A flue gas early warning method based on edge computing

Abstract

The present application relates to the technical field of smoke early warning, and particularly relates to a smoke early warning method based on edge computing, which comprises the following steps: data acquisition; risk index calculation; abnormal node determination; accumulation behavior and static wind low identification; deviation determination; dynamic intervention; and early warning. In the process of night tunnel construction, the present application introduces multi-dimensional parameters closely related to smoke movement, such as plate depth, wind speed and welding frequency, and further combines smoke particle concentration to determine double factors, further identifies smoke accumulation trend based on the concentration gradient between abnormal nodes and welding frequency, and analyzes and infers static wind low area in combination with the trend of terrain depression and airflow slowing down, thereby effectively solving the problems that early warning response lags due to single data, transmission delay and centralized processing, and further causing smoke abnormalities that cannot be discovered in time and safety hazards that are aggravated.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of smoke early warning, and particularly relates to a smoke early warning method based on edge computing. BACKGROUND

[0002] With the continuous construction of urban underground transportation infrastructure, nighttime tunnel construction has become an important way to ensure traffic operation and construction progress. However, in the middle section of the tunnel, due to the long distance from the ventilation opening, strong structural sealing, and frequent construction machinery and welding operations, smoke is prone to accumulate in the local space, leading to reduced visibility and increased concentration of harmful particulate matter. This not only endangers the health of workers, but also increases the risk of fire and explosion, posing higher challenges to real-time monitoring and rapid response capabilities of the tunnel construction environment.

[0003] Chinese patent application publication No. CN118731281A discloses a smoke detection and analysis method and system. The method includes: obtaining a set of monitoring information on a plurality of sampling detection nodes. The set of monitoring information reported by each sampling detection node is obtained by aggregating and processing the monitoring information reported by a plurality of monitoring sensors within the coverage range of each sampling detection node. According to a pre-set sampling monitoring task, it is determined whether all sampling detection nodes reporting the set of sampling information cover the corresponding monitoring sampling section. When it is determined that all sampling detection nodes cover the corresponding monitoring sampling section, the data monitoring platform determines whether the set of sampling information reported by each sampling detection node includes the sensing monitoring information reported by all data collection sensors within the coverage range of each sampling detection node. When it is determined that the set of sensing monitoring information reported by each sampling detection node includes the sensing monitoring information reported by all monitoring sensors within the coverage range of each sampling detection node, the data monitoring platform analyzes the monitoring state of the monitoring sampling section based on the set of sensing monitoring information reported by all sampling detection nodes and a pre-trained sampling state analysis model.

[0004] Therefore, the smoke detection and analysis method and system have the following problems: the monitoring range of the sampling detection node and the sensor coverage have blind spots, resulting in incomplete monitoring information; the method relies on pre-set sampling monitoring tasks, which lacks flexibility and is difficult to adapt to dynamic changes in the construction environment; the data aggregation processing of the method is centralized on the platform, lacking real-time preprocessing on site, resulting in high response delay; the monitoring state analysis of the method highly depends on the training model, and the model generalization ability and accuracy are limited; the method lacks comprehensive analysis of construction behaviors (such as welding frequency, wind speed, and other multi-dimensional parameters), the analyzed data is single, and the transmission delay and centralized processing lead to delayed early warning response. SUMMARY

[0005] To this end, the application provides a flue gas early warning method based on edge computing, which overcomes the problem of delayed early warning response and intensified safety hazards caused by the inability to timely discover flue gas abnormalities due to single data and transmission delay and centralized processing in the prior art by collecting and processing multi-source construction environment data in real time through tunnel internal edge computing nodes.

[0006] To achieve the above-mentioned purpose, the application provides a flue gas early warning method based on edge computing, comprising:

[0007] Real-time acquisition of the drop plate depth, flue gas particle concentration, welding frequency and wind speed at each edge node uniformly arranged in the middle section of the tunnel during nighttime tunnel construction;

[0008] Calculation of the static wind accumulation risk index according to the drop plate depth, the wind speed and the welding frequency;

[0009] Abnormality determination according to the flue gas particle concentration, the static wind accumulation risk index and a preset risk threshold to determine a plurality of abnormal nodes;

[0010] Determination of a plurality of accumulation behavior nodes according to the flue gas particle concentration and the welding frequency between any two adjacent abnormal nodes, and determination of a plurality of static wind low-lying behavior nodes according to the change trend of the drop plate depth and the wind speed of any two adjacent abnormal nodes;

[0011] Deviation determination according to the overlap of the accumulation behavior nodes and the static wind low-lying behavior nodes to obtain an aggregation point deviation result;

[0012] Adjustment of the preset exhaust fan speed or the preset risk threshold at the corresponding abnormal nodes according to the aggregation point deviation result;

[0013] Issuance of an early warning to the abnormal nodes re-acquired after adjusting the exhaust fan speed or the preset risk threshold.

[0014] Further, the process of deviation determination according to the overlap of the accumulation behavior nodes and the static wind low-lying behavior nodes to obtain an aggregation point deviation result comprises:

[0015] Determination of the overlap degree according to all the accumulation behavior nodes and all the static wind low-lying behavior nodes;

[0016] Determination of the distribution density according to all the time stamps in which the overlap degree is greater than a preset overlap degree threshold within a preset deviation determination period;

[0017] Flue gas aggregation point position deviation determination based on the comparison result of the distribution density and a preset distribution density range to obtain the aggregation point deviation result.

[0018] Further, the process of adjusting the preset exhaust fan rotating speed or the preset risk threshold at the corresponding abnormal node according to the aggregation point deviation result comprises:

[0019] When the distribution density is greater than the maximum value of the preset distribution density range, increasing the preset exhaust fan rotating speed according to the relative deviation of the distribution density and the maximum value of the preset distribution density range;

[0020] When the distribution density is less than the minimum value of the preset distribution density range, decreasing the preset risk threshold according to the relative deviation of the minimum value of the preset distribution density range and the distribution density.

[0021] Further, the process of determining a number of accumulation behavior nodes according to the smoke particle concentration and the welding frequency between each two adjacent abnormal nodes comprises:

[0022] Obtaining the number of edge nodes between each two adjacent abnormal nodes to obtain a non-abnormal point number;

[0023] Based on the comparison result of the non-abnormal point number and a preset point number threshold, determining a rapid accumulation state index according to the smoke particle concentration and the welding frequency;

[0024] Determining a number of accumulation behavior nodes according to the comparison result of the rapid accumulation state index and a preset accumulation state index threshold.

