A safety step distance monitoring method and system for tunnel safety

Abstract

The present invention provides a safety step distance monitoring method and system for tunnel safety, including: deploying an edge computing module at the sensor node to preprocess and compress the collected monitoring data, reducing the data transmission volume through data dimensionality reduction and feature extraction, alleviating the data transmission pressure, and improving the data transmission efficiency and real-time performance; adopting an incremental learning algorithm to perform online analysis on the monitoring data, dynamically updating the anomaly detection model according to the newly collected data samples, adaptively adjusting the warning threshold, improving the accuracy and real-time performance of anomaly recognition, and reducing false alarms and missed alarms; comprehensively applying the multi-sensor data fusion technology to perform correlation analysis on the sensor data of different types and different positions, improving the reliability and accuracy of the monitoring data through data redundancy and complementarity, and realizing a comprehensive assessment of the safety status of tunnel construction.

Description

Technical Field

[0001] The present invention belongs to the technical field of tunnel safety, and particularly relates to a safety step distance monitoring method and system for tunnel safety. Background Art

[0002] During the tunnel construction process, it is necessary to monitor the structural parameters such as displacement and strain inside the tunnel and the environmental parameters such as temperature and humidity in real time to ensure the construction safety and quality. However, the internal environment of the tunnel is complex, the layout position and quantity of sensors are limited, and data transmission faces problems such as signal attenuation and interference. At the same time, tunnel construction is a dynamic process, and the monitoring system needs to adjust the sensor layout and data transmission path in real time according to the construction progress and environmental changes to ensure the timeliness and accuracy of the monitoring data.

[0003] In addition, the processing and analysis of massive monitoring data also pose challenges to the computing power and storage capacity of the system.

[0004] How to optimize the sensor layout scheme under the condition of limited sensor resources, establish a stable and reliable data transmission channel, and quickly identify abnormal situations from massive monitoring data is the key technical problem faced by the tunnel construction safety monitoring system. Summary of the Invention

[0005] The purpose of the present invention is to propose a safety step distance monitoring method and system for tunnel safety. By applying technologies such as multi-sensor data fusion, adaptive routing algorithm, edge computing, and machine learning, an intelligent tunnel construction safety monitoring system is constructed to provide reliable technical guarantees for the safety and quality of tunnel construction.

[0006] To achieve the above purpose, in the first aspect of the present invention, a safety step distance monitoring method for tunnel safety is provided. The method includes:

[0007] S1. Dynamically optimize the sensor layout scheme according to the tunnel construction progress and environmental changes. Design a multi-objective optimization algorithm based on the layout position, layout quantity, and data transmission path of the sensors, and adjust the sensor layout in real time according to the optimization result;

[0008] S2. According to the optimized sensor layout, for the wireless transmission environment between the edge sensor devices inside the tunnel and the server, adopt an adaptive routing algorithm to dynamically select the optimal data transmission path. By evaluating the transmission link quality in real time, predicting the signal attenuation and interference situation, and selecting the transmission path with the strongest signal and the smallest attenuation, ensure the stability and reliability of data transmission;

[0009] S3. Deploy an edge computing module at the sensor edge node, and preprocess, compress, reduce the dimension, and extract features from the collected monitoring data in the edge computing module;

[0010] S4. Construct a digital twin model of the tunnel environment and equipment status based on the extracted features. Through the real-time interaction of physical information and virtual information, achieve a comprehensive perception and dynamic tracking of the tunnel construction process, conduct a fusion analysis of the monitoring data of the sensors, promptly detect abnormal situations, and provide a basis for safety early warning;

[0011] S5. According to the detection data of the sensors used in the digital twin model, adopt an incremental learning algorithm to conduct an online analysis of the monitoring data of the sensors. According to the newly collected data samples, dynamically update the anomaly detection model and adaptively adjust the early warning threshold;

[0012] S6. Comprehensively apply multi-sensor data fusion technology to conduct a correlation analysis of the sensor data of different types and at different positions in the tunnel. Through data redundancy and complementarity, conduct a comprehensive assessment of the safety status of tunnel construction;

[0013] S7. Establish a data processing architecture based on cloud-edge collaboration, store and calculate the massive monitoring data in layers. The edge side is responsible for the real-time processing and analysis of the data, and the cloud side is responsible for the long-term storage and mining of the data. Through cloud-edge collaboration, achieve the efficient management and utilization of the monitoring data, and provide comprehensive data support for the safety of tunnel construction.

[0014] Further, the S1 specifically includes:

[0015] Obtain the tunnel construction progress and environmental change data, and judge whether it is necessary to adjust the sensor layout plan according to the preset evaluation rules. The evaluation rules are whether the preset indicators exceed the preset thresholds;

[0016] If adjustment is required, trigger the dynamic optimization process of the sensor layout plan;

[0017] Through a multi-objective optimization algorithm, with the sensor layout position, layout quantity, and data transmission path as optimization variables, and the monitoring data quality, transmission delay, and energy consumption as optimization objectives, establish a sensor layout optimization model;

[0018] Use the particle swarm optimization algorithm to solve the sensor layout optimization model and output the optimized sensor layout plan; according to the optimized sensor layout plan, generate a sensor layout adjustment instruction and send the instruction to the corresponding sensor nodes:

[0019] For the sensor nodes that need to be added, control their deployment and data acquisition behaviors;

[0020] For the sensor nodes that need to be removed, control them to stop data acquisition and remove them from the network;

[0021] For the sensor nodes that need to adjust their positions or data transmission paths, control them to perform the corresponding adjustment operations;

[0022] During the tunnel construction and environmental monitoring process, the monitoring data collected by each sensor node is obtained, and the data is preprocessed. Through data processing, the adaptability of the current sensor deployment scheme is evaluated:

[0023] If it is found during the monitoring process that the data quality deteriorates, the transmission delay increases, or the energy consumption is abnormal, that is, the data quality index is lower than the preset threshold, the transmission delay exceeds the preset threshold, or the energy consumption exceeds the preset threshold, then the sensor deployment optimization process is triggered again, and the sensor deployment is dynamically adjusted to ensure the reliable operation of the monitoring system;

[0024] Track the tunnel construction progress and environmental changes. According to the changes in monitoring requirements, trigger the sensor deployment optimization once every certain period of time. At the same time, when it is found during the monitoring process that the sensor deployment does not meet the current requirements, trigger the sensor deployment optimization;

[0025] Through regular and irregular optimizations, dynamically adapt to the sensor deployment scheme.

[0026] Furthermore, the steps of the particle swarm optimization algorithm include:

[0027] Initialize the particle swarm, calculate the fitness of each particle, and update the position and velocity of the particle until the termination condition is met;

[0028] The position vector of the particle represents the sensor deployment scheme, and the fitness function is designed according to the optimization goal;

[0029] Through iterative optimization, obtain the optimal sensor deployment scheme that meets the monitoring requirements, and determine the deployment position, deployment quantity, and data transmission path of each sensor.

