Intelligent tunnel construction safety monitoring and early warning system and method

Abstract

The invention discloses an intelligent tunnel construction safety monitoring and early warning system and method, and belongs to the technical field of tunnel engineering monitoring. The method comprises the steps that construction quality data and environment data in the tunnel construction process are collected; performing data preprocessing on the data, extracting key features from the preprocessed data for analysis, and generating a daily data analysis report; generating a dynamic threshold of the construction quality data based on a dynamic adaptive threshold optimization model DATOM; and adjusting an early warning trigger condition according to the real-time construction stage label, and sending alarm information to terminal equipment when the construction quality data and the environment data deviate from the threshold value and reach the early warning condition. According to the method, the requirements of different construction stages on dynamic change of the monitoring parameter threshold can be met, and the safety monitoring and early warning precision is relatively high.

Description

Technical Field

[0001] The present invention belongs to the technical field of tunnel engineering monitoring, and specifically relates to an intelligent tunnel construction safety monitoring and early warning system and method. Background Art

[0002] Traditional tunnel engineering monitoring has long relied on manual inspections and fixed threshold alarm mechanisms, resulting in inherent flaws such as delayed response, data silos, and high false alarm rates. While existing monitoring solutions employ sensors and wireless transmission technologies, their technical bottleneck remains the conflict between static warning thresholds and dynamic construction requirements. Existing systems use fixed thresholds, making them incapable of adapting to the quality control requirements of different stages of tunnel excavation, support, and lining. Furthermore, environmental interference (such as temperature and humidity fluctuations) can easily lead to misjudgments.

[0003] In recent years, the development of LSTM neural networks, Transformers, edge computing, and blockchain technologies has provided new solutions to these problems. While LSTMs can capture short-term temporal features, they lack the ability to model long-range dependencies. Meanwhile, the Transformer's self-attention mechanism, while adept at analyzing global correlations, lacks sensitivity to local details. Therefore, effectively improving multi-scale feature extraction capabilities in complex construction environments like tunneling, while also addressing the need for dynamic threshold optimization of monitoring parameters at different stages of construction, is crucial for ensuring tunnel construction safety. Summary of the Invention The purpose of the present invention is to provide an intelligent tunnel construction safety monitoring and early warning system and method, so as to solve the problem that the existing technology cannot effectively extract multi-scale features in complex construction environments such as tunnel construction, especially cannot meet the requirements of dynamic changes in monitoring parameter thresholds in different construction stages, resulting in relatively low accuracy of tunnel construction safety monitoring.

[0004] In order to achieve the above object, the technical solution provided by the present invention is as follows: A first aspect of the present invention provides a smart tunnel construction safety monitoring and early warning method, comprising: Collect construction quality data and environmental data during tunnel construction, convert the data format and transmit it to the server; Preprocess the data transmitted to the server, extract key features from the preprocessed data, and generate daily data analysis reports; Generate thresholds for construction quality data based on the dynamic adaptive threshold optimization model DATOM; Adjust the warning trigger conditions based on real-time construction stage tags, and send alarm information to the terminal device when the construction quality data and environmental data deviate from the threshold and reach the warning conditions; The dynamic threshold output by the dynamic adaptive threshold optimization model DATOM integrates the local features extracted by LSTM and the feature vector extracted by Transformer to capture global dependencies , the dynamic threshold calculation formula is as follows: , UB t 、 LB t are the upper and lower thresholds, respectively. is the weight matrix, is the bias vector, which is obtained through model training optimization; for and The fusion feature vector of It is the historical benchmark average for this construction phase.

[0005] According to the technical solution described in the first aspect of the present invention, the construction quality data includes structural mechanical parameters, specifically one or more of steel bar stress, concrete strain, tunnel surrounding rock displacement, and crack growth rate; and the environmental data includes one or more of ambient temperature, humidity, gas concentration, and groundwater level. Structural mechanical parameters serve as primary monitoring indicators, while environmental data serve as auxiliary monitoring indicators.

[0006] According to the technical solution described in the first aspect of the present invention, the construction quality data and environmental data are transmitted to the server through an Internet of Things wireless transmission device group. The Internet of Things wireless transmission device group includes a 5G / 4G / WiFi DTU device and a built-in edge computing module. The edge computing module uses a Kalman filtering algorithm to perform data denoising and detects outliers through an isolation forest algorithm. The edge computing module compresses and encrypts the data before transmission, and only transmits abnormal data that exceeds the baseline by 20% to 30%.

