Signal alarm system for gas in tunnel
By constructing a vertical concentration gradient model library and conducting time-series analysis, the problem of full vertical height coverage in traditional tunnel gas monitoring systems has been solved, enabling accurate monitoring and efficient early warning of gases within tunnels and improving the risk management capabilities for tunnel safety.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-30
- Publication Date
- 2026-03-31
AI Technical Summary
Traditional gas monitoring systems in tunnels cannot achieve concentration coverage across the entire vertical height, resulting in the failure to detect leaks of light gases at the top and the failure to promptly alarm for deposits of heavy gases at the bottom, creating blind spots in safety protection and making it impossible to accurately determine the gas risk status.
A vertical concentration gradient model library coupling gas type, vertical altitude, and environmental parameters is constructed. Multi-altitude data is collected by drones and mobile robots, and real-time corrected concentration values at all altitudes are output. Combined with time series analysis and early warning time prediction modules, high-risk growth trends are identified and early warning times are predicted.
It achieves full-height gas concentration coverage within the tunnel, reducing over-warning and under-warning, improving risk management efficiency and safety, adapting to environmental changes in different tunnel working conditions, and ensuring monitoring accuracy and early warning reliability.
Smart Images
Figure CN121768152A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of signal alarm technology, specifically a signal alarm system for gas in tunnels. Background Technology
[0002] The tunnel gas alarm system is a safety monitoring system specifically designed to monitor the concentration of harmful gases or oxygen content in tunnels and trigger an alarm when the concentration exceeds the standard. Its core function is to ensure the safety of personnel (such as maintenance personnel and passers-by) and equipment in the tunnel in real time and prevent accidents such as poisoning and explosions caused by gas leaks, vehicle emissions or geological factors.
[0003] However, long tunnels are prone to vertical stratification due to airflow resistance and temperature gradients: light gases (methane, hydrogen) accumulate at the tunnel top (4-6m above the ground), while heavy gases (CO, NO2, hydrogen sulfide, slightly denser than air) are deposited in the middle and lower parts (1-2m above the ground). Traditional fixed sensors are mostly placed at eye level (1.5-2m), which can lead to missed leaks of light gases at the top and delayed alarms due to gases at the bottom not diffusing to the sensors (such as CO leaks from the bottom of a tool cart during maintenance). They cannot fully cover the vertical stratification of light gases at the top and heavy gases at the bottom within the tunnel based on the concentration data collected by fixed sensors. This is particularly problematic for light gases like methane and hydrogen. The leakage of gases at the top of the tunnel (4-6m) cannot be detected by sensors at a height of 1.5-2m, leading to missed explosion risks. For heavy gases such as CO and hydrogen sulfide, if the leakage source is located in low-height areas such as the bottom of a tool vehicle or ground cracks, the gas needs to diffuse for a long time to reach "eye level" before it can be detected, causing a delay in alarms for potential poisoning. Furthermore, alarms need to be transmitted via mobile data communication services. This limitation of single-point fixed-height monitoring makes it impossible for traditional systems to form concentration monitoring coverage across the entire vertical height, thus making it impossible to accurately determine the gas risk status at different heights. Ultimately, this results in blind spots in safety protection within the tunnel, making it difficult to effectively ensure the safety of personnel and equipment.
[0004] Therefore, the present invention provides a signal alarm system for gas inside a tunnel. Summary of the Invention
[0005] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.
[0006] The technical solution adopted by this invention to solve its technical problem is: a signal alarm system for gas in a tunnel, the system specifically comprising:
[0007] Model library construction module: By acquiring historical concentration data of various types of gases in the tunnel, synchronously linking working condition information of different vertical heights and different environmental parameters, and based on data correlation analysis, constructing a vertical concentration gradient model library that couples gas type, vertical height, and environmental parameters;
[0008] Corrected concentration output module: Real-time acquisition of gas concentration data and environmental parameters, matching of vertical concentration gradient model, output of corrected concentration values at different heights, and extraction of corrected concentration values within a preset key height range to obtain a sequence of corrected concentration values.
[0009] Time series analysis module: acquires the corrected concentration value sequence at different collection time points, performs time series trend analysis on the corrected concentration values corresponding to the same vertical height, and extracts the vertical height value corresponding to the high-risk growth trend as the vertical height to be warned;
[0010] Early warning time prediction module: Based on the growth and change pattern of the vertical height to be warned, the module predicts the warning time when the corrected concentration at the vertical height to be warned reaches the alarm value through trend fitting and alarm value extrapolation, providing a time basis for subsequent early warning intervention.