[0030] Furthermore, the S2 specifically includes:

[0031] According to the wireless signal transmission environment parameters inside the tunnel, obtain multiple real-time link quality evaluation indicators;

[0032] Preprocess the obtained link quality evaluation index data to improve the data quality;

[0033] Input the preprocessed link quality evaluation index data into a pre-constructed link quality prediction model based on a long short-term memory neural network, and calculate the predicted link quality values of each candidate transmission path for a period of time in the future through this model;

[0034] According to the predicted link quality values of each candidate transmission path, combined with the preset link quality threshold, use the fuzzy analytic hierarchy process to comprehensively consider the link bandwidth and delay, and score the comprehensive transmission performance of each candidate transmission path to obtain the comprehensive score value of each candidate transmission path;

[0035] Sort the candidate transmission paths in descending order according to the comprehensive score value, and select the candidate transmission path with the highest comprehensive score value as the current optimal data transmission path;

[0036] If the comprehensive score value of the current optimal transmission path is lower than the preset path switching threshold, trigger the path switching mechanism, select the candidate transmission path with the second highest comprehensive score in the sorting result as the new optimal data transmission path, and switch the data transmission to this path;

[0037] During the data transmission process, monitor the transmission quality of the current link and obtain real-time link quality parameters:

[0038] When the link quality drops below the preset threshold and lasts for more than a certain time, re-trigger the adaptive routing algorithm, update the link quality prediction model, and dynamically select the optimal transmission path;

[0039] At the same time, for the wireless signal interference factors existing inside the tunnel, adopt the least mean square error adaptive filtering algorithm to adjust the equalization coefficient of the received signal in real time.

[0040] Further, the S3 specifically includes:

[0041] If the amount of monitoring data collected by the sensor node is large, deploy the edge computing module at the sensor node, preprocess and compress the monitoring data, and reduce the data transmission volume through data dimensionality reduction and feature extraction:

[0042] First, according to the characteristics of the preprocessed monitoring data, use the principal component analysis method for data dimensionality reduction, extract the main features of the data, remove redundant information, and reduce the data dimension;

[0043] Then, take the main features extracted by the principal component analysis as the input, and compress and encode the data through the LZW lossless compression algorithm;

[0044] Dynamically adjust the number of principal components of the principal component analysis and the dictionary size of the LZW compression algorithm according to the CPU and memory resources of the edge computing module and the real-time requirements of the monitoring data, and minimize the data transmission volume while meeting the real-time requirements;

[0045] If the characteristics of the monitoring data are relatively complex, adopt the convolutional neural network algorithm for feature extraction, and automatically learn the high-level features of the data through convolutional and pooling operations to improve the effect of data dimensionality reduction and compression:

[0046] Fuse the high-level feature map extracted by the convolutional neural network with the result of the principal component analysis as the input of the LZW compression algorithm;

[0047] Adaptive adjust the compression ratio, fragment size, and retransmission times of data transmission according to the parameters of the sensor nodes;

[0048] Through task offloading and load balancing between the edge computing module and the cloud server, offload computationally intensive tasks to the cloud to reduce the processing pressure on the edge computing module. At the same time, place storage-intensive tasks on the edge side to reduce the storage overhead of the cloud and improve the real-time performance and scalability of the overall system.

[0049] Further, the S4 specifically includes:

[0050] Deploy various sensor devices at the tunnel construction site to collect the physical information of the tunnel in real time and obtain the virtual information of the tunnel at the same time;

[0051] Match and associate the collected physical information with the virtual information to construct a digital twin model of the tunnel construction process; wherein, the digital twin model includes the three-dimensional geometric information, environmental parameters, and equipment status of the tunnel;

[0052] Based on the constructed digital twin model, perform fusion analysis on various monitoring data;

[0053] Adopt the Kalman filter algorithm, set corresponding filtering parameters according to the specific frequency range of the tunnel environment, remove noise from the monitoring data, and at the same time extract the key features in the data and perform feature enhancement;

[0054] Use the association rule mining algorithm to analyze the association rules between the tunnel environment parameters and the equipment status;

[0055] By setting the support and confidence thresholds, mine the frequently occurring parameter combination patterns to construct a tunnel safety status evaluation model; if the monitoring data exceeds the preset threshold, it is judged as an abnormal situation;

[0056] Dynamically update the digital twin model according to the tunnel construction progress plan to achieve real-time tracking of the tunnel construction process;

[0057] Through three-dimensional visualization technology, visually display the construction status and environmental information of the tunnel to provide intuitive decision-making support for construction management personnel;

[0058] When abnormal situations are found through monitoring data analysis, trigger the early warning mechanism in a timely manner;

[0059] Adopt the decision tree algorithm, and through the branch structure of the decision tree, automatically generate corresponding emergency plans according to the early warning mechanism; wherein, the emergency plans include the processing procedures for abnormal situations, the required resources, and the personnel division of labor;

[0060] Push the early warning information and emergency plans to relevant personnel in a timely manner to assist in the tunnel safety management work. At the same time, feedback the detailed information of abnormal situations to the digital twin model, and display the abnormal situations through the visual interface of the model to provide an intuitive basis for subsequent construction decisions.

[0061] Further, the step S5 specifically includes:

[0062] Obtain real-time monitoring data, and use the real-time monitoring data as new data samples for the support vector machine incremental learning algorithm to dynamically update the parameters of the SVM anomaly detection model;

[0063] Use the updated SVM anomaly detection model to perform online analysis on the real-time monitoring data to identify the abnormal data therein;

[0064] For the identified abnormal data, use the statistical process control method to adaptively adjust the early warning threshold, and improve the accuracy of anomaly identification by dynamically calculating new control limits;

[0065] Use the confusion matrix to evaluate the anomaly identification performance of the adjusted early warning threshold. If both the accuracy rate and the recall rate meet the requirements, apply the adjusted early warning threshold to subsequent anomaly detection;

[0066] If the performance still does not meet the requirements, add the misidentified samples to the incremental training set of the SVM, and optimize the anomaly detection model through incremental learning;

[0067] Iterate the above steps to continuously improve the performance of the anomaly detection system until the actual application requirements are met;

[0068] Finally, deploy the optimized SVM anomaly detection model to the production environment to perform online anomaly analysis and early warning on the real-time monitoring data, and timely discover and handle abnormal situations in the production process.

[0069] Further, the step S6 specifically includes:

[0070] Obtain the monitoring data collected by tunnel sensors of different types and at different locations, and clean the obtained monitoring data;

[0071] Perform exploratory analysis on the cleaned data, and analyze the distribution characteristics and trend changes of the data through visualization and other methods, and perform feature selection and feature extraction;

[0072] Analyze the correlation between the monitoring data of different sensors, and use the association rule mining algorithm to discover the frequent patterns and association rules in the monitoring data, and mine the hidden safety state features in the data;

[0073] Based on the results of correlation analysis and association rule mining, select machine learning algorithms such as decision trees or support vector machines to establish a tunnel construction safety status evaluation model, and use methods such as cross-validation to train and optimize the model;

[0074] Normalize the monitoring data of different sensors, and then fuse multi-source heterogeneous data;

[0075] On the basis of fusion, use the Kalman filter algorithm to process the data and dynamically track the change trend of the tunnel safety status;

[0076] According to historical data and expert experience, preset the threshold of the tunnel construction safety status;

[0077] Compare the output of the safety status evaluation model with the preset threshold. When the evaluation result exceeds the threshold, trigger the warning mechanism;

[0078] Send the warning information to relevant personnel to prompt safety hazard investigation and handling.

