Tunnel construction gas intelligent monitoring method, device and system

Abstract

The invention discloses an intelligent monitoring method, device and system for tunnel construction gas, and the method comprises the steps: collecting multi-source gas data in real time through a mixed gas monitoring network disposed in a tunnel construction environment, the mixed gas monitoring network comprises a fixed gas monitoring node, a mobile gas monitoring node and a personnel wearable monitoring node; based on the multi-source gas data, constructing a three-dimensional gas concentration field dynamic model covering the whole tunnel construction area by using a data fusion and machine learning algorithm; on the basis of the three-dimensional gas concentration field dynamic model, the spatio-temporal evolution trend of the gas concentration in a future preset time period is predicted, at least one control instruction is generated and executed according to the prediction result, and the control instruction comprises an intelligent early warning instruction and / or a ventilation system adjusting instruction. According to the intelligent monitoring method, device and system, traditional post-event alarm is converted into pre-warning and active intervention, the safety of tunnel construction is remarkably improved, and the emergency response speed is remarkably increased.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel construction safety monitoring technology, and in particular to a method, device and system for intelligent monitoring of gas in tunnel construction. Background Art

[0002] As an important transportation infrastructure, tunnel construction is usually carried out in a narrow, closed, semi-enclosed space. During construction activities such as excavation, blasting, lining, welding, and mechanical equipment operation, a variety of harmful gases are inevitably generated or accumulated, such as carbon monoxide (CO), methane (CH4), hydrogen sulfide (H2S), nitrogen oxides (NO x ) and other harmful gases. Construction activities also consume significant amounts of oxygen (O2), leading to localized hypoxia. The accumulation of these harmful gases or the reduction of oxygen concentrations poses a serious threat to the lives of tunnel workers and can lead to serious safety accidents such as poisoning, fire, and explosion. This presents a critical challenge that urgently needs to be addressed in tunnel construction safety management.

[0003] Existing tunnel construction gas monitoring technology primarily relies on the deployment of fixed-point gas sensors at key locations within the tunnel. While this approach enables real-time monitoring of gas concentrations at specific points, its limitations are significant. First, the limited number of sensors deployed and their fixed locations prevent them from covering the entire tunnel space. This is particularly true in areas of dynamic construction, such as the tunnel face and temporary support structures, which can easily create monitoring blind spots. Second, traditional monitoring systems typically only have over-limit alarm functions, issuing alarms only when concentrations exceed the standard. This lacks forward-looking predictive capabilities and fails to provide sufficient pre-response time for personnel to evacuate hazards and respond to emergencies. Third, existing ventilation systems often utilize timed or manual control modes, failing to accurately and on-demand adjust the system based on the real-time, dynamic distribution of gas concentrations within the tunnel. This often results in over-ventilation or under-ventilation in critical areas, resulting in energy waste and persistent safety risks. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems in the related art to a certain extent. To this end, the present invention aims to provide a method, device and system for intelligent monitoring of gas in tunnel construction.

[0005] To achieve the above objectives, in a first aspect, a method for intelligently monitoring gas in tunnel construction according to an embodiment of the present invention includes:

[0006] Real-time collection of multi-source gas data via a mixed gas monitoring network deployed within the tunnel construction environment, comprising fixed gas monitoring nodes, mobile gas monitoring nodes, and wearable monitoring nodes.

[0007] Based on the multi-source gas data, a three-dimensional gas concentration field dynamic model covering the entire tunnel construction area is constructed using data fusion and machine learning algorithms;

[0008] Based on the three-dimensional gas concentration field dynamic model, the spatiotemporal evolution trend of the gas concentration within a preset time period in the future is predicted, and at least one control instruction is generated and executed according to the prediction result. The control instruction includes an intelligent early warning instruction and / or a ventilation system adjustment instruction.

[0009] According to one embodiment of the present invention, the mobile gas monitoring node includes at least one of an unmanned aerial vehicle monitoring node, a crawler or wheeled robot monitoring node, and a vehicle-mounted monitoring node installed on a tunnel construction vehicle;

[0010] The mobile gas monitoring node performs dynamic path planning according to a preset inspection path or according to model building requirements to actively detect monitoring blind spots or key areas of concern.

[0011] According to one embodiment of the present invention, in the step of collecting multi-source gas data in real time, three-dimensional spatial coordinate information and timestamp information are synchronously added to the gas concentration data collected by each monitoring node;

[0012] Among them, the coordinates of the fixed gas monitoring node are preset values, the coordinates of the mobile gas monitoring node are obtained through the positioning module, and the coordinates of the wearable monitoring node are obtained through its integrated UWB or Bluetooth AOA positioning module to form a four-dimensional data set with spatiotemporal tags.

[0013] According to one embodiment of the present invention, constructing a three-dimensional gas concentration field dynamic model covering the entire tunnel construction area based on the multi-source gas data using data fusion and machine learning algorithms includes:

[0014] De-noising, standardization, and spatiotemporal alignment are performed on the collected multi-source gas data, converting the sparse, heterogeneous data from fixed gas monitoring nodes, mobile gas monitoring nodes, and wearable gas monitoring nodes into a structured dataset in a unified spatiotemporal reference system.

[0015] Based on the processed structured data set, a spatial interpolation algorithm is used to calculate the gas concentration values ​​at locations where sensors are not deployed in the tunnel space, and the interpolation calculation results are obtained.

[0016] By introducing timing constraints and physical diffusion model constraints, the interpolation results are constrained and adjusted to ensure that the changes in the gas concentration field between adjacent time frames conform to the physical laws of gas diffusion, thereby forming three-dimensional gas concentration field data;

[0017] The three-dimensional gas concentration field data is spatially registered and fused with the geometric model of the tunnel construction environment to construct a continuous and smooth three-dimensional gas concentration field dynamic model covering the entire tunnel.

[0018] According to one embodiment of the present invention, predicting the spatiotemporal evolution trend of gas concentration within a future preset time period based on the three-dimensional gas concentration field dynamic model includes:

[0019] A three-dimensional gas concentration field dynamic model based on a continuous time series is used as basic input data, and known airflow field data in the tunnel is collected at the same time, wherein the airflow field data includes at least wind speed, wind direction and turbulence characteristics;

[0020] Use the long short-term memory network model or Kalman filter algorithm to build a prediction model of the spatiotemporal evolution of the gas concentration field;

[0021] Based on the spatiotemporal evolution prediction model, a four-dimensional data set of the harmful gas concentration field within a preset time period in the future is calculated and generated;

[0022] Extract and quantify the diffusion range, movement path, and concentration peak changes of harmful gases from the predicted four-dimensional data set;

[0023] Based on the hazardous threshold standards for harmful gases, a graded risk assessment is conducted on the predicted concentration field to identify high-risk areas and time windows that may appear in the future.

[0024] According to one embodiment of the present invention, after constructing the three-dimensional gas concentration field dynamic model covering the entire tunnel construction area, the method further includes:

[0025] When the gas concentration value in the three-dimensional gas concentration field dynamic model exceeds a preset threshold, the maximum possible leakage source or abnormal accumulation source of the harmful gas is traced back and calibrated in a three-dimensional visual manner by real-time analysis of the gradient vector of the three-dimensional gas concentration field or application of a computational fluid dynamics inverse diffusion model.

[0026] According to one embodiment of the present invention, the generation and execution of the intelligent early warning instruction includes:

[0027] Based on the predicted distribution of hazardous gas concentrations and personnel location information, the system automatically calculates the spatiotemporal relationship between personnel and high-risk areas, generating differentiated graded warning strategies for different groups of personnel.

