Health limit-based foundation pit gas film tail gas dust linkage control method and system

Abstract

The present application relates to a health limit-based foundation pit gas film tail gas dust linkage control method and system, belonging to the field of foundation pit construction environment control and occupational health protection technology. It includes: obtaining the concentration-thermal plume coupling characteristics and spatial distribution characteristics of pollutants, outputting the pollution event type, pollution source target area and confidence level through the pollution event identification model; constructing a dynamic risk assessment vector and comparing it with a health double threshold system in multiple dimensions to determine the governance urgency and governance level; generating a linkage control strategy through multi-objective optimization to dynamically determine the start timing, spatial direction and power allocation of the governance action; collecting concentration response data after governance execution, and revising the power allocation parameters when the response characteristics deviate from the expected. The present application realizes the collaborative governance and closed-loop adaptive control of the tail gas and dust double pollution sources in the foundation pit gas film through multi-physical field coupling sensing and model-driven identification, taking the legal occupational health exposure limit as the core control basis.

Description

Technical Field

[0001] This invention belongs to the field of environmental control and occupational health protection technology for foundation pit construction, specifically relating to a method and system for joint control of exhaust gas dust from foundation pit air film based on health limits. Background Technology

[0002] Air-supported membrane structures for foundation pits are a green construction technology that uses the pressure difference between the inside and outside of the pit to provide structural support. By leveraging this pressure difference, it creates large-span, fully open, and barrier-free construction spaces without the need for any beams or columns. In recent years, air-supported membrane technology for foundation pits has been widely used in projects such as subways, utility tunnels, and deep foundation pits. By constructing an inflatable membrane structure above the foundation pit, a relatively enclosed construction work space is formed, effectively suppressing dust emissions, reducing construction noise, and improving working conditions under adverse weather conditions.

[0003] In terms of environmental governance, existing technologies mainly focus on dust control within the air-supported membrane structure. Representative solutions include circulating guide pipes combined with nozzles to create directional airflow for dust collection, top-mounted pump-assisted filtration devices to absorb rising dust, cyclone filters combined with water mist dust suppression for dust filtration, and multiple sets of dust suppression components arranged along the length of the foundation pit for spray dust suppression. Regarding intelligent monitoring, existing patents involve visual monitoring systems for foundation pit air-supported membrane structures and intelligent adaptive monitoring and automatic repair systems. By deploying a multi-node sensor network, integrating internal pressure control, air supply monitoring, and environmental monitoring modules, risk warning and remote visual monitoring are achieved. In terms of structural optimization, existing research mainly focuses on improving the ventilation holes at the top of the air-supported membrane structure, optimizing the constant pressure dust removal structure, and physical isolation and wet dust suppression solutions such as the combination of dustproof canopies and atomizing nozzles.

[0004] However, existing air-supported membrane environmental remediation technologies for foundation pits still have three core limitations: First, the pollution control targets are singular. Existing technologies primarily focus on dust control, neglecting the accumulation of exhaust gases (containing carbon monoxide, nitrogen oxides, diesel particulates, and polycyclic aromatic hydrocarbons) generated by diesel-powered construction machinery operating continuously in enclosed spaces. While existing control systems within the air-supported membrane can monitor dust content and harmful gas concentrations and continuously supply fresh air via fans, a synergistic control system addressing the combined pollution of exhaust gases and dust has not yet been established. The combined pollution from exhaust gases and dust in enclosed spaces exponentially increases the health hazards to workers.

[0005] Secondly, the control logic lacks a compliance basis. The start and stop thresholds of existing intelligent monitoring systems are mostly based on experience or are set manually, without being directly linked to statutory occupational health standards. This results in a lack of clear compliance drivers for governance measures, failing to ensure that workers are always in a healthy and safe working environment.

[0006] Third, the methods of dust control are extensive. Most existing dust suppression devices are fixed and spray all over the area, lacking intelligent control capabilities. They cannot effectively adapt to changes in construction conditions and cannot achieve precise tracking and control based on the location of mobile pollution sources.

[0007] Therefore, there is an urgent need for a method and system for the coordinated control of air-supported membrane structures in foundation pits that can use statutory occupational health standards as the core control basis, achieve accurate identification and coordinated treatment of exhaust gas and dust, and have intelligent adaptive regulation capabilities. Summary of the Invention

[0008] To address the aforementioned problems in the existing technology, this invention provides a method and system for coordinated control of dust generation from exhaust gas from foundation pits based on health limits. The objective of this invention can be achieved through the following technical solutions: The method for controlling dust emissions from excavated air-supported membrane structures in foundation pits based on health limits includes the following steps: The concentration-thermal plume coupling characteristics and spatial distribution characteristics of pollutants are obtained, input into a pre-trained pollution event identification model, and the pollution event type, pollution source target area and its confidence level are output. A dynamic risk assessment vector is constructed by combining the real-time concentration of pollutants. The dynamic risk assessment vector is compared with the preset health dual threshold system in multiple dimensions to determine the urgency and level of treatment; a linkage control strategy is generated through multi-objective optimization, with the weighted minimization of exhaust gas treatment energy consumption and dust treatment energy consumption as the objective and the compliance of pollutant concentration as the constraint, to dynamically determine the start-up sequence, spatial orientation and power allocation of exhaust gas and dust treatment actions. The system issues control commands corresponding to the linkage control strategy, collects concentration response data after the treatment is executed, identifies the response characteristics of the concentration decay curve, and when the response characteristics deviate from the expectation, it inversely corrects the power allocation parameters of the linkage control strategy.

