+230 of a certain tunnel are: surrounding rock grade IV, rock mass integrity coefficient 0.62, and fault density 0.3 faults / m.

3. The method and system for comprehensive risk assessment of tunnel collapse disasters according to claim 1, characterized in that: The construction dynamic data in step 102 includes the following indicators: the arch settlement rate, the surrounding convergence value, and the support stress, which are monitored in real time by fiber optic grating sensors and convergence meters. The monitoring data for the K10+230 section of the tunnel are: arch settlement 1.8 mm / d, convergence value 3.2 mm, and support stress 18 MPa.

4. The method and system for comprehensive assessment of tunnel collapse disaster risk according to claim 1, characterized in that: The environmental data in step 103 focuses on collecting daily rainfall, groundwater level, and air humidity using rain gauges and water level sensors. The daily rainfall in this monitoring section is 25 mm, and the groundwater level is 6 m deep. On-site monitoring specifications are followed.

5. The method and system for comprehensive risk assessment of tunnel collapse disasters according to claim 1, characterized in that: Step 104: UAV 3D data: focusing on rock mass fracture width and tunnel face deformation, using a point cloud density of 50 points / cm. 2 The laser radar drone was used for detection; the measured width of the crack in this section was 0.8 mm and the local protrusion on the face of the tunnel was 2 cm, in accordance with the non-contact measurement standard.

6. The method and system for comprehensive risk assessment of tunnel collapse disasters according to claim 1, characterized in that: Step 2: Data preprocessing; Outlier removal: The 3σ criterion is used to filter the data. For example, if the threshold for the arch subsidence rate is set to 5 mm / d, and a monitoring point initially collects data of 6.2 mm / d, it is determined to be an outlier and removed. Normalization: The Z-score algorithm is used to standardize data of different dimensions. The algorithm formula is: Z=(x-μ) / Ï; (Where x is the original data, μ is the mean, and σ is the standard deviation), taking a support stress of 18MPa as an example, the standardized value is calculated to be 0.72; Data completion: KNN algorithm is used to interpolate for missing data (neighborhood parameter k=5). For a certain period of time, the groundwater level data is missing. By calling the synchronous data of 5 surrounding monitoring points, the interpolation is completed to 6.1m.

7. The method and system for comprehensive assessment of tunnel collapse disaster risk according to claim 1, characterized in that: Step 3: Construction of a dynamic indicator system; The indicator system uses "collapse risk level" as the target layer, with three criterion layers and corresponding indicator layers. The weights and indicator thresholds for each layer are as follows: Step 301 Criterion Layer 1: Geological Factors, basic weight 0.45, includes three indicator layers: Surrounding rock grade: dynamic weight 0.18, threshold range is II-V; Rock mass integrity coefficient: dynamic weight 0.12, threshold range 0.3-0.8; Fault density: dynamic weight 0.08, threshold range 0-0.5 faults / m; Step 302, Criterion Layer Two: Construction Factors, with a basic weight of 0.35, includes three indicator layers: The rate of arch subsidence is: dynamic weight 0.15, threshold range 0-5mm / d (where >2mm / d triggers an early warning). Support stress: dynamic weight 0.10, threshold range 0-30MPa (where >25MPa triggers an early warning); Vibration frequency of construction equipment: dynamic weight 0.06, threshold range 0-10Hz (where >8Hz triggers an early warning); Step 303, Criterion Layer 3: Environmental Factors, with a base weight of 0.20, includes three indicator layers: Daily rainfall: dynamic weight 0.09, threshold range 0-50mm (where >20mm triggers an alert); Groundwater level depth: dynamic weight 0.07, threshold range 0-10m (where <5m increases the risk level); Rock mass fracture width: dynamic weight 0.04, threshold range 0-2mm (where >1mm triggers an early warning).