[0079] Further, the S7 specifically includes:

[0080] According to the characteristics of the monitoring data of tunnel construction, design a cloud-edge collaborative data processing architecture and divide the responsibilities of the cloud side and the edge side;

[0081] Deploy a real-time data processing module on the edge side, use Apache Flink to clean and filter the collected monitoring data, and use Kafka for data stream processing to extract key features;

[0082] Judge whether there is an abnormal situation through preset threshold rules: if an abnormality is detected, immediately trigger a warning and upload the relevant data to the cloud through the MQTT protocol for further analysis;

[0083] The cloud receives the monitoring data uploaded by the edge side, combines historical data and the expert knowledge base, and uses the isolation forest algorithm to detect anomalies in the data, identify hidden abnormal patterns and trends, and evaluate the safety risks of tunnel construction;

[0084] According to the results of the cloud analysis, send the optimized data processing parameters to the edge side through the RESTful API to dynamically adjust the real-time data analysis strategy and improve the efficiency and accuracy of edge computing;

[0085] Use visualization tools to present the analysis results of the cloud to users in the form of a dashboard to assist in safety decision-making at the construction site;

[0086] For a large amount of monitoring data, the InfluxDB time series database is used for storage. By utilizing the writing and querying capabilities, efficient management and retrieval of the monitoring data are achieved, and Prometheus is used for real-time data collection and monitoring.

[0087] In the second aspect of the present invention, a safety step distance monitoring system for tunnel safety is provided. The system includes:

[0088] A sensor layout optimization module, which is used to dynamically optimize the sensor layout plan according to the tunnel construction progress and environmental changes. By designing a multi-objective optimization algorithm based on the layout position, layout quantity, and data transmission path of the sensors, and adjusting the sensor layout in real time according to the optimization results;

[0089] An adaptive routing module, which is used to adopt an adaptive routing algorithm for the wireless transmission environment between the edge sensor devices inside the tunnel and the server according to the optimized sensor layout, dynamically select the optimal data transmission path. By evaluating the transmission link quality in real time, predicting signal attenuation and interference situations, and selecting the transmission path with the strongest signal and the least attenuation, the stability and reliability of data transmission are ensured;

[0090] An edge computing module, which is used to deploy an edge computing module at the sensor edge node, preprocess and compress the collected monitoring data in the edge computing module, and perform data dimensionality reduction and feature extraction;

[0091] A digital twin module, which is used to construct a digital twin model of the tunnel environment and equipment status according to the extracted features. Through the real-time interaction of physical information and virtual information, comprehensive perception and dynamic tracking of the tunnel construction process are realized, the monitoring data of the sensors are fused and analyzed, abnormal situations are discovered in time, and a basis for safety early warning is provided;

[0092] An incremental learning module, which is used to perform online analysis of the monitoring data of the sensors by adopting an incremental learning algorithm according to the detection data of the sensors used in the digital twin model. According to the newly collected data samples, the abnormal detection model is dynamically updated, and the warning threshold is adaptively adjusted;

[0093] A safety assessment module, which is used to comprehensively apply multi-sensor data fusion technology, conduct correlation analysis on the sensor data of different types and positions inside the tunnel, and comprehensively evaluate the safety status of tunnel construction through data redundancy and complementarity;

[0094] A cloud-edge collaboration module, which is used to establish a data processing architecture based on cloud-edge collaboration, store and calculate the large amount of monitoring data in layers. The edge side is responsible for real-time data processing and analysis, and the cloud side is responsible for long-term data storage and mining. Through cloud-edge collaboration, efficient management and utilization of the monitoring data are realized, providing comprehensive data support for tunnel construction safety.

[0095] The beneficial technical effects of the present invention are at least as follows:

[0096] By dynamically optimizing the sensor layout scheme, the present invention adopts an adaptive routing algorithm to select the optimal data transmission path, and deploys an edge computing module at the sensor node for data preprocessing and compression. At the same time, a digital twin model of the tunnel environment and equipment status is constructed to achieve comprehensive perception and dynamic tracking. An incremental learning algorithm is used to perform online analysis on the monitoring data and dynamically update the anomaly detection model. The multi-sensor data fusion technology is comprehensively applied to improve the reliability and accuracy of the monitoring data. A data processing architecture based on cloud-edge collaboration is established to achieve efficient management and utilization of the monitoring data. The present invention can adapt to the tunnel construction progress and environmental changes, optimize the sensor layout and data transmission, improve the accuracy and real-time performance of anomaly recognition, and provide comprehensive data support and early warning basis for tunnel construction safety. Through organic combination, the present invention forms an efficient, accurate and reliable tunnel safety step distance monitoring system, providing comprehensive technical support for tunnel construction safety. Description of the Drawings

[0097] The present invention is further described with reference to the drawings. However, the embodiments in the drawings do not constitute any limitation to the present invention. For those of ordinary skill in the art, other drawings can also be obtained according to the following drawings without creative efforts.

[0098] Figure 1 It is a flowchart of a safety step distance monitoring method for tunnel safety according to the present invention.

[0099] Figure 2 It is a schematic diagram of a safety step distance monitoring method and system for tunnel safety according to the present invention.

[0100] Figure 3 It is another schematic diagram of a safety step distance monitoring method and system for tunnel safety according to the present invention.

[0101] Figure 4 It is a schematic diagram of the structure of a safety step distance monitoring system for tunnel safety according to the present invention. Detailed Embodiments

[0102] The embodiments of the present invention are described in detail below. The examples of the embodiments are shown in the drawings, where the same or similar reference numerals denote the same or similar elements or elements with the same or similar functions throughout. The embodiments described below by referring to the drawings are exemplary and are only used to explain the present invention, and should not be construed as a limitation to the present invention.

[0103] As Figures 1 - 3 shown, a safety step distance monitoring system for tunnel safety provided by an embodiment of the present invention may specifically include:

[0104] S101. Dynamically optimize the sensor layout plan according to the tunnel construction progress and environmental changes. Through a multi-objective optimization algorithm, comprehensively consider factors such as the sensor layout position, layout quantity, and data transmission path to obtain the optimal sensor layout plan that meets the monitoring requirements, and adjust the sensor layout in real time according to the optimization results.

[0105] Obtain the tunnel construction progress and environmental change data, and judge whether it is necessary to adjust the sensor layout plan according to the preset evaluation rules. The evaluation rules include whether indicators such as monitoring data quality, transmission delay, and energy consumption exceed the preset thresholds.

[0106] If adjustment is required, trigger the dynamic optimization process of the sensor layout plan.

[0107] Among them, the monitoring data quality is measured by the accuracy Q and integrity C of the data. The transmission delay is measured by the average transmission time T of the data from the sensor node to the data center. The energy consumption is measured by the average power consumption E of the sensor node. It is expressed as follows:

[0108]

[0109] Among them, D i and respectively represent the actual value and predicted value of the i-th data point, N represents the total number of data points, and t i represents the transmission time of the i-th data point, and e i represents the power consumption of the i-th sensor node.

[0110] Through a multi-objective optimization algorithm, with the sensor layout position, layout quantity, and data transmission path as the optimization variables and the monitoring data quality, transmission delay, and energy consumption as the optimization objectives, establish a sensor layout optimization model. Among them, the monitoring data quality can be measured by indicators such as the accuracy and integrity of the data, the transmission delay can be measured by the average transmission time of the data from the sensor node to the data center, and the energy consumption can be measured by the average power consumption of the sensor node.

[0111] Among them, set the optimization variables: the sensor layout position P, the layout quantity N, and the data transmission path R.

[0112] Optimization objectives: monitoring data quality Q, transmission delay T, and energy consumption E. It is expressed as follows:

[0113]

[0114] Among them, α, β, and γ are weight coefficients used to balance the importance of different optimization objectives.