[0007] According to the technical solution described in the first aspect of the present invention, the data preprocessing includes one or more of data cleaning, format conversion, and outlier processing; the daily data analysis report contains the statistical distribution, trend forecast and comparative analysis of each monitoring data with historical data, and the construction quality data is superimposed on the tunnel BIM model through a three-dimensional heat map, and high-risk areas are marked in real time.

[0008] According to the technical solution of the first aspect of the present invention, the threshold value of the construction quality data generated based on the dynamic adaptive threshold optimization model DATOM specifically includes: The LSTM module processes the local window data and outputs the hidden state feature vector representing the short-term mutation ; Process global window data through the Transformer module and output a multi-head attention feature vector that captures long-term dependencies ; The hidden state feature vector output by the LSTM module Multi-head attention feature vector output by Transformer module According to the trainable fusion weight coefficient α, weighted fusion is performed to obtain the fusion feature : ,α∈[0,1]; The fusion features Input the fully connected layer, through the weight matrix and the bias vector Calculate the dynamic adjustment amount and superimpose the historical benchmark mean of the construction stage , thereby outputting the final threshold pair.

[0009] According to the technical solution of the first aspect of the present invention, the process of generating the weight coefficient α is as follows: (1) Parameter initialization: At the beginning of the DATOM model construction, α is initialized to a neutral value of 0.5 as a trainable scalar variable and embedded in the neural network weight matrix; (2) Dual-branch feature extraction and fusion: When the sensor data stream is input in real time, the model performs two types of feature analysis in parallel through the LSTM branch and the Transformer branch, and fuses the two types of features to obtain the fused feature ; (3) Loss feedback and gradient generation: After the fused features are output through the fully connected layer, the dynamic threshold prediction value is compared with the ideal threshold marked by the engineer to calculate the Huber loss function; the gradient of α is automatically solved through backpropagation: ,When the LSTM feature has a better ability to reduce the loss than the Transformer, the gradient is negative, driving α to increase; (4) Dynamic optimization iteration: The Adam optimizer is used to update the α value. The amplitude of each update is determined by the gradient direction and the historical update momentum, and the Sigmoid function is used to strictly constrain α to the interval [0,1].

[0010] According to the technical solution described in the first aspect of the present invention, the dynamic adaptive threshold optimization model DATOM uses new data for incremental training every 24 hours, and during the incremental training every 24 hours, the model inherits the α value converged on the previous day as the initial condition.

[0011] According to the technical solution described in the first aspect of the present invention, a hierarchical risk warning is performed by combining the over-scale Ex of a single construction quality data and the comprehensive risk index Rtotal. The warning rules are as follows: Blue warning: The exceedance of a single construction quality parameter is 10% ≤ Ex < 20% or the crack growth rate is 0.05mm / d ≤ v < 0.1mm / d; Yellow warning: A single construction quality parameter exceeds the standard by 20% ≤ Ex < 30%; or at least two construction quality parameters exceed the standard by 15% ≤ Ex < 20%; or the crack growth rate is 0.1mm / d ≤ v < 0.15mm / d, and the comprehensive risk index is 20% ≤ Rtotal < 40%; Orange warning: A single construction quality parameter exceeds the standard by 30% ≤ Ex < 50%, or the crack growth rate is 0.15mm / d ≤ v < 0.2mm / d; Red warning: A single construction quality parameter exceeds the standard Ex ≥ 50% or the crack expansion rate v ≥ 0.2 mm / d or the comprehensive risk index Rtotal ≥ 50%; The comprehensive risk index Rtotal is calculated as follows: , In the above formula: : Construction stage weight, excavation stage is 0.6, support stage is 0.3, lining stage is 0.1; : No. i Overscaling of parameters; : No. i Critical overscaling threshold of the parameter; : Environmental correction coefficient, customized.

[0012] According to the technical solution described in the first aspect of the present invention, when the dynamic threshold generated by the DATOM model triggers an early warning three times in a row, the system automatically raises the early warning level, generates an emergency task work order and assigns it to a designated person.

