Tunnel construction dust intelligent control method and system based on Internet of Things

Abstract

The invention discloses a tunnel construction dust intelligent control method and system based on the Internet of Things, and relates to the field of intelligent control, and the method comprises the specific steps: collecting dust parameters, environment parameters, tunnel images and working condition data of any tunnel, and carrying out the preprocessing; performing type matching on any tunnel by using a dual verification method; constructing a dynamic obstacle three-dimensional model based on the position data and the working condition data of the construction machinery; cFD simulation is carried out based on the tunnel preprocessing data of any tunnel and the dynamic obstacle three-dimensional model, a grid node information matrix and an edge weight matrix are constructed, and wind flow area division is carried out on the tunnel space of any tunnel; calculating the critical settling velocity of dust in each air flow area; and calculating the concentration over-limit probability of each node, and performing staged regulation and control based on the critical settling velocity and the concentration over-limit probability. According to the method, multiple types of data are integrated, tunnel dust pollution and environment conditions are comprehensively reflected, and reliable data support is provided for accurate regulation and control.

Description

Technical Field

[0001] This application relates to the field of intelligent control, and in particular to an intelligent control method and system for tunnel construction dust based on the Internet of Things. Background Technology

[0002] Currently, drilling and blasting operations, shield tunneling, and material transportation all generate large amounts of dust during tunnel construction. Prolonged exposure to excessive dust concentrations can easily lead to pneumoconiosis among construction workers, while also reducing visibility and increasing the risk of safety accidents. To control dust concentration, existing technologies typically rely on manual sampling or fixed-point sensor dust monitoring, combined with ventilation systems based on preset airflow and pressure for dust removal and dilution.

[0003] However, existing dust control technologies have the following limitations: First, monitoring is lagging, and traditional methods have long data update cycles, making it difficult to reflect the dynamic changes of dust in key areas such as the tunnel face and transportation channels in real time; second, control parameters are fixed, and the ventilation system is not dynamically adjusted according to the actual shape of the tunnel and construction conditions, which can easily lead to energy waste or local dust accumulation; third, the coupling effect of multiple factors is ignored, and dust concentration is used as the basis for control only, without comprehensively considering the interference of wind speed field, harmful gases and dynamic obstacles on dust diffusion, resulting in insufficient control accuracy.

[0004] Therefore, how to improve the real-time performance, adaptability, and accuracy of dust control is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides an intelligent control method and system for tunnel construction dust based on the Internet of Things, which overcomes the above-mentioned defects.

[0006] To achieve the above objectives, this application provides the following solution: Firstly, this application provides an intelligent dust control method for tunnel construction based on the Internet of Things, the specific steps of which are as follows: Collect dust parameters, environmental parameters, tunnel images, and operating condition data for any tunnel; and preprocess the dust parameters, environmental parameters, and tunnel images to generate dust preprocessing data, environmental preprocessing data, and tunnel preprocessing data. The dust pretreatment data and environmental pretreatment data are matched with a multi-dimensional tunnel shape standard database using a dual verification method to obtain the tunnel type and associated first ventilation parameter for any tunnel; a dynamic obstacle 3D model is constructed based on the location data of the construction machinery and the working condition data. CFD simulation is performed based on the tunnel preprocessing data of any tunnel and the three-dimensional model of the dynamic obstacle. The tunnel space of any tunnel is divided into grids according to the simulation results, a grid node information matrix is ​​constructed, and a nonlinear edge weight calculation function is used to generate an edge weight matrix. The tunnel space of any tunnel is divided into airflow regions based on the preset knowledge graph and the tunnel flow field map generated by the CFD simulation. Calculate the critical settling velocity of dust in each airflow region; input the grid node information matrix and the edge weight matrix into the trained spatiotemporal graph neural network model, output the dust concentration prediction value and calculate the concentration exceedance probability of each node, and perform phased regulation based on the critical settling velocity and the concentration exceedance probability.

[0007] Optionally, the preprocessing steps for the dust parameters and the environmental parameters are as follows: Missing values ​​are identified, supplemented, or labeled for the dust parameters and the environmental parameters, respectively. The dust parameters and environmental parameters that have been processed for missing values ​​are respectively subjected to outlier identification and optimization. Overlapping moving average filtering is applied to the dust parameters and environmental parameters after outlier processing to obtain dust preprocessing data and environmental preprocessing data, respectively.

[0008] Optionally, the preprocessing steps for the tunnel image are as follows: Calculate the dark channel image based on the normalized tunnel image; A preset proportion of pixels are selected in the dark channel image and positioned at the corresponding position in the tunnel image. After optimization using guided filtering, the dehazed image is reconstructed. The dehazed image is denoised using Gaussian filtering, and contour curvature features and cross-sectional size features are extracted based on the denoised dehazed image to generate tunnel preprocessing data.