[0009] Specifically, the extraction method of the concentration-thermal plume coupling feature is as follows: taking the spatial distribution of thermal plume at the exhaust port of engineering machinery captured by the thermal imaging node as a reference, the concentration gradient tensor of exhaust pollutants and the thermal plume vector field are spatiotemporally aligned to generate a temperature-weighted joint feature vector of concentration-thermal plume.

[0010] Specifically, the method for extracting the spatial distribution characteristics of pollutants is as follows: calculate the angle between the anisotropic diffusion principal axis of the pollutant concentration field and the tangent direction of the nearest mechanical trajectory in the positioning field of the engineering machinery; when the angle is in a non-orthogonal interval, the machinery is determined to be a suspected pollution source; and combine the concentration gradient consistency verification of multiple sensor nodes to determine the target area of ​​the pollution source.

[0011] Specifically, the pollution event identification model uses historical concentration-thermal plume coupling feature vectors and pollutant spatial distribution features as training samples, and pollution event types and pollution source target areas as labels, and is trained through supervised learning; the confidence level is the posterior probability output by the model, used to characterize the reliability of the judgment result.

[0012] Specifically, the dynamic risk assessment vector is constructed by integrating real-time pollutant concentration, concentration change trend, spatial diffusion trend and confidence level, and generating a multi-dimensional risk assessment vector according to preset weights. The multi-dimensional comparison includes: the deviation between the concentration of each pollutant and the warning threshold, the similarity between the concentration change trend and the historical typical pollution pattern, and the coupling degree between the spatial diffusion trend and the airflow field inside the air film.

[0013] Specifically, the urgency of governance is determined based on the weighted fusion result of risk values ​​in each dimension of the dynamic risk assessment vector; the governance level includes pre-start level, preventive governance level and mandatory governance level, which correspond to different start-up sequences and power combinations of exhaust gas and dust governance actions in the linkage control strategy.

[0014] Specifically, the objective function of the multi-objective optimization also includes: the convergence time for pollutant concentrations to drop below the warning threshold; the constraints also include that the concentrations of each pollutant do not exceed the action threshold, that the spatial orientation of the treatment devices does not conflict, and that priority is given to the area where the operators are located.

[0015] Specifically, the dynamic determination of the activation sequence of exhaust gas and dust control actions is carried out as follows: the pollution event types include exhaust gas-dominated, dust-dominated, and coupled superimposed types; when it is determined to be a coupled superimposed type and the confidence level is higher than a preset threshold, exhaust gas control and dust control are activated simultaneously; when it is determined to be an exhaust gas-dominated or dust-dominated type and the confidence level is lower than a preset threshold, the dominant type of control is activated first, the auxiliary type of control is activated after a preset delay, and the delay time is dynamically adjusted according to the concentration response data.

[0016] Specifically, the concentration decay curve response characteristics include: decay onset delay, decay slope, overshoot, and steady-state residual concentration; the inversion correction method is as follows: taking the deviation between the actual response characteristics and the expected decay model as input, the power allocation parameters are iteratively optimized through gradient descent until the actual response characteristics fall within the confidence interval of the expected decay model; wherein, the expected decay model is pre-calibrated according to the pollution event type, treatment level, and power allocation parameters.

[0017] Specifically, the health dual threshold system includes an action threshold and an early warning threshold. The action threshold is the statutory occupational health exposure limit. The early warning threshold is adaptively drifted according to the background concentration: the statistical average of pollutant concentration during periods without mechanical operation is used as the background baseline, and the early warning threshold is the background baseline plus a preset ratio of the action threshold. The background baseline is updated periodically with the season, meteorological conditions, and the intensity of operations inside the air-supported membrane.

[0018] Specifically, it also includes: tracking the location of workers, calculating the time-weighted average exposure concentration based on the pollutant concentration in the area and the duration of stay, and comparing it with the legally permissible concentration to determine the personnel exposure level; when the personnel exposure level exceeds the treatment level corresponding to the regional pollutant concentration, embedding the personnel exposure level into the constraints of the multi-objective optimization, prioritizing the allocation of treatment power in the area where the personnel are located, and sending out adjustment warnings or evacuation signals in a graded manner.

[0019] The beneficial effects of this invention are as follows: Achieving precise and coordinated treatment of exhaust gas and dust fills a technological gap in the industry: Existing pit air-supported membrane environmental treatment technologies mainly target dust pollution, failing to incorporate exhaust gas (containing CO, NO2, diesel particulates, and polycyclic aromatic hydrocarbons) generated by diesel construction machinery operating continuously in enclosed spaces into the treatment system. Fine particles can penetrate deep into the alveoli, causing respiratory diseases. This invention acquires concentration-thermal plume coupling characteristics and pollutant spatial distribution characteristics, utilizing the physical property of exhaust gas emissions carrying heat (thermal plume) to add a "thermal fingerprint" dimension to exhaust gas identification. This is input into a pre-trained pollution event identification model to accurately distinguish the sources of exhaust gas and dust pollution. Furthermore, differentiated coordinated treatment schemes are designed for the different diffusion characteristics of exhaust gas and dust, effectively filling the gap in existing technologies for the treatment of complex pollution.

[0020] Driven by statutory occupational health standards, this invention constructs a compliant and precise control system. Existing intelligent monitoring systems often use start / stop thresholds based on experience or manually set values, failing to establish a direct link with statutory occupational health standards. This invention uses statutory occupational health exposure limits as the core triggering basis for system control, constructing a dual-threshold system of "early warning threshold + action threshold." This provides the control logic with a clear compliance basis, elevating control logic from experience-based judgment to compliance-based judgment. Simultaneously, the early warning threshold adaptively drifts based on background concentration, adapting to dynamic changes in seasons, weather, and work intensity, further enhancing the accuracy and environmental adaptability of control.