[0028] For people about to enter predicted high-risk areas, structured warning information containing dangerous gas types, concentrations, risk levels, and optimal avoidance routes will be pushed to their wearable devices;

[0029] According to the risk level and urgency, select and activate the corresponding alarm mode, which includes wearable device vibration reminder, sound and light alarm, regional broadcast or full tunnel evacuation instruction;

[0030] Track and record the movement trajectory and response behavior of personnel after the warning, evaluate the warning effect, and use the warning effect as feedback information to optimize the hierarchical warning strategy.

[0031] According to one embodiment of the present invention, the generation and execution of the ventilation system adjustment instruction includes:

[0032] Based on the predicted three-dimensional gas concentration field and airflow field data, a multi-objective optimization model is constructed, which simultaneously considers the three objectives of rapid removal efficiency of harmful gases, minimization of energy consumption, and environmental comfort;

[0033] Setting constraints for the multi-objective optimization model, including performance boundaries of each ventilation device, safe gas concentration thresholds for each area of ​​the tunnel space, and energy consumption limits;

[0034] By solving a multi-objective optimization model, a spatiotemporal dynamic ventilation control strategy is generated. This ventilation control strategy assigns optimal operating parameters to ventilation equipment in different zones and time periods within the tunnel. The ventilation control strategy is transmitted to the controllers of each ventilation equipment via an industrial field bus or wireless communication network.

[0035] The changes in the gas concentration field after ventilation adjustment are monitored in real time, the ventilation effect is evaluated, and the evaluation results are fed back to the multi-objective optimization model.

[0036] In a second aspect, an intelligent tunnel construction gas monitoring device according to an embodiment of the present invention includes:

[0037] A collection unit, configured to collect multi-source gas data in real time via a mixed gas monitoring network deployed within the tunnel construction environment, the mixed gas monitoring network comprising fixed gas monitoring nodes, mobile gas monitoring nodes, and wearable gas monitoring nodes;

[0038] A concentration field model construction unit is used to construct a three-dimensional gas concentration field dynamic model covering the entire tunnel construction area based on the multi-source gas data using data fusion and machine learning algorithms;

[0039] A prediction control unit is used to predict the spatiotemporal evolution trend of gas concentration within a preset time period in the future based on the three-dimensional gas concentration field dynamic model, and to generate and execute at least one control instruction based on the prediction result, wherein the control instruction includes an intelligent early warning instruction and / or a ventilation system adjustment instruction.

[0040] In a third aspect, an intelligent tunnel construction gas monitoring system according to an embodiment of the present invention includes:

[0041] A mixed gas monitoring network for real-time collection of multi-source gas data, comprising multiple fixed gas monitoring nodes deployed in the tunnel, at least one mobile gas monitoring node, and multiple wearable gas monitoring nodes;

[0042] A server is communicatively connected to the mixed gas monitoring network, and is configured to: receive and process the multi-source gas data; construct a three-dimensional gas concentration field dynamic model covering the entire tunnel construction area based on the multi-source gas data using data fusion and machine learning algorithms; and predict the spatiotemporal evolution trend of gas concentration within a future preset time period based on the three-dimensional gas concentration field dynamic model, and generate and execute at least one control instruction based on the prediction result, the control instruction including an intelligent early warning instruction and / or a ventilation system adjustment instruction.

[0043] According to the intelligent tunnel construction gas monitoring method, device and system provided by the embodiments of the present invention, a mixed gas monitoring network formed by fixed gas monitoring nodes, mobile gas monitoring nodes and wearable monitoring nodes is used to realize the continuous collection of gas data in the entire tunnel, and a three-dimensional gas concentration field dynamic model covering the entire tunnel construction area is constructed to predict the diffusion trend of harmful gases. Finally, intelligent early warning instructions and ventilation system adjustment instructions are generated and executed based on the prediction results. In this way, the traditional "post-event alarm" is transformed into "pre-event warning and active intervention", which significantly improves the safety, emergency response speed and energy utilization efficiency of tunnel construction.

[0044] Additional aspects and advantages of the present invention will be set forth in part in the description which follows and, in part, will be obvious from the description which follows, or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying any creative work.

[0046] Figure 1 This is a flow chart of an embodiment of the intelligent monitoring method for tunnel construction gas according to the present invention;

[0047] Figure 2 This is a flow chart of step S102 in the tunnel construction gas intelligent monitoring method of the present invention;

[0048] Figure 3This is a flowchart of step S103 of the intelligent monitoring method for tunnel construction gas of the present invention, which predicts the spatiotemporal evolution trend of gas concentration within a preset time period in the future based on the three-dimensional gas concentration field dynamic model;

[0049] Figure 4 This is a flow chart of the generation and execution of the intelligent early warning instruction in step S103 of the intelligent monitoring method for tunnel construction gas of the present invention;

[0050] Figure 5 This is a flow chart of the generation and execution of ventilation system adjustment instructions in step S103 of the intelligent monitoring method for tunnel construction gas of the present invention;

[0051] Figure 6 It is a structural diagram of an embodiment of the intelligent gas monitoring device for tunnel construction of the present invention.

[0052] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0053] The following describes embodiments of the present invention in detail. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended only to explain the present invention and are not to be construed as limiting the present invention.

[0054] Reference Figure 1 As shown, Figure 1 The following is a flowchart of an embodiment of a method for intelligently monitoring gas in tunnel construction according to an embodiment of the present invention. For ease of description, only the portion relevant to the embodiment of the present invention is shown. Specifically, the method for intelligently monitoring gas in tunnel construction can be executed by a server and specifically includes:

[0055] S101. Real-time multi-source gas data is collected through a mixed gas monitoring network deployed within the tunnel construction environment. The mixed gas monitoring network includes fixed gas monitoring nodes, mobile gas monitoring nodes, and wearable monitoring nodes. In other words, the mixed gas monitoring network is a three-dimensional, multi-layered sensing network that reduces monitoring blind spots and acquires data with high temporal and spatial resolution.

[0056] In specific application scenarios, fixed gas monitoring nodes are deployed along the longitudinal direction of the tunnel at preset intervals (for example, every 50-100 meters). The deployment location is preferably in the safe area behind the tunnel face, near large equipment (such as substations, pump rooms), areas with dense construction personnel, and specific sections with complex geological structures and easy escape or accumulation of harmful gases. Each fixed gas monitoring node can integrate multiple sensors and can simultaneously monitor carbon monoxide (CO), methane (CH4), hydrogen sulfide (H2S), nitrogen oxides (NO x ) and the volume fraction of oxygen (O2). Preferably, to improve the accuracy of subsequent models, fixed gas monitoring nodes can also integrate temperature, humidity, air pressure, and wind speed and direction sensors to collect environmental parameters that affect gas diffusion.

[0057] Mobile gas monitoring nodes can fill the monitoring gaps between fixed gas monitoring nodes and focus on coverage of dynamically changing construction areas. Preferably, the mobile gas monitoring nodes include at least one of drone monitoring nodes, crawler or wheeled robot monitoring nodes, and vehicle-mounted monitoring nodes installed on tunnel construction vehicles. In one embodiment, the mobile gas monitoring nodes can be installed on construction vehicles (such as slag trucks, concrete mixers, loaders) or automated equipment (such as tunnel boring machines) that move frequently in the tunnel. These nodes move with the equipment to form a dynamic monitoring track and collect gas data along the way in real time. In another embodiment, a dedicated remote-controlled or autonomously navigated monitoring robot can also be used as a carrier of the mobile gas monitoring node, and it will go to a specific area for detailed inspection according to a preset inspection path or according to server instructions. The mobile node is usually battery-powered and sends data to the server via wireless communication technologies such as 5G, Wi-Fi or LoRa.