[0011] The beneficial effects of this invention are as follows:
[0012] This invention constructs a vertical concentration gradient model library that couples gas type, vertical height, and environmental parameters. Taking into account the characteristics of light gases being enriched at the top and heavy gases being deposited at the bottom, it divides the vertical height bidirectionally with a reference height as the core. It combines the collection of historical data from multiple heights by UAVs and mobile robots, and then outputs the full-height corrected concentration through real-time model matching, thereby improving the full-height gas concentration coverage monitoring of the tunnel's vertical space.
[0013] This invention uses a dual logic of determining continuous growth and identifying key growth phases through the proportion of growth windows to distinguish different growth types. For continuous growth scenarios, it ensures the reliability of long-term trend judgment by emphasizing stable growth intensity. For phased growth scenarios, it focuses on extracting key growth rates during effective growth periods to reduce the interference of fluctuations during non-risk periods. At the same time, it combines growth intensity verification to filter out low-growth invalid growth, which reduces the waste of resources due to excessive warnings and also reduces the safety risks of missing short-term concentrated risks, making the extraction of the vertical height of warning more in line with actual risks.
[0014] This invention uses a linear extrapolation of the warning time based on the stable growth intensity in a continuously growing scenario, and a conservative prediction based on the growth rate of the fastest period in a phased key growth scenario. It outputs the time when the corrected concentration reaches the alarm value, providing clear guidance for emergency intervention. This allows tunnel gas safety management to shift from passive alarm to proactive prediction, significantly improving the efficiency and safety of risk handling.
[0015] This invention uses Pearson correlation coefficient to screen strongly correlated environmental parameters of gas during the model building stage, and then uses cluster analysis to classify working conditions, so that the model can adapt to changes in the tunnel environment. During real-time monitoring, the standardization process is used to match the distance to the cluster center, accurately locate the current working condition and call the corresponding model, reducing the concentration calculation deviation caused by environmental fluctuations. Thus, it can maintain stable monitoring accuracy and early warning reliability under different operating scenarios, and is suitable for gas safety management needs of various long tunnels and complex working conditions. Attached Figure Description
[0016] The invention will now be further described with reference to the accompanying drawings.
[0017] Figure 1 This is an architectural diagram of a signal alarm system for gas inside a tunnel according to the present invention;
[0018] Figure 2 This is a flowchart of the steps of a signal alarm method for gas in a tunnel according to the present invention;
[0019] Figure 3 This is a flowchart of some steps in the timing analysis module of a signal alarm system for gas inside a tunnel according to the present invention. Detailed Implementation
[0020] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.
[0021] Example 1
[0022] Please see Figure 1 and Figure 3 As shown in the embodiment of the present invention, a signal alarm system for gas in a tunnel includes:
[0023] Model library construction module: By acquiring historical concentration data of various types of gases in the tunnel, synchronously linking working condition information of different vertical heights and different environmental parameters, and based on data correlation analysis, constructing a vertical concentration gradient model library that couples gas type, vertical height, and environmental parameters;
[0024] In this module, firstly, for various types of gases within the tunnel, those skilled in the art will screen out the gases requiring key monitoring based on the tunnel's risk scenarios, including but not limited to:
[0025] Light gases: gases with a density less than that of air, such as methane and hydrogen; Heavy gases: gases with a density greater than that of air, such as CO, NO2, and hydrogen sulfide.
[0026] The preset height step is based on the fixed sensor installation position as the reference height, and the vertical height is divided into several steps according to the tunnel height.
[0027] For example, if the tunnel height is 6m (from 0m on the ground to 6m on the arch), the installation height of the fixed sensor is 1.8m (reference height), and the preset height step is 0.6m (the step selection is based on the need to cover the bottom heavy gas deposition area → the middle monitoring area → the top light gas enrichment area, and the 0.6m step can balance accuracy and data volume). The vertical height is divided in both directions: upward (arch direction) and downward (ground direction).
[0028] The downward division (surface side, covering heavy gas deposition area): starting from the reference height of 1.8m, decrease downwards in steps of 0.6m until approaching the ground, resulting in the following vertical heights in the downward direction: 0.3m, 0.6m, 1.2m;
[0029] Divide upwards (on the arch side, covering the light gas enrichment zone): Starting from the reference height of 1.8m, increase upwards in steps of 0.6m until approaching the arch. The vertical heights in the upward direction are: 2.4m, 3.0m, 3.6m, 4.2m.