[0021] Achieving multi-objective dynamic optimization significantly reduces pollution control energy consumption: Existing pollution control devices are mostly fixed, covering the entire area with spraying, lacking intelligent control capabilities. This makes it difficult to achieve precise control based on the location and severity of pollution sources, resulting in high energy consumption and low efficiency. This invention generates a linkage control strategy through multi-objective optimization. With the goal of minimizing the weighted average energy consumption of exhaust gas treatment and dust control, and with pollutant concentration compliance as a constraint, it dynamically determines the start-up sequence, spatial orientation, and power allocation of treatment actions. This achieves precise control of "where the pollution is, at what intensity, and when to treat it," avoiding energy waste from comprehensive treatment and significantly reducing pollution control energy consumption.

[0022] Establishing a closed-loop feedback correction mechanism to achieve adaptive control: Existing governance measures are mostly open-loop control, lacking the ability to evaluate effects and make adaptive adjustments. This invention collects concentration response data after governance implementation, extracts response characteristics of the concentration decay curve (decay onset delay, decay slope, overshoot, and steady-state residual concentration), and when the actual response characteristics deviate from the expected decay model, iteratively optimizes the power allocation parameters through gradient descent, forming a complete closed-loop control link of "perception-decision-execution-feedback-correction". This enables the system to have self-learning and adaptive capabilities, and can dynamically optimize the control strategy based on real-time feedback of the governance effect. Attached Figure Description

[0023] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0024] Figure 1 This is an overall flowchart of the method of the present invention; Figure 2 This is a flowchart for determining the type of pollution event according to the present invention; Figure 3 This is a schematic diagram of the thermal plume coupling sensing of the present invention. Detailed Implementation

[0025] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided.

[0026] Example 1: Environmental linkage control for deep foundation pit air-supported membrane construction of a subway station See Figure 1 , Figure 2 and Figure 3This embodiment takes a deep foundation pit project of a subway station in a certain city as an example. The foundation pit is 165 meters long, 80 meters wide and 18 meters deep. It adopts an air-supported foundation pit air membrane for fully enclosed construction, with the air membrane covering an area of ​​approximately 13,200 square meters. Four non-road mobile machinery are installed in the foundation pit, including one 400-type excavator (National IV emission standard), one 240-type excavator (National IV emission standard), one 150-type excavator (National IV emission standard) and one ZL50 loader. At the same time, about 15 to 20 workers are working in different areas inside the air membrane to carry out earthwork excavation, support installation and excavated soil transportation.

[0027] (I) Deployment of Multi-Physics Sensing Networks A multi-physics coupled sensing network is deployed within the air-supported membrane structure of the foundation pit, including a gas concentration sensing array, thermal imaging nodes, UWB positioning base stations, and edge computing gateways.

[0028] Gas concentration sensor array deployment: A monitoring section is set up every 20 meters along the length of the foundation pit, for a total of 8 monitoring sections. Each section has one concentration sensor array deployed at the bottom, middle, and top of the foundation pit, for a total of 24 sensor nodes. The concentration sensor array includes exhaust gas pollutant sensing channels and dust pollutant sensing channels: the exhaust gas pollutant sensing channel is equipped with a CO sensor (range 0~200mg / m³). 3 (Accuracy ±2%FS), NO2 sensor (range 0~50mg / m³) 3 (Accuracy ±3%FS) and diesel particulate sensor (range 0~1000μg / m 3 Accuracy ±5μg / m 3 The dust pollutant sensing channel is equipped with a TSP sensor (range 0~10000μg / m). 3 PM 10 Sensors and PM2.5 sensors (accuracy ±10μg / m³) 3 Each sensor continuously collects concentration time-series data at a preset sampling period (10 seconds in this embodiment).

[0029] Thermal imaging node deployment: Six infrared thermal imaging sensors are deployed at the four corners and center of the air-supported membrane structure in the foundation pit, covering the entire working area. These sensors are used to capture the spatial distribution of the thermal plume from the exhaust vents of the construction machinery in real time. The thermal imaging sensors are configured with a temperature resolution better than 0.05℃, a spatial resolution of 640×512 pixels, and a frame rate of 30Hz, enabling continuous acquisition of exhaust vent temperature field distribution and thermal gas diffusion trajectory data.

[0030] UWB positioning base station deployment: UWB positioning base stations are installed at the four corners and the center of the air-supported membrane structure in the foundation pit to construct an indoor positioning network. UWB positioning tags are installed on the roof of each piece of construction machinery and on the safety helmets of each worker, achieving a positioning accuracy better than 30cm. The UWB positioning system continuously uploads the real-time location coordinates of the construction machinery and workers at a frequency of 10Hz.

[0031] Edge computing gateway deployment: Data from various sensors, thermal imaging data, and UWB positioning data are aggregated through an edge computing gateway deployed on-site. The edge computing gateway performs spatiotemporal alignment on the data from each channel—using a unified timestamp and spatial coordinate system as a reference, it synchronizes data from different sampling frequencies and deployment locations; then, it uses a Kalman filter algorithm for noise reduction preprocessing to generate a multi-dimensional state vector sequence with spatiotemporal stamps, which is then uploaded to the cloud control platform via a 5G communication module.

[0032] (II) Concentration-thermal plume coupling feature extraction and pollution event identification In this embodiment, the extraction method of concentration-thermal plume coupling features is as follows: taking the spatial distribution of thermal plume at the exhaust port of engineering machinery captured by the thermal imaging node as the benchmark, the concentration gradient tensor of exhaust pollutants is spatiotemporally aligned with the thermal plume vector field to generate a temperature-weighted joint feature vector of concentration-thermal plume.