[0013] The second aspect of the present invention further provides a smart tunnel construction safety monitoring and early warning system, comprising: Data acquisition unit, used to collect construction quality data and environmental data during tunnel construction, and convert the data format and transmit it to the server; The data preprocessing unit is used to preprocess the data transmitted to the server, extract key features from the preprocessed data for analysis, and generate daily data analysis reports; A dynamic threshold generation unit, which generates a dynamic threshold for construction quality data based on a dynamic adaptive threshold optimization model DATOM; and The early warning unit sends an alarm message to the terminal device when the construction quality data and environmental data deviate from the threshold and reach the early warning condition; The dynamic adaptive threshold optimization model DATOM includes an LSTM module, a Transformer module, a feature fusion module and a fully connected layer. The LSTM module is used to extract local features. , the Transformer module is used to extract feature vectors that capture global dependencies , the feature fusion module is used to and Perform weighted fusion to generate fusion feature vector ; The fully connected layer is based on the fusion feature vector , through the weight matrix and the bias vector Calculate the dynamic adjustment amount and superimpose the historical benchmark mean of the construction stage , thereby outputting the final threshold pair.

[0014] Compared with the prior art, the present invention can achieve the following beneficial effects: (1) The present invention collects construction quality parameters and environmental data in real time during tunnel construction, takes construction quality parameters as the main monitoring indicators, and generates dynamic warning thresholds for construction quality parameters based on the dynamic adaptive threshold optimization model DATOM that integrates LSTM and Transformer. This model can not only effectively extract multi-scale features in complex construction environments such as tunnel construction, but also effectively improve the reliability and accuracy of risk warnings.

[0015] (2) When the present invention generates the dynamic warning threshold of construction quality data through the dynamic adaptive threshold optimization model DATOM, the local features extracted by LSTM (such as short-term stress fluctuations) and the global dependency (such as cross-stage stress trends) feature vectors captured by Transformer are weightedly fused, and the value of the fusion weight α is back-propagated and updated during the model training process, so that the model can automatically optimize the fusion ratio of local and global features according to the characteristics of the data, which is conducive to further ensuring the rationality of the warning threshold and thus improving the accuracy of subsequent risk warnings.

[0016] (3) The present invention combines the over-scale Ex of a single construction quality data and the comprehensive risk index Rtotal to perform graded risk warning, thereby effectively avoiding the occurrence of risk omissions and false detections; at the same time, different warning thresholds can be further generated according to different construction stages, thereby meeting the needs of dynamic changes in monitoring parameter thresholds in different construction stages and having good dynamic adaptability.

[0017] (4) The present invention superimposes construction quality data such as stress and strain onto the tunnel BIM model through a three-dimensional heat map, marking the spatial distribution of high-risk areas in real time. When the dynamic threshold generated by the DATOM model triggers an early warning three times in a row, the system automatically raises the warning level, generates an emergency task work order, and assigns it to a designated person, thus achieving a closed-loop management of "risk identification-level determination-task assignment." BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In the drawings, the sizes and proportions do not represent the sizes and proportions of actual products. The drawings are merely illustrative and some non-essential elements or features are omitted for clarity.

[0019] Figure 1 is a flow chart of the early warning method of the present invention; Figure 2 Schematic diagram of the structure of the LSTM module and the Transformer module in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The present invention is described in detail below with reference to specific embodiments.

[0021] Combine Figure 1 As shown, an embodiment of the present invention provides a smart tunnel construction safety monitoring and early warning method, including the following steps: Step 1: Use sensors to collect construction quality data and environmental data in real time during tunnel construction, convert the data format and transmit it to the server.

[0022] Construction quality data serves as primary safety warning monitoring data, while environmental data serves as auxiliary monitoring data. Specifically, in an embodiment of the present invention, the construction quality data includes structural mechanics parameters, including one or more of steel bar stress, concrete strain, tunnel surrounding rock displacement, and crack growth rate data. Environmental data includes, but is not limited to, ambient temperature, humidity, gas concentration, and groundwater level.

[0023] Specifically, the aforementioned steel bar stress, concrete strain, and tunnel surrounding rock displacement are monitored using stress sensors, strain sensors, and displacement sensors, respectively. The crack growth rate is determined based on the strain changes on both sides of the crack as monitored by strain gauges, combined with time data. (In actual engineering, stress and strain sensors can be deployed immediately after tunnel excavation, along with strain gauges in potential cracking areas and existing apparent cracks. As cracks grow, the resistance of the strain gauges changes, and the change in resistance can be converted into the change in crack width.) The ambient temperature, humidity, gas concentration, and groundwater level are collected using temperature sensors, humidity sensors, gas concentration sensors, and liquid level sensors, respectively.