[0058] Wearable monitoring nodes are used to ensure individual safety. Every person entering a tunnel is required to wear a wearable monitoring node, which can be integrated into a helmet, worn as an armband, or mounted on a badge. These nodes monitor gas concentrations in the wearer's microenvironment (near the breathing zone) in real time, particularly CO and O2. In addition to data upload capabilities, these nodes also feature local audio, visual, and vibration alarms, alerting the wearer immediately if danger is detected.

[0059] S102. Based on the multi-source gas data, a three-dimensional gas concentration field dynamic model covering the entire tunnel construction area is constructed using data fusion and machine learning algorithms. The purpose of this step is to convert data from discrete monitoring points of different types and different precisions into a continuous, dynamic four-dimensional model that can accurately describe the distribution and changes of gas concentrations in the entire tunnel space. Exemplarily, this step usually first pre-processes the received multi-source heterogeneous data (i.e., multi-source gas data). Subsequently, a data fusion algorithm, such as a Kalman filter or a Bayesian inference model, is used to effectively fuse the high-precision, low-density data provided by the fixed gas monitoring nodes with the highly dynamic, possibly noisy data provided by the mobile gas monitoring nodes and the wearable monitoring nodes of personnel, to obtain a set of more reliable and comprehensive data points. Finally, a three-dimensional gas concentration field is constructed using a machine learning algorithm.

[0060] S103. Based on the three-dimensional gas concentration field dynamic model, predict the spatiotemporal evolution trend of the gas concentration within a preset time period in the future, and generate and execute at least one control instruction according to the prediction result, wherein the control instruction includes an intelligent early warning instruction and / or a ventilation system adjustment instruction.

[0061] Specifically, the system uses the current real-time three-dimensional gas concentration field as input to predict the evolution of the gas concentration distribution throughout the tunnel space over a preset time period (e.g., the next 15 or 30 minutes). The prediction results are presented as a series of three-dimensional gas concentration fields at future moments. Based on these predictions, intelligent warning instructions can be generated and executed. These intelligent warning instructions are forward-looking, not backward-looking. For example, even if the current CO concentration in a certain area is 25 ppm (below the alarm threshold), if the model predicts that the concentration in that area will likely exceed the dangerous threshold of 50 ppm in 30 minutes, a graded warning will be immediately triggered. For example, a level 1 warning (yellow alert) will be generated and notified via the central control room's large screen or the manager's mobile app, stating, "CO concentration in area A is rapidly increasing. Please pay close attention." If the concentration is predicted to exceed the limit within 15 minutes, a level 2 warning (orange alert) will be triggered, and a targeted alert will be sent to the wearable devices of personnel in the area, instructing them to prepare for evacuation. This trend-based warning mechanism provides valuable lead time for personnel to avoid danger and for managers to make decisions.

[0062] At the same time, ventilation system adjustment commands can be generated and executed, enabling intelligent, on-demand ventilation. Based on the predicted spatial and temporal distribution of future gas concentrations, areas of impending or already established high concentrations of harmful gases can be identified. This information is then used to automatically generate precise control commands for the tunnel's intelligent ventilation system. For example, if a rapid accumulation of CH4 concentrations near the tunnel face is predicted within 15 minutes, a command is automatically sent to the jet fans or forced-in duct controllers in that area to increase their speed by 30% or activate backup fans for localized enhanced ventilation. If construction activity is predicted to cease in a particular section of the tunnel and gas concentrations are well below safe levels, the fan power in that area can be appropriately reduced, thereby avoiding energy waste. This predictive control approach not only ensures effective ventilation in critical areas but also achieves overall energy savings and consumption reductions, transforming the ventilation system's operation from a crude, intensive mode to a precise, differentiated one.

[0063] According to the intelligent tunnel construction gas monitoring method, device and system provided by the embodiments of the present invention, a mixed gas monitoring network formed by fixed gas monitoring nodes, mobile gas monitoring nodes and wearable monitoring nodes is used to realize the continuous collection of gas data in the entire tunnel, and a three-dimensional gas concentration field dynamic model covering the entire tunnel construction area is constructed to predict the diffusion trend of harmful gases. Finally, intelligent early warning instructions and ventilation system adjustment instructions are generated and executed based on the prediction results. In this way, the traditional "post-event alarm" is transformed into "pre-event warning and active intervention", which significantly improves the safety, emergency response speed and energy utilization efficiency of tunnel construction.

[0064] Preferably, the mobile gas monitoring nodes perform dynamic path planning according to preset inspection paths or according to model building requirements to actively detect monitoring blind spots or key areas of concern. The administrator can pre-set one or more inspection paths covering the main areas of the tunnel, such as paths that go back and forth along the center line of the tunnel, cover the cross section of the tunnel in an "S" or "Z" shape, or surround key equipment (such as main ventilation fans, refuge chamber entrances). The mobile gas monitoring nodes will autonomously execute these paths according to the set schedule, thereby ensuring periodic and basic data coverage of the vast area between the fixed monitoring nodes. The data collected in this mode provides a stable and regular background data source for the construction of the three-dimensional gas concentration field dynamic model, ensuring the global continuity and basic accuracy of the model.

[0065] Furthermore, a dynamic path planning mechanism based on model building requirements enables on-demand allocation and intelligent scheduling of monitoring resources. During the continuous construction and updating of the three-dimensional gas concentration field dynamic model, not only are concentration predictions calculated for each location, but the uncertainty or confidence level of these predictions can also be quantified. When the prediction uncertainty for a local area within the tunnel exceeds a preset threshold, this typically indicates a lack of sufficient and timely field data support, creating a potential "monitoring blind spot." In this situation, the server generates a high-priority dynamic detection task and issues it to the mobile gas monitoring node closest to the target area or in a suitable state. Upon receiving the task, the mobile gas monitoring node initiates a dynamic path planning algorithm to calculate an optimal path that allows it to quickly and safely reach the target area and efficiently cover it. Upon reaching the target area, the mobile gas monitoring node conducts detailed detection according to a predefined scanning pattern (such as a grid or spiral pattern), transmitting the collected high-density, timely gas data back to the server in real time. The server uses this new data to instantly update the model, rapidly reducing the prediction uncertainty in that area and narrowing the monitoring blind spot. In addition, when the model predicts that a certain area has a very high probability of becoming a "key focus area" where harmful gases accumulate in a short period of time in the future, this dynamic path planning mechanism can also be triggered to dispatch mobile gas monitoring nodes to the area in advance for verification and key monitoring.

[0066] In one embodiment of the present invention, in the step of collecting multi-source gas data in real time, three-dimensional spatial coordinate information and time stamp information are synchronously added to the gas concentration data collected by each monitoring node.

[0067] Among them, the coordinates of the fixed gas monitoring node are preset values, the coordinates of the mobile gas monitoring node are obtained through the positioning module, and the coordinates of the wearable monitoring node are obtained through its integrated UWB or Bluetooth AOA positioning module to form a four-dimensional data set with spatiotemporal tags.

[0068] In other words, the three-dimensional spatial coordinates and timestamp information are added to each piece of gas concentration data collected by each monitoring node in the mixed gas monitoring network, transforming isolated and disordered data into a structured four-dimensional dataset with spatiotemporal labels.