[0025] Further, the process of determining a rapid accumulation state index according to the smoke particle concentration and the welding frequency based on the comparison result of the non-abnormal point number and a preset point number threshold comprises:

[0026] When the non-abnormal point number is greater than the preset point number threshold, determining a particle concentration gradient according to the smoke particle concentration of the two adjacent abnormal nodes corresponding to each time within the preset state determination period, and the smoke particle concentration at the edge node closest to the midpoint of the two adjacent abnormal nodes;

[0027] When the non-abnormal point number is less than or equal to the preset point number threshold, determining the particle concentration gradient according to the smoke particle concentration of the two adjacent abnormal nodes corresponding to each time within the preset state determination period;

[0028] Determining a rapid accumulation state index according to the particle concentration gradient within the preset state determination period and the welding frequency.

[0029] Further, the process of determining a rapid accumulation state index according to the particle concentration gradient within the preset state determination period and the welding frequency comprises:

[0030] determining a frequency gradient according to the welding frequencies of two adjacent abnormal nodes corresponding to each time within a same preset state determination period;

[0031] determining the fast accumulation state index according to the particle concentration gradient and the frequency gradient.

[0032] Further, the process of determining a number of static wind low behavior nodes according to the change trend of the drop plate depth and the wind speed of each arbitrary two adjacent abnormal nodes includes:

[0033] determining a number of depth gradients according to the drop plate depths of each arbitrary two adjacent abnormal nodes corresponding to each time from the initial time to the current time within a same preset state determination period;

[0034] determining a number of wind speed gradients according to the wind speeds of each arbitrary two adjacent abnormal nodes corresponding to each time from the initial time to the current time within a same preset state determination period;

[0035] determining a trend consistency degree according to all the depth gradients and all the wind speed gradients;

[0036] determining a number of the static wind low behavior nodes according to a comparison result of the trend consistency degree and a preset consistency degree threshold.

[0037] Further, the process of calculating a static wind accumulation risk index according to the drop plate depth, the wind speed and the welding frequency includes:

[0038] determining the static wind accumulation risk index according to a ratio of the drop plate depth to the wind speed, a preset retention weight, the welding frequency and a preset welding frequency weight.

[0039] Further, the process of determining a number of abnormal nodes according to the flue gas particle concentration, the static wind accumulation risk index and a preset risk threshold for abnormality determination includes:

[0040] determining a number of flue gas steady state values according to the flue gas particle concentrations from the initial time to the current time within a preset abnormality determination period;

[0041] determining a number of risk index steady state values according to the static wind accumulation risk indexes from the initial time to the current time within the preset abnormality determination period;

[0042] determining a number of the abnormal nodes according to abnormality determination of all the flue gas steady state values, all the risk index steady state values and the preset risk threshold.

[0043] Further, the process of determining a number of the abnormal nodes according to abnormality determination of all the flue gas steady state values, all the risk index steady state values and the preset risk threshold includes:

[0044] determining a steady state similarity according to all the steady state values of the smoke and all the steady state values of the risk index;

[0045] determining whether the edge node is abnormal according to a comparison result of the steady state similarity and the preset risk threshold value, to determine a number of abnormal nodes.

[0046] Compared with the prior art, the beneficial effects of the present application are that, in view of the actual problem that smoke retention is more likely to occur in the middle section of the tunnel during nighttime tunnel construction, especially in the region where the work is frequent and the ventilation blind area is significant, by introducing multi-dimensional parameters such as plate lowering depth, wind speed and welding frequency that are closely related to smoke movement, a static wind accumulation risk index is constructed to effectively represent the combined effect of the specific structure of the tunnel and the construction behavior on smoke retention; then, a double-factor determination is performed in combination with the smoke particle concentration to improve the accuracy of abnormal node identification, and further based on the concentration gradient and welding frequency between abnormal nodes, the smoke accumulation trend is identified, and the static wind low-lying area is inferred in combination with the trend analysis of the terrain depression and airflow slowing down, so that the smoke gathering point positioning is more accurate. Through dynamic analysis of the overlap degree of the above two types of key behavior nodes, the false alarm and missed alarm problems caused by the traditional early warning relying on static concentration threshold are avoided; at the same time, according to the deviation result, the local exhaust intensity or the risk threshold value is dynamically adjusted to realize feedback type risk intervention, enhance the system adaptability and control accuracy, realize the accurate identification and response control of smoke anomaly in complex construction environment, and effectively solve the problems of delayed early warning response caused by single data and transmission delay and centralized processing, and further intensify the problem of smoke anomaly that cannot be discovered in time and safety hazards.

[0047] Further, by introducing two types of dynamic parameters such as the ratio of plate lowering depth to wind speed and welding frequency that are closely related to smoke migration behavior, and then performing normalization processing after assigning weights, the static wind accumulation risk index can reflect the inhibition effect of the tunnel structure terrain on airflow flow and the enhancement effect of high-frequency construction work on continuous release of smoke, thereby effectively representing the comprehensive risk degree of smoke retention in the specific spatial section of the tunnel under nighttime operation state, and reflecting the objective trend that gas is more likely to accumulate in low-lying static wind area and the frequency of heat source increases to aggravate smoke accumulation in the physical level. It can dynamically perceive the local smoke anomaly caused by the coordinated evolution of structure environment and construction behavior, and provide more accurate and more physically interpretable risk factor support for subsequent abnormal identification and early warning determination.

[0048] Further, by introducing the standard deviation calculation of the flue gas particle concentration and the static wind accumulation risk index within a certain period, the concentration fluctuation and retention trend of each edge node in the time dimension can be effectively quantified, and the high-risk area with stable flue gas accumulation but small airflow disturbance can be identified. This judgment mechanism from the perspective of steady-state fluctuation not only captures the progressive risk caused by poor flue gas diffusion, but also reflects the local flue gas fluctuation characteristics caused by changes in construction intensity. By combining the flue gas steady-state value with the risk index steady-state value and combining the preset risk threshold for constraint, the timeliness and pertinence of abnormal node identification can be effectively enhanced, avoiding the disadvantages of traditional single-point concentration over-limit judgment method which is easily disturbed by instantaneous fluctuation, and realizing more stable and more predictive risk warning response.

[0049] Further, by maximum-minimum normalization processing of all flue gas steady-state values and risk index steady-state values, and introducing the Pearson correlation coefficient as the calculation index of steady-state similarity, the strength of the linkage relationship between local flue gas concentration change and construction ventilation environment can be effectively reflected. When the two are highly positively correlated, that is, the steady-state similarity is high, it means that the welding behavior and the poor airflow environment together contribute to flue gas retention, and there is a substantial risk at the edge node. This judgment mechanism not only avoids over-reliance on a single parameter, but also improves the sensitivity and accuracy of abnormal identification, which can more comprehensively capture the combined effects of construction disturbance on the local atmospheric environment, effectively supporting high-confidence identification of abnormal nodes.