[0033] Specifically, using a 60-second sliding time window, within each time window: Step 1: Calculate the concentration change rate of exhaust gas pollutants and dust pollutants at each monitoring section.

[0034] Concentration change rate = (Concentration at the end of the window - Concentration at the beginning of the window) / 60 seconds The units are mg / (m 3 •s (gaseous pollutants) or μg / (m 3 •s) (particulate matter).

[0035] Step 2: Extract thermal plume characteristic parameters.

[0036] The temperature gradient, hot gas rise velocity, and thermal diffusion radius of the thermal plume core region of each construction machinery exhaust outlet were extracted from thermal imaging data. The thermal plume core region was defined as a continuous area with a temperature more than 5°C higher than the ambient temperature.

[0037] Step 3: Generate concentration-thermal plume coupled feature vector.

[0038] The concentration values ​​at the same spatiotemporal location are correlated with the thermal plume intensity values. Taking the spatial location of the thermal plume core area as a reference, the exhaust gas pollutant concentration value at that location is weighted and fused with the thermal plume temperature gradient value to generate a concentration-thermal plume coupled feature vector.

[0039] Examples of exhaust-driven pollution incidents: Suppose that at a certain moment, the CO sensor data in the monitoring area at the northwest corner of the gas-supported membrane shows that the concentration increases from 3.5 mg / m³ within 60 seconds. 3 Rising to 8.2 mg / m³ 3The rate of ascent was 0.078 mg / (m²). 3 (·s), exceeding the CO rate threshold of 0.02 mg / (m³). 3 ·s); NO2 sensor from 0.8 mg / m 3 Rising to 2.1 mg / m³ 3 The rate of ascent was 0.022 mg / (m²). 3 (·s), exceeding the NO2 rate threshold of 0.005 mg / (m³). 3 Diesel particulate matter from 15 μg / m³ 3 Rise to 38 μg / m 3 The ascent rate was 0.38 μg / (m²). 3 •s), exceeding the diesel particulate rate threshold of 0.1 μg / (m 3 Meanwhile, thermal imaging data in the area showed a distinct thermal plume core region (temperature gradient ≥ 8℃ / m, thermal diffusion radius approximately 1.2m) at the exhaust outlet of a 400-type excavator, while TSP and PM in this area were... 10 The concentration rise rates of both PM2.5 sensors were lower than their respective rate thresholds.

[0040] The aforementioned concentration-thermal plume coupling feature vectors were input into a pre-trained pollution event identification model. This model is a multi-input neural network model, trained using historical concentration-thermal plume coupling features and pollutant spatial distribution features as training samples, and manually labeled pollution event types as labels, through supervised learning. The model outputs the pollution event type as "exhaust gas-dominated type" with a confidence level of 0.92 (the posterior probability output by the model).

[0041] Examples of dust-dominated pollution events: Suppose that at another moment, the TSP sensor in the monitoring area at the southeast corner of the air film detects a concentration of 1200 μg / m³ within 60 seconds. 3 Rising to 3800 μg / m 3 The ascent rate was 43.3 μg / (m 3 (·s), exceeding the TSP rate threshold of 1.0 μg / (m 3 ·s); PM 10 From 850 μg / m 3 Rise to 2600 μg / m 3 The ascent rate was 29.2 μg / (m²). 3 ·s), exceeding PM 10 Rate threshold 0.8 μg / (m 3The concentrations of CO, NO2, and diesel particulate matter all rose at rates below their corresponding rate thresholds. Thermal imaging data in this area showed no significant thermal plume core (temperature gradient <2℃ / m). Meanwhile, the AI ​​video recognition unit detected a loader loading construction waste, generating significant dust. The model output the pollution event type as "dust-dominated" with a confidence level of 0.88.

[0042] Example of a coupled and superimposed pollution event: Assuming that at the third moment, the CO and TSP concentrations in the monitoring area of ​​the gas film rise synchronously—CO from 4.0 mg / m³. 3 Rising to 9.5 mg / m³ 3 The rate of ascent was 0.092 mg / (m²). 3 ·s), TSP from 1100 μg / m 3 Rising to 4200 μg / m 3 The ascent rate was 51.7 μg / (m²). 3 Both the ·s and ·s values ​​exceeded their respective rate thresholds. Thermal imaging data showed that an excavator was breaking rocks in the area, with a hot plume core area (temperature gradient ≥7℃ / m) at the exhaust outlet. Simultaneously, AI video recognition captured significant dust generated by the rock-breaking operation. The model output pollution event type as "coupled superposition type" with a confidence level of 0.85.

[0043] (III) Extraction of spatial distribution characteristics of pollutants and identification of target areas of pollution sources In this embodiment, the spatial distribution characteristics of pollutants are extracted by calculating the angle between the anisotropic diffusion principal axis of the pollutant concentration field and the tangent direction of the nearest mechanical trajectory in the positioning field of the engineering machinery.

[0044] Specifically, spatial interpolation (using the Kriging interpolation method) is performed based on the concentration data from each monitoring section to generate a pollutant concentration field distribution map. The anisotropic diffusion principal axis of the concentration field is calculated—on a two-dimensional plane, the concentration diffuses from the center of high values ​​outwards, and the direction of the fastest diffusion is the principal axis direction. Simultaneously, the real-time location and movement trajectory of each piece of engineering machinery are obtained from UWB positioning data, and the tangent direction of the nearest machinery trajectory is calculated.

[0045] When the angle between the principal axis of anisotropic diffusion in the concentration field and the tangent direction of the machine trajectory is in a non-orthogonal range (angle <30° or >150°), the machine is identified as a suspected pollution source. Combined with the concentration gradient consistency verification of multiple sensor nodes—that is, when the concentration gradient directions of multiple adjacent monitoring sections all point to the same machine location—the machine is confirmed as a pollution source, and the target area of ​​the pollution source is identified.