[0030] For each type of gas, historical concentration values corresponding to different vertical heights are obtained, and historical environmental parameter data are obtained simultaneously.
[0031] The historical concentration values corresponding to different vertical heights were obtained as follows: the historical concentration values at the reference height were obtained by monitoring with fixed sensors, while the historical concentration values at other heights (other than the reference height) were collected by a combination of drones and tracked mobile robots.
[0032] For example, a drone equipped with sensors hovers at a preset height to collect samples; a mobile robot equipped with the same sensors collects samples longitudinally along the tunnel.
[0033] Environmental parameter data includes, but is not limited to: temperature data, wind speed data, humidity data, and air pressure data.
[0034] Based on any combination of environmental parameter and any gas, the correlation between the environmental parameter and the concentration gradient is analyzed by calculating the Pearson correlation coefficient. The process is as follows:
[0035] Based on the density characteristics of the gases, the gases are divided into light gas types and heavy gas types, and the historical concentration gradient is calculated based on the type.
[0036] For light gases: calculate the difference between the historical concentration value at the highest vertical height at the top and the historical concentration value at the reference height, and use it as the historical concentration gradient;
[0037] For heavy gases: calculate the difference between the historical concentration value at the lowest vertical height at the bottom and the historical concentration value at the reference height, and use it as the historical concentration gradient;
[0038] Obtain the historical environmental parameter values and historical concentration gradient values corresponding to the same sampling point at different timestamps to obtain the corresponding historical environmental parameter value sequence and historical concentration gradient value sequence;
[0039] Calculate the Pearson correlation coefficient between the historical environmental parameter value series and the historical concentration gradient value series, and take the absolute value as the correlation coefficient;
[0040] Combinations with correlation coefficients greater than a correlation coefficient threshold are identified as strongly correlated combinations.
[0041] Combinations with correlation coefficients less than or equal to a correlation coefficient threshold are identified as weakly correlated combinations.
[0042] Among them, the correlation coefficient threshold is used to quantify the correlation strength between environmental parameters and gas concentration gradient. In essence, it is to distinguish the threshold between strongly correlated parameters that have a significant impact on the concentration gradient and weakly correlated parameters that have a weak correlation. This helps to screen key variables for model input and reduce noise interference.
[0043] Obtain all strongly correlated combinations and organize them. For any gas, extract the environmental parameters from all corresponding strongly correlated combinations to obtain strongly correlated environmental parameter combinations.
[0044] Obtain the combination of strongly correlated environmental parameters for each gas.
[0045] Based on any single gas, cluster analysis is performed on the numerical values corresponding to strongly correlated environmental parameter combinations to obtain multiple operating condition categories. The process is as follows:
[0046] All historical environmental parameter values in strongly correlated environmental parameter combinations are standardized to obtain standardized historical environmental parameter value data. Min-Max normalization is then used to normalize the parameter values to the [0,1] interval.
[0047] The optimal number of clusters K is determined by combining the elbow method with the profile coefficient method.
[0048] Specifically, the range of values for K is preset (e.g., K=2,3,4,5,6); for each K value, the K-means algorithm is used to perform clustering, and the sum of squared errors (SSE) (i.e., the sum of squared distances from each sample to its cluster center) is calculated.
[0049] Plot the SSE-K value curve. The K value corresponding to the elbow (the point where the rate of SSE decreases significantly) in the curve is a candidate value (e.g., when K=3, SSE drops sharply from 120 at K=2 to 80, and only drops to 75 at K=4, with the elbow being K=3).
[0050] Calculate the silhouette coefficient for candidate K values, and select the K value with the highest silhouette coefficient as the final number of clusters;
[0051] Cluster analysis is performed based on the K-means algorithm;
[0052] Specifically, K samples are randomly selected from the standardized parameter values as initial cluster centers. For example, when K=3, three sets of standardized values of [wind speed, air pressure] are selected as centers.
[0053] Sample assignment: Calculate the Euclidean distance between each sample and the K cluster centers, and assign the sample to the category of the nearest cluster center;
[0054] Update centers: Based on the assigned samples, recalculate the mean of each category (i.e., the new cluster centers).
[0055] Repeat the sample allocation and center update steps until the change in cluster centers is less than a preset threshold (e.g., change < 0.001) or the maximum number of iterations (e.g., 500 times) is reached, then stop the iteration and obtain multiple cluster centers;
[0056] For each cluster category, the mean and range of the raw (unstandardized) values of the strongly correlated environmental parameters are calculated:
[0057] The cluster categories correspond to the operating condition categories;
[0058] For example, Category 1: Low wind speed-high pressure condition; Category 2: Medium wind speed-normal pressure condition.