[0050] Further, by introducing the number of non-abnormal edge nodes between adjacent abnormal nodes as a structural reference, and combining the flue gas particle concentration and welding frequency in this section, the rapid accumulation state index is calculated, which can dynamically identify whether there is a persistent and centripetal flue gas accumulation trend between abnormal nodes. Not only does it consider the spatial variation of local concentration, but also takes into account the heat contribution of welding operations, making the identification of local flue gas accumulation behavior more sensitive and accurate. By utilizing the interaction between multiple data sources, the trend analysis capability is enhanced, thereby accurately extracting potential accumulation areas and effectively supporting subsequent deviation evaluation and response deployment of ventilation control strategies.

[0051] Further, by introducing the concentration difference and spatial distance of non-abnormal nodes near the midpoint when there are more non-abnormal points, the transverse accumulation trend of local air masses in the tunnel cross-section is reflected. When there are fewer non-abnormal points, the concentration difference and distance between adjacent abnormal nodes are directly used to quickly capture the longitudinal concentration distribution change, thereby achieving adaptive quantification of the particle concentration gradient under different node distribution conditions. Combined with the welding frequency information, the particle generation rate and spatial diffusion condition are coupled and evaluated, making the rapid accumulation state index reflect not only the source intensity change but also the diffusion obstruction degree, which helps to accurately identify the abnormal accumulation trend of flue gas in complex construction sections and improve the sensitivity and reliability of early warning judgment.

[0052] Further, by introducing particle concentration gradient and welding frequency gradient at the same time, the frequency gradient model is established combined with the spatial distance factor, and the concentration gradient and frequency gradient at the end time are normalized to maximum-minimum value in a unified preset state determination period, so that the parameters of different dimensions have comparability, and the change trend of another index is avoided to be covered by a single index due to dimensional difference or too large absolute value; then the normalized results are weighted by using the preset concentration gradient weight and frequency gradient weight, which can accurately reflect the superposition effect of welding operation intensity and smoke diffusion trend on accumulation rate, so as to identify the hidden danger of rapid accumulation of smoke in the early stage, and realize more timely and accurate risk response.

[0053] Further, by using the standard deviation of the lowered plate depth to reflect the amplitude of the concave change of the tunnel structure, and using the standard deviation of the wind speed to represent the local airflow stability, and then normalizing to eliminate the dimensional difference, and then calculating the trend consistency of the two, the accurate identification of the area where the low-lying terrain and airflow slowing exist at the same time in the construction section is realized; when the synchronization of the two types of characteristics in time change is high, the potential dangerous position prone to static wind retention can be effectively locked, thereby improving the accuracy of the determination of the static wind low-lying behavior node, and providing a more reliable basis for the subsequent smoke accumulation point positioning and exhaust strategy optimization.

[0054] Further, by the joint determination method of overlap degree and distribution density, the spatial consistency of smoke accumulation trend and airflow low-lying distribution can be comprehensively evaluated in the time dimension, so as to more accurately identify the real position change of the smoke accumulation point. The calculation of overlap degree directly reflects the matching degree of the two types of key behavior nodes in space, and the distribution density quantifies the persistence and frequency of this high matching state in a determination period, so that accidental and short-time fluctuations will not be misjudged as position deviation. When the distribution density deviates from the preset range, it means that the coupling characteristics of smoke accumulation and static wind low-lying in the construction environment have abnormal changes, which may cause the accumulation point position to migrate to the non-expected area, so as to trigger the positioning correction and risk intervention in advance, effectively reducing the false alarm and false alarm and improving the pertinence of response.

[0055] Further, by dynamically comparing the distribution density with the upper and lower limits of the preset distribution density range, adaptive optimization of the exhaust fan speed and the risk threshold is realized: when the distribution density is higher than the upper limit of the range, the system proportionally increases the exhaust fan speed according to the exceeding amplitude, so that the ventilation intensity and the smoke continuous aggregation trend are positively linked, and the retention effect of the high concentration area is quickly weakened; when the distribution density is lower than the lower limit of the range, the risk threshold is proportionally reduced according to the low degree, so as to improve the sensitivity of abnormal triggering and prevent low-frequency but high-risk smoke aggregation from being ignored. This method can balance the ventilation energy consumption and the early warning accuracy under different working conditions, so that the control strategy can not only respond to high-risk aggregation in time, but also maintain sufficient detection sensitivity in the low-risk stage, significantly improving the adaptability and safety level of the system to the dynamic changes of smoke in complex construction environment. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 a flowchart of the smoke early warning method based on edge computing of the embodiment;

[0057] Figure 2 a determination logic diagram for determining whether an edge node is abnormal in the embodiment;

[0058] Figure 3 a determination logic diagram for determining whether a static wind low-lying behavior node is in the embodiment;

[0059] Figure 4 a determination logic diagram for determining whether the position deviation of the smoke aggregation point occurs in the embodiment. DETAILED DESCRIPTION

[0060] In order to make the purpose and advantages of the present application clearer and more apparent, the present application will be further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present application, and do not limit the present application.

[0061] The preferred embodiments of the present application will be described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the present application, and are not intended to limit the protection scope of the present application.

[0062] Please refer to Figure 1 shown, which is a flowchart of the smoke early warning method based on edge computing of the embodiment, the embodiment provides a smoke early warning method based on edge computing, which comprises:

[0063] Real-time acquisition of the plate lowering depth, smoke particle concentration, welding frequency and wind speed at each edge node uniformly arranged in the middle section of the tunnel during night tunnel construction;

[0064] Calculating a static wind accumulation risk index according to the plate lowering depth, the wind speed and the welding frequency;

[0065] abnormality determination is performed according to the smoke particle concentration, the static wind accumulation risk index and a preset risk threshold value, to determine a plurality of abnormal nodes;

[0066] a plurality of accumulation behavior nodes are determined according to the smoke particle concentration and the welding frequency between any two adjacent abnormal nodes, and a plurality of static wind depression behavior nodes are determined according to the change trend of the plate depth and the wind speed of any two adjacent abnormal nodes;

[0067] deviation determination is performed according to the overlap of the accumulation behavior nodes and the static wind depression behavior nodes, to obtain an aggregation point deviation result;

[0068] According to the aggregation point deviation result, the preset exhaust fan speed at the corresponding abnormal node or the preset risk threshold value is adjusted;

[0069] After adjusting the exhaust fan speed or the preset risk threshold value, the abnormal nodes obtained again are warned.