[0046] Target area identification for exhaust gas-dominated pollution sources: For the aforementioned exhaust gas-dominated pollution event, the anisotropic diffusion axis of the concentration field is oriented northwest-southeast (consistent with the location of the 400-type excavator and the core area of ​​the thermal plume), and the angle between it and the tangent direction of the 400-type excavator's trajectory is 12° (within the non-orthogonal range). The consistency verification of the concentration gradient at multiple sensor nodes passed. The system identified the pollution source target area as within an 8-meter radius of the northwest corner monitoring section.

[0047] Target areas identified as dust-dominant pollution sources: For the aforementioned dust-dominated pollution event, the anisotropic diffusion axis of the concentration field radiates outwards from the loader's operating point, with an angle of 8° to the tangent of the loader's trajectory. The consistency verification of the concentration gradient at multiple sensor nodes passed. The system pinpointed the pollution source target area to within a 6-meter radius of the loader's operating point in the southeast corner.

[0048] Coupled and superimposed pollution source target area identification: For the aforementioned coupled and superimposed pollution events, the system separately identifies the dust area (within a 5-meter radius of the excavator rock-breaking operation point) and the exhaust gas area (within an 8-meter radius of the monitoring section where the sensor node with the largest CO concentration gradient is located), and determines the union of the two as the target area of ​​the pollution source.

[0049] (iv) Comparison of dynamic risk assessment vector construction with the dual health threshold system The cloud-based control platform incorporates a dual-threshold health system. Based on GBZ 2.1-2019 "Occupational Exposure Limits for Hazardous Factors in the Workplace Part 1: Chemical Hazardous Factors", the permissible short-term exposure concentration for CO is 30 mg / m³. 3 The system sets this as the action threshold; the permissible short-term exposure concentration of NO2 is 10 mg / m³. 3 The time-weighted average permissible concentration of silica dust (total dust) is 8 mg / m³. 3 (i.e., 8000 μg / m) 3 The warning threshold adaptively shifts based on background concentration: the background baseline is the statistical average of pollutant concentrations during periods without mechanical operation (such as nighttime shutdown periods), and the warning threshold = background baseline + action threshold × 50%. The background baseline is updated weekly, based on the average concentration during the most recent week during periods without mechanical operation.

[0050] The dynamic risk assessment vector is constructed by integrating real-time pollutant concentration, concentration change trend, spatial diffusion trend, and confidence level of model output, and generating a multi-dimensional risk assessment vector according to preset weights.

[0051] Taking the aforementioned exhaust-dominated pollution incident as an example (CO concentration 18 mg / m³) 3 It is at the warning threshold of 15 mg / m³ 3 With action threshold 30mg / m 3between): Real-time concentration deviation: 18 / 15 = 1.20 (deviation between concentration and warning threshold) Concentration change trend: The rate of increase was 0.078 mg / (m³). 3 The similarity between the emission pattern (·s) and typical historical exhaust gas pollution patterns is 0.85. Spatial diffusion trend: The concentration gradient direction points towards the northwest corner, with a coupling degree of 0.78 with the main airflow direction (southeast-northwest) within the gas film. Model confidence level: 0.92 The system performs weighted fusion on the above four-dimensional data (with weights of 0.35, 0.25, 0.20, and 0.20 respectively) to generate a dynamic risk assessment vector with a comprehensive score of 0.75 (out of 1.0).

[0052] Multi-dimensional comparisons included: the deviation of each pollutant concentration from the warning threshold (1.20), the similarity of concentration change trends to historical typical pollution patterns (0.85), and the coupling degree between spatial diffusion trends and the airflow field within the air film (0.78). Based on the comprehensive score, the urgency of treatment was determined to be "moderately high," and the treatment level was "preventive treatment level."

[0053] (V) Generation of multi-objective optimization and linkage control strategies The system generates a coordinated control strategy through multi-objective optimization based on the urgency of pollution control and the type of pollution event. The goal is to minimize the weighted average energy consumption for exhaust gas treatment and dust control, while the constraint is to achieve pollutant concentration standards.

[0054] For the aforementioned exhaust-dominated pollution event (with a treatment level of preventative treatment), the multi-objective optimization model is as follows: Objective function: min J = w1·E_tail + w2·E_dust Where E_tail is the energy consumption for exhaust gas treatment (which depends on the power of the suction fan and the running time), E_dust is the energy consumption for dust control (which depends on the power of the diversion device and the spray system), and w1 and w2 are weighting coefficients (in this embodiment, w1=0.7, w2=0.3, and the weight of exhaust gas treatment energy consumption is higher).

[0055] Constraints: Pollutant concentrations do not exceed the action threshold (CO ≤ 30 mg / m³). 3 ); The total power shall not exceed the air-supported membrane power supply capacity (200kW in this embodiment). The power of each treatment device shall not be less than the minimum effective power (exhaust fan ≥ 15kW, spray system ≥ 5kW). The optimization solution employs a multi-objective particle swarm optimization algorithm with a particle swarm size of 50 and 200 iterations. The optimization result is as follows: Start-up sequence: Start exhaust gas treatment immediately (delay 0 seconds), dust control will not start temporarily (because it is determined to be exhaust gas-dominant). Spatial orientation: The air intake of the top-mounted suction hood points towards the exhaust port of the 400-type excavator in the northwest corner (horizontal angle 315°, elevation angle 25°). Power distribution: The suction fan operates at 60% of its rated power (36kW), and the spray system operates at standby power (2kW); Startup timing control in low-confidence scenarios: Assuming the confidence level for determining the aforementioned exhaust gas-dominated pollution event is 0.65 (below the preset threshold of 0.80), the system, according to the startup sequence logic described in claim 8, prioritizes exhaust gas treatment (dominant type), followed by dust control (auxiliary type) after a 60-second delay. During the delay period, the system continuously monitors CO concentration response data: if the CO concentration shows a significant decreasing trend within 30 seconds (change rate < -0.05 mg / (m³)... 3 If the CO concentration continues to rise, the delay time will be shortened from 60 seconds to 30 seconds, and auxiliary treatment will be initiated in advance to consolidate the treatment effect. If the CO concentration continues to rise, auxiliary treatment will be initiated immediately, and the confidence level judgment result will be marked as "requires manual verification," and management personnel will be notified to verify the source of pollution on-site. The dynamic adjustment relationship between concentration response data and delay time is realized by a fuzzy control rule base, and the adjustment amount is determined based on the deviation between the real-time concentration change rate and the expected change rate.