[0115] Establish a sensor layout optimization model, which is expressed as follows:

[0116]

[0117] The particle swarm optimization algorithm is used to solve the above optimization model. The key steps of the particle swarm optimization algorithm include:

[0118] Initialize the particle swarm, calculate the fitness of each particle, and update the position and velocity of the particles until the termination condition is met. The particle position x i =(P i , N i , R i ), the particle velocity v i , and the fitness function:

[0119] f(x i )=(αQ(x i )+βT(x i )+γE(x i ))

[0120] The position vector of the particle represents the sensor layout scheme, and the fitness function is designed according to the optimization goal. The update rules are as follows:

[0121] v i (t + 1)=ωv i (t)+c1r1(p i -x i (t))+c2r2(p best -x i (t))

[0122] x i (t + 1)=x i (t)+v i (t + 1)

[0123] where ω represents the inertia weight, c1 and c2 are learning factors, r1 and r2 are random numbers, p i is the best position of particle i, and p best is the global best position.

[0124] Through iterative optimization, an optimal sensor layout scheme that meets the monitoring requirements is obtained, and the layout position, layout quantity, and data transmission path of each sensor are determined.

[0125] According to the optimized sensor layout scheme, generate sensor layout adjustment instructions and send the instructions to the corresponding sensor nodes.

[0126] For the sensor nodes that need to be added, control their deployment and data collection behaviors; for the sensor nodes that need to be removed, control them to stop data collection and remove them from the network; for the sensor nodes that need to adjust their positions or data transmission paths, control them to perform the corresponding adjustment operations.

[0127] During the tunnel construction and environmental monitoring process, continuously obtain the monitoring data collected by each sensor node, and clean, fuse, and analyze the data.

[0128] The methods of data cleaning include outlier detection, noise filtering, etc., the methods of data fusion include Kalman filtering, Bayesian estimation, etc., and the methods of data analysis include statistical analysis, trend analysis, etc.

[0129] Through data processing, evaluate the adaptability of the current sensor deployment scheme.

[0130] If problems such as data quality degradation, increased transmission delay, or abnormal energy consumption are found during the monitoring process, that is, the data quality index is lower than the preset threshold, the transmission delay exceeds the preset threshold, or the energy consumption exceeds the preset threshold, then re-trigger the sensor deployment optimization process, dynamically adjust the sensor deployment, and ensure the reliable operation of the monitoring system.

[0131] Continuously track the tunnel construction progress and environmental changes. According to the changes in monitoring requirements, trigger the sensor deployment optimization once every certain period (such as one week or one month), and at the same time, when it is found during the monitoring process that the sensor deployment does not meet the current requirements, also trigger the sensor deployment optimization. Through regular and irregular optimizations, the sensor deployment scheme can dynamically adapt to the characteristics of each stage of tunnel construction and provide continuous and effective monitoring services.

[0132] S102. For the complex and changeable wireless transmission environment inside the tunnel, adopt an adaptive routing algorithm to dynamically select the optimal data transmission path. By real-time evaluating the link quality, predicting the signal attenuation and interference conditions, select the transmission path with the strongest signal and the smallest attenuation to ensure the stability and reliability of data transmission.

[0133] According to the wireless signal transmission environment parameters inside the tunnel, obtain multiple real-time link quality evaluation indicators X such as signal strength S, signal-to-noise ratio SNR, and bit error rate BER.

[0134] Preprocess the obtained link quality evaluation index data, including data cleaning, denoising, normalization, etc., to improve the data quality.

[0135] Input the preprocessed link quality evaluation index data into a pre-constructed link quality prediction model based on a long short-term memory neural network (LSTM), and calculate the predicted link quality values of each candidate transmission path for a future period of time through this model. It is expressed as follows:

[0136]

[0137] Among them, represents the predicted link quality value for a future period of time, Represents the link quality assessment index data after preprocessing; Represents the link quality assessment index data from time t - n to time t; f LSTM Represents the prediction function of the LSTM model.

[0138] According to the predicted link quality values of each candidate transmission path, combined with a preset link quality threshold, using the fuzzy analytic hierarchy process (FAHP), comprehensively considering factors such as link bandwidth and delay, score the comprehensive transmission performance of each candidate transmission path to obtain the comprehensive score value of each candidate transmission path. It is expressed as follows:

[0139]

[0140] Among them, w1, w2, and w3 are weight coefficients, representing the importance of link bandwidth B, delay D, and link quality prediction value respectively.

[0141] Sort the candidate transmission paths in descending order according to the comprehensive score value, and select the candidate transmission path with the highest comprehensive score value as the current optimal data transmission path. If the comprehensive score value of the current optimal transmission path is lower than the preset path switching threshold (such as 6), trigger the path switching mechanism, select the candidate transmission path with the second - highest comprehensive score in the sorting result as the new optimal data transmission path, and switch the data transmission to this path.

[0142] During the data transmission process, continuously monitor the transmission quality of the current link to obtain real - time link quality parameters.

[0143] When the link quality drops below the preset threshold (such as the signal - to - noise ratio is lower than 10 dB) and lasts for more than a certain time (such as 5 seconds), re - trigger the adaptive routing algorithm, update the link quality prediction model, and dynamically select the optimal transmission path to ensure transmission stability.

[0144] Aiming at wireless signal interference factors such as Doppler frequency shift and near - far effect existing inside the tunnel, use the least mean square (LMS) adaptive filtering algorithm to adjust the equalization coefficient of the received signal in real time, reduce signal distortion, improve the signal - to - noise ratio, and enhance the reliability of data transmission. It is expressed as follows:

[0145] w(n + 1) = w(n)+μ·e(n)·x(n)

[0146] Among them, w(n) is the equalization coefficient vector, μ is the step - size parameter, e(n) is the error signal, and x(n) is the input signal vector;

[0147] The error signal is calculated as follows:

[0148] e(n) = d(n)-w T(n)·x(n)

[0149] Among them, d(n) is the desired signal.

[0150] S103. Deploy an edge computing module at the sensor node to preprocess and compress the collected monitoring data. Through data dimensionality reduction and feature extraction, reduce the data transmission volume, relieve the data transmission pressure, and improve the data transmission efficiency and real-time performance.

[0151] If the amount of monitoring data collected by the sensor node is large, deploy the edge computing module at the sensor node to preprocess and compress the monitoring data. Through data dimensionality reduction and feature extraction, reduce the data transmission volume and relieve the data transmission pressure:

[0152] First, according to the characteristics of the preprocessed monitoring data, use the principal component analysis method for data dimensionality reduction, extract the main features of the data, remove redundant information, and reduce the data dimension.

[0153] Then, take the main features extracted by the principal component analysis as the input, and compress and encode the data through the LZW lossless compression algorithm. While ensuring the data integrity, minimize the data transmission volume and improve the data transmission efficiency.

[0154] Among them, the LZW lossless compression algorithm takes the main features extracted by the principal component analysis as the input, compresses and encodes the data through the LZW lossless compression algorithm. While ensuring the data integrity, minimizes the data transmission volume and improves the data transmission efficiency. The dictionary size dynamically adjusts the dictionary size of the LZW compression algorithm according to the CPU and memory resources of the edge computing module and the real-time requirements of the monitoring data.

[0155] Dynamically adjust the number of principal components of the principal component analysis and the dictionary size of the LZW compression algorithm according to the CPU and memory resources of the edge computing module and the real-time requirements of the monitoring data, and minimize the data transmission volume while meeting the real-time requirements.