[0070] In this embodiment, for the risk that smoke is prone to stay during nighttime tunnel construction, a plurality of edge nodes are arranged in the middle section of the tunnel along the construction direction, each edge node is arranged at a distance of 5-10 meters, preferably 7 meters, at the junction of the tunnel side wall and the vault, and is fixed to the inner wall of the tunnel lining by an anchor support. Each edge node is provided with a laser ranging sensor for obtaining the change value of the vertical distance from the ground to the tunnel vault at the corresponding position of the node in real time, to represent the change of the local plate depth caused by construction; a laser particulate matter sensor is configured to obtain the smoke particle concentration of PM2.5 and PM10 in the current space; a hot film or ultrasonic wind speed sensor is configured to obtain the instantaneous wind speed value at the node; and an electromagnetic interference detection module and a high-frequency sound signal recognition module (with an FFT spectrum analysis algorithm) are also configured to obtain the welding frequency near the node. The above edge nodes are integrated with a low-power edge computing module for local processing of the collected data, including data preprocessing, abnormality pre-screening and feature parameter extraction, and a wireless communication module (such as a 5G communication module or a LVRa communication module) is arranged inside the tunnel to establish a data transmission link with the central control system, to realize real-time dynamic perception of the smoke diffusion environment in the middle section of the tunnel during nighttime tunnel construction.

[0071] In this embodiment, the edge node includes a small edge computing unit integrating a smoke sensor, a wind speed probe and a construction behavior monitoring module, which can locally calculate intermediate indicators including but not limited to a static wind accumulation risk index, a concentration gradient and a frequency gradient in real time, complete preliminary abnormal node identification and behavior node determination without uploading all original data, reduce transmission pressure and response delay, and ensure the timeliness of the warning and the robustness of the system in the nighttime tunnel construction environment.

[0072] In this embodiment, two adjacent abnormal nodes refer to two abnormal nodes that do not have other abnormal nodes between them in the order of spatial position arrangement among all edge nodes arranged along the construction axis in the middle section of the tunnel.

[0073] In this embodiment, the exhaust fans are arranged at the top of the middle section of the tunnel and are spaced apart along the construction axis with a spacing of preferably 15-20 meters, and are staggered with the arrangement position of the edge nodes to ensure that air flow intervention can be performed in the shortest path when an abnormal node or aggregation point deviation result is monitored. Each exhaust fan is connected to the main tunnel ventilation pipe through a high-temperature and dust-proof pipe, and is equipped with a variable frequency speed regulation module, which can automatically adjust the speed according to the real-time risk index uploaded by the edge computing unit to realize on-demand partitioned exhaust. This arrangement can maintain the uniformity of the overall ventilation while focusing on covering the smoke-prone vault and the work-intensive area, reducing the risk of ventilation blind area, and reducing unnecessary energy consumption.

[0074] In this embodiment, issuing a warning refers to issuing a prompt signal through an audible and visual alarm device, a tunnel broadcasting system or a construction personnel terminal, and can be pushed to the construction dispatching platform at the same time to realize instant reminding and personnel evacuation guidance for high-risk areas, so as to complete effective intervention before the smoke concentration reaches the harm threshold, thereby ensuring the safety of construction personnel and reducing the probability of accidents.

[0075] The preset risk threshold is a judgment criterion for determining whether the edge node is in an abnormal state, and its value is determined by factors such as the steady-state distribution law of smoke particle concentration and static wind accumulation risk index, tunnel ventilation conditions, work type, and nighttime atmospheric pressure stability. It is usually set to 0.6-0.8 to achieve a balance between sensitive identification of edge node smoke abnormalities and fault tolerance control. In this embodiment, the preset risk threshold is set to 0.72, which can effectively filter false positives caused by short-term fluctuations, improve the stability and accuracy of the warning trigger, and ensure that pre-intervention is completed before the smoke accumulates.

[0076] The preset exhaust fan speed is a key control parameter for regulating the exhaust efficiency of the tunnel middle section, and its value depends on factors such as the power level of the exhaust system, the cross-sectional area of the tunnel, the wind resistance structure, the degree of equipment aging, and the environmental cleanliness level required for nighttime construction. It is usually set to 800-1500 rpm to meet the ventilation and air exchange needs under different intensity work conditions. In this embodiment, the preset exhaust fan speed is set to 1200 rpm, which can maintain the thermal comfort of construction personnel while quickly diluting the local smoke concentration and effectively alleviating the accumulation trend of high-risk nodes.

[0077] Aiming at the actual problem that smoke is more likely to accumulate in the middle section of the tunnel during night-time tunnel construction, especially in the area where the work is frequent and the ventilation blind area is significant, by introducing multi-dimensional parameters such as the lowering plate depth, wind speed and welding frequency closely related to smoke movement, a static wind accumulation risk index is constructed to effectively represent the combined effect of the specific structure of the tunnel and the construction behavior on smoke accumulation. Then, in combination with the smoke particle concentration, the double-factor judgment is carried out to improve the accuracy of the identification of abnormal nodes. Further, based on the concentration gradient between abnormal nodes and the welding frequency, the smoke accumulation trend is identified, and the static wind low-lying area is analyzed and inferred in combination with the trend of terrain depression and airflow slowing down, so that the positioning of the smoke accumulation point is more accurate. Through dynamic analysis of the overlap degree of the above two types of key behavior nodes, the false alarm and missed alarm problems caused by the traditional warning relying on static concentration threshold are avoided. At the same time, according to the deviation result, the local exhaust intensity or risk threshold is dynamically adjusted to realize feedback risk intervention, enhance the system adaptability and control accuracy, and realize the accurate identification and response control of smoke abnormalities in complex construction environment, effectively solving the problem that the warning response lags due to single data and transmission delay and centralized processing, and further causing the smoke abnormalities cannot be discovered in time and the safety hazards are intensified.

[0078] Specifically, the process of calculating the static wind accumulation risk index according to the lowering plate depth, the wind speed and the welding frequency includes: determining the static wind accumulation risk index according to the ratio of the lowering plate depth to the wind speed, a preset accumulation weight, the welding frequency and a preset welding frequency weight, Q = a × G / G0 + b × J / J0, wherein Q is the static wind accumulation risk index, G is the ratio of the lowering plate depth to the wind speed, a is the preset accumulation weight, b is the preset welding frequency weight, G0 is a preset accumulation maximum safety value, and J0 is a preset welding frequency maximum safety value.

[0079] The preset accumulation weight is a weight factor representing the contribution degree of the construction area terrain to the airflow accumulation risk, and its value depends on parameters such as the tunnel section shape, the lowering plate structure influence range, the wind flow interference intensity and the wind speed stability during construction period. It is usually set between 0.4 and 0.7 to regulate the dominant degree of the ratio of the lowering plate depth to the wind speed in the risk index. In this embodiment, it is set to 0.6 to enhance the response sensitivity to local static wind phenomenon, so that the system can accurately reflect the possible gas accumulation trend even in a light wind condition.