[0009] Optionally, the execution steps of the dual verification method are as follows: Calculate the SSIM value of the tunnel preprocessing data and any standard tunnel type in the multi-dimensional tunnel shape standard database; The feature point matching pass rate of the tunnel preprocessing data is calculated. Determine whether the matching success rate of the SSIM value and the feature point is greater than or equal to a preset threshold. If so, the matching is considered successful.

[0010] Optionally, the expression for the nonlinear edge weight calculation function is: ; In the formula, The dust diffusion coefficient; Real-time wind speed at the node; For nodes and distance, This is the spatial attenuation coefficient; The obstacle influence coefficient; This refers to the obstacle influence factor.

[0011] Optionally, the step of obtaining the probability of exceeding the concentration limit is as follows: the spatiotemporal graph neural network model is propagated multiple times using the Monte Carlo random dropout technique to obtain multiple sets of dust concentration prediction values; and the dust concentration prediction values ​​are processed based on the concentration probability density function to obtain the dust concentration exceeding the limit probability of each node.

[0012] Optionally, the expression for the probability of dust concentration exceeding the limit is: ; In the formula, Let be the concentration probability density function. This is the safety threshold.

[0013] Optionally, the phased regulation is as follows: In the first stage of regulation, if the critical settling velocity of any of the airflow areas exceeds the preset critical settling velocity threshold, the first ventilation parameter is adjusted in a gradient until the preset standard value is reached. In the second stage of regulation, if the dust concentration exceeds the preset exceedance probability threshold, the first ventilation parameter is adjusted in stages according to the risk level. In the third stage of regulation, when the rate of change of dust concentration is continuously lower than the stable threshold for a fixed period of time, the adjusted ventilation parameters are gradually reduced using an energy-saving gradient until the dust concentration approaches the safe threshold.

[0014] Secondly, this application provides an intelligent control system for tunnel construction dust based on the Internet of Things, including: The multi-source data acquisition module is used to collect dust parameters, environmental parameters, tunnel images, and operating condition data for any tunnel. The analysis and dynamic modeling module is used to match the dust pre-processing data, the environmental pre-processing data, and the multi-dimensional tunnel shape standard database using a dual verification method to obtain the tunnel type and associated first ventilation parameter of any tunnel; construct a dynamic obstacle 3D model based on the location data of construction machinery and the working condition data; perform CFD simulation based on the tunnel pre-processing data of any tunnel and the dynamic obstacle 3D model, divide the tunnel space of any tunnel into grids according to the simulation results, construct a grid node information matrix, and generate an edge weight matrix using a nonlinear edge weight calculation function; and divide the tunnel space of any tunnel into airflow regions based on a preset knowledge graph and the tunnel flow field map generated by CFD simulation. The dynamic threshold and intelligent control module is used to calculate the critical settling velocity of dust in each airflow area; it inputs the grid node information matrix and edge weight matrix into the trained spatiotemporal graph neural network model, outputs the dust concentration prediction value and calculates the concentration exceedance probability of each node, and performs phased control based on the critical settling velocity and the concentration exceedance probability.

[0015] Optionally, it also includes a feedback optimization module, used to acquire feedback data after regulation, evaluate the regulation effect based on the feedback data, and optimize the phased regulation based on the evaluation results.

[0016] According to the specific embodiments provided in this application, this application has the following technical effects: 1. Integrating four types of data—images, environment, operating conditions, and equipment—covering multiple dimensions such as dust concentration, particle size distribution, temperature and humidity, and airflow status, it comprehensively reflects tunnel dust pollution and environmental conditions, providing reliable data support for precise control.

[0017] 2. The three-tiered control strategy (pre-control + real-time control + predictive control) achieves full-cycle management of "early prevention - rapid response - early intervention," solving the problem of lagging technology. When dust concentration rises rapidly after blasting, pre-control is initiated in advance, real-time control provides a rapid response, and predictive control avoids the risk of future exceedances, ensuring that dust concentration remains within a safe range.