[0069] Specifically, the way of attaching the spatiotemporal tag varies according to the type of monitoring node. For the fixed gas monitoring node, its three-dimensional spatial coordinates are constant preset values. During the deployment phase, the position of each fixed gas monitoring node is measured to obtain its coordinates in the unified coordinate system of the tunnel. For the mobile gas monitoring node, since its position is constantly changing dynamically, it must rely on its own integrated positioning module to obtain coordinates in real time. Exemplarily, the positioning module can be a simultaneous localization and mapping (SLAM) technology based on a lidar or a depth camera. SLAM technology enables mobile gas monitoring nodes to build an environmental map in an unknown environment in real time, and use the map to synchronously estimate their own posture, thereby continuously outputting three-dimensional coordinates. For the personnel wearable monitoring node, a more lightweight and accurate positioning module can be selected for integration. For example, ultra-wideband (UWB) positioning technology can be used, or Bluetooth angle of arrival (Bluetooth AOA) technology can be used.

[0070] This embodiment simultaneously attaches 3D spatial coordinates and timestamps to every piece of data generated by fixed, mobile, and wearable monitoring nodes. This allows for the integration of fragmented information from diverse sources and properties into a unified, standardized, and analyzable 4D dataset. Each record in this dataset includes time, location, and gas concentration information, providing essential, high-quality data input for the subsequent construction of a high-fidelity 3D gas concentration field dynamic model, accurate spatiotemporal evolution trend prediction, and closed-loop intelligent control.

[0071] Reference Figure 2 As shown, in one embodiment of the present invention, step S102 may specifically include:

[0072] S201. De-noise, standardize, and spatiotemporally align the collected multi-source gas data, and convert the sparse, heterogeneous data from fixed gas monitoring nodes, mobile gas monitoring nodes, and wearable gas monitoring nodes into a structured data set under a unified spatiotemporal reference system.

[0073] Specifically, the server first performs a denoising operation on the received gas concentration data. For example, algorithms such as median filtering or wavelet denoising are used. The denoised data is then normalized. Because the brands, models, and accuracies of sensors used in fixed, mobile, and wearable gas monitoring nodes may vary, resulting in differences in the dimensions and numerical ranges of their output data, methods such as min-max scaling or Z-score normalization are used to map all concentration data to a unified, dimensionless interval. This provides a foundation for subsequent multi-source data fusion and stable training of machine learning models. Finally, each processed concentration data item is bound to its 3D spatial coordinate information and timestamp information provided at the time of acquisition by a positioning module such as UWB, Bluetooth AOA, or SLAM. This integrates the originally fragmented and heterogeneous data streams into a structured 4D dataset based on a unified spatiotemporal reference system, containing 3D spatial coordinates, time information, and concentration data.

[0074] S202 : Based on the processed structured data set, a spatial interpolation algorithm is used to calculate the gas concentration values ​​at locations where no sensors are deployed in the tunnel space to obtain interpolation calculation results.

[0075] That is to say, after obtaining the structured data set, a spatial interpolation calculation is performed, the purpose of which is to infer the continuous gas concentration distribution of the entire tunnel space from the discrete monitoring point data. Since the monitoring nodes are essentially sparse sampling points no matter how they are deployed, they cannot directly cover every corner of the tunnel. In one example, the Kriging interpolation method (Kriging) can be used. This method is a geostatistical optimal interpolation technique based on spatial autocorrelation. It not only considers the distance relationship between the interpolation point and the surrounding known monitoring points, but more importantly, it quantifies and utilizes the spatial correlation structure of gas concentration through variance function analysis, that is, the closer the point is, the more similar its concentration value is. By solving the Kriging equations, the concentration estimate can be calculated for the location where the sensor is not deployed in the tunnel space, and the variance of the estimate, that is, the uncertainty, can be given at the same time.

[0076] S203. By introducing timing constraints and physical diffusion model constraints, the interpolation results are constrained and adjusted to ensure that the changes in the gas concentration field between adjacent time frames conform to the physical laws of gas diffusion, so as to form three-dimensional gas concentration field data.

[0077] The interpolation results between two adjacent time points may exhibit physically unreasonable, drastic jumps. This step introduces timing constraints and physical diffusion model constraints to constrain and adjust the interpolation results. The introduction of timing constraints can be implemented using a state-space model (such as a Kalman filter). The optimized gas concentration field at the previous moment is used as the prior prediction for the current moment, and the spatial interpolation result at the current moment is then used as the observation. The Kalman filter update step corrects the prior prediction, resulting in a concentration field that smoothly evolves over time, taking into account the continuity of historical states and incorporating current real-time observations. Furthermore, to ensure that the spatial distribution of the concentration field conforms to physical reality, this step introduces a simplified physical diffusion model as a physical constraint. This physical diffusion model can be a simplified computational fluid dynamics (CFD) model based on the Navier-Stokes equations. This model can be used to determine whether there are regions in the interpolation results that violate basic gas diffusion laws (such as gas appearing out of thin air or diffusing rapidly against strong winds). The concentration values ​​in these regions are then adjusted to more accurately reflect the actual physical process. After the dual constraint adjustment, the final three-dimensional gas concentration field data is not only continuous in space, but also smooth in time evolution and conforms to physical logic.

[0078] S204. Spatially align and fuse the three-dimensional gas concentration field data with the geometric model of the tunnel construction environment to construct a continuous and smooth three-dimensional gas concentration field dynamic model covering the entire tunnel.

[0079] To transform abstract concentration field data into a tool that managers can intuitively understand and interact with, this step spatially registers and deeply fuses the constraint-adjusted 3D gas concentration field data with a pre-built, high-precision 3D geometric model of the tunnel construction environment (e.g., a Building Information Model (BIM)). Spatial registration ensures precise alignment of the gas concentration data coordinate system with the BIM model, accurately mapping each concentration data point to its actual physical location within the BIM model. The fusion process uses visualization rendering technology to overlay the 3D gas concentration field on the tunnel BIM model in the form of a color cloud, isosurface, or volume rendering. For example, a color spectrum from blue (safe) to red (dangerous) can be used to represent gas concentration levels. This creates a visually continuous, smooth, and reliable 3D dynamic model of the gas concentration field covering the entire tunnel. Within this dynamically updated model, managers can zoom, rotate, and slice to visually visualize the gas distribution within the cross-section and observe the accumulation, diffusion, and flow trends of hazardous gases within the tunnel, providing reliable and intuitive decision support for subsequent forecasting, early warning, and intelligent control.

[0080] Through the above steps, this embodiment converts raw, sparse, and heterogeneous data from sensors of different types and precision into a high-quality structured data set that has undergone quality control and is based on a unified spatiotemporal benchmark, providing a reliable data foundation for all subsequent analysis and modeling. By introducing Kriging interpolation, time series constraints, and physical model constraints, the constructed three-dimensional gas concentration field not only achieves spatial coverage and continuity, but also ensures its smoothness in time evolution and rationality in physical processes, greatly improving the fidelity and credibility of the model. In addition, by integrating with the BIM model to build a digital twin, the complex and abstract four-dimensional data is transformed into an intuitive, interactive, dynamic visualization model, greatly enhancing the interpretability and operability of the data.

[0081] Reference Figure 3 As shown, in one embodiment of the present invention, in step S103, based on the three-dimensional gas concentration field dynamic model, the temporal and spatial evolution trend of the gas concentration within a preset time period in the future is predicted, including:

[0082] S301. A three-dimensional gas concentration field dynamic model on a continuous time series is used as basic input data, and known airflow field data in the tunnel is collected at the same time. The airflow field data at least includes wind speed, wind direction, and turbulence characteristics.

[0083] Specifically, the temporally continuous sequence of three-dimensional gas concentration field dynamic models generated in the previous stage is used as the core basic input data. This sequence is a series of gas concentration distributions throughout the tunnel space, ordered by time. To make the prediction more consistent with physical reality, known real-time and historical airflow field data within the tunnel are simultaneously collected and integrated. These airflow field data are physical quantities that affect gas diffusion. They include wind speed and direction information measured by wind speed and direction sensors deployed in the tunnel, as well as quantitative descriptions of turbulent characteristics, such as turbulent kinetic energy and turbulent dissipation rate. These parameters facilitate the accurate simulation of the mixing and dilution process of gases in complex airflows. These airflow field data are aligned in time and space with the three-dimensional gas concentration field data, together forming the composite input of the prediction model.