[0080] The preset welding frequency weight is a parameter for measuring the influence degree of welding work on local smoke accumulation, which depends on the construction process density, the material type, the welding method (such as manual arc welding / arc welding), and the night work frequency fluctuation range. The weight is generally set between 0.3 and 0.6, and in this embodiment, it is set to 0.4, which can moderately reflect the smoke increment brought by welding activities, while avoiding excessive disturbance to the overall risk assessment result, thereby improving the model stability and generalization ability.

[0081] The preset maximum safe value of the static wind accumulation risk is an upper limit reference value for normalization in the static wind accumulation risk assessment, represents an allowed safe boundary of the ratio of the drop plate depth to the wind speed under normal ventilation conditions, and depends on the tunnel ventilation design standard, the exhaust flow rate, the drop plate structure scale, and historical safety data. It is usually set between 8 m·s / m³ and 12 m·s / m³. In this embodiment, it is set to 10 m·s / m³, which can reasonably define a high-risk threshold area when the wind speed is low and provide a physically interpretable normalization scale for the risk index.

[0082] The preset maximum safe value of the welding frequency is a maximum acceptable welding frequency per unit time, is used to measure the potential contribution of the work intensity to the smoke concentration, depends on the construction organization plan, the number of welders, the welding time, and the night work system, and is usually set between 30 times / min and 60 times / min. In this embodiment, it is set to 45 times / min, which can reasonably cover most standard construction workloads and accurately reflect the risk state of the high-frequency welding section without causing false positives.

[0083] By introducing two types of dynamic parameters related to the smoke migration behavior, namely the ratio of the drop plate depth to the wind speed and the welding frequency, and normalizing them by assigning weights, the static wind accumulation risk index can reflect both the inhibitory effect of the tunnel structure terrain on airflow flow and the enhancing effect of high-frequency construction work on the continuous release of smoke, thereby effectively characterizing the comprehensive risk degree of smoke accumulation in the tunnel under night work conditions. It reflects the objective trend that gas is more likely to accumulate in low-lying static wind areas and that the increase in heat source frequency exacerbates smoke accumulation, can dynamically perceive the local smoke anomalies caused by the coordinated evolution of the structure environment and construction behavior, and provides more accurate and physically interpretable risk factor support for subsequent anomaly identification and early warning determination.

[0084] Specifically, the process of determining the abnormal nodes according to the smoke particle concentration, the static wind accumulation risk index, and a preset risk threshold value includes: calculating the standard deviation of the smoke particle concentration from the initial time to each time in a preset abnormal determination period to determine a plurality of smoke steady-state values; calculating the standard deviation of the static wind accumulation risk index from the initial time to each time in the preset abnormal determination period to determine a plurality of risk index steady-state values; and determining a plurality of the abnormal nodes according to all the smoke steady-state values, all the risk index steady-state values, and the preset risk threshold value.

[0085] The preset abnormality determination period refers to a time interval for calculating the standard deviation of the smoke particle concentration and the static wind accumulation risk index, which depends on the duration of welding work in tunnel nighttime construction, the frequency of local ventilation and the dynamic change law of smoke diffusion stability, and is usually set between 30 seconds and 5 minutes, and is set to 90 seconds in this embodiment. It can effectively smooth the interference caused by instantaneous fluctuations while ensuring timely response, making the extracted steady-state value more representative, thereby improving the stability and accuracy of abnormal node identification.

[0086] By introducing the standard deviation calculation of the smoke particle concentration and the static wind accumulation risk index within a certain period, the concentration fluctuation and retention trend of each edge node in the time dimension can be effectively quantified, and the high-risk area with stable smoke accumulation and small airflow disturbance can be identified. This judgment mechanism from the perspective of steady-state fluctuation not only captures the progressive risk caused by poor smoke diffusion, but also reflects the local smoke fluctuation characteristics caused by changes in construction intensity. By jointly determining the smoke steady-state value and the risk index steady-state value, and combining the preset risk threshold for constraint, the timeliness and pertinence of abnormal node identification can be effectively enhanced, avoiding the disadvantages of traditional single-point concentration overrun judgment method which is easily disturbed by instantaneous fluctuations, and realizing more stable and more predictive risk warning response.

[0087] Please refer to Figure 2 As shown in the drawings, the determination logic diagram for determining whether the edge node is abnormal in this embodiment includes: performing maximum-minimum value normalization processing on all smoke steady-state values to obtain a smoke normalized value set, and performing maximum-minimum value normalization processing on all risk index steady-state values to obtain a risk normalized value set; calculating the Pearson correlation coefficient of the smoke normalized value set and the risk normalized value set to determine the steady-state similarity; and determining that the edge node is abnormal when the steady-state similarity is greater than the preset risk threshold to determine the number of abnormal nodes.

[0088] The preset risk threshold is the lower limit of the steady-state similarity for determining whether the edge node is abnormal, which depends on the tunnel ventilation conditions, construction intensity and the correlation distribution characteristics in historical smoke accumulation events, and is usually set between 0.6 and 0.9, and is set to 0.75 in this embodiment. It can effectively balance the risk of false alarm and false alarm, and ensure that only when the smoke concentration change and the static wind accumulation trend are highly coupled, the abnormality determination is triggered, thereby improving the reliability of abnormal identification and the effectiveness of early warning triggering.

[0089] By maximum-minimum normalization processing of the full flue gas steady-state value and the risk index steady-state value, the Pearson correlation coefficient is introduced as the calculation index of the steady-state similarity, which can effectively reflect the strength of the linkage relationship between the local flue gas concentration change and the construction ventilation environment. When the two are highly positively correlated, that is, the steady-state similarity is high, it means that the welding behavior superimposes the poor airflow environment to jointly cause the flue gas retention, and there is substantial risk at the edge node. This determination mechanism not only avoids excessive dependence on a single parameter, but also improves the sensitivity and accuracy of abnormal identification, can more comprehensively capture the combined influence of construction disturbance on the local atmospheric environment, and effectively support high-confidence identification of abnormal nodes.

[0090] Specifically, the process of determining a plurality of accumulation behavior nodes according to the flue gas particle concentration and the welding frequency between each two adjacent abnormal nodes includes: obtaining the number of edge nodes between each two adjacent abnormal nodes, obtaining the number of non-abnormal points; determining a rapid accumulation state index according to the flue gas particle concentration and the welding frequency based on the comparison result of the number of non-abnormal points and the preset point number threshold; determining a plurality of accumulation behavior nodes according to the comparison result of the rapid accumulation state index and the preset accumulation state index threshold.