[0018] 3. It is compatible with various tunnel types such as straight, spiral, and bifurcated tunnels, as well as various construction stages such as blasting, tunneling, support, and cleaning. It can be widely used in construction scenarios such as highway tunnels, railway tunnels, and mine roadways, and has strong practicality and promotional value. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the overall method flow provided in one embodiment of this application; Figure 2 This is a schematic diagram of the system structure provided in an embodiment of this application. Detailed Implementation

[0021] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0023] This embodiment discloses an intelligent dust control method for tunnel construction based on the Internet of Things, such as... Figure 1 As shown, the specific steps are as follows: Step 1: Collect dust parameters, environmental parameters, tunnel images, and operating condition data for any tunnel; and preprocess the dust parameters, environmental parameters, and tunnel images to generate dust preprocessing data, environmental preprocessing data, and tunnel preprocessing data. Step 2: Using a dual verification method, the dust pretreatment data, the environmental pretreatment data, and the multi-dimensional tunnel shape standard database are matched to obtain the tunnel type and associated first ventilation parameter for any tunnel; a dynamic obstacle 3D model is constructed based on the location data of the construction machinery and the working condition data. Step 3: Perform CFD simulation based on the tunnel preprocessing data of any tunnel and the dynamic obstacle 3D model. Based on the simulation results, divide the tunnel space of any tunnel into grids, construct a grid node information matrix, and generate an edge weight matrix using a nonlinear edge weight calculation function. Divide the tunnel space of any tunnel into airflow regions based on the preset knowledge graph and the tunnel flow field map generated by the CFD simulation. Step 4: Calculate the critical settling velocity of dust in each airflow region; input the grid node information matrix and edge weight matrix into the trained spatiotemporal graph neural network model, output the dust concentration prediction value and calculate the concentration exceedance probability of each node, and perform phased regulation based on the critical settling velocity and the concentration exceedance probability.

[0024] Furthermore, this embodiment establishes an Internet of Things (IoT) monitoring network to collect real-time dust parameters (including dust concentration, particle diameter distribution, etc.) and environmental parameters (including wind speed, wind direction, CO volume concentration, NO) during tunnel construction. xThe system uses data such as dust concentration, temperature, and humidity, tunnel images, and operating conditions. It combines tunnel shape recognition and dynamic obstacle modeling to construct a dust diffusion simulation model. Using Stokes' theorem and spatiotemporal neural network, it calculates dust settling velocity and predicts the probability of dust concentration exceeding limits. Finally, through a phased control strategy, it dynamically optimizes ventilation parameters to ensure that dust concentration remains stable within a safe threshold while reducing energy consumption.

[0025] Furthermore, the data acquisition devices are specifically deployed as follows: around the tunnel face (0-50m range), every 30m along the transport corridor, and at the exit of the return airway, laser scattering dust sensors (monitoring dust concentration and particle diameter distribution), ultrasonic anemometers (monitoring wind speed and direction), and gas sensors (monitoring CO volume concentration and NO) are respectively installed. x The system includes sensors for dust concentration and temperature / humidity; a high-definition industrial camera (resolution ≥ 2K, frame rate ≥ 15fps) is installed at the top of the tunnel face to acquire real-time images of the tunnel face (i.e., tunnel images), ensuring a clear and unobstructed tunnel face outline; GPS positioning modules and operating condition sensors (monitoring engine speed and operating status) are installed on construction machinery (such as excavators and loaders) and transport vehicles; all equipment is connected to the IoT platform via a LoRaWAN / 5G industrial gateway to record real-time construction condition data (such as blasting charge, tunneling machine working time, and equipment operation data). The monitoring frequency for each parameter is as follows: dust parameters 10s / time, environmental parameters 10s / time, image data 1min / frame, positioning and operating condition data 5s / time.

[0026] In one embodiment, the raw sensor data is subjected to outlier removal (using the 3σ criterion) and smoothing (using moving average filtering with a window length of 5); the face image is subjected to dehazing and noise reduction, and the shape features of the image (such as the face contour curvature, cross-sectional dimensions, etc.) are extracted.

[0027] In one embodiment, the preprocessing steps for the dust parameters and the environmental parameters are as follows: Missing values ​​are identified, supplemented, or labeled for the dust parameters and the environmental parameters, respectively. The dust parameters and environmental parameters that have been processed for missing values ​​are respectively subjected to outlier identification and optimization. Overlapping moving average filtering is applied to the dust parameters and environmental parameters after outlier processing to obtain dust preprocessing data and environmental preprocessing data, respectively.

[0028] Furthermore, the steps for processing the sensor's raw data are as follows: The system iterates through time-series data such as dust concentration, temperature, and humidity, identifying missing values, invalid identifiers, and other missing data. Single missing values ​​are filled using linear interpolation (based on the mean of the three preceding and following valid data points). If more than five consecutive missing values ​​are found, the time period is marked and a sensor self-check is triggered. Assuming the data sequence approximates a normal distribution, the mean μ and standard deviation σ are first calculated. If the data points satisfy |x i -μ|>3σ was identified as an outlier and replaced with a weighted average of the two valid data points before and after it (weight coefficients 0.1, 0.4, 0.4, 0.1) to ensure temporal continuity. Then, an overlapped moving average filter with a window length of 5 was used, selecting two data points before and after the current data point to form a window, and calculating the arithmetic mean as the output. The first and last two data points of the sequence were padded with a half-window to avoid distortion, filter out random noise, and preserve real-time performance.