[0024] It should be noted that the specific types of sensors in the present invention can be selected and combined according to actual needs. For example, the vibration of the support structure can also be collected, and the sensor data collection interval for each project can be dynamically adjusted according to the construction stage. For example, data collection can be performed every 5 minutes during the excavation stage, increased to every 2 minutes during the support stage, and restored to every 5 minutes during the lining stage.

[0025] The data collected by the sensors is preferably transmitted to the server via an IoT wireless transmission device set, which includes a 5G / 4G / WiFi DTU device and a built-in edge computing module. The edge computing module uses a Kalman filter algorithm to denoise the data and an isolation forest algorithm to detect outliers.

[0026] Specifically, in the embodiment of the present invention, when the Kalman filter algorithm is used for data denoising, the state equation is described as the relationship between the current state and the previous state, and the observation equation relates the actual measurement value to the predicted value; the process noise and the observation noise are both assumed to be Gaussian distributed, and the initial covariance matrix is ​​set to a diagonal matrix to simplify the calculation. ， Denoise the raw sensor data.

[0027] More preferably, the edge computing module compresses and encrypts the data (for example, using the AES-256 algorithm) before transmitting it to the cloud, and only transmits abnormal data that exceeds the baseline by 20% to 30%, thereby saving bandwidth and solving the problem of a surge in bandwidth pressure caused by full sensor data transmission.

[0028] Step 2: Preprocess the data transmitted to the server, extract key features from the preprocessed construction quality data, and generate a daily data analysis report.

[0029] The preprocessing includes one or more of data cleaning, format conversion, and outlier processing. Data cleaning uses a sliding window algorithm (window length 30 minutes, step length 5 minutes) to detect outliers, and combines construction stage tags to correct sensor drift errors caused by temperature.

[0030] The method extracts key feature analysis (mean, variance, extreme values) from the preprocessed data to generate a daily data analysis report, which includes the statistical distribution of the monitoring data, trend prediction (using the ARIMA model), and comparative analysis with historical data. Furthermore, the construction quality data is displayed in multiple dimensions in the form of graphs, tables, and heat maps to achieve real-time monitoring and collection, and the corresponding data is superimposed on the tunnel BIM model through a three-dimensional heat map to mark high-risk areas in real time.

[0031] Step 3: Generate the threshold of construction quality data based on the dynamic adaptive threshold optimization model DATOM; In this embodiment of the present invention, the Dynamic Adaptive Threshold Optimization Model (DATOM) is constructed based on the fusion of LSTM and Transformer. By combining the advantages of LSTM and Transformer, it can effectively capture the multi-scale characteristics of time series data, thereby facilitating the precise generation of dynamic thresholds. The LSTM, through its gating mechanism, is able to effectively capture local short-term dependencies in the data, while the Transformer's self-attention mechanism can directly capture the relationship between any two moments in the sequence, excelling at capturing global dependencies. Combining these two approaches comprehensively captures the characteristics of time series and provides a more accurate basis for the generation of dynamic thresholds.

[0032] Specifically, the construction process of the dynamic adaptive threshold optimization model (DATOM) includes: (1) Construction of LSTM module The core of LSTM is the gating mechanism, which includes input gate, forget gate and output gate, to control the information flow and effectively alleviate the gradient vanishing problem.

[0033] LSTM module uses local window data As input data, at any time t , the input gate calculation formula is , the forget gate is , the output gate is , the candidate memory is , update the memory unit to , the hidden state is updated to .in, represents the sigmoid activation function, represents element-wise multiplication, W is the weight matrix, and b is the bias term.

[0034] (2) Construction of the Transformer module Transformer is mainly composed of self-attention mechanism, multi-head attention, feedforward neural network, residual connection and layer normalization.

[0035] The Transformer module uses global window data As input data, in the self-attention mechanism, the query, key, and value matrices are first generated through linear transformation, that is, ,in 、 、 is a learnable weight matrix. Then calculate the attention score , d k is the dimension of the key.

[0036] Multi-head attention maps the input to multiple subspaces and calculates the attention separately, and then splices them together. The calculation of each head is , the multi-head attention output is ,in Respectively i The parameters of the head and the output projection matrix.

[0037] In addition, the Transformer module adds positional encoding to retain sequence position information and improves training stability through feedforward networks, residual connections, and layer normalization.

[0038] (3) Construction of feature fusion module The feature fusion module is used to generate local features of the LSTM module. Output of Transforme module Different from the serial fusion method, the embodiment of the present invention preferably adopts a weighted fusion method, assuming that the fusion output is , where α∈[0,1] is a trainable fusion weight that can adaptively balance local and global features. During model training, the value of α is adjusted through backpropagation, allowing the model to automatically optimize the fusion ratio of local and global features based on data characteristics.