[0156] If the characteristics of the monitoring data are relatively complex, use the convolutional neural network algorithm for feature extraction, and automatically learn the high-level features of the data through convolutional and pooling operations to improve the effect of data dimensionality reduction and compression. It is expressed as follows:

[0157] O = σ(X * W + b)

[0158] Among them, X is the input data, W is the convolutional kernel, b is the bias term, σ is the activation function, and O is the output of the convolutional layer;

[0159] Pooling operation, P cnn is the pooling output:

[0160] P cnn= Pool(O)

[0161] Among them, Pool is a pooling function, such as max pooling or average pooling.

[0162] Fuse the high-level feature maps extracted by the convolutional neural network with the results of principal component analysis as the input of the LZW compression algorithm. It is expressed as follows:

[0163] F fusion = concat(X pca , P cnn )

[0164] Among them, X pca represents the data after dimensionality reduction, F fusion represents the fused data, and concat represents the feature fusion operation.

[0165] According to parameters such as the network bandwidth, signal strength, and transmission delay of the sensor node, adaptively adjust the compression ratio, fragment size, and retransmission times of data transmission, while ensuring the reliability of data transmission, improving the data transmission efficiency and real-time performance.

[0166] Through task offloading and load balancing between the edge computing module and the cloud server, offload computationally intensive tasks such as data compression and feature extraction to the cloud to reduce the processing pressure on the edge computing module.

[0167] At the same time, place storage-intensive tasks such as data decompression and storage on the edge side to reduce the storage overhead of the cloud and improve the real-time performance and scalability of the overall system.

[0168] S104. Build a digital twin model of the tunnel environment and equipment status. Through the real-time interaction of physical information and virtual information, achieve a comprehensive perception and dynamic tracking of the tunnel construction process, conduct a fusion analysis of the monitoring data, promptly discover abnormal situations, and provide a basis for safety early warning.

[0169] By deploying various sensor devices at the tunnel construction site, collect physical information such as tunnel environment parameters and equipment operation status in real time, and at the same time obtain virtual information such as BIM models and construction progress plans.

[0170] Match and associate the collected physical information with virtual information to build a digital twin model of the tunnel construction process. The model contains multi-dimensional data such as the three-dimensional geometric information, environmental parameters, and equipment status of the tunnel.

[0171] Based on the built digital twin model, conduct a fusion analysis of various monitoring data.

[0172] Adopt the Kalman filter algorithm, set corresponding filtering parameters according to the specific frequency range of the tunnel environment, and remove noise from the monitoring data.

[0173] Meanwhile, extract key features in the data, such as temperature, humidity, vibration frequency, etc., and perform feature enhancement to provide more effective data support for subsequent analysis.

[0174] Use association rule mining algorithms, such as the Apriori algorithm, to analyze the association rules between tunnel environmental parameters and equipment status. By setting support and confidence thresholds, mine frequently occurring parameter combination patterns and construct a tunnel safety status assessment model.

[0175] If the monitored data exceeds the preset threshold, it is judged as an abnormal situation. According to the tunnel construction progress plan, dynamically update the digital twin model to achieve real-time tracking of the tunnel construction process.

[0176] Through 3D visualization technology, intuitively display the construction status and environmental information of the tunnel, providing intuitive decision-making support for construction management personnel.

[0177] When abnormal situations are found through monitoring data analysis, trigger the early warning mechanism in a timely manner.

[0178] Adopt decision tree algorithms, such as the ID3 algorithm, comprehensively consider features such as the type, severity, and duration of abnormal situations, and automatically generate corresponding emergency plans through the branch structure of the decision tree. The emergency plan includes the handling process of abnormal situations, required resources, personnel division of labor, etc.

[0179] Push the early warning information and emergency plan to relevant personnel in a timely manner, send notifications via text messages, emails, etc., and assist in carrying out tunnel safety management work.

[0180] Meanwhile, feedback the detailed information of abnormal situations, such as abnormal types, occurrence times, locations, etc., to the digital twin model, and display the abnormal situations through the visualization interface of the model, providing an intuitive basis for subsequent construction decisions.

[0181] S105. Adopt an incremental learning algorithm to perform online analysis on the monitoring data. According to the newly collected data samples, dynamically update the anomaly detection model, adaptively adjust the warning threshold, improve the accuracy and real-time performance of anomaly recognition, and reduce false alarms and missed alarms.

[0182] Obtain real-time monitoring data and use it as new data samples for the incremental learning algorithm of support vector machine (SVM) to dynamically update the parameters of the SVM anomaly detection model.

[0183] Using the updated SVM anomaly detection model, online analysis is performed on real-time monitoring data to identify abnormal data therein. For the identified abnormal data, the statistical process control (SPC) method is used to adaptively adjust the warning threshold, and the accuracy of anomaly identification is improved by dynamically calculating new control limits. Among them, the control limits are calculated as follows:

[0184]

[0185] Among them, is the sample mean, s is the sample standard deviation, and k is the control limit coefficient (usually taken as 3).

[0186] The adaptive adjustment is as follows:

[0187] New Threshold=AdjustT hreshold(UCL,LCL,X new )

[0188] Among them, X new is the real-time monitoring data, X new ={x1,x2,...,x new}。

[0189] The confusion matrix is used to evaluate the anomaly identification performance of the adjusted warning threshold. If both the accuracy rate and the recall rate meet the requirements, the adjusted warning threshold is applied to subsequent anomaly detection; if the performance still does not meet the requirements, the misidentified samples are added to the incremental training set of SVM, and the anomaly detection model is further optimized through incremental learning. The confusion matrix is expressed as follows:

[0190]

[0191] Among them, TP is the true positive, FP is the false positive, FN is the false negative, and TN is the true negative.

[0192] The above steps are continuously iterated to continuously improve the performance of the anomaly detection system until the actual application requirements are met. Finally, the optimized anomaly detection system is deployed to the production environment to perform online anomaly analysis and warning on real-time monitoring data, timely discover and handle abnormal situations in the production process, and ensure production safety and product quality.

[0193] S106. Comprehensively apply multi-sensor data fusion technology to conduct correlation analysis on sensor data of different types and at different positions, and improve the reliability and accuracy of monitoring data through data redundancy and complementarity to achieve a comprehensive assessment of the safety status of tunnel construction.

[0194] Obtain the monitoring data collected by tunnel sensors of different types and at different locations. For the obtained monitoring data, adopt data cleaning techniques such as median filtering and outlier detection to remove noise and outliers in the monitoring data, and improve the quality and reliability of the monitoring data.

[0195] Conduct exploratory analysis on the cleaned data. Through visualization and other means, understand the distribution characteristics, trend changes, etc. of the data, and perform feature selection and feature extraction to lay a foundation for subsequent analysis.

[0196] Use indicators such as Pearson correlation coefficient to analyze the correlation between the monitoring data of different sensors, and adopt association rule mining algorithms such as Apriori to discover frequent patterns and association rules in the monitoring data, and mine the hidden safety state characteristics in the data.

[0197] According to the results of correlation analysis and association rule mining, select machine learning algorithms such as decision trees or support vector machines to establish a tunnel construction safety state assessment model, and use methods such as cross-validation to train and optimize the model to improve the accuracy and reliability of safety state assessment.

[0198] Normalize the monitoring data of different sensors to eliminate the dimension difference, and then adopt data fusion methods such as weighted average or principal component analysis to fuse multi-source heterogeneous data. Through fusion, the complementarity of different sensor data can be fully utilized, the uncertainty of single-sensor data can be reduced, and more comprehensive and reliable tunnel safety state information can be obtained. On the basis of fusion, use the Kalman filter algorithm to further process the data and dynamically track the change trend of the tunnel safety state. Through the Kalman filter, the safety state can be predicted and updated in real time, providing a basis for timely warning.