[0029] In one embodiment, the preprocessing steps for the tunnel image are as follows: Calculate the dark channel image based on the normalized tunnel image; A preset proportion of pixels are selected in the dark channel image and positioned at the corresponding position in the tunnel image. After optimization using guided filtering, the dehazed image is reconstructed. The dehazed image is denoised using Gaussian filtering, and contour curvature features and cross-sectional size features are extracted based on the denoised dehazed image to generate tunnel preprocessing data.

[0030] Further, the processing steps for the tunnel face image are as follows: the tunnel RGB image (i.e., the tunnel image) is normalized to the [0, 1] interval, and the dark channel image is calculated using a 15×15 window; the 0.1% pixel with the highest brightness in the dark channel image is selected to correspond to the original image position, and after guided filtering optimization, the dehazed image is reconstructed based on the atmospheric scattering model, and finally the pixel values ​​are corrected to the 0-255 interval; then, a 5×5 Gaussian filter is used to check the noise reduction of the dehazed grayscale image to ensure that the tunnel face contour is clear, and a binarized edge image is obtained through Canny edge detection (low threshold 50, high threshold 150). Based on connected component analysis, the main contour that conforms to the geometric features of the tunnel face is selected, and the pixel coordinates are converted into physical coordinates. Then, the contour coordinate sequence is smoothed, the curvature value of each contour point is solved, and the mean, maximum, minimum and variance of curvature are extracted as features (i.e., contour curvature features). Finally, using the center vertical line of the smallest circumscribed rectangle of the face as the reference axis, five characteristic sections are selected along the axis at 1m intervals. The width and maximum longitudinal height of each section are measured, and the average width, extreme value, and longitudinal height are calculated as dimensional features (i.e., cross-sectional dimensional features).

[0031] In one embodiment, the execution steps of the dual verification method are as follows: Calculate the SSIM value of the tunnel preprocessing data and any standard tunnel type in the multi-dimensional tunnel shape standard database; The feature point matching pass rate of the tunnel preprocessing data is calculated. Determine whether the matching success rate of the SSIM value and the feature point is greater than or equal to a preset threshold. If so, the matching is considered successful.

[0032] Furthermore, a multi-dimensional tunnel shape standard database is used to store core data for typical types such as straight lines, spiral lines, and bifurcated tunnels, including high-definition standard images of the tunnel face and corresponding ventilation parameters (air volume, air pressure, ventilation duct layout and installation angle), ensuring comprehensive matching dimensions.

[0033] Furthermore, a dual verification scheme of "Structural Similarity (SSIM) + Feature Point Matching Pass Rate" is adopted for matching. Specifically, for the preprocessed face image, the SIFT algorithm is used to extract key features: the scale space extrema are detected by the difference of Gaussian pyramid, and the position and orientation of the feature points are determined after thresholding and precise localization; a feature vector set of the image to be identified is constructed based on the gradient histogram of the neighboring pixels of the feature points. The ratio of the number of valid matching point pairs to the total number of feature points in the image to be identified (match pass rate) is calculated, and a threshold of ≥75% is set to confirm the validity of the match. When both verifications meet the threshold requirements, the match is considered successful, and the corresponding standard tunnel type and associated initial ventilation parameters (i.e., the first ventilation parameter) are retrieved from the multi-dimensional tunnel shape standard database. If the requirements are not met, a full database search is triggered for secondary verification. If a match is still not found, a "pending manual assistance" signal is output to ensure the reliability and engineering practicality of the matching results.

[0034] In one embodiment, a dynamic obstacle 3D model is constructed in the BIM platform based on the GPS positioning data of the construction machinery (i.e., the location data of the construction machinery) and the working condition sensor data. The model includes the equipment's external dimensions, real-time location, and working status (such as tunneling and muck removal). The dynamic obstacle 3D model is input as a boundary condition into the computational fluid dynamics (CFD) simulation system. Combined with the tunnel's 3D geometric parameters (i.e., contour curvature characteristics and cross-sectional size characteristics) (length, cross-sectional area, radius of curvature), the simulated boundaries of airflow and dust diffusion are corrected.

[0035] In one embodiment, a critical area for dust diffusion (20m around the working face and the intersection of the transport channel) is identified, and a high-resolution grid (grid size 0.5m×0.5m×0.5m) is set; a low-resolution grid (grid size 2m×2m×2m) is set in non-critical areas (such as the middle of the straight section); a grid node information matrix is ​​constructed, which includes the location coordinates (x, y, z), initial dust concentration, wind speed, and temperature of each node.