[0091] The preset point number threshold refers to the minimum edge node number limit for determining whether there are enough non-abnormal nodes between adjacent abnormal nodes for trend identification, which depends on the density of edge nodes in the middle section of the tunnel, the regional scale, and the spatial resolution requirement of flue gas diffusion. It is usually set between 2 and 6, and in this embodiment, it is set to 4, which can ensure the continuity of data distribution while excluding occasional false positives caused by isolated abnormal nodes, thereby improving the stability and robustness of accumulation trend identification.

[0092] The preset accumulation state index threshold refers to the state quantity lower limit for determining whether the edge node has significant flue gas accumulation behavior, which depends on the change range of the particle concentration gradient, the typical disturbance level of the welding frequency, and the desired early warning sensitivity. It is usually set between 0.35 and 0.7, and in this embodiment, it is set to 0.5, which can effectively distinguish between short-term concentration disturbance and persistent accumulation trend, avoid false positives for atypical welding activities or low disturbance gas changes, and enhance the ability to accurately identify high-risk aggregation areas.

[0093] By introducing the number of non-exceptional edge nodes between adjacent exceptional nodes as a structural reference, and combining the smoke particle concentration and welding frequency within the section, the fast accumulation state index is calculated comprehensively, which can dynamically identify whether there is a persistent and centripetal smoke accumulation trend between the exceptional nodes, not only considering the spatial variation of local concentration, but also combining the heat source contribution of welding operation, making the identification of local smoke accumulation behavior more sensitive and accurate, and strengthening the trend analysis ability by using the interaction between multiple data sources, thereby accurately extracting the potential accumulation area and effectively supporting the subsequent deviation evaluation of the accumulation point and the response deployment of the ventilation control strategy.

[0094] Specifically, based on the comparison result of the non-exceptional point number and the preset point number threshold, the process of determining the fast accumulation state index according to the smoke particle concentration and the welding frequency includes: when the non-exceptional point number is greater than the preset point number threshold, determining the particle concentration gradient according to the smoke particle concentration of the two adjacent exceptional nodes corresponding to each time within the preset state determination period, the smoke particle concentration at the edge node closest to the midpoint of the two adjacent exceptional nodes, Δc=︱Cm-(Ca+Cb) / 2︱ / Dm, wherein Δc is the particle concentration gradient, Cm is the smoke particle concentration at the edge node closest to the midpoint of the two adjacent exceptional nodes, Ca and Cb are the smoke particle concentrations of the two adjacent exceptional nodes, and Dm is the distance from the midpoint of the two adjacent exceptional nodes to the closest edge node; when the non-exceptional point number is less than or equal to the preset point number threshold, determining the particle concentration gradient according to the smoke particle concentration of the two adjacent exceptional nodes corresponding to each time within the preset state determination period, Δc=︱Ca-Cb︱ / L, wherein L is the distance between the two adjacent exceptional nodes; determining the fast accumulation state index according to the particle concentration gradient and the welding frequency within the preset state determination period.

[0095] The preset state determination period is a time window for statistical analysis of smoke particle concentration gradient and welding frequency variation, which depends on the dynamic characteristics of smoke diffusion and accumulation in the tunnel construction process and the response speed of the monitoring node, and is usually set between 30 seconds and 5 minutes. In this embodiment, it is set to 1 minute, which can capture the concentration change and welding operation fluctuation in a short time while ensuring sufficient data sampling, and realize sensitive identification and immediate response to fast accumulation state.

[0096] By introducing the concentration difference and spatial distance of non-anomalous nodes near the midpoint when the number of non-anomalous points is large, the lateral accumulation trend of local air masses in the tunnel cross-section is reflected; when the number of non-anomalous points is small, the concentration difference and distance between adjacent anomalous nodes are directly used to quickly capture the longitudinal concentration distribution change, thereby realizing the adaptive quantification of the particle concentration gradient under different node distribution conditions; combined with the welding frequency information, the particle generation rate and spatial diffusion condition are coupled to evaluate, so that the rapid accumulation state index not only reflects the source intensity change, but also takes into account the diffusion obstruction degree, which helps to accurately identify the abnormal accumulation trend of smoke in the complex construction section, and improves the sensitivity and reliability of the early warning judgment.

[0097] Specifically, the process of determining the rapid accumulation state index according to the particle concentration gradient and the welding frequency in the preset state determination period includes: determining the frequency gradient according to the welding frequencies of two adjacent anomalous nodes corresponding to each time in the same preset state determination period, Δf=︱Fa-Fb︱ / L, where Δf is the frequency gradient, Fa and Fb are the welding frequencies of the two adjacent anomalous nodes; performing maximum-minimum value normalization processing on the particle concentration gradient at the end time according to all particle concentration gradients in the preset state determination period to obtain a concentration normalized value, and performing maximum-minimum value normalization processing on the frequency gradient at the end time according to all frequency gradients in the preset state determination period to obtain a frequency normalized value; determining the rapid accumulation state index according to the concentration normalized value and the frequency normalized value, W=p×c’+m×f’, where W is the rapid accumulation state index, p is a preset concentration gradient weight, m is a preset frequency gradient weight, c’ is the concentration normalized value, and f’ is the frequency normalized value.

[0098] The preset concentration gradient weight is a coefficient for measuring the influence degree of particle concentration change on the rapid accumulation state index, which depends on the sensitivity of tunnel construction stage smoke particle diffusion to visibility and air quality, and is usually set between 0.4 and 0.7. In this embodiment, it is set to 0.6, which can increase the index value in advance when the concentration change is small but close to the critical state, thereby realizing early warning of smoke accumulation risk.

[0099] The preset frequency gradient weight is a coefficient for measuring the influence degree of welding frequency change on the rapid accumulation state index, which depends on the correlation between welding operation intensity and smoke generation rate, and is usually set between 0.3 and 0.6. In this embodiment, it is set to 0.4, which can amplify the contribution proportion of frequency change when welding operation is concentrated for a short time and smoke dispersion condition is insufficient, thereby improving the response speed to sudden high-risk accumulation events.

[0100] By introducing the particle concentration gradient and the welding frequency gradient at the same time, the frequency gradient model is established combined with the spatial distance factor, and the concentration gradient and the frequency gradient at the end time are normalized by maximum-minimum value in the preset state determination period, so that the parameters of different dimensions have comparability, and the change trend of another index is avoided to be covered by a single index due to the dimensional difference or the absolute value being too large; then the normalization result is weighted by using the preset concentration gradient weight and the frequency gradient weight, which can accurately reflect the superposition of welding work intensity and smoke diffusion trend on the accumulation rate, so as to identify the hidden danger of rapid smoke accumulation in the early stage, and realize more timely and accurate risk response.