8. The method and system for comprehensive risk assessment of tunnel collapse disasters according to claim 1, characterized in that: Step 4: Calculation of the triple fusion evaluation model; Step 401: Level 1 Assessment: Fuzzy Comprehensive Evaluation; Establish a rating scale: low risk (0-0.3), medium risk (0.3-0.6), high risk (0.6-1.0); Constructing a fuzzy matrix: Taking geological factors in segment K10+230 as an example; Level 1 output result: Overall evaluation score of 0.58, corresponding to medium risk level; Step 402 Secondary Optimization: Backpropagation Neural Network; Training data support: The model was trained using 1200 historical landslide examples, including 320 small landslides, 580 medium landslides, and 300 large landslides. Network structure design: The input layer contains 12 indicators, the hidden layer has 8 neurons, and the output layer corresponds to 3 risk levels; Optimization results: The fuzzy comprehensive evaluation result of 0.58 was corrected to 0.63, corresponding to the high-risk level, and the model test accuracy reached 89.7%. Step 403: Level 3 Validation: Monte Carlo Simulation; Uncertainty parameters selected: surrounding rock grade (value range IV-V), crown settlement rate (value range 1.8-2.5mm / d); Simulation parameter settings: After 1000 simulations, the risk probability distribution was found to be 78.3% high risk and 21.7% medium risk. Final assessment result: High risk level.

9. The method and system for comprehensive assessment of tunnel collapse disaster risk according to claim 1, characterized in that: Step 5: Risk level output and verification; Low risk: The comprehensive evaluation value ranges from 0 to 0.

3. The recommended treatment is routine monitoring (monitoring frequency once every 2 hours). The verification case is a tunnel section at K8+450, with an evaluation value of 0.

27. No collapse occurred during construction. Medium risk: The comprehensive evaluation value ranges from 0.3 to 0.

6. The recommended treatment is to increase the frequency of monitoring (once every 30 minutes) and strengthen the support. The verification case is a local area of ​​section K8+450 of a tunnel, where the evaluation value was 0.

45. After the support was reinforced, the evaluation value dropped to 0.

29. High risk: The comprehensive evaluation value ranges from 0.6 to 1.

0. The recommended treatment is to stop work and evacuate + special reinforcement (advanced small pipe support); the verification case is a tunnel section K10+230, with an evaluation value of 0.

63. After special reinforcement, the evaluation value dropped to 0.

25. Verification data source: 30 monitoring sections of the Xinjiang Kashuang Water Diversion Tunnel. The accuracy rate of this method reached 92.3%, which is 15.6% higher than the traditional fuzzy evaluation method.

10. A comprehensive risk assessment system for tunnel collapse disasters, characterized in that: Includes the following system modules; Multi-source data acquisition module: Its core function is to receive four types of data: geological, construction, environmental, and UAV data, and it supports the transmission of three protocols: RS485, 5G, and IPsec. Hardware and software support includes fiber optic grating sensors, lidar UAVs, and data acquisition gateways. Data preprocessing module: Core functions include outlier removal, normalization, and data completion, with a processing latency of ≤5s; hardware and software support is provided by edge computing nodes; Indicator system management module: Core functions include custom indicator configuration, threshold parameter setting, and dynamic weight algorithm integration; hardware and software support is a visual configuration platform; Dynamic weight calculation module: Its core function is to realize the collaborative calculation of AHP algorithm and entropy weight algorithm; the hardware and software support is the Python programming environment; Triple fusion evaluation module: The core function is to integrate three algorithms: fuzzy evaluation, BP neural network and Monte Carlo simulation; hardware and software support includes GPU server, TensorFlow framework and MATLAB tool. Risk output and early warning module: The core functions are risk level visualization display, risk zoning map generation, sound and light alarm and SMS notification, with an early warning delay of ≤10s; Hardware and software support includes industrial tablets, SMS modems, and LED warning screens; Database module: Its core functions are to store historical data, real-time monitoring data, and evaluation result data, and it supports MySQL database and Redis caching technology; the hardware and software support is cloud server.

Images