[0036] Furthermore, a nonlinear edge weight calculation function is constructed to characterize the dust interaction intensity between grid nodes: ; In the formula, This is the dust diffusion coefficient (calculated based on the diameter of dust particles, such as 0.15 m² / s for PM10). Real-time wind speed at the node; For nodes and distance, The spatial attenuation coefficient is taken as 5m. The obstacle influence coefficient is taken as 1.2. The obstacle influence factor is calculated based on the spatial overlap rate between node connections and obstacles, ranging from [0, 1]; the edge weight matrix is ​​generated based on the above function (N is the total number of grid nodes).

[0037] In one embodiment, a CFD simulation is run to obtain a tunnel flow field map; the airflow region is divided according to a preset knowledge graph, and the steps for building the preset knowledge graph are as follows: The project collects engineering parameters (cross-sectional dimensions, length, slope, support type, etc.) for typical types (straight lines, spiral lines, bifurcated tunnels), as well as equipment and operating parameters (ventilation duct diameter, installation height, design wind pressure, etc.), historical CFD simulation data (wind speed, wind direction, vortex distribution) under different working conditions, field measured flow field data, and corresponding area labeling results. The parameter units are standardized, and outliers are removed using the Grubbs criterion to form a structured dataset.

[0038] The threshold values ​​for the three types of regions are stored as basic rules. A mapping relationship of "tunnel type - ventilation parameters - flow field index" is constructed based on CFD simulation and measured data, and "division index" and "flow field region" are associated; the problem of overlapping divisions under special working conditions is solved.

[0039] The initial graph is stored using the Neo4j graph database, and real-time CFD simulation and experimental data are integrated. Entity attributes and relationships are updated through incremental learning. The division accuracy is checked regularly, and when it is lower than 92%, the rule base is optimized in reverse (such as adjusting the distance threshold of the backflow area) to ensure that the graph is adapted to the actual engineering situation.

[0040] In one embodiment, the airflow zone is specifically divided as follows: Jet zone: The angle between the airflow direction and the axis of the air duct is ≤30°, and the wind speed is ≥70% of the initial wind speed; Vortex region: The airflow direction is in a rotating state (vortex quantity ≥ 5s) -1 ); Return flow zone: ≥30m from the working face, and the wind direction is opposite to that of the jet flow zone (wind speed ≤-1m / s).

[0041] In one embodiment, the step of obtaining the dust settling velocity is as follows: Retrieve the Stokes formula and input the dust parameters (dust density). air density Aerodynamic viscosity Dust particle diameter ): ; In the formula, The acceleration due to gravity is 9.8 m / s². The dust settling velocity in each airflow region is sorted in descending order, and the maximum value is taken as the critical settling velocity for that region. ).

[0042] In one embodiment, the step of obtaining the probability of dust concentration exceeding the limit is as follows: the spatiotemporal graph neural network model is propagated multiple times using the Monte Carlo random dropout technique to obtain multiple sets of dust concentration prediction values; and the dust concentration prediction values ​​are processed based on the concentration probability density function to obtain the dust concentration exceeding the limit probability of each node.

[0043] Furthermore, the grid node information matrix and edge weight matrix are input into the optimized spatiotemporal graph neural network (ST-GNN) model: A tunnel topology adaptive preprocessing module is added to the input layer. This module embeds curvature correction factors and branch flow distribution coefficients by combining the physical coordinates of grid nodes with the tunnel type (straight / spiral / bifurcation tunnel). It integrates historical data from the past 5 minutes (dust concentration, wind speed, temperature) with CFD flow field topology information to construct a "spatiotemporal-physical" dual-dimensional input matrix, addressing the problem of insufficient adaptation to complex tunnel structures. The core layer uses multi-scale spatiotemporal separation convolutions instead of ordinary convolutions and GRUs: the spatial branch dynamically generates adaptive topology convolution kernels (1×3 convolution for straight tunnels, 3×3 dynamic convolution for spiral / bifurcation tunnels), and the number of convolutional layers is adjusted according to complexity (2 layers for straight tunnels, 3 layers for spiral tunnels, 4 layers for bifurcation tunnels). The number of neurons is optimized to a gradient structure of 80-128-128-80. The temporal branch introduces dilation rates of 2, 4, and 8 with dilated TCNs to enhance the capture of dependencies at different time scales. An innovative two-level attention mechanism replaces the fixed-weight module: the first-level modal attention dynamically learns feature weights through a fully connected layer (dust concentration weights can be adaptively adjusted from 0.5 to 0.8), while the second-level spatial attention combines the flow field weight matrix to strengthen the correlation of airflow influence. A multi-path residual connection (short-circuit + cross-layer fusion) is designed to improve the efficiency of deep feature transfer. The output layer adds a concentration spatiotemporal change rate feature, forming a two-dimensional output with the concentration mean. The Monte Carlo random dropout technique employs a layered sampling strategy (150 samplings in the eddy / recirculation zone and 100 samplings in the ordinary region), and calibrates the probability density function using the Platt scaling method to improve the accuracy of exceeding-limit probability calculations. This optimization improves the model's prediction accuracy by more than 15% under complex tunnel conditions, increases the advance warning time for exceeding limits by 2-3 minutes, and significantly enhances its adaptability and engineering practicality.