[0199] According to historical data and expert experience, preset the threshold of the tunnel construction safety state. Compare the output of the safety state assessment model with the preset threshold. When the assessment result exceeds the threshold, trigger the warning mechanism. The warning information can be sent to relevant personnel in a timely manner by means of text messages, emails, etc., prompting them to conduct safety hazard investigation and handling.

[0200] During the entire safety state assessment process, adopt data visualization techniques such as dashboards and heat maps to intuitively display the real-time safety state of the tunnel, the change trend of sensor data, warning information, etc., providing intuitive decision-making support for managers. Visualization should run through all aspects of data analysis, modeling, warning, etc. to improve the usability and interpretability of the system. Through visualization, managers can comprehensively master the safety dynamics of tunnel construction, discover abnormalities in a timely manner, and make correct decisions.

[0201] S107. Establish a data processing architecture based on cloud-edge collaboration, store and calculate massive monitoring data in layers. The edge side is responsible for real-time data processing and analysis, and the cloud side is responsible for long-term data storage and mining. Through cloud-edge collaboration, achieve efficient management and utilization of monitoring data, and provide comprehensive data support for tunnel construction safety.

[0202] According to the characteristics of the monitoring data of tunnel construction, design a data processing architecture for cloud-edge collaboration, and reasonably divide the responsibilities of the cloud side and the edge side.

[0203] Deploy a real-time data processing module on the edge side, use Apache Flink to clean and filter the collected monitoring data, and use Kafka for data stream processing to extract key features.

[0204] Through preset threshold rules, judge whether there are abnormal situations. If an anomaly is detected, immediately trigger an alarm, and upload the relevant data to the cloud side through the MQTT protocol for further analysis.

[0205] The cloud side receives the monitoring data uploaded by the edge side, combines historical data and the expert knowledge base, and uses the Isolation Forest algorithm to perform anomaly detection on the data, identify hidden anomaly patterns and trends, and evaluate the safety risks of tunnel construction.

[0206] According to the results of the cloud-side analysis, issue the optimized data processing parameters to the edge side through the RESTful API, dynamically adjust the strategy of real-time data analysis, and improve the efficiency and accuracy of edge computing.

[0207] Use visualization tools such as Grafana to present the analysis results of the cloud side to users in the form of an intuitive dashboard to assist in safety decision-making at the construction site.

[0208] For massive monitoring data, use the InfluxDB time series database for storage, and utilize its high-performance writing and query capabilities to achieve efficient management and retrieval of monitoring data. At the same time, use Prometheus for real-time data collection and monitoring to ensure the reliability of the data processing process.

[0209] Through the architecture design of cloud-edge collaboration, combined with technologies such as stream data processing, machine learning algorithms, and time series databases, build a set of efficient, intelligent, and reliable tunnel construction safety monitoring data processing systems. This system can timely detect abnormal situations, provide accurate safety warnings, and provide visual decision-making support for the construction site, ensuring the safety and efficiency of tunnel construction.

[0210] Specifically, the edge side is responsible for real-time data processing, cleaning, filtering, and preliminary anomaly detection to ensure the real-time nature and preliminary accuracy of the data.

[0211] The cloud is responsible for in-depth data analysis, historical data management, expert knowledge base application, and optimization parameter distribution, providing advanced analysis and decision-making support.

[0212] Data visualization uses tools such as Grafana to present the analysis results to users in an intuitive form, assisting in safety decisions at the construction site.

[0213] Data storage and monitoring adopt InfluxDB and Prometheus to achieve efficient management and real-time monitoring of monitoring data, ensuring the reliability of the data processing process.

[0214] To better implement the above method, as Figure 4 shown, the present invention also proposes a safety step distance monitoring system for tunnel safety, and the system includes:

[0215] A sensor layout optimization module 101, which is used to dynamically optimize the sensor layout plan according to the tunnel construction progress and environmental changes, design a multi-objective optimization algorithm based on the layout position, layout quantity, and data transmission path of the sensors, and adjust the sensor layout in real time according to the optimization results;

[0216] An adaptive routing module 102, which is used to adopt an adaptive routing algorithm for the wireless transmission environment between the edge sensor devices inside the tunnel and the server according to the optimized sensor layout, dynamically select the optimal data transmission path, select the transmission path with the strongest signal and the smallest attenuation by evaluating the transmission link quality in real time and predicting signal attenuation and interference conditions, and ensure the stability and reliability of data transmission;

[0217] An edge computing module 103, which is used to deploy an edge computing module at the sensor edge node, preprocess and compress the collected monitoring data in the edge computing module, and perform data dimensionality reduction and feature extraction;

[0218] A digital twin module 104, which is used to construct a digital twin model of the tunnel environment and equipment status according to the extracted features, realize comprehensive perception and dynamic tracking of the tunnel construction process through real-time interaction of physical information and virtual information, perform fusion analysis on the monitoring data of the sensors, discover abnormal situations in time, and provide a basis for safety early warning;

[0219] An incremental learning module 105, which is used to perform online analysis on the monitoring data of the sensors by adopting an incremental learning algorithm according to the detection data of the sensors used in the digital twin model, dynamically update the anomaly detection model according to the newly collected data samples, and adaptively adjust the early warning threshold;

[0220] The safety assessment module 106 is used to comprehensively apply multi-sensor data fusion technology to conduct correlation analysis on sensor data of different types and at different positions in the tunnel, and through data redundancy and complementarity, comprehensively evaluate the safety status of tunnel construction;

[0221] The cloud-edge collaboration module 107 is used to establish a data processing architecture based on cloud-edge collaboration, store and calculate massive monitoring data in layers. The edge side is responsible for real-time processing and analysis of data, and the cloud side is responsible for long-term storage and mining of data. Through cloud-edge collaboration, efficient management and utilization of monitoring data are achieved, providing comprehensive data support for tunnel construction safety.

[0222] In summary, the data acquisition and layout optimization of the present invention provides a basic data source for the system, ensuring the comprehensiveness and reliability of the data. Based on the optimized layout scheme, the stability and reliability of data transmission are ensured. The data is preprocessed and compressed at the edge node to reduce the data transmission volume and improve the system efficiency. Using the preprocessed data, a digital twin model is constructed to achieve comprehensive perception and dynamic tracking. Based on the digital twin model, online analysis and anomaly detection are carried out to improve the accuracy and real-time performance of early warning. By integrating multi-sensor data, the reliability and accuracy of monitoring data are improved. Through cloud-edge collaboration, efficient management and utilization of data are achieved, ensuring the overall performance and reliability of the system. Through the organic combination of these steps, an efficient, accurate and reliable tunnel safety step monitoring system is formed, providing comprehensive technical support for tunnel construction safety.

[0223] Unless otherwise specifically stated, the relative steps, numerical expressions and values of the components and steps set forth in these embodiments do not limit the scope of the present application.

[0224] If the described functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium executable by a processor. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art or the part of the current technical solution, can be embodied in the form of a software product. The current computer software product is stored in a storage medium, including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present application. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disks or optical discs that can store program codes.

[0225] In the description of the present application, it should be noted that the orientation or positional relationship indicated by terms such as "upper" and "lower" is based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship in which the current invention product is usually placed during use. It is only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation to the present application.

[0226] In the description of the present application, it should also be noted that unless otherwise clearly specified and defined, the terms "arranged", "installed", "connected" and "coupled" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood according to specific circumstances.