[0084] S302: Using a long short-term memory network model or a Kalman filter algorithm, a spatiotemporal evolution prediction model of the gas concentration field is constructed.

[0085] In this step, after the input data is prepared, a time series prediction algorithm is used to construct a prediction model for the spatiotemporal evolution of the gas concentration field. For example, a long short-term memory (LSTM) model is used. The input to this LSTM model is continuous three-dimensional gas concentration field data within a time window and the corresponding airflow field data. By training on a large amount of historical data, the LSTM model can autonomously learn the complex nonlinear laws of the time evolution of gas concentration under airflow conditions. Alternatively, a Kalman filter algorithm can be used. This algorithm treats the entire three-dimensional gas concentration field as a high-dimensional system state and establishes a state transition equation based on physics or empirical statistics. This equation describes how the concentration field evolves from the current moment to the next, with the airflow field data serving as the driving parameter of this equation.

[0086] S303: Based on the spatiotemporal evolution prediction model, a four-dimensional data set of the harmful gas concentration field within a preset time period in the future is calculated and generated.

[0087] This step performs predictive calculations based on the trained spatiotemporal evolution prediction model to generate a harmful gas concentration field within a preset time period in the future (for example, the next 30 minutes). Specifically, the most recent historical data is used to predict the three-dimensional gas concentration field of the first time step in the future (for example, t+1 minute); then, this newly predicted concentration field will be used as part of the input to predict the concentration field of the second time step (t+2 minutes). This cycle is repeated until the frame-by-frame prediction of the entire preset time period is completed. The final output of this process is a four-dimensional data set that includes spatial and temporal dimensions, depicting the changes in gas concentration at any location in the tunnel over a period of time in the future.

[0088] S304. Extract and quantify the diffusion range, movement path, and concentration peak change of the harmful gas from the predicted four-dimensional data set.

[0089] Since the original four-dimensional prediction data set is complete but has too high an information density, it can be further analyzed and refined to extract key indicators that have direct guiding significance for safety management. Therefore, the predicted four-dimensional data set is subjected to automated post-processing analysis to quantitatively extract the evolution characteristics of harmful gases. Specifically, by identifying and tracking high-concentration gas clouds in the prediction results, its estimated diffusion range in the future time period is determined, that is, how the boundary of the area where the gas concentration exceeds the standard will expand or shrink. At the same time, by calculating the displacement of the gas cloud center of mass or concentration peak point on a continuous time frame, its expected movement path can be depicted. It is also possible to continuously track the concentration peak within the gas cloud and generate a curve of the peak concentration changing with time, so as to intuitively judge whether the danger is rapidly accumulating, stabilizing, or dissipating.

[0090] S305. Conduct a graded risk assessment of the predicted concentration field based on the hazardous gas threshold standards to identify high-risk areas and time windows that may appear in the future.

[0091] For example, the danger threshold standards for different harmful gases (such as CO, CH4, etc.) are preset according to safety regulations. Then, the predicted concentration values ​​at various future time and space points are compared with these danger thresholds. Through comparison, it is possible to accurately identify in which specific time window in the future, in which specific section of the tunnel, the concentration of a certain gas will most likely exceed a certain risk level. The result of this graded risk assessment is no longer a simple "yes or no" judgment, but a detailed and forward-looking description of future risks in time, space and intensity, providing a clear and unambiguous decision-making basis for the subsequent initiation of graded warnings and the implementation of proactive intervention.

[0092] Through the above steps, this embodiment greatly improves the accuracy and physical authenticity of the prediction of the spatiotemporal evolution trend of gases by integrating the gas concentration field and key airflow field data and utilizing algorithms such as LSTM or Kalman filtering. In addition, it can convert complex model outputs into a series of intuitive and quantifiable key risk indicators, such as diffusion range, movement path, and concentration peak changes, so that the understanding of future risks is no longer a vague qualitative judgment, but an accurate quantitative analysis. In addition, by performing a graded risk assessment on the prediction results, potential high-risk areas and time windows can be identified in advance, and the passive response mode can be transformed into an active prevention mode, thereby gaining valuable advance time to protect the lives of tunnel construction workers and optimize the allocation of emergency resources, and realizing intelligent and forward-looking safety management of tunnel construction.

[0093] In one embodiment of the present invention, after step S102, the method further includes:

[0094] When the gas concentration value in the three-dimensional gas concentration field dynamic model exceeds a preset threshold, the maximum possible leakage source or abnormal accumulation source of the harmful gas is traced back and calibrated in a three-dimensional visual manner by real-time analysis of the gradient vector of the three-dimensional gas concentration field or application of a computational fluid dynamics inverse diffusion model.

[0095] Specifically, when the gas concentration at any location in the three-dimensional gas concentration field dynamic model exceeds a pre-set safety threshold, the server initiates a source tracing analysis process. This aims to shift the focus of safety management from post-incident emergency response to rapid root cause diagnosis, thereby providing accurate decision-making support for eliminating potential hazards. Specifically, real-time analysis of the three-dimensional gas concentration field gradient vector can be used to quickly locate the source. Upon detecting an excessive concentration event, the gradient of the three-dimensional gas concentration field at that moment is immediately calculated across the entire tunnel space. The gradient vector at any point points in the direction of the fastest increase in gas concentration at that point. Therefore, the diffusion path of hazardous gases from the source can be viewed macroscopically as following the gradient field. Conversely, tracing the source requires starting from the point or region with the highest concentration and performing path integration or iterative tracing in the opposite direction of the gradient vector. For example, one or more "virtual tracer particles" are automatically initialized at the peak concentration point and allowed to move within a negative gradient field. The trajectories of these particles retrace the gas diffusion path, ultimately converging or terminating at one or more regions with extremely low gradient values. These regions are then identified as the most likely leak source or starting point of abnormal accumulation. This method has fast calculation speed, low requirements on computing resources, and can provide near real-time preliminary source pointing for emergency response.

[0096] In addition, the reverse diffusion model of computational fluid dynamics (CFD) can also be applied to trace the source in reverse. This method is suitable for scenarios that require higher positioning accuracy or when the airflow environment in the tunnel is complex. The reverse diffusion model is an "inverse problem" solving process. The currently known, high-precision three-dimensional gas concentration field distribution is used as the "result", and the real-time monitored airflow field data such as wind speed, wind direction, temperature, and air pressure in the tunnel are used as the "known conditions". Then, the "cause" that can lead to this result is reversely solved, that is, the location, shape, and intensity of the leakage source. This method is computationally intensive and relatively time-consuming, but its results have higher physical authenticity and positioning accuracy.

[0097] By tracing back to the source of gas, this embodiment enables root cause analysis of safety incidents. Using 3D visualization, it quickly and accurately pinpoints the source of leaks or abnormal accumulations, significantly reducing the time it takes for emergency responders to troubleshoot and locate faults. This provides valuable time for emergency response and reduces the risk of secondary incidents. Furthermore, it provides managers with a more targeted basis for decision-making, making emergency response more accurate and effective.

[0098] Reference Figure 4 As shown, in one embodiment of the present invention, the generation and execution of the intelligent early warning instruction includes:

[0099] S401. Based on the predicted distribution of hazardous gas concentrations and personnel location information, the temporal and spatial relationships between personnel and high-risk areas are automatically calculated, and differentiated graded warning strategies are generated for different groups of personnel.