[0044] The specific implementation steps are as follows: Input historical monitoring data (dust concentration, wind speed, temperature) for the past 5 minutes; output the spatiotemporal dynamic characteristics of dust concentration for each grid node within the next 10 minutes; use Monte Carlo random dropout technique to perform 100 forward propagations for each prediction time step to obtain multiple sets of dust concentration prediction values; use kernel density estimation method to construct a concentration probability density function, integrate it over the area above the safety threshold (e.g., PM10 ≤ 2 mg / m³), and obtain the probability of dust concentration exceeding the limit for each node. : ; In the formula, Let be the concentration probability density function. This is the safety threshold.

[0045] In one embodiment, the phased regulation specifically involves: In the first stage of regulation, if the critical settling velocity of any of the airflow areas exceeds the preset critical settling velocity threshold, the first ventilation parameter is adjusted in a gradient until the preset standard value is reached. In the second stage of regulation, if the dust concentration exceeds the preset exceedance probability threshold, the first ventilation parameter is adjusted in stages according to the risk level. In the third stage of regulation, when the dust concentration change rate remains below a stable threshold for a fixed period of time, the adjusted ventilation parameters are gradually reduced using an energy-saving gradient until the dust concentration approaches a safe threshold. Furthermore, the first stage of regulation (settlement velocity driven): Set a critical settlement velocity threshold (e.g., set the threshold to 0.05 m / s): If a certain airflow area The first control signal is issued; the gradient is changed by 50m with the air volume. 3 The airflow rate is adjusted in increments of 50 Pa, continuously adjusting the first ventilation parameter until all areas are covered, thus obtaining the second ventilation parameter. Simultaneously, gas parameters are monitored; if CO concentration > 30 ppm or O2 concentration < 19.5%, the airflow rate is further increased (in increments of 30 Pa). 3 ( / min) until the gas parameters meet the standard requirements.

[0046] Second-stage regulation (driven by excessive probability): Set an over-limit probability threshold (e.g., 5%): If a certain grid node Triggering tiered regulation: a. Low risk ( Adjust the angle of the ventilation duct (±5° each time) to enhance local ventilation; b. Medium risk ( Increase the fan speed (5% of rated speed each time) to increase the total air volume; c. High risk ( ): Start the standby fan and switch to the "main fan + standby fan" combined ventilation mode.

[0047] Third-stage regulation (energy-saving optimization): Monitor the rate of change of dust concentration and gas parameters after adjustment at 10-minute intervals: ; when When the dust concentration reaches a stable state after three consecutive intervals, the time point is recorded as the transition time point. After the transition time point, the ventilation parameters are continuously reduced in a gradient of 50 m³ / min air volume and 30 Pa air pressure until the dust concentration approaches the safe threshold (2 mg / m³). The ventilation parameters at this time are the optimal energy-saving parameters, and constant ventilation is carried out using these parameters.

[0048] In one embodiment, the IoT platform compiles and archives historical monitoring data (dust concentration, ventilation parameters, energy consumption) monthly and caches local control records. It also optimizes the weight parameters of the spatiotemporal graph neural network using the gradient descent method. Based on the tunnel shape and working condition data of the new construction section, it updates the tunnel shape database and the three-dimensional model parameters of dynamic obstacles to ensure the adaptability of the control strategy.

[0049] This embodiment also discloses an intelligent control system for tunnel construction dust based on the Internet of Things, such as... Figure 2 As shown, it includes: The multi-source data acquisition module is used to collect dust parameters, environmental parameters, tunnel images, and operating condition data for any tunnel. The analysis and dynamic modeling module is used to match the dust pre-processing data, the environmental pre-processing data, and the multi-dimensional tunnel shape standard database using a dual verification method to obtain the tunnel type and associated first ventilation parameter of any tunnel; construct a dynamic obstacle 3D model based on the location data of construction machinery and the working condition data; perform CFD simulation based on the tunnel pre-processing data of any tunnel and the dynamic obstacle 3D model, divide the tunnel space of any tunnel into grids according to the simulation results, construct a grid node information matrix, and generate an edge weight matrix using a nonlinear edge weight calculation function; and divide the tunnel space of any tunnel into airflow regions based on a preset knowledge graph and the tunnel flow field map generated by CFD simulation. The dynamic threshold and intelligent control module is used to calculate the critical settling velocity of dust in each airflow area; it inputs the grid node information matrix and edge weight matrix into the trained spatiotemporal graph neural network model, outputs the dust concentration prediction value and calculates the concentration exceedance probability of each node, and performs phased control based on the critical settling velocity and the concentration exceedance probability.