[0039] (4) Model training and dynamic threshold generation Training data: The training data of the DATOM model includes historical monitoring data of construction quality data, construction stage labels S t (such as excavation / support / lining), environmental parameters E t These data provide rich information for the model to learn the characteristics and patterns of time series under different conditions.

[0040] Incremental training: The model is incrementally trained every 24 hours using new data. During training, the LSTM module continues to learn local feature changes in the new data, while the Transformer module further captures global dependencies in the data. By continuously updating model parameters, the model can adapt to dynamic changes in the data.

[0041] Dynamic threshold generation: After training, the DATOM model generates its dynamic threshold curve (upper threshold UB(t) and lower threshold LB(t)) for each sensor.

[0042] Combine Figure 2 As shown in Figure 2, the process of generating dynamic thresholds using the DATOM model is as follows: First, obtain the real-time sensor data stream: Specifically, in the embodiment of the present invention, extract the local window data with a length of 10 and global window data of length 100 , and load the current construction stage label St and environmental parameters Et at the same time.

[0043] Feature extraction is then performed in parallel: the LSTM module is gated by a gating mechanism (including input gate i t , Forget Gate f t , memory unit Ct ) processes local data and outputs a feature vector representing short-term mutations The Transformer module analyzes global data through the self-attention mechanism (including query matrix Q, key matrix K, value matrix V) to generate feature vectors that capture long-term dependencies. .

[0044] The two types of features enter the adaptive fusion layer, and the weighted synthesis is performed by the weight coefficient α obtained by training to obtain the fusion feature : ; Fusion Features Input the fully connected layer, through the weight matrix and the bias vector Calculate the dynamic adjustment amount and superimpose the historical benchmark mean of the construction stage , the final output threshold pair: ; Among them, the weight matrix and the bias vector Updated by Adam optimizer; It is the historical benchmark average for this construction phase.

[0045] It is further preferred to use an LSTM neural network to extract time series features, incrementally update the network weights every 24 hours, and dynamically adjust the threshold curve so that the threshold deviation tolerance decreases linearly with the construction progress. Furthermore, as a preferred embodiment, the generation process of α is as follows: (1) Parameter initialization: Initially, α is initialized to a neutral value of 0.5 as a trainable scalar variable and embedded in the neural network weight matrix. At this point, the system has not yet learned any data patterns, and LSTM and Transformer features are given equal weight.

[0046] (2) Dual-branch feature extraction: When the sensor data stream is input in real time, the system performs two types of feature analysis in parallel: The LSTM branch focuses on a local window, such as a local window of 10 time steps (t1=10), and captures transient anomalies through a gating mechanism; The Transformer branch analyzes a global window, such as a global window of 100 time steps (t2=100), and uses the self-attention mechanism to model long-term patterns.

[0047] (3) Differentiable fusion operation: The system performs core fusion calculations: Here, α is used as a lever for adjustment. If its value is 0.7, the LSTM features are given 70% weight, and the Transformer features only account for 30%.

[0048] (4) Loss feedback and gradient generation: After the fused features are output through the fully connected layer, the system compares the dynamic threshold prediction value with the ideal threshold marked by the engineer and calculates the Huber loss function. The gradient of α is automatically solved through backpropagation: The gradient directly reflects the contribution efficiency of the two types of features: when the LSTM feature is better than the Transformer in reducing the loss, the gradient is negative, driving α to increase.

[0049] (5) Dynamic optimization iteration: The Adam optimizer is used to update the α value, with the magnitude of each update determined by the gradient direction and the momentum of the historical update. This process is repeated during training, allowing α to gradually evolve from an initial value of 0.5 to a state that is suitable for the scene.

[0050] (6) Engineering constraint guarantee: The Sigmoid function strictly constrains α to the range [0, 1] to prevent weight loss. During incremental training every 24 hours, the system inherits the α value that converged the previous day as the initial condition, and new data only causes limited drift.

[0051] Ultimately, α converges to a data-driven intelligent weight over several gradient updates. Its value quantifies the relative importance of local sudden risk and global evolutionary patterns in real time. By optimizing the generation process of α, different construction quality thresholds can be met at different stages, thereby improving early warning accuracy.

[0052] Step 4: Adjust the warning trigger conditions according to the real-time construction stage label, and send an alarm message to the terminal device when the construction quality data and environmental data deviate from the threshold and reach the warning condition.