[0101] Please refer to Figure 3 As shown in FIG. 6, which is a determination logic diagram for determining the static wind low behavior node in the embodiment, in the embodiment, the process of determining a plurality of static wind low behavior nodes according to the change trend of the drop plate depth and the wind speed of each arbitrary two adjacent abnormal nodes includes: calculating the standard deviation of the drop plate depth of each arbitrary two adjacent abnormal nodes corresponding to each time from the initial time to each time in the same preset state determination period to determine a plurality of depth gradients; calculating the standard deviation of the wind speed of each arbitrary two adjacent abnormal nodes corresponding to each time from the initial time to each time in the same preset state determination period to determine a plurality of wind speed gradients; maximum-minimum value normalization is performed on all depth gradients to obtain a depth normalized value set, and maximum-minimum value normalization is performed on all wind speed gradients to obtain a wind speed normalized value set, the Pearson correlation coefficient of the depth normalized value set and the wind speed normalized value set is calculated to determine the trend consistency; when the trend consistency is greater than a preset consistency threshold, the abnormal node is determined to be a static wind low behavior node to determine a plurality of static wind low behavior nodes.

[0102] The preset consistency threshold is a numerical limit for determining the synchronization of the change trend of the drop plate depth and the change trend of the wind speed, which depends on the structural difference of the tunnel section, the construction process and the local airflow disturbance intensity, and is usually set between 0.6 and 0.9, and is set to 0.8 in the embodiment, which can effectively reduce the misjudgment caused by accidental fluctuations while ensuring the identification sensitivity, so that the determination of the static wind low behavior node is more stable and reliable.

[0103] By using the standard deviation of the drop plate depth to reflect the concave change amplitude of the tunnel structure and the standard deviation of the wind speed to represent the local airflow stability, and then performing normalization to eliminate the dimensional difference, and then calculating the trend consistency of the two, the accurate identification of the area where the low-lying terrain and airflow slowing down exist at the same time in the construction section is realized; when the synchronization of the two types of characteristics in time change is high, the potential dangerous position prone to static wind retention can be effectively locked, thereby improving the determination accuracy of the static wind low behavior node and providing a more reliable basis for subsequent smoke gathering point positioning and exhaust strategy optimization.

[0104] Please refer to Figure 4 As shown in the figure, it is a determination logic diagram for determining the position deviation of the smoke accumulation point in the embodiment. In the embodiment, the deviation is determined according to the overlap of the accumulation behavior node and the static wind low behavior node, and the process of obtaining the accumulation point deviation result includes: determining the overlap degree according to all the accumulation behavior nodes and all the static wind low behavior nodes, V = (|Sa|∩|Sb|) / min(|Sa|, |Sb|), wherein V is the overlap degree, Sa is the set of all accumulation behavior nodes, and Sb is the set of all static wind low behavior nodes; determining the distribution density according to all the time stamps of the overlap degree greater than the preset overlap degree threshold in the preset deviation determination period, wherein D = n / N, D is the distribution density, n is the number of times that the time stamp of the overlap degree greater than the preset overlap degree threshold appears, and N is the number of times that the comparison overlap degree and the preset overlap degree threshold are compared; determining that the position deviation of the smoke accumulation point occurs when the distribution density is not within the preset distribution density range, and obtaining the accumulation point deviation result.

[0105] The preset overlap degree threshold is a reference ratio for determining whether the accumulation behavior node and the static wind low behavior node have high consistency in spatial distribution. It depends on the airflow organization characteristics of the tunnel section, the node layout density and the historical construction monitoring data, and is usually set between 0.4 and 0.7. In the embodiment, it is set to 0.6, which can effectively balance the false negatives caused by too strict and the false positives caused by too loose, and improve the reliability of the accumulation point position deviation determination.

[0106] The preset distribution density range is used to define the reasonable frequency interval of high overlap state within one deviation determination period. It depends on the operation type, ventilation mode and tunnel section shape at the construction stage, and is usually set between [0.3, 0.6]. In the embodiment, it is set to [0.4, 0.55], which can accurately distinguish between occasional high overlap and persistent high overlap, and ensure that position deviation warning is triggered only when a real accumulation trend is formed, thereby reducing unnecessary intervention actions.

[0107] Through the joint determination method of overlap degree and distribution density, the spatial consistency of smoke accumulation trend and airflow low distribution can be comprehensively evaluated in the time dimension, so as to more accurately identify the real position change of the smoke accumulation point. The calculation of the overlap degree directly reflects the matching degree of the two types of key behavior nodes in space, and the distribution density quantifies the persistence and frequency of this high matching state within one determination period, so that occasional and short-time fluctuations are not misjudged as position deviation. When the distribution density deviates from the preset range, it means that the coupling characteristics of smoke accumulation and static wind low in the construction environment have changed abnormally, which may cause the accumulation point position to migrate to an unexpected area, thereby triggering position correction and risk intervention in advance, effectively reducing false negatives and false positives and improving the relevance of the response.

[0108] Specifically, according to the aggregation point deviation result, the process of adjusting the preset exhaust fan rotating speed or the preset risk threshold at the corresponding abnormal node comprises: when the distribution density is greater than the maximum value of the preset distribution density range, increasing the preset exhaust fan rotating speed according to the relative deviation of the distribution density and the maximum value of the preset distribution density range, R' = R × [1 + k × (D-Dmax) / Dmax], wherein R' is the increased preset exhaust fan rotating speed, R is the preset exhaust fan rotating speed before increasing, k is a preset proportion coefficient, and Dmax is the maximum value of the preset distribution density range; when the distribution density is less than the minimum value of the preset distribution density range, decreasing the preset risk threshold according to the relative deviation of the minimum value of the preset distribution density range and the distribution density, U' = U × [1-k × (Dmin-D) / D], wherein U' is the decreased preset risk threshold, U is the preset risk threshold before decreasing, and Dmin is the minimum value of the preset distribution density range.

[0109] The preset proportion coefficient is a proportional adjustment factor for adjusting the exhaust fan rotating speed or the risk threshold according to the distribution density deviation amplitude, is determined by the response sensitivity of the ventilation system, the fan power margin, and the safety plan requirements of the construction site, and is usually set to be between 0.1 and 0.3, and is set to 0.2 in the embodiment, so that stable adjustment can be realized when the deviation is small, and the risk intervention intensity can be quickly enhanced when the deviation is large, so that the system stability and the emergency response speed are considered.