[0050] In one embodiment, a feedback optimization module is further included, which is used to acquire feedback data after regulation, evaluate the regulation effect based on the feedback data, and optimize the phased regulation based on the evaluation results.

[0051] Furthermore, the system in this embodiment is used to execute the above method, and the functions and architecture of each module are as follows: Multi-source data acquisition module: Composed of a high-definition camera, a multi-parameter sensor array (dust concentration sensor, particle size analyzer, temperature and humidity sensor, airflow sensor, gas sensor), construction equipment sensors, and equipment operation sensors, supporting multiple communication methods such as LoRa / WiFi / 4G / 5G; responsible for collecting tunnel face images, environmental data, construction condition data, and equipment operation data, converting the data into digital signals and transmitting them to edge computing nodes; adaptable to various complex tunnel environments.

[0052] The analysis and dynamic modeling module can identify tunnel shape, construction stage, and face advancement status based on collected data, analyze dust physical and diffusion characteristics, establish a dust diffusion model, and perform rapid inference of deep learning models.

[0053] The dynamic threshold and intelligent control module communicates in real time with the analysis and dynamic modeling module to acquire operating condition and dust characteristic data, and provides dynamic thresholds to the intelligent control module. It can store basic safety thresholds and calculate correction coefficients based on operating conditions, dust characteristics, and area characteristics, outputting dynamic thresholds and updating them in real time. This enables the execution of a three-level control strategy, dynamically adjusting ventilation, spray, and airflow parameters to ensure that the control parameters meet equipment safety and energy-saving constraints; it also directly controls the operation of fans, spray devices, and airflow devices.

[0054] The feedback optimization module consists of a fan control unit, a spray control unit, a flow guiding control unit, a linkage logic unit, a feedback data acquisition unit, an effect evaluation unit, and a reinforcement learning optimization unit. It collects feedback data after regulation, evaluates the regulation effect and energy consumption, and continuously optimizes the regulation parameters through a reinforcement learning model to achieve coordinated linkage of the fan, spray, and flow guiding equipment. Based on the coordinated logic, it synchronously adjusts the parameters of each device to improve control efficiency.

[0055] It also includes an edge computing module and a cloud management platform, consisting of an edge server, a local database, an audible and visual alarm, a communication module, a model training server, a monitoring interface, and a report generation unit; it can store the local optimization dataset for the most recent month, and can process data, make local decisions, and execute control commands in real time in a multi-threaded parallel manner, ensuring response speed and achieving long-term optimization and global control.

[0056] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0057] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A smart dust control method for tunnel construction based on the Internet of Things, characterized in that, The specific steps are as follows: Collect dust parameters, environmental parameters, tunnel images, and operating condition data for any tunnel; and preprocess the dust parameters, environmental parameters, and tunnel images to generate dust preprocessing data, environmental preprocessing data, and tunnel preprocessing data. The dust pretreatment data and environmental pretreatment data are matched with a multi-dimensional tunnel shape standard database using a dual verification method to obtain the tunnel type and associated first ventilation parameter for any tunnel; a dynamic obstacle 3D model is constructed based on the location data of the construction machinery and the working condition data. CFD simulation is performed based on the tunnel preprocessing data of any tunnel and the three-dimensional model of the dynamic obstacle. The tunnel space of any tunnel is divided into grids according to the simulation results, a grid node information matrix is ​​constructed, and a nonlinear edge weight calculation function is used to generate an edge weight matrix. Based on the tunnel flow field map generated by the preset knowledge graph and CFD simulation, the tunnel space of any tunnel is divided into airflow regions. Calculate the critical settling velocity of dust in each airflow zone; The grid node information matrix and the edge weight matrix are input into the trained spatiotemporal graph neural network model, the dust concentration prediction value is output, and the concentration exceedance probability of each node is calculated. Based on the critical settling velocity and the concentration exceedance probability, phased regulation is performed.