[0100] Specifically, this step first performs a spatiotemporal correlation analysis based on the future spatiotemporal distribution of the harmful gas concentration field predicted in the previous step, and combines the location information uploaded in real time by the personnel's wearable devices (such as UWB or Bluetooth AOA positioning modules). Calculate the individual risk exposure of each person in the tunnel. Determine whether the person is currently in a high-concentration area. By comparing the real-time motion vector of the person with the predicted gas cloud diffusion and movement path. Determine the probability of the person intersecting with the high-risk area within the preset time period in the future, the estimated entry time, etc., to form a risk assessment result. Based on the individual risk assessment results, generate and match differentiated graded warning strategies for personnel in different situations, different types of work, and different groups.

[0101] S402: For personnel who are about to enter a predicted high-risk area, structured warning information including the type, concentration, risk level and optimal avoidance path of hazardous gases is pushed to the personnel's wearable device.

[0102] In other words, when it's determined that a person or individuals are about to enter a predicted high-risk area, a structured warning message is immediately triggered. This information is sent via a wireless communication network to the wearable monitoring node worn by the targeted individuals. This structured warning information includes information such as the type of hazardous gas, concentration, risk level, and optimal avoidance path, helping relevant personnel accurately understand the current dangerous situation.

[0103] S403. Select and activate a corresponding alarm method based on the risk level and urgency, including wearable device vibration reminder, sound and light alarm, regional broadcast or full tunnel evacuation instruction.

[0104] For lower-level risks (for example, concentrations are on an upward trend but will not exceed the standard in the short term), only a vibration reminder on the wearable device may be triggered, or a notice of concern may be pushed to the management terminal of the regional supervisor to avoid unnecessary interference with normal construction. As the risk level increases, the alarm method will be gradually upgraded, for example, activating the audio and visual alarm of the wearable device (beeping and flashing at a specific frequency). When an extremely high risk is predicted or a dangerous situation has actually occurred, the highest level of alarm will be activated. In addition to the sounding of all individual and regional alarms, the regional broadcast system in the tunnel will also be controlled to broadcast dangerous situations and evacuation instructions by voice.

[0105] S404: Track and record personnel's movement trajectories and response behaviors after the warning, evaluate the warning effectiveness, and use this as feedback to optimize the tiered warning strategy. In other words, quantitative behavioral data such as personnel's movement trajectories and response behaviors will be input into the warning strategy model as feedback. By analyzing this data, the system can self-learn and adjust, thereby optimizing the tiered warning strategy.

[0106] Through this differentiated, hierarchical early warning process, this embodiment accurately maps complex future risk scenarios to individual safety guidance, enabling refined management that matches people to risks. This not only enables high-risk population identification, route planning, and multimodal alerting, but also enables continuous self-optimization of early warning strategies through real-time performance evaluation, significantly improving the pertinence, timeliness, and enforceability of early warnings.

[0107] Reference Figure 5 As shown, in one embodiment of the present invention, the generation and execution of the ventilation system adjustment instruction includes:

[0108] S501. Based on the predicted three-dimensional gas concentration field and airflow field data, a multi-objective optimization model is constructed, wherein the model simultaneously considers three objectives: rapid removal efficiency of harmful gases, minimization of energy consumption, and environmental comfort.

[0109] This step constructs a multi-objective optimization model whose input data are the future three-dimensional gas concentration field predicted in the previous step and the real-time airflow field data. The model seeks a balance between three interrelated objectives. The first objective is "rapid removal efficiency of harmful gases." Its quantitative indicators can be the shortest time required to reduce the predicted high-concentration area to below the safety threshold while meeting safety constraints, or the maximum average rate of decrease in harmful gas concentration per unit time. The second objective is "minimization of energy consumption," which directly corresponds to the total electrical energy consumption of the entire tunnel ventilation system (including all jet fans, axial flow fans, etc.) during the execution of the control task. The third objective is "environmental comfort." This objective focuses on the quality of the working environment for construction workers. Its quantitative indicators can include controlling the wind speed in the work area within the acceptable comfort range for humans and controlling the noise generated by the fan operation below the limit specified by occupational health standards.

[0110] S502. Setting constraints for the multi-objective optimization model. The agreed conditions include the performance boundaries of each ventilation device, the safe gas concentration thresholds of each area in the tunnel space, and the upper limit of energy consumption.

[0111] "Safety gas concentration threshold" means that at any time and in any area of ​​the tunnel where people are active, the predicted concentration of harmful gases needs to be lower than the safety upper limit specified by the state or industry to ensure the safety of personnel. "Performance boundary of each ventilation equipment" is a constraint that takes the physical limitations of each fan into consideration, including its maximum and minimum operating power, rated air volume, speed adjustment range, etc. It ensures that the model will not generate control instructions that exceed the actual capabilities of the equipment. In addition, an "energy consumption upper limit" can also be set as a constraint. For example, the total energy consumption in a single day or a specific operating cycle must not exceed the preset budget value, or the instantaneous total power cannot exceed the load limit of the substation. These constraints together define a multi-dimensional feasible solution space, in which the optimal operating strategy will be sought.

[0112] S503. Generate a spatiotemporal dynamic ventilation control strategy by solving a multi-objective optimization model. The ventilation control strategy allocates optimal operating parameters to ventilation equipment in different zones and time periods within the tunnel. The ventilation control strategy is transmitted to the controller of each ventilation equipment via an industrial field bus or a wireless communication network.

[0113] For example, an optimization algorithm such as the non-dominated sorting genetic algorithm (NSGA-II) can be used for the solution. The ventilation control strategy assigns optimal operating parameters, such as fan start / stop status, operating speed percentage, or target air volume, to the ventilation equipment in each ventilation zone within the tunnel (e.g., the tunnel face, secondary lining trolley, entrance / exit area, etc.) for different future time periods. These ventilation control strategies are transmitted to the controllers of each ventilation device within the tunnel via an industrial fieldbus or industrial-grade wireless communication network, which directly drives the devices to perform the corresponding adjustment actions.

[0114] S504: Monitor changes in the gas concentration field after ventilation adjustments in real time, evaluate ventilation effectiveness, and feed the evaluation results back to the multi-objective optimization model. This feedback mechanism allows the model to dynamically modify its internal parameters. For example, it can adjust the evaluation of a particular fan's performance under specific operating conditions or optimize the estimation of local wind resistance within the tunnel, thereby achieving self-calibration and performance iteration of the model.

[0115] Through the above-mentioned multi-objective optimization-driven ventilation control process, this embodiment can achieve on-demand, zoned, and differentiated operation of the ventilation system while ensuring tunnel construction safety, significantly shortening the time required to remove harmful gases and reducing overall energy consumption while maintaining a comfortable working environment for personnel. The closed-loop feedback mechanism ensures that the strategy is synchronized with actual on-site conditions, improving the real-time, accuracy, and energy-saving benefits of ventilation adjustment.

[0116] Reference Figure 6 As shown, Figure 6The following is a schematic diagram of the structure of an embodiment of an intelligent gas monitoring device for tunnel construction provided by an embodiment of the present invention. For ease of description, only the parts related to the embodiment of the present invention are shown. Specifically, the intelligent gas monitoring device for tunnel construction includes:

[0117] The collection unit 601 is used to collect multi-source gas data in real time through a mixed gas monitoring network deployed in the tunnel construction environment. The mixed gas monitoring network includes fixed gas monitoring nodes, mobile gas monitoring nodes and wearable monitoring nodes.

[0118] The concentration field model construction unit 602 is used to construct a three-dimensional gas concentration field dynamic model covering the entire tunnel construction area based on the multi-source gas data using data fusion and machine learning algorithms.