[0053] Specifically, as one of the implementation methods, when the construction quality data or environmental data deviates from the threshold and reaches the set overscale, an early warning is activated, wherein the threshold of the environmental data is directly set in advance manually, and the threshold of the environmental data can be adaptively adjusted in different construction stages. For example, in the excavation stage, the gas concentration threshold can be appropriately increased.

[0054] The definition of over-scale is as follows, which is divided into two cases: over-scale and under-scale: .

[0055] As a further preferred embodiment, combining the overscaling of single construction quality data Ex , and introduces the comprehensive risk index Rtotal for graded risk warning. The warning rules are as follows: Blue warning: The exceedance of a single construction quality parameter is 10% ≤ Ex < 20% or the crack growth rate is 0.05mm / d ≤ v < 0.1mm / d; Yellow warning: A single construction quality parameter exceeds the standard by 20% ≤ Ex < 30%; or at least two construction quality parameters exceed the standard by 15% ≤ Ex < 20%; or the crack growth rate is 0.1mm / d ≤ v < 0.15mm / d, and the comprehensive risk index is 20% ≤ Rtotal < 40%; Orange warning: The excess of a single construction quality parameter is 30% ≤ Ex < 50%, or the crack growth rate is 0.15 mm / d ≤ v < 0.2 mm / d; Red warning: Ex of a single construction quality parameter exceeds 50% or the crack growth rate v exceeds 0.2 mm / d or the comprehensive risk index Rtotal exceeds 50%; The comprehensive risk index Rtotal is calculated as follows: , In the above formula: : Construction stage weight, excavation stage is 0.6, support stage is 0.3, lining stage is 0.1; : No.i Overscaling of parameters; : No. i The critical over-scale threshold of each parameter is set in advance; : Environmental correction coefficient, customized, for example, rainy season = 1.2, fire warning triggered = 2.0, normal = 1.0.

[0056] Since multiple factors may cause missed detections and false detections if warnings are issued solely based on over-scale indicators, the embodiment of the present invention further introduces a comprehensive risk index Rtotal to effectively integrate the over-scale of construction quality parameters and environmental parameters, thereby helping to reduce the risk of false detections and missed detections.

[0057] Furthermore, the coupling discrimination rules of natural disasters can be introduced: (1) Fire risk index: , in, T 0 is the temperature safety threshold, CO 0 is the safety threshold of CO concentration.

[0058] (2) Water inrush risk index: , Among them, h is the height of the groundwater level, h0 is the safe water level; Q is the water inflow, and Q0 is the safe water inflow.

[0059] Further preferably, when the dynamic threshold generated by the DATOM model triggers an early warning three times in a row, the system automatically raises the early warning level, generates an emergency task work order and assigns designated personnel.

[0060] As a further preferred implementation scheme, when an early warning occurs in step three, the early warning linkage mechanism is triggered: The system automatically retrieves the global feature weight matrix output by Transformer and locates the parameters that contribute most to the risk (e.g., stress and vibration). The weight increased significantly); Combined with BIM model 3D positioning, areas with abnormal associated parameters are highlighted in the tunnel thermal map, thereby accurately locating risk areas; Generate emergency work orders including parameter correlation analysis (such as "Stress-vibration coupling abnormality in K10+500 section, it is recommended to immediately check the steel arch connection") and push them synchronously via SMS, email and platform.

[0061] Furthermore, in the embodiment of the present invention, the alarm information is stored through blockchain technology. The blockchain adopts a consortium chain architecture and deploys 3 consensus nodes. The PBFT mechanism requires more than 2 / 3 of the nodes to reach a consensus; the data block hash check uses SHA-256, the block generation interval is 10 minutes, and the log is stored encrypted with AES-256, thereby ensuring the non-tamperability of the warning data and forming a complete intelligent monitoring-risk assessment-emergency response closed-loop system.