[0227] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A safe step distance monitoring method for tunnel safety, characterized in that: The method comprises: S1. Dynamically optimize the sensor layout plan according to the tunnel construction progress and environmental changes, design a multi-objective optimization algorithm based on the sensor layout location, layout quantity, and data transmission path, and adjust the sensor layout in real time according to the optimization results; S2. Based on the optimized sensor layout, an adaptive routing algorithm is used to dynamically select the optimal data transmission path for the wireless transmission environment between the edge sensor devices and the server inside the tunnel. By real-time evaluation of the transmission link quality, signal attenuation and interference are predicted, and the transmission path with the strongest signal and the smallest attenuation is selected to ensure the stability and reliability of data transmission. S3. Deploy an edge computing module at the sensor edge node, preprocess and compress the collected monitoring data in the edge computing module, and perform data dimension reduction and feature extraction; S4. Build a digital twin model of the tunnel environment and equipment status based on the extracted features. Through real-time interaction between physical and virtual information, realize comprehensive perception and dynamic tracking of the tunnel construction process, conduct fusion analysis of the sensor monitoring data, discover abnormal situations in time, and provide a basis for safety warning. S5. Based on the detection data of the sensors used in the digital twin model, an incremental learning algorithm is used to analyze the monitoring data of the sensors online, and the anomaly detection model is dynamically updated based on the newly collected data samples, and the warning threshold is adaptively adjusted; S6. Comprehensively apply multi-sensor data fusion technology to correlate and analyze sensor data of different types and locations in the tunnel, and comprehensively evaluate the safety status of tunnel construction through data redundancy and complementarity; S7. Establish a data processing architecture based on cloud-edge collaboration to store and calculate massive monitoring data in layers. The edge side is responsible for real-time processing and analysis of data, and the cloud side is responsible for long-term storage and mining of data. Through cloud-edge collaboration, efficient management and utilization of monitoring data can be achieved, providing comprehensive data support for tunnel construction safety.

2. A tunnel safety-oriented safety step distance monitoring method according to claim 1, characterized in that: Said S1 specifically includes: Obtain tunnel construction progress and environmental change data, and determine whether the sensor deployment plan needs to be adjusted based on preset evaluation rules. The evaluation rules are whether the preset indicators exceed the preset thresholds. If adjustment is needed, the dynamic optimization process of the sensor deployment plan is triggered; Through a multi-objective optimization algorithm, a sensor deployment optimization model is established with sensor deployment location, deployment quantity and data transmission path as optimization variables, and monitoring data quality, transmission delay and energy consumption as optimization targets. The particle swarm optimization algorithm is used to solve the sensor layout optimization model and output the optimized sensor layout plan. According to the optimized sensor layout plan, the sensor layout adjustment instructions are generated and sent to the corresponding sensor nodes: For sensor nodes that need to be added, control their deployment and data collection behavior; For the sensor nodes that need to be removed, control them to stop data collection and remove them from the network; For sensor nodes that need to adjust their positions or data transmission paths, control them to perform corresponding adjustment operations; During the tunnel construction and environmental monitoring process, the monitoring data collected by each sensor node is obtained, the data is cleared and processed, and the adaptability of the current sensor deployment plan is evaluated through data processing: If the data quality is found to be degraded, the transmission delay is increased, or the energy consumption is abnormal during the monitoring process, that is, the data quality index is lower than the preset threshold, the transmission delay exceeds the preset threshold, or the energy consumption exceeds the preset threshold, the sensor deployment optimization process is re-triggered to dynamically adjust the sensor deployment to ensure the reliable operation of the monitoring system; Track the progress of tunnel construction and environmental changes, and trigger sensor deployment optimization at regular intervals according to changes in monitoring needs. At the same time, if it is found during the monitoring process that the sensor deployment does not meet the current needs, it will also trigger sensor deployment optimization; Dynamically adapt the sensor deployment plan through regular and irregular optimization.

3. A tunnel safety-oriented safety step distance monitoring method according to claim 2, characterized in that: The steps of the particle swarm optimization algorithm include: Initialize the particle swarm, calculate the fitness of each particle, and update the position and velocity of the particle until the termination condition is met; The particle position vector represents the sensor deployment scheme, and the fitness function is designed according to the optimization goal; Through iterative optimization, the optimal sensor deployment plan that meets monitoring needs is obtained, and the deployment location, deployment quantity and data transmission path of each sensor are determined.

4. The method for monitoring safe step distance for tunnel safety according to claim 1, characterized in that: The S2 specifically includes: According to the wireless signal transmission environment parameters inside the tunnel, multiple real-time link quality evaluation indicators are obtained; Preprocess the acquired link quality assessment indicator data to improve data quality; The preprocessed link quality evaluation index data is input into a pre-built link quality prediction model based on a long short-term memory neural network, and the predicted link quality value of each candidate transmission path in the future period of time is calculated by the model; According to the predicted link quality value of each candidate transmission path, combined with the preset link quality threshold, the fuzzy hierarchical analysis method is used to comprehensively consider the link bandwidth and delay, and the comprehensive transmission performance of each candidate transmission path is scored to obtain the comprehensive score value of each candidate transmission path; Sort the candidate transmission paths from high to low according to the comprehensive score, and select the candidate transmission path with the highest comprehensive score as the current optimal data transmission path; If the comprehensive score of the current optimal transmission path is lower than the preset path switching threshold, the path switching mechanism is triggered, and the candidate transmission path with the second highest comprehensive score in the sorting result is selected as the new optimal data transmission path, and data transmission is switched to this path; During data transmission, monitor the transmission quality of the current link and obtain real-time link quality parameters: When the link quality drops below the preset threshold for more than a certain period of time, the adaptive routing algorithm is re-triggered to update the link quality prediction model and dynamically select the optimal transmission path; At the same time, in response to the wireless signal interference factors inside the tunnel, the minimum mean square error adaptive filtering algorithm is used to adjust the equalization coefficient of the received signal in real time.

5. The method for monitoring safe step distance for tunnel safety according to claim 1, characterized in that: The S3 specifically includes: If the amount of monitoring data collected by the sensor node is large, the edge computing module is deployed at the sensor node to preprocess and compress the monitoring data, and reduce the amount of data transmission through data dimension reduction and feature extraction: First, according to the characteristics of the preprocessed monitoring data, the principal component analysis method is used to reduce the data dimension, extract the main features of the data, remove redundant information, and reduce the data dimension; Then, the main features extracted by principal component analysis are used as input, and the data is compressed and encoded using the LZW lossless compression algorithm; According to the CPU and memory resources of the edge computing module and the real-time requirements of the monitoring data, the number of principal components of the principal component analysis and the dictionary size of the LZW compression algorithm are dynamically adjusted to minimize the amount of data transmission while meeting the real-time requirements; If the features of the monitoring data are more complex, the convolutional neural network algorithm is used for feature extraction. The high-level features of the data are automatically learned through convolution and pooling operations to improve the effect of data dimensionality reduction and compression: The high-level feature map extracted by the convolutional neural network is fused with the results of principal component analysis as the input of the LZW compression algorithm; Adaptively adjust the data transmission compression ratio, fragment size and retransmission times according to the parameters of the sensor node; Through task offloading and load balancing between the edge computing module and the cloud server, computing-intensive tasks are offloaded to the cloud, reducing the processing pressure of the edge computing module. At the same time, storage-intensive tasks are performed on the edge side to reduce storage overhead in the cloud and improve the real-time and scalability of the overall system.