[0059] The process of constructing a vertical concentration gradient model library that couples gas type, vertical height, and environmental parameters is as follows:
[0060] Based on gas type, vertical height, and operating condition category, the preprocessed historical data (including historical concentration values and historical strongly correlated environmental parameter values at all vertical heights) is split into independent data subsets.
[0061] For example, the first subset is: gas: methane (light gas), vertical height: 4.2m (maximum height at the top), and operating condition category: low wind speed - high pressure;
[0062] Ensemble learning algorithms such as random forest are used to construct vertical concentration gradient models for different gases, different operating conditions, and different vertical heights.
[0063] The historical concentration values at the reference height and the historical environmental parameter values with strong correlation to the working condition category are used as model inputs, and the concentration values corresponding to different vertical heights are used as model outputs.
[0064] For example, the input variables for the methane-4.2m-low wind speed-high pressure subset are: base height (1.8m) methane concentration, wind speed, and air pressure (both of which are strongly correlated environmental parameters of methane); the outputs correspond to different vertical height concentration values.
[0065] Each data subset is divided into a training set and a validation set in a 7:3 ratio. The training set is used to train the model, and the validation set is used to validate the model.
[0066] For each data subset, the hyperparameters of the random forest algorithm are optimized using a combination of grid search and 5-fold cross-validation.
[0067] Specifically, the hyperparameters to be optimized include: number of decision trees (range 100-500 trees, step size 50), maximum tree depth (range 5-15 layers, step size 2), minimum number of sample splits (range 2-10, step size 1), and minimum number of sample leaf nodes (range 1-5, step size 1).
[0068] The optimal hyperparameter combination for each subset is selected with the minimum mean absolute error (MAE) of the validation set as the optimization objective.
[0069] Input the training set into the random forest model with the optimal hyperparameters and start iterative training:
[0070] The maximum number of iterations is set to 500. After each iteration, the goodness of fit (R²) of the training set and the MAE of the validation set are calculated. When the change in MAE of the validation set is <0.00 in 10 consecutive iterations, or when the maximum number of iterations is reached, the model training is considered to have converged, the iteration is stopped, and the current model parameters are saved.
[0071] Calculate the error index between the model-predicted concentration values and the measured concentration values in the validation set. It is required that the prediction error of more than 90% of the samples is <10%, and the overall MAE and root mean square error (RMSE) meet the preset accuracy threshold.
[0072] All validated models are stored in a three-level directory structure of gas type → operating condition category → vertical height to build a vertical concentration gradient model library;
[0073] The primary category is divided by gas type;
[0074] The secondary categories are divided into operating conditions under each gas type;
[0075] The three-level directory is divided according to vertical height under each working condition category.
[0076] This module constructs a vertical concentration gradient model library, which allows for the synchronous output of gas concentration values at different vertical heights when collecting gas concentration data and environmental parameters in real time. This reduces the limitations of fixed sensors monitoring gas concentration at fixed heights, and the impact of environmental fluctuations can be reduced through cluster analysis of operating conditions.
[0077] Corrected concentration output module: Real-time acquisition of gas concentration data and environmental parameters, matching the corresponding vertical concentration gradient model, outputting corrected concentration values at different heights, and extracting corrected concentration values within a preset height range to obtain a sequence of corrected concentration values;
[0078] In this module, the real-time concentration value of the gas is collected (obtained by a fixed sensor), and the corresponding real-time environmental parameters are collected simultaneously.
[0079] For any gas, extract the corresponding real-time strongly correlated environmental parameter values and perform standardization processing (using the Min-Max normalization method consistent with the model library construction stage) to obtain standardized real-time strongly correlated environmental parameter values.
[0080] Calculate the standardized real-time strongly correlated environmental parameter values, and then calculate the Euclidean distance between the real-time strongly correlated environmental parameter values and the cluster centers of all operating condition categories corresponding to the gas (corresponding to the cluster centers in the aforementioned model library construction module). Select the operating condition to which the cluster center with the smallest Euclidean distance belongs as the current real-time operating condition.