[0119] The prediction control unit 603 is used to predict the spatiotemporal evolution trend of the gas concentration within a preset time period in the future based on the three-dimensional gas concentration field dynamic model, and generate and execute at least one control instruction according to the prediction result, and the control instruction includes an intelligent early warning instruction and / or a ventilation system adjustment instruction.

[0120] The intelligent tunnel construction gas monitoring device provided by the embodiment of the present invention utilizes a mixed gas monitoring network formed by fixed gas monitoring nodes, mobile gas monitoring nodes and wearable monitoring nodes to realize the continuous collection of gas data in the entire tunnel, and construct a three-dimensional gas concentration field dynamic model covering the entire tunnel construction area, thereby predicting the diffusion trend of harmful gases. Finally, based on the prediction results, intelligent early warning instructions and ventilation system adjustment instructions are generated and executed, thereby realizing the transformation of traditional "post-event alarm" into "pre-event warning and active intervention", significantly improving the safety, emergency response speed and energy utilization efficiency of tunnel construction.

[0121] In one embodiment of the present invention, the mobile gas monitoring node includes at least one of an unmanned aerial vehicle monitoring node, a crawler or wheeled robot monitoring node, and a vehicle-mounted monitoring node installed on a tunnel construction vehicle.

[0122] The mobile gas monitoring node performs dynamic path planning according to a preset inspection path or according to model building requirements to actively detect monitoring blind spots or key areas of concern.

[0123] In one embodiment of the present invention, the acquisition unit is also used to synchronously add three-dimensional spatial coordinate information and timestamp information to the gas concentration data collected by each monitoring node; wherein, the coordinates of the fixed gas monitoring node are preset values, the coordinates of the mobile gas monitoring node are obtained through the positioning module, and the coordinates of the personnel wearable monitoring node are obtained through its integrated UWB or Bluetooth AOA positioning module to form a four-dimensional data set with time and space tags.

[0124] In one embodiment of the present invention, the concentration field model building unit includes:

[0125] The preprocessing module is used to denoise, standardize and align the collected multi-source gas data in time and space, and convert the sparse and heterogeneous data from fixed gas monitoring nodes, mobile gas monitoring nodes and wearable gas monitoring nodes into a structured data set under a unified time and space reference system.

[0126] The interpolation calculation module is used to calculate the gas concentration value at the location where no sensor is deployed in the tunnel space based on the processed structured data set using a spatial interpolation algorithm to obtain the interpolation calculation result.

[0127] The adjustment module is used to adjust the interpolation results by introducing timing constraints and physical diffusion model constraints to ensure that the changes in the gas concentration field between adjacent time frames conform to the physical laws of gas diffusion, so as to form three-dimensional gas concentration field data.

[0128] The first construction module is used to spatially align and fuse the three-dimensional gas concentration field data with the geometric model of the tunnel construction environment to construct a continuous and smooth three-dimensional gas concentration field dynamic model covering the entire tunnel.

[0129] In one embodiment of the present invention, the prediction control unit includes:

[0130] The data collection module is used to use the three-dimensional gas concentration field dynamic model on the continuous time series as basic input data, and at the same time collect known airflow field data in the tunnel, wherein the airflow field data at least includes wind speed, wind direction and turbulence characteristics.

[0131] The second building module is used to construct a spatiotemporal evolution prediction model of the gas concentration field by using a long short-term memory network model or a Kalman filter algorithm.

[0132] The calculation module is used to calculate and generate a four-dimensional data set of the harmful gas concentration field within a preset time period in the future based on the spatiotemporal evolution prediction model.

[0133] The extraction module is used to extract and quantify the diffusion range, movement path and concentration peak change of harmful gases from the predicted four-dimensional data set.

[0134] The assessment module is used to conduct a graded risk assessment of the predicted concentration field based on the hazardous threshold standards of harmful gases and identify high-risk areas and time windows that may appear in the future.

[0135] In one embodiment of the present invention, the apparatus further comprises:

[0136] The tracing unit is used to trace back and locate the maximum possible leakage source or abnormal accumulation source of harmful gas in a three-dimensional visual manner by real-time analysis of the gradient vector of the three-dimensional gas concentration field or application of a computational fluid dynamics inverse diffusion model when the gas concentration value in the three-dimensional gas concentration field dynamic model exceeds a preset threshold.

[0137] In one embodiment of the present invention, the prediction control unit further includes:

[0138] The graded warning module is used to automatically calculate the spatiotemporal relationship between personnel and high-risk areas based on the predicted distribution of harmful gas concentration fields and personnel location information, and generate differentiated graded warning strategies for different groups of personnel.

[0139] The push module is used to push structured warning information containing the type, concentration, risk level and optimal avoidance path of hazardous gases to people who are about to enter predicted high-risk areas through their wearable devices.

[0140] The alarm module is used to select and activate the corresponding alarm mode according to the risk level and urgency. The alarm mode includes wearable device vibration reminder, sound and light alarm, regional broadcast or full tunnel evacuation instruction.

[0141] The optimization module is used to track and record the movement trajectory and response behavior of personnel after the warning, evaluate the warning effect, and use the warning effect as feedback information to optimize the hierarchical warning strategy.

[0142] In one embodiment of the present invention, the prediction control unit includes:

[0143] The third construction module is used to construct a multi-objective optimization model based on the predicted three-dimensional gas concentration field and airflow field data. The model simultaneously considers the three objectives of rapid removal efficiency of harmful gases, minimization of energy consumption and environmental comfort.

[0144] A setting module is used to set constraints for the multi-objective optimization model. The agreed conditions include the performance boundaries of each ventilation device, the safe gas concentration thresholds of each area in the tunnel space, and the upper limit of energy consumption;

[0145] The solution module is used to generate a spatiotemporal dynamic ventilation control strategy by solving a multi-objective optimization model. The ventilation control strategy assigns optimal operating parameters to ventilation equipment in different zones and time periods within the tunnel. The ventilation control strategy is transmitted to the controller of each ventilation equipment via an industrial field bus or a wireless communication network.

[0146] The feedback module is used to monitor the changes in the gas concentration field after ventilation adjustment in real time, evaluate the ventilation effect, and feed back the evaluation results to the multi-objective optimization model.

[0147] Reference Figure 6 As shown, Figure 6 The following is a schematic diagram of the structure of an embodiment of the tunnel construction gas intelligent monitoring system provided by an embodiment of the present invention. For ease of description, only the parts related to the embodiment of the present invention are shown. Specifically, the tunnel construction gas intelligent monitoring system includes:

[0148] The mixed gas monitoring network is used to collect multi-source gas data in real time. It includes multiple fixed gas monitoring nodes deployed in the tunnel, at least one mobile gas monitoring node and multiple wearable monitoring nodes.

[0149] A server is communicatively connected to the mixed gas monitoring network, and is configured to: receive and process the multi-source gas data; construct a three-dimensional gas concentration field dynamic model covering the entire tunnel construction area based on the multi-source gas data using data fusion and machine learning algorithms; and predict the spatiotemporal evolution trend of gas concentration within a future preset time period based on the three-dimensional gas concentration field dynamic model, and generate and execute at least one control instruction based on the prediction result, the control instruction including an intelligent early warning instruction and / or a ventilation system adjustment instruction.

[0150] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between the various embodiments can be referred to in conjunction with each other. For device or system embodiments, since they are generally similar to method embodiments, their description is relatively simple, and relevant parts can be referred to the partial description of the method embodiments.

[0151] It should also be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or device comprising the element.