6. The method for monitoring safe step distance for tunnel safety according to claim 1, characterized in that: The S4 specifically includes: By deploying various sensor devices at the tunnel construction site, the physical information of the tunnel is collected in real time, and the virtual information of the tunnel is obtained at the same time; Matching and associating the collected physical information with the virtual information to construct a digital twin model of the tunnel construction process; wherein the digital twin model includes the three-dimensional geometric information, environmental parameters and equipment status of the tunnel; Based on the constructed digital twin model, various monitoring data are integrated and analyzed; The Kalman filter algorithm is used to set corresponding filter parameters according to the specific frequency range of the tunnel environment to remove noise from the monitoring data. At the same time, key features in the data are extracted and enhanced. The association rules mining algorithm is used to analyze the association rules between tunnel environmental parameters and equipment status; By setting support and confidence thresholds, frequently occurring parameter combination patterns are mined and a tunnel safety status assessment model is constructed; if the monitoring data exceeds the preset threshold, it is judged as an abnormal situation; According to the tunnel construction schedule, the digital twin model is dynamically updated to achieve real-time tracking of the tunnel construction process; Through 3D visualization technology, the construction status and environmental information of the tunnel can be directly displayed, providing intuitive decision-making support for construction managers; When abnormal conditions are found in monitoring data analysis, the early warning mechanism is triggered in time; Adopting the decision tree algorithm, through the branch structure of the decision tree, the corresponding emergency plan is automatically generated according to the early warning mechanism; wherein, the emergency plan includes the handling process of abnormal situations, the required resources and the division of labor of personnel; Early warning information and emergency plans are pushed to relevant personnel in a timely manner to assist in tunnel safety management. At the same time, detailed information of abnormal situations is fed back to the digital twin model, and the abnormal situations are displayed through the model's visual interface, providing an intuitive basis for subsequent construction decisions.

7. The method for monitoring safe step distance for tunnel safety according to claim 1, characterized in that: The S5 specifically includes: Acquire real-time monitoring data and use the real-time monitoring data as new data samples for the support vector machine incremental learning algorithm to dynamically update the parameters of the SVM anomaly detection model; Use the updated SVM anomaly detection model to analyze the real-time monitoring data online and identify abnormal data; For the identified abnormal data, the statistical process control method is used to adaptively adjust the warning threshold, and the accuracy of abnormal identification is improved by dynamically calculating new control limits; The confusion matrix is ​​used to evaluate the anomaly recognition performance of the adjusted warning threshold. If both the precision and recall rates meet the requirements, the adjusted warning threshold is applied to subsequent anomaly detection. If the performance still does not meet the requirements, the samples with identification errors are added to the incremental training set of SVM, and the anomaly detection model is optimized through incremental learning; Iterate the above steps to continuously improve the performance of the anomaly detection system until it meets actual application requirements; Finally, the optimized SVM anomaly detection model is deployed in the production environment to perform online anomaly analysis and early warning on real-time monitoring data, so as to timely discover and handle abnormal situations in the production process.

8. The method for monitoring safe step distance for tunnel safety according to claim 1, characterized in that: The S6 specifically includes: Acquire monitoring data collected by tunnel sensors of different types and locations, and clean the acquired monitoring data; Conduct exploratory analysis on the cleaned data, analyze the distribution characteristics and trend changes of the data through visualization and other methods, and perform feature selection and feature extraction; Analyze the correlation between monitoring data from different sensors, and use association rule mining algorithms to discover frequent patterns and association rules in monitoring data, and mine the hidden security status characteristics in the data; According to the results of correlation analysis and association rule mining, a machine learning algorithm such as decision tree or support vector machine is selected to establish a tunnel construction safety status assessment model, and the model is trained and optimized using methods such as cross-validation. Normalize the monitoring data from different sensors and then fuse multi-source heterogeneous data; On the basis of fusion, the Kalman filter algorithm is used to process the data and dynamically track the changing trend of the tunnel safety status; Preset thresholds for tunnel construction safety status based on historical data and expert experience; Compare the output of the security status assessment model with the preset threshold. When the assessment result exceeds the threshold, trigger the early warning mechanism. Send early warning information to relevant personnel, prompting them to check and deal with safety hazards.

9. The method for monitoring safe step distance for tunnel safety according to claim 1, characterized in that: The S7 specifically includes: Based on the characteristics of monitoring data from tunnel construction, a cloud-edge collaborative data processing architecture is designed to divide the responsibilities of the cloud and edge sides; Deploy a real-time data processing module on the edge side, use Apache Flink to clean and filter the collected monitoring data, and use Kafka to process data streams and extract key features; Through the preset threshold rules, determine whether there is an abnormal situation: if an abnormality is detected, an early warning is triggered immediately, and the relevant data is uploaded to the cloud through the MQTT protocol for further analysis; The cloud receives monitoring data uploaded by the edge side, combines historical data with expert knowledge base, and uses the isolation forest algorithm to detect anomalies in the data, identify hidden abnormal patterns and trends, and assess the safety risks of tunnel construction; According to the results of cloud analysis, the optimized data processing parameters are sent to the edge side through RESTful API, and the real-time data analysis strategy is dynamically adjusted to improve the efficiency and accuracy of edge computing; Using visualization tools, the analysis results in the cloud are presented to users in the form of dashboards to assist safety decision-making at the construction site; InfluxDB time series database is used to store massive monitoring data. Writing and querying capabilities are used to achieve efficient management and retrieval of monitoring data. Prometheus is used for real-time data collection and monitoring.

10. A safety step distance monitoring system for tunnel safety, characterized in that: The system comprises: The sensor layout optimization module is used to dynamically optimize the sensor layout plan according to the tunnel construction progress and environmental changes. It designs a multi-objective optimization algorithm based on the sensor layout location, layout quantity, and data transmission path, and adjusts the sensor layout in real time according to the optimization results. The adaptive routing module is used to dynamically select the optimal data transmission path based on the optimized sensor layout and the wireless transmission environment between the edge sensor devices inside the tunnel and the server using an adaptive routing algorithm. It evaluates the transmission link quality in real time, predicts signal attenuation and interference, and selects the transmission path with the strongest signal and the least attenuation to ensure the stability and reliability of data transmission. An edge computing module is used to deploy the edge computing module at the edge node of the sensor, pre-process and compress the collected monitoring data in the edge computing module, and perform data dimension reduction and feature extraction; The digital twin module is used to build a digital twin model of the tunnel environment and equipment status based on the extracted features. Through the real-time interaction of physical and virtual information, it can realize the comprehensive perception and dynamic tracking of the tunnel construction process, conduct fusion analysis on the monitoring data of the sensors, timely discover abnormal situations, and provide a basis for safety warning; The incremental learning module is used to analyze the sensor monitoring data online based on the detection data of the sensor used in the digital twin model using an incremental learning algorithm, dynamically update the anomaly detection model based on the newly collected data samples, and adaptively adjust the warning threshold; The safety assessment module is used to comprehensively apply multi-sensor data fusion technology to correlate and analyze sensor data of different types and locations in the tunnel, and comprehensively evaluate the safety status of tunnel construction through data redundancy and complementarity; The cloud-edge collaboration module is used to establish a data processing architecture based on cloud-edge collaboration, which stores and calculates massive monitoring data in layers. The edge side is responsible for real-time processing and analysis of data, and the cloud side is responsible for long-term storage and mining of data. Through cloud-edge collaboration, efficient management and utilization of monitoring data can be achieved, providing comprehensive data support for tunnel construction safety.

Images