2. The intelligent dust control method for tunnel construction based on the Internet of Things according to claim 1, characterized in that, The preprocessing steps for the dust parameters and the environmental parameters are as follows: Missing values ​​are identified, supplemented, or labeled for the dust parameters and the environmental parameters, respectively. The dust parameters and environmental parameters that have been processed for missing values ​​are respectively subjected to outlier identification and optimization. Overlapping moving average filtering is applied to the dust parameters and environmental parameters after outlier processing to obtain dust preprocessing data and environmental preprocessing data, respectively.

3. The intelligent dust control method for tunnel construction based on the Internet of Things according to claim 1, characterized in that, The preprocessing steps for the tunnel image are as follows: Calculate the dark channel image based on the normalized tunnel image; A preset proportion of pixels are selected in the dark channel image and positioned at the corresponding position in the tunnel image. After optimization using guided filtering, the dehazed image is reconstructed. The dehazed image is denoised using Gaussian filtering, and contour curvature features and cross-sectional size features are extracted based on the denoised dehazed image to generate tunnel preprocessing data.

4. The intelligent dust control method for tunnel construction based on the Internet of Things according to claim 1, characterized in that, The execution steps of the dual verification method are as follows: Calculate the SSIM value of the tunnel preprocessing data and any standard tunnel type in the multi-dimensional tunnel shape standard database; The feature point matching pass rate of the tunnel preprocessing data is calculated. Determine whether the matching success rate of the SSIM value and the feature point is greater than or equal to a preset threshold. If so, the matching is considered successful.

5. The intelligent dust control method for tunnel construction based on the Internet of Things according to claim 1, characterized in that, The expression for the nonlinear edge weight calculation function is: ； In the formula, The dust diffusion coefficient; Real-time wind speed at the node; For nodes and distance, This is the spatial attenuation coefficient; The obstacle influence coefficient; This refers to the obstacle influence factor.

6. The intelligent dust control method for tunnel construction based on the Internet of Things according to claim 1, characterized in that, The steps for obtaining the probability of dust concentration exceeding the limit are as follows: Monte Carlo random dropout technique is used to perform multiple forward propagations on the spatiotemporal graph neural network model to obtain multiple sets of dust concentration prediction values; and the dust concentration prediction values ​​are processed based on the concentration probability density function to obtain the dust concentration exceeding the limit probability of each node.

7. The intelligent dust control method for tunnel construction based on the Internet of Things according to claim 6, characterized in that, The expression for the probability of dust concentration exceeding the limit is: ； In the formula, Let be the concentration probability density function. This is the safety threshold.

8. The intelligent dust control method for tunnel construction based on the Internet of Things according to claim 1, characterized in that, The phased regulation is specifically as follows: In the first stage of regulation, if the critical settling velocity of any of the airflow areas exceeds the preset critical settling velocity threshold, the first ventilation parameter is adjusted in a gradient until the preset standard value is reached. In the second stage of regulation, if the dust concentration exceeds the preset exceedance probability threshold, the first ventilation parameter is adjusted in stages according to the risk level. In the third stage of regulation, when the rate of change of dust concentration is continuously lower than the stable threshold for a fixed period of time, the adjusted ventilation parameters are gradually reduced using an energy-saving gradient until the dust concentration approaches the safe threshold.

9. An intelligent control system for dust in tunnel construction based on the Internet of Things, characterized in that, include: The multi-source data acquisition module is used to collect dust parameters, environmental parameters, tunnel images, and operating condition data for any tunnel. The analysis and dynamic modeling module is used to match the dust pre-processing data, the environmental pre-processing data, and the multi-dimensional tunnel shape standard database using a dual verification method to obtain the tunnel type and associated first ventilation parameter of any tunnel; and to construct a dynamic obstacle 3D model based on the location data of the construction machinery and the working condition data. CFD simulation is performed based on the tunnel preprocessing data of any tunnel and the three-dimensional model of the dynamic obstacle. The tunnel space of any tunnel is divided into grids according to the simulation results, a grid node information matrix is ​​constructed, and a nonlinear edge weight calculation function is used to generate an edge weight matrix. Based on the tunnel flow field map generated by the preset knowledge graph and CFD simulation, the tunnel space of any tunnel is divided into airflow regions. The dynamic threshold and intelligent control module is used to calculate the critical settling velocity of dust in each airflow area; The grid node information matrix and edge weight matrix are input into the trained spatiotemporal graph neural network model, which outputs the dust concentration prediction value and calculates the concentration exceedance probability of each node. Based on the critical settling velocity and the concentration exceedance probability, phased regulation is carried out.

10. The intelligent control system for tunnel construction dust based on the Internet of Things according to claim 9, characterized in that, It also includes a feedback optimization module, which is used to acquire feedback data after regulation, evaluate the regulation effect based on the feedback data, and optimize the phased regulation based on the evaluation results.

Images