[0110] By dynamically comparing the distribution density with the upper and lower limits of the preset distribution density range, adaptive optimization of the exhaust fan rotating speed and the risk threshold is realized: when the distribution density is higher than the upper limit of the range, the system increases the exhaust fan rotating speed in proportion to the exceeding amplitude, so that the ventilation intensity and the smoke continuous aggregation trend are positively linked, and the retention effect of the high concentration area is quickly weakened; when the distribution density is lower than the lower limit of the range, the risk threshold is proportionally reduced according to the low degree, so that the sensitivity of the abnormal trigger is improved, and the low-frequency but high-risk smoke aggregation is prevented from being ignored. The method can balance the ventilation energy consumption and the early warning accuracy under different working conditions, so that the control strategy can timely respond to high-risk aggregation and maintain sufficient detection sensitivity at the low-risk stage, and the adaptability and safety level of the system to the dynamic changes of the smoke under complex construction environment are significantly improved.

[0111] The above only describes the preferred embodiments of the present application and is not used to limit the present application; for those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A flue gas early warning method based on edge computing, characterized in that, include: Real-time data acquisition of the drop depth, flue gas particle concentration, welding frequency, and wind speed at each edge node uniformly set in the middle section of the tunnel during nighttime tunnel construction; The static wind accumulation risk index is calculated based on the lowering depth, the wind speed, and the welding frequency. Anomalies are determined based on the flue gas particle concentration, the still wind accumulation risk index, and a preset risk threshold to identify several abnormal nodes. Several accumulation behavior nodes are determined based on the flue gas particle concentration and welding frequency between any two adjacent abnormal nodes, and several calm wind depression behavior nodes are determined based on the changing trend of the drop plate depth and wind speed between any two adjacent abnormal nodes. The deviation is determined based on the overlap between the accumulation behavior nodes and the calm wind low-lying behavior nodes to obtain the accumulation point deviation result; Based on the aggregation point deviation results, adjust the preset exhaust fan speed or the preset risk threshold at the corresponding abnormal node. An early warning is issued for the abnormal node that is obtained again after adjusting the exhaust fan speed or the preset risk threshold.

2. The flue gas early warning method based on edge computing according to claim 1, characterized in that, The process of determining the aggregation point deviation result based on the overlap between the aggregation behavior nodes and the calm wind low-lying behavior nodes includes: The degree of overlap is determined based on all the accumulation behavior nodes and all the calm wind depression behavior nodes; The distribution density is determined based on all timestamps within a preset deviation determination period where the overlap is greater than a preset overlap threshold. Based on the comparison between the distribution density and the preset distribution density range, the deviation of the flue gas accumulation point is determined, and the accumulation point deviation result is obtained.

3. The flue gas early warning method based on edge computing according to claim 2, characterized in that, The process of adjusting the preset exhaust fan speed or the preset risk threshold at the corresponding abnormal node based on the aggregation point deviation result includes: When the distribution density is greater than the maximum value of the preset distribution density range, the preset exhaust fan speed is increased according to the relative deviation between the distribution density and the maximum value of the preset distribution density range; When the distribution density is less than the minimum value of the preset distribution density range, the preset risk threshold is reduced based on the minimum value of the preset distribution density range and the relative deviation of the distribution density.

4. The flue gas early warning method based on edge computing according to claim 3, characterized in that, The process of determining several accumulation behavior nodes based on the flue gas particle concentration and the welding frequency between any two adjacent anomalous nodes includes: Obtain the number of edge nodes between any two adjacent abnormal nodes to get the number of non-abnormal points; Based on the comparison results between the number of non-abnormal points and the preset point threshold, the rapid accumulation state index is determined according to the flue gas particle concentration and the welding frequency. Several accumulation behavior nodes are determined based on the comparison between the rapid accumulation state index and the preset accumulation state index threshold.

5. The flue gas early warning method based on edge computing according to claim 4, characterized in that, Based on the comparison between the number of non-abnormal points and the preset point threshold, the process of determining the rapid accumulation state index according to the flue gas particle concentration and the welding frequency includes: When the number of non-abnormal points is greater than the preset number of points threshold, the particle concentration gradient is determined by determining the concentration of flue gas particles at the two adjacent abnormal nodes corresponding to each time point within the period and the concentration of flue gas particles at the edge node closest to the midpoint of the two adjacent abnormal nodes, based on the preset state. When the number of non-abnormal points is less than or equal to the preset number of points threshold, the particle concentration gradient is determined by the concentration of flue gas particles at two adjacent abnormal nodes corresponding to each moment within the period based on the preset state. The rapid accumulation state index is determined based on the particle concentration gradient and welding frequency within the predetermined state during the cycle.

6. The flue gas early warning method based on edge computing according to claim 5, characterized in that, The process of determining the rapid accumulation state index based on the particle concentration gradient and welding frequency within the preset state includes: The frequency gradient is determined based on the welding frequencies of two adjacent abnormal nodes at different times within the same preset state. The rapid accumulation state index is determined based on the particle concentration gradient and the frequency gradient.

7. The flue gas early warning method based on edge computing according to claim 6, characterized in that, The process of determining several calm wind depression behavior nodes based on the changing trend of the drop depth and wind speed between any two adjacent abnormal nodes includes: Based on the same preset state, determine several depth gradients based on the drop depth of any two adjacent abnormal nodes within the same preset state period, from the initial time to each time. Based on the same preset state, determine several wind speed gradients for the wind speeds of any two adjacent abnormal nodes within a period from the initial time to each time. The degree of trend consistency is determined based on all the depth gradients and all the wind speed gradients. Several calm wind low-lying behavior nodes are determined based on the comparison results between the trend consistency and the preset consistency threshold.

8. The flue gas early warning method based on edge computing according to claim 7, characterized in that, The process of calculating the still wind accumulation risk index based on the lowering depth, the wind speed, and the welding frequency includes: The still wind accumulation risk index is determined based on the ratio of the lowering plate depth to the wind speed, the preset retention weight, the welding frequency, and the preset welding frequency weight.

9. The flue gas early warning method based on edge computing according to claim 8, characterized in that, The process of determining several abnormal nodes by analyzing the flue gas particle concentration, the still wind accumulation risk index, and a preset risk threshold includes: Based on the concentration of flue gas particles from the initial time to each time within the preset anomaly determination period, several flue gas steady-state values ​​are determined. Based on the static wind accumulation risk index from the initial time to each time within the preset anomaly determination period, a number of steady-state values ​​of risk index are determined. Anomalies are determined based on all the steady-state values ​​of the flue gas, all the steady-state values ​​of the risk index, and the preset risk threshold, in order to identify several abnormal nodes.

10. The flue gas early warning method based on edge computing according to claim 9, characterized in that, The process of determining several abnormal nodes by analyzing all the steady-state values ​​of the flue gas, all the steady-state values ​​of the risk index, and the preset risk threshold includes: Steady-state similarity is determined based on all the flue gas steady-state values ​​and all the risk index steady-state values; The edge node is determined to be abnormal based on the comparison result between the steady-state similarity and the preset risk threshold, thereby identifying a number of abnormal nodes.

Images