[0056] (vi) Governance Implementation and Feedback Correction The system issues control commands corresponding to the linkage control strategy. The exhaust gas treatment command controls the steering mechanism of the steerable top-suction exhaust gas collection hood to drive the intake port to rotate with multiple degrees of freedom. Based on the 400-type excavator's real-time UWB positioning data, the intake port direction is dynamically adjusted to ensure the intake port normal vector is collinear with the excavator's exhaust port direction. The suction fan is started at 60% of its rated power, guiding the exhaust-containing hot gas flow into the catalytic oxidation filtration channel. The catalytic oxidation filtration channel includes a diesel oxidation catalyst (DOC) and a selective catalytic reduction (SCR) connected in series—the DOC oxidizes CO and THC in the exhaust gas to CO2 and H2O, while the SCR reduces NOx to N2 and water. The purified gas is then discharged through the exhaust port at the top of the gas film.

[0057] The system synchronously collects concentration response data after the treatment is implemented. It continuously monitors CO concentration changes in the northwest corner with a sampling period of 10 seconds. The CO concentration ranged from 18 mg / m³. 3 As the descent begins, record the following response characteristics: Attenuation initiation delay: The time difference between the start of treatment and the beginning of concentration decline, which is 45 seconds in this embodiment; Attenuation slope: the average rate of decrease during the concentration reduction phase, which in this example is -0.12 mg / (m²).3 ·s); Overshoot: The minimum value below the final steady-state value during the concentration decrease process; in this example, it is 11 mg / m³. 3 ; Steady-state residual concentration: The final value after the concentration stabilizes; in this example, it is 12 mg / m³. 3 ; The system compares the actual response characteristics described above with the expected decay model. The expected decay model is a first-order exponential decay model. Where C0 = 6 mg / m 3 (Initial overshoot portion), C∞=12mg / m 3 (Steady-state residual concentration), λ is the attenuation coefficient, obtained by fitting historical treatment data. Different λ value libraries correspond to different pollution event types and treatment levels: exhaust gas-dominated preventive treatment corresponds to λ=0.08s. -1 (In this embodiment), the exhaust gas-dominated mandatory treatment corresponds to λ=0.12s. -1 The dust-dominant type corresponds to λ=0.05~0.07s. -1 The coupled superposition type corresponds to λ = 0.09~0.14s. -1 The λ value library is continuously updated and optimized as actual governance data accumulates.

[0058] The actual attenuation slope is -0.12 mg / (m²). 3 ·s), compared to the theoretical decay slope of the expected decay model -0.10 mg / (m 3 •s) There is a deviation (deviation amount 20%). The system determines that the response characteristics deviate from the expectation and initiates inversion correction.

[0059] The inversion correction method involves using the deviation between the actual response characteristics and the expected attenuation model as input, and iteratively optimizing the power allocation parameters using the gradient descent method. Specifically, a deviation response matrix is ​​constructed, and regularized least squares is used to solve it in reverse to determine the adjustment direction and amount of the exhaust fan power parameters. The adjustment step size does not exceed the preset upper limit (5% of the rated power in this embodiment), and the adjusted power does not exceed the physical rated range. After inversion correction and iterative optimization: after the first iteration, the exhaust fan power is adjusted from 36kW to 39kW, and the deviation between the actual response characteristics and the expected attenuation model decreases from 20% to 12%; after the second iteration, the power is adjusted to 41kW, and the deviation decreases to 8%; after the third iteration, the power is adjusted to 42kW, and the deviation decreases to 4%, the actual response characteristics fall within the 95% confidence interval of the expected attenuation model, and the system locks the current power allocation parameters.

[0060] (vii) Personnel Exposure Linkage Control The system tracks the location of each worker in real time using UWB positioning tags. Taking worker A as an example, A was located in the northwest corner area (where the average CO concentration was 12 mg / m³) for the first two hours. 3 Two hours later, the CO concentration was moved to the central region (where the average CO concentration was 8 mg / m³). 3 The time-weighted average CO exposure concentration for component A is calculated as follows: TWA_CO ​​= (12mg / m 3 × 120min + 8mg / m 3 (× 120min) / 240min = 10mg / m 3 , The time-weighted average permissible concentration of CO in GBZ 2.1-2019 is 20 mg / m³. 3 A's TWA_CO ​​is 10 mg / m³ 3 The exposure level of the personnel is "normal".

[0061] Assume another worker, B, is working at a CO concentration of 18 mg / m³. 3 After 3 hours of continuous operation in the area, the TWA_CO ​​was calculated to be 15 mg / m³. 3 The concentration reached the legally permissible level of 20 mg / m³. 3 75%. The system determines the personnel exposure level as "moderate" and embeds multi-objective optimization constraints: prioritize the allocation of treatment power in the area where B is located, and at the same time send a rest warning signal to B's smart bracelet - "CO exposure concentration has reached the warning value (75% of the legally permissible concentration), it is recommended to suspend work and rest."