[0081] Based on the gas type matching of the first-level directory in the vertical concentration gradient model library, and based on the current real-time operating conditions matching of the second-level directory in the vertical concentration gradient model library;
[0082] Call the vertical concentration gradient models for all corresponding vertical heights in the second-level directory;
[0083] For any given vertical height, the real-time concentration value and the real-time strongly correlated environmental parameter value are used as inputs to the vertical concentration gradient model, and the corresponding corrected concentration value is output.
[0084] Sort all vertical height correction concentration values from low to high to form a table showing the correspondence between vertical height and correction concentration, thus completing full multi-height concentration coverage;
[0085] Based on the tunnel safety monitoring requirements (personnel activity, equipment distribution), preset height ranges are established (such as "personnel maintenance height range 0.3-1.2m", "top equipment safety range 3.6-4.2m", "middle passageer breathing height range 1.5-2.4m"). The preset height range range is set by those skilled in the art in combination with the tunnel scenario.
[0086] Based on a preset height range, the correction concentration values for all heights within the preset height range are extracted from the correspondence table between vertical height and correction concentration to obtain a sequence of correction concentration values.
[0087] It should be noted that the reason for extracting the corrected concentration corresponding to the preset height range is as follows: the vertical height coverage of the tunnel is wide, but the safety risks and monitoring priorities of different heights are significantly different. For example, personnel maintenance is mostly active at a height of 0.3-1.2m, the breathing of passers-by is concentrated at a height of 1.5-2.4m, and the enrichment of light gases at a height of 3.6-4.2m is the focus for top equipment (such as cables). Extracting the corrected concentration of the preset height range can directly filter out height data with no personnel / equipment distribution and extremely low risk, reducing the computational power occupied by redundant data. At the same time, it allows subsequent trend analysis and early warning judgment to focus on the core areas that truly affect personnel safety and equipment stability, thereby improving the system response efficiency.
[0088] This module outputs corrected concentration values at multiple vertical heights and extracts corrected concentration values within a preset height range. This allows it to filter out height data with no personnel / equipment distribution and extremely low risk, reducing redundant data consumption and enabling subsequent time-series analysis modules to focus on core areas that truly affect personnel safety and equipment stability. This reduces interference from non-critical data and improves the system's response efficiency for trend analysis and early warning judgment of high-risk areas.
[0089] Time series analysis module: acquires the corrected concentration value sequence at different collection time points, performs time series trend analysis on the corrected concentration values corresponding to the same vertical height, and extracts the vertical height value corresponding to the high-risk growth trend as the vertical height to be warned;
[0090] In this module, a preset acquisition period is provided, which is set by those skilled in the art in combination with the characteristics of tunnel gas concentration changes.
[0091] For each vertical height within the preset height range, obtain the corrected concentration values at all collection time points within the collection period, and sort them in ascending order according to the timestamp to obtain the time-series concentration sequence;
[0092] For any time-series concentration sequence, two adjacent acquisition time points are used as a calculation window. The difference between adjacent corrected concentration values within the calculation window is calculated to obtain the concentration deviation.
[0093] Extract the calculation window with a concentration deviation greater than zero as the growth window, and count the number of growth windows.
[0094] The ratio of the number of growth windows to the total number of calculation windows is used to obtain the proportion of growth windows;
[0095] If the proportion of the growth window is greater than the persistence threshold, the time series concentration sequence is determined to have a continuous growth trend.
[0096] If the growth window percentage is less than or equal to the sustainability threshold, then further identification is needed to determine whether there is a key growth trend in a given phase.
[0097] Among them, the stability of the growth trend of the time-series concentration sequence is quantified by the persistence threshold. Essentially, it is a threshold that distinguishes between a continuous growth trend with a stable upward pattern and a non-continuous growth caused by random fluctuations, thus providing a basis for the preliminary judgment of risk trends.
[0098] Traverse all calculation windows, treat a single existing growth window as an isolated growth window, and integrate consecutive adjacent growth windows as a growth period;
[0099] The number of computation windows within each growth period is used as the number of consecutive windows.
[0100] If the number of consecutive windows is greater than the minimum consecutive window threshold, it is determined to be an effective growth period;
[0101] If the number of consecutive windows is less than or equal to the minimum consecutive window threshold, it is determined to be an invalid growth period;
[0102] Among them, the minimum continuous window threshold quantifies the risk effectiveness and intervention feasibility of the growth period. Essentially, it distinguishes between the threshold of the continuous growth period that corresponds to real sudden risks and can reserve time for operation and maintenance response and the threshold of isolated interference windows caused by random fluctuations. This avoids missing short-term concentrated risks and ensures the time matching between early warning and operation and maintenance handling.