[0152] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0153] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for intelligent monitoring of gas in tunnel construction, characterized in that: include: Real-time collection of multi-source gas data via a mixed gas monitoring network deployed within the tunnel construction environment, comprising fixed gas monitoring nodes, mobile gas monitoring nodes, and wearable monitoring nodes. Based on the multi-source gas data, a three-dimensional gas concentration field dynamic model covering the entire tunnel construction area is constructed using data fusion and machine learning algorithms; Based on the three-dimensional gas concentration field dynamic model, the spatiotemporal evolution trend of the gas concentration within a preset time period in the future is predicted, and at least one control instruction is generated and executed according to the prediction result. The control instruction includes an intelligent early warning instruction and / or a ventilation system adjustment instruction.

2. The intelligent monitoring method for tunnel construction gas according to claim 1, characterized in that: The mobile gas monitoring node includes at least one of an unmanned aerial vehicle (UAV) monitoring node, a crawler or wheeled robot monitoring node, and a vehicle-mounted monitoring node installed on a tunnel construction vehicle; The mobile gas monitoring node performs dynamic path planning according to a preset inspection path or according to model building requirements to actively detect monitoring blind spots or key areas of concern.

3. The intelligent monitoring method for tunnel construction gas according to claim 1, characterized in that: In the step of collecting multi-source gas data in real time, three-dimensional spatial coordinate information and timestamp information are synchronously added to the gas concentration data collected by each monitoring node; Among them, the coordinates of the fixed gas monitoring node are preset values, the coordinates of the mobile gas monitoring node are obtained through the positioning module, and the coordinates of the wearable monitoring node are obtained through its integrated UWB or Bluetooth AOA positioning module to form a four-dimensional data set with spatiotemporal tags.

4. The intelligent monitoring method for tunnel construction gas according to claim 1, characterized in that: The method of constructing a three-dimensional gas concentration field dynamic model covering the entire tunnel construction area based on the multi-source gas data and utilizing data fusion and machine learning algorithms includes: De-noising, standardization, and spatiotemporal alignment are performed on the collected multi-source gas data, converting the sparse, heterogeneous data from fixed gas monitoring nodes, mobile gas monitoring nodes, and wearable gas monitoring nodes into a structured dataset in a unified spatiotemporal reference system. Based on the processed structured data set, a spatial interpolation algorithm is used to calculate the gas concentration values ​​at locations where sensors are not deployed in the tunnel space, and the interpolation calculation results are obtained; By introducing timing constraints and physical diffusion model constraints, the interpolation results are constrained and adjusted to ensure that the changes in the gas concentration field between adjacent time frames conform to the physical laws of gas diffusion, thereby forming three-dimensional gas concentration field data; The three-dimensional gas concentration field data is spatially registered and fused with the geometric model of the tunnel construction environment to construct a continuous and smooth three-dimensional gas concentration field dynamic model covering the entire tunnel.

5. The intelligent monitoring method for tunnel construction gas according to claim 1, characterized in that: The predicting of the spatiotemporal evolution trend of the gas concentration within a future preset time period based on the three-dimensional gas concentration field dynamic model includes: A three-dimensional gas concentration field dynamic model based on a continuous time series is used as basic input data, and known airflow field data in the tunnel is collected at the same time, wherein the airflow field data includes at least wind speed, wind direction and turbulence characteristics; Use the long short-term memory network model or Kalman filter algorithm to build a prediction model of the spatiotemporal evolution of the gas concentration field; Based on the spatiotemporal evolution prediction model, a four-dimensional data set of the harmful gas concentration field within a preset time period in the future is calculated and generated; Extract and quantify the diffusion range, movement path, and concentration peak changes of harmful gases from the predicted four-dimensional data set; Based on the hazardous threshold standards for harmful gases, a graded risk assessment is conducted on the predicted concentration field to identify high-risk areas and time windows that may appear in the future.

6. The intelligent monitoring method for tunnel construction gas according to claim 1, characterized in that: After constructing the three-dimensional gas concentration field dynamic model covering the entire tunnel construction area, the method further includes: When the gas concentration value in the three-dimensional gas concentration field dynamic model exceeds a preset threshold, the maximum possible leakage source or abnormal accumulation source of the harmful gas is traced back and calibrated in a three-dimensional visual manner by real-time analysis of the gradient vector of the three-dimensional gas concentration field or application of a computational fluid dynamics inverse diffusion model.

7. The intelligent monitoring method for tunnel construction gas according to claim 1, characterized in that: The generation and execution of the intelligent early warning instruction includes: Based on the predicted distribution of hazardous gas concentrations and personnel location information, the system automatically calculates the spatiotemporal relationship between personnel and high-risk areas, generating differentiated graded warning strategies for different groups of personnel. For people about to enter predicted high-risk areas, structured warning information containing dangerous gas types, concentrations, risk levels, and optimal avoidance routes will be pushed to their wearable devices; According to the risk level and urgency, select and activate the corresponding alarm mode, which includes wearable device vibration reminder, sound and light alarm, regional broadcast or full tunnel evacuation instruction; Track and record the movement trajectory and response behavior of personnel after the warning, evaluate the warning effect, and use the warning effect as feedback information to optimize the hierarchical warning strategy.

8. The intelligent monitoring method for tunnel construction gas according to claim 1, characterized in that: The generation and execution of the ventilation system adjustment instruction includes: Based on the predicted three-dimensional gas concentration field and airflow field data, a multi-objective optimization model is constructed, which simultaneously considers the three objectives of rapid removal efficiency of harmful gases, minimization of energy consumption, and environmental comfort; Setting constraints for the multi-objective optimization model, including performance boundaries of each ventilation device, safe gas concentration thresholds for each area of ​​the tunnel space, and energy consumption limits; By solving a multi-objective optimization model, a spatiotemporal dynamic ventilation control strategy is generated. This ventilation control strategy assigns optimal operating parameters to ventilation equipment in different zones and time periods within the tunnel. The ventilation control strategy is transmitted to the controllers of each ventilation equipment via an industrial field bus or wireless communication network. The changes in the gas concentration field after ventilation adjustment are monitored in real time, the ventilation effect is evaluated, and the evaluation results are fed back to the multi-objective optimization model.

9. An intelligent gas monitoring device for tunnel construction, characterized in that: include: A collection unit, configured to collect multi-source gas data in real time via a mixed gas monitoring network deployed within the tunnel construction environment, the mixed gas monitoring network comprising fixed gas monitoring nodes, mobile gas monitoring nodes, and wearable gas monitoring nodes; A concentration field model construction unit is used to construct a three-dimensional gas concentration field dynamic model covering the entire tunnel construction area based on the multi-source gas data using data fusion and machine learning algorithms; A prediction control unit is used to predict the spatiotemporal evolution trend of gas concentration within a preset time period in the future based on the three-dimensional gas concentration field dynamic model, and to generate and execute at least one control instruction based on the prediction result, wherein the control instruction includes an intelligent early warning instruction and / or a ventilation system adjustment instruction.

10. An intelligent gas monitoring system for tunnel construction, characterized in that: include: A mixed gas monitoring network for real-time collection of multi-source gas data, comprising multiple fixed gas monitoring nodes deployed in the tunnel, at least one mobile gas monitoring node, and multiple wearable gas monitoring nodes; A server is communicatively connected to the mixed gas monitoring network, and is configured to: receive and process the multi-source gas data; construct a three-dimensional gas concentration field dynamic model covering the entire tunnel construction area based on the multi-source gas data using data fusion and machine learning algorithms; and predict the spatiotemporal evolution trend of gas concentration within a future preset time period based on the three-dimensional gas concentration field dynamic model, and generate and execute at least one control instruction based on the prediction result, the control instruction including an intelligent early warning instruction and / or a ventilation system adjustment instruction.

Images