[0062] Assume that worker C's TWA_CO ​​reaches 100% of the legally permissible concentration (20 mg / m³). 3 The system determined the person's exposure level to be "high risk" and sent a mandatory evacuation signal to Person C's smart bracelet: "CO exposure concentration has reached the limit, please evacuate immediately." Simultaneously, the system synchronized this information to the management console, notifying managers to verify and arrange for personnel replacement.

[0063] Example 2: Air-supported membrane environment control for a linear foundation pit of a utility tunnel This embodiment uses a comprehensive utility tunnel project in the core area of ​​a city as an example. The foundation pit is linearly arranged, 215 meters long, 18 meters wide, and 9 meters deep, and is constructed using a sliding rail air-ribbed membrane structure for enclosure. Unlike Embodiment 1, the foundation pit in this embodiment is linear and elongated, with construction machinery moving along the long side of the pit and workers distributed across multiple work surfaces.

[0064] (a) Deployment of the sensing network A monitoring section is set up every 15 meters along the length of the foundation pit, for a total of 15 monitoring sections. Each section has a concentration sensor array deployed at both the bottom and top of the pit. A thermal imaging node is deployed every 30 meters along the long side of the pit, for a total of 8 nodes. UWB positioning base stations are deployed at 20-meter intervals on both sides of the pit, for a total of 24 stations. Other deployment methods are the same as in Example 1.

[0065] (II) Differentiated Spatial Feature Extraction and Region Locking Since the foundation pit in this embodiment is linear in shape and the construction machinery moves along the long side, the extraction of the spatial distribution characteristics of pollutants needs to consider the one-dimensional spatial continuity.

[0066] For exhaust gas-dominated pollution events, the sensor peak location considers the concentration gradient between adjacent sections. If the CO concentration rise rate is the highest at the 6th monitoring section, and the CO concentration at the 5th and 7th monitoring sections shows a decreasing trend (the concentration gradient direction points to the 6th section), then the area where the 6th monitoring section is located (within a 5-meter range before and after, i.e., 10 meters in length × 18 meters in width) is locked as the target area of ​​the pollution source.

[0067] For dust-dominated pollution events, if AI video recognition and UWB positioning data indicate that the pollution source is located between the 5th and 6th monitoring sections, then the middle area (8 meters long × 18 meters wide) between these two monitoring sections will be locked as the target area of ​​the pollution source.

[0068] For coupled and superimposed pollution events, the system separately identifies the dust area and the exhaust gas area, and determines the union of the two as the target area of ​​the pollution source.

[0069] (III) Differentiated handling of multi-objective optimization constraints Due to the narrow and elongated space of linear foundation pits, spatial orientation conflict constraints of remediation devices are more important. In the constraints of multi-objective optimization, the weight of the spatial orientation non-overlapping constraint is increased—when the included angle between the spatial orientations of two remediation devices is less than 15°, it is determined to be an orientation conflict, and the optimization algorithm needs to reallocate the spatial orientations.

[0070] In this embodiment, the number of construction machines is relatively small (2 excavators + 1 loader), but the work areas are dispersed. The optimized solution is as follows: the exhaust gas treatment device points to the excavator in the 6th cross-section area, and the dust control device points to the loader in the 8th cross-section area. The spatial pointing angle between the two is 35°, which is non-conflicting.

[0071] (iv) Modular adaptation The size of the foundation pit in this embodiment (215m×18m×9m) differs significantly from that in Embodiment 1 (165m×80m×18m). The system adopts a modular design, flexibly adjusting the number of sensor nodes, treatment devices, and UWB positioning base stations according to the actual size of the foundation pit, the number of construction machinery, and the number of workers. The adjustment rules are as follows: sensor nodes are configured at a rate of one set per 100-200 square meters; thermal imaging nodes are configured at a rate of no less than one unit per 30 meters of foundation pit length; top-suction collection hoods are configured at a rate of no less than one per construction machinery; and circulating air guiding devices and micro-mist nozzles are configured at a rate of no less than one set per 30 meters of foundation pit length.

[0072] In this embodiment, a total of 30 sets of sensor nodes (15 cross sections × 2 layers), 8 thermal imaging nodes, 3 top suction collection hoods, 6 sets of circulating air guiding devices, 6 sets of micro-mist nozzles, and 24 UWB positioning base stations are configured.

[0073] (v) Feedback and Correction Mechanism The feedback correction mechanism in this embodiment is the same as in Embodiment 1. After the treatment is executed, the system synchronously collects concentration response data of the target area, identifies the response characteristics of the concentration decay curve (decay start delay, decay slope, overshoot, and steady-state residual concentration), and compares it with the expected decay model. When the actual response characteristics deviate from the expectation, the power allocation parameters are iteratively optimized using the gradient descent method.

[0074] Due to the narrow and elongated space of the linear foundation pit, the airflow field distribution differs from that in Example 1. Therefore, the λ value library for the expected attenuation model is calibrated differently from that in Example 1: the exhaust gas-dominated preventative treatment corresponds to λ=0.06s. -1 (less than 0.08s in Example 1) -1 (Due to the slightly lower air exchange efficiency in linear foundation pits), the dust-dominated type corresponds to λ=0.04~0.06s. -1 The λ value library was independently fitted based on the historical governance data of this embodiment.

[0075] All hardware devices can be installed and debugged without altering the main structure of the air-supported membrane structure, resulting in low modification costs and a short implementation period. The system can seamlessly integrate with the project's existing smart construction site platform and occupational health management platform, enabling data sharing and unified management.