[0103] If there is at least one effective growth period, the time series concentration sequence is determined to have a stage-specific key growth trend;
[0104] If there is no effective growth period, then the time series concentration sequence is determined to lack a key growth trend in stages;
[0105] It needs to be explained that the role of identifying the existence of a key growth trend in a phase is as follows: First, it reduces the risk of missing short-term concentrated growth. Such short-term concentrated growth may correspond to sudden risks (such as short-term equipment leakage or temporary vehicle congestion). If it is not identified, it will be missed because the overall proportion does not meet the standard. Identifying key growth trends in a phase can accurately capture such risks.
[0106] Secondly, it balances the sensitivity and stability of early warning. If the growth window ratio is used as the only criterion, local high-risk growth may be ignored because the overall ratio does not meet the standard. The identification of key growth trends in a phase is supplemented by the number of consecutive windows. This reduces the rigidity of a single ratio standard and filters out the interference of isolated growth windows through the minimum consecutive window threshold (ensuring the stability of early warning). A balance is found between not issuing excessive warnings and not omitting risks.
[0107] Third, it provides a more refined basis for graded early warning. A continuous growth trend usually corresponds to long-term, stable risks (such as chronic leakage), while a phased key growth trend corresponds to short-term, sudden risks (such as a sudden increase in concentration caused by temporary operations). By distinguishing between these two trends, a long-term monitoring and gradual intervention strategy can be adopted for the former, and an immediate early warning and rapid response strategy can be adopted for the latter, thereby improving the pertinence and effectiveness of the early warning system.
[0108] For time-series concentration sequences that are determined to be either a sustained growth trend or a key phase of growth trend, calculate the growth intensity;
[0109] The concentration deviation within the growth window is extracted and averaged to obtain the growth intensity;
[0110] If the growth intensity is greater than the minimum effective growth rate threshold, then mark the corresponding vertical height as a high-risk growth trend;
[0111] If the growth intensity is less than or equal to the minimum effective growth rate threshold, even if the growth window ratio meets the standard, but the growth rate is slow, the corresponding vertical height is marked as low-risk slow growth.
[0112] Traverse all vertical heights within the preset height range, extract the vertical heights marked as high-risk growth trends, and use them as the vertical heights to be warned.
[0113] Among them, the minimum effective growth rate threshold quantifies the urgency of the risk of gas concentration growth. It is a threshold that distinguishes between high-risk growth rates that may reach the alarm value in the short term and low-risk growth rates that will not exceed the standard in the long term. The core is to combine the safety red line of the gas alarm threshold with the hazard rate of toxic diffusion to provide a quantitative basis for risk intervention priority classification.
[0114] This module uses a dual logic of determining continuous growth based on the proportion of growth windows and identifying key growth periods based on effective growth periods. It can distinguish different types of concentration growth risks. By determining the growth trend and filtering the growth intensity, it extracts only the vertical height corresponding to the high-risk growth trend as the vertical height to be warned. This allows the subsequent warning time prediction module to focus on the core risk height, reduce interference from non-critical height data, and direct warning resources to key areas that affect personnel safety and equipment stability, thereby improving the targeting of system risk management.
[0115] Early warning time prediction module: Based on the growth and change pattern of the vertical height to be warned, through trend fitting and alarm value deduction, predicts the warning time when the corrected concentration at the vertical height to be warned reaches the alarm value, providing a time basis for subsequent early warning intervention;
[0116] In this module, the vertical height to be warned and the corresponding time-series concentration sequence are obtained;
[0117] For any given vertical height to be warned, the time-series concentration sequence
[0118] If the time-series concentration sequence shows a continuous increasing trend, obtain the last corrected concentration value of the time-series concentration sequence;
[0119] The difference between the alarm value and the corrected concentration value is calculated to obtain the concentration value that still needs to be increased;
[0120] The ratio of the required concentration increase to the increase intensity is used to calculate the number of calculation windows.
[0121] Since the calculation window is the interval between two adjacent data collection time points, the warning time is obtained by multiplying the number of calculation windows with the interval between the two data collection time points.
[0122] It should be noted that if, during the aforementioned process, the corrected concentration value exceeds the alarm value, an immediate warning will be issued.
[0123] If the time-series concentration sequence shows a key growth trend in stages, the calculation should focus on the effective growth period;
[0124] Extract all effective growth periods from the time-series concentration sequence. For each effective growth period, calculate the average concentration deviation corresponding to all calculation windows within the growth period as the period growth rate. Extract the maximum value of