[0062] The embodiment of the present invention further provides a smart tunnel construction safety monitoring and early warning system, including: The data acquisition unit is used to collect construction quality data, environmental data and equipment status parameters during the tunnel construction process, and convert the data format and transmit it to the server; The data preprocessing unit is used to preprocess the data transmitted to the server, extract key features from the preprocessed data for analysis, and generate daily data analysis reports; A dynamic threshold generation unit, which generates a dynamic threshold for construction quality data based on a dynamic adaptive threshold optimization model DATOM; and The early warning unit sends an alarm message to the terminal device when the construction quality data and environmental data deviate from the threshold and reach the early warning condition; The dynamic adaptive threshold optimization model DATOM includes an LSTM module, a Transformer module, a feature fusion module and a fully connected layer. The LSTM module is used to extract local features. , the Transformer module is used to extract feature vectors that capture global dependencies , the feature fusion module is used to and Perform weighted fusion to generate fusion feature vector ; The fully connected layer is based on the fusion feature vector , through the weight matrix and the bias vector Calculate the dynamic adjustment amount and superimpose the historical benchmark mean of the construction stage , thereby outputting the final threshold pair.

[0063] The system can monitor the environmental parameters and equipment status in the tunnel in real time. Through intelligent data processing and analysis, it can promptly detect problems and issue early warnings. It can also provide customized content based on the actual situation of tunnel construction to ensure safe and efficient operation of the tunnel.

[0064] Furthermore, the early warning system is linked to the status of each sensor. If the sensor is operating normally, it will be displayed as "online", and if the device is disconnected, it will be displayed as "offline". An offline timeout of 30 minutes will trigger a device maintenance alarm.

[0065] The scope of protection of the present invention is limited only by the claims. Thanks to the teachings of the present invention, those skilled in the art will readily recognize that alternative structures to the structures disclosed in the present invention can be used as viable alternative embodiments, and that the embodiments disclosed in the present invention can be combined to produce new embodiments, which also fall within the scope of the appended claims.

Claims

1. A smart tunnel construction safety monitoring and early warning method, characterized in that: include: Collect construction quality data and environmental data during tunnel construction, convert the data format and transmit it to the server; Preprocess the data transmitted to the server, extract key features from the preprocessed data, and generate daily data analysis reports; Generate thresholds for construction quality data based on the dynamic adaptive threshold optimization model DATOM; Adjust the warning trigger conditions based on real-time construction stage tags, and send alarm information to the terminal device when the construction quality data and environmental data deviate from the set threshold and reach the warning conditions; The dynamic threshold output by the dynamic adaptive threshold optimization model DATOM integrates the local features extracted by LSTM and the feature vector extracted by Transformer to capture global dependencies , the dynamic threshold calculation formula is as follows: ， in, UB t 、 LB t are the upper and lower thresholds, respectively. is the weight matrix, is the bias vector, which is obtained through model training optimization; for and The fusion feature vector of It is the historical benchmark average for this construction phase.

2. The smart tunnel construction safety monitoring and early warning method according to claim 1 is characterized in that: The construction quality data includes structural mechanics parameters, which include one or more of steel bar stress, concrete strain, tunnel surrounding rock displacement, and crack propagation rate; the environmental data includes one or more of ambient temperature, humidity, gas concentration, and groundwater level.

3. The smart tunnel construction safety monitoring and early warning method according to claim 1 is characterized in that: The construction quality data and environmental data are transmitted to the server via an IoT wireless transmission device group, which includes a 5G / 4G / WiFi DTU device and a built-in edge computing module. The edge computing module uses a Kalman filter algorithm to denoise the data and detects outliers through an isolation forest algorithm. The edge computing module compresses and encrypts the data before transmission, and only transmits abnormal data that exceeds the baseline by 20% to 30%.

4. The smart tunnel construction safety monitoring and early warning method according to claim 1 is characterized in that: The data preprocessing includes one or more of data cleaning, format conversion, and outlier processing; the daily data analysis report contains the statistical distribution, trend forecast, and comparative analysis of each monitoring data with historical data, and superimposes the construction quality data onto the tunnel BIM model through a three-dimensional heat map, marking high-risk areas in real time.

5. The smart tunnel construction safety monitoring and early warning method according to any one of claims 1 to 4, characterized in that: The thresholds for generating construction quality data based on the dynamic adaptive threshold optimization model DATOM specifically include: The LSTM module processes the local window data and outputs the hidden state feature vector representing the short-term mutation ; Process global window data through the Transformer module and output a multi-head attention feature vector that captures long-term dependencies ; The hidden state feature vector output by the LSTM module Multi-head attention feature vector output by Transformer module According to the trainable fusion weight coefficient α, weighted fusion is performed to obtain the fusion feature : ,α∈[0,1]; The fusion features Input the fully connected layer through the weight matrix and the bias vector Calculate the dynamic adjustment amount and superimpose the historical benchmark mean of the construction stage , thereby outputting the final threshold pair.

6. The smart tunnel construction safety monitoring and early warning method according to claim 5 is characterized in that: The generation process of the weight coefficient α is as follows: (1) Parameter initialization: At the beginning of the DATOM model construction, α is initialized to a neutral value of 0.5 as a trainable scalar variable and embedded in the neural network weight matrix; (2) Dual-branch feature extraction and fusion: When the sensor data stream is input in real time, the model performs two types of feature analysis in parallel through the LSTM branch and the Transformer branch, and fuses the two types of features to obtain the fused feature ; (3) Loss feedback and gradient generation: After the fused features are output through the fully connected layer, the dynamic threshold prediction value is compared with the ideal threshold marked by the engineer to calculate the Huber loss function; the gradient of α is automatically solved through backpropagation: ,When the LSTM feature has a better ability to reduce the loss than the Transformer, the gradient is negative, driving α to increase; (4) Dynamic optimization iteration: The Adam optimizer is used to update the α value. The amplitude of each update is determined by the gradient direction and the historical update momentum, and the Sigmoid function is used to strictly constrain α to the interval [0,1].

7. The smart tunnel construction safety monitoring and early warning method according to claim 6 is characterized in that: The dynamic adaptive threshold optimization model DATOM performs incremental training using newly added data every 24 hours, and during the incremental training every 24 hours, the model inherits the α value converged on the previous day as the initial condition.

8. The smart tunnel construction safety monitoring and early warning method according to claim 7 is characterized in that: Combined with the Excessive Scale Ex of a single construction quality data and the comprehensive risk index Rtotal, a graded risk warning is performed. The warning rules are as follows: Blue warning: The exceedance of a single construction quality parameter is 10% ≤ Ex < 20% or the crack growth rate is 0.05mm / d ≤ v < 0.1mm / d; Yellow warning: A single construction quality parameter exceeds the standard by 20% ≤ Ex < 30% or at least two construction quality parameters exceed the standard by 15% ≤ Ex < 20% or the crack growth rate is 0.1mm / d ≤ v < 0.15mm / d, and the comprehensive risk index is 20% ≤ Rtotal < 40%; Orange warning: A single construction quality parameter exceeds the standard by 30% ≤ Ex < 50%, or the crack growth rate is 0.15mm / d ≤ v < 0.2mm / d; Red warning: A single construction quality parameter exceeds the standard Ex ≥ 50% or the crack expansion rate v ≥ 0.2 mm / d or the comprehensive risk index Rtotal ≥ 50%; The comprehensive risk index Rtotal is calculated as follows: ， In the above formula: : Construction stage weight, excavation stage is 0.6, support stage is 0.3, lining stage is 0.1; : No. i Overscaling of parameters; : No. i Critical overscaling threshold of the parameter; : Environmental correction coefficient, customized.

9. A smart tunnel construction safety monitoring and early warning system, characterized in that: include: The data acquisition unit is used to collect construction quality data, environmental data and equipment status parameters during the tunnel construction process, and convert the data format and transmit it to the server; The data preprocessing unit is used to preprocess the data transmitted to the server, extract key features from the preprocessed data for analysis, and generate daily data analysis reports; Dynamic threshold generation unit, which generates dynamic thresholds for construction quality data based on the dynamic adaptive threshold optimization model DATOM; as well as The early warning unit sends an alarm message to the terminal device when the construction quality data and environmental data deviate from the threshold and reach the early warning condition; The dynamic adaptive threshold optimization model DATOM includes an LSTM module, a Transformer module, a feature fusion module and a fully connected layer. The LSTM module is used to extract local features. , the Transformer module is used to extract feature vectors that capture global dependencies , the feature fusion module is used to and Perform weighted fusion to generate fusion feature vector ; The fully connected layer is based on the fusion feature vector , through the weight matrix and the bias vector Calculate the dynamic adjustment amount and superimpose the historical benchmark mean of the construction stage , thereby outputting the final threshold pair.

10. The intelligent tunnel construction safety monitoring and early warning system according to claim 9 is characterized in that: The dynamic threshold output by the dynamic adaptive threshold optimization model DATOM integrates the local features extracted by LSTM and the feature vector extracted by Transformer to capture global dependencies , the dynamic threshold calculation formula is as follows: ， in, UB t 、 LB t are the upper and lower thresholds, respectively. is the weight matrix, is the bias vector, which is obtained through model training optimization; for and The fusion feature vector of It is the historical benchmark average for this construction phase.

Images