[0076] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for controlling dust pollution from exhaust gas from a foundation pit using a geotextile membrane structure based on health limits, characterized in that: Includes the following steps: The concentration-thermal plume coupling characteristics and spatial distribution characteristics of pollutants are obtained, input into a pre-trained pollution event identification model, and the pollution event type, pollution source target area and its confidence level are output. A dynamic risk assessment vector is constructed by combining the real-time concentration of pollutants. The dynamic risk assessment vector is compared with the preset health dual threshold system in multiple dimensions to determine the urgency and level of treatment; a linkage control strategy is generated through multi-objective optimization, with the weighted minimization of exhaust gas treatment energy consumption and dust treatment energy consumption as the objective and the compliance of pollutant concentration as the constraint, to dynamically determine the start-up sequence, spatial orientation and power allocation of exhaust gas and dust treatment actions. The system issues control commands corresponding to the linkage control strategy, collects concentration response data after the treatment is executed, identifies the response characteristics of the concentration decay curve, and when the response characteristics deviate from the expectation, it inversely corrects the power allocation parameters of the linkage control strategy.

2. The method according to claim 1, characterized in that, The extraction method of the concentration-thermal plume coupling feature is as follows: taking the spatial distribution of thermal plume at the exhaust port of engineering machinery captured by the thermal imaging node as the benchmark, the concentration gradient tensor of exhaust pollutants is spatiotemporally aligned with the thermal plume vector field to generate a temperature-weighted joint feature vector of concentration-thermal plume.

3. The method according to claim 1, characterized in that, The method for extracting the spatial distribution characteristics of pollutants is as follows: calculate the angle between the anisotropic diffusion principal axis of the pollutant concentration field and the tangent direction of the nearest mechanical trajectory in the positioning field of the engineering machinery. When the angle is in the non-orthogonal interval, the machinery is determined to be a suspected pollution source. Combined with the concentration gradient consistency verification of multiple sensor nodes, the target area of ​​the pollution source is determined.

4. The method according to claim 1, characterized in that, The pollution event identification model is trained using historical concentration-thermal plume coupling feature vectors and pollutant spatial distribution features as training samples, and pollution event type and pollution source target area as labels, through supervised learning training; the confidence level is the posterior probability output by the model, used to characterize the reliability of the judgment result.

5. The method according to claim 1, characterized in that, The dynamic risk assessment vector is constructed by integrating real-time pollutant concentration, concentration change trend, spatial diffusion trend and confidence level, and generating a multi-dimensional risk assessment vector according to preset weights. The multi-dimensional comparison includes: the deviation between the concentration of each pollutant and the warning threshold, the similarity between the concentration change trend and the historical typical pollution pattern, and the coupling degree between the spatial diffusion trend and the airflow field inside the air film.

6. The method according to claim 1, characterized in that, The urgency of the governance is determined based on the weighted fusion result of the risk values ​​of each dimension in the dynamic risk assessment vector; the governance level includes the pre-start level, the preventive governance level, and the mandatory governance level, which correspond to different start-up sequences and power combinations of exhaust gas and dust governance actions in the linkage control strategy.

7. The method according to claim 1, characterized in that, The objective function of the multi-objective optimization also includes: the convergence time for pollutant concentrations to drop below the warning threshold; the constraints also include that the concentrations of each pollutant do not exceed the action threshold, that the spatial orientation of the treatment devices does not conflict, and that priority is given to the area where the operators are located.

8. The method according to claim 1, characterized in that, The method for dynamically determining the activation sequence of exhaust gas and dust control actions is as follows: the pollution event types include exhaust gas-dominated, dust-dominated, and coupled superimposed types. When the system is determined to be a coupled superposition type and the confidence level is higher than the preset threshold, exhaust gas treatment and dust control are started simultaneously; when the system is determined to be exhaust gas-dominated or dust-dominated and the confidence level is lower than the preset threshold, the dominant type of treatment is started first, and the auxiliary type of treatment is started after a preset delay, and the delay time is dynamically adjusted according to the concentration response data.

9. The method according to claim 1, characterized in that, The concentration decay curve response characteristics include: decay onset delay, decay slope, overshoot, and steady-state residual concentration; the inversion correction method is as follows: taking the deviation between the actual response characteristics and the expected decay model as input, the power allocation parameters are iteratively optimized through gradient descent until the actual response characteristics fall within the confidence interval of the expected decay model; wherein, the expected decay model is pre-calibrated according to the pollution event type, treatment level, and power allocation parameters.

10. The method according to claim 1, characterized in that, The health dual threshold system includes an action threshold and an early warning threshold. The action threshold is the legal occupational health exposure limit. The early warning threshold is adaptively drifted according to the background concentration: the statistical average of pollutant concentration during periods without mechanical operation is used as the background baseline, and the early warning threshold is the background baseline plus a preset ratio of the action threshold. The background baseline is updated periodically with the season, meteorological conditions, and the intensity of operations inside the air-supported membrane.

11. The method according to claim 1, characterized in that, Also includes: Track the location of workers, calculate the time-weighted average exposure concentration based on the concentration of pollutants in the area and the duration of their stay, and compare it with the legally permissible concentration to determine the exposure level of the workers; When the personnel exposure level exceeds the treatment level corresponding to the regional pollutant concentration, the personnel exposure level is embedded into the constraints of the multi-objective optimization, priority is given to ensuring the treatment power allocation of the area where the personnel are located, and a graded adjustment warning or evacuation signal is sent.

12. A pit air-supported membrane exhaust gas dust control system based on health limits, characterized in that, It includes a processor for performing the method as described in any one of claims 1 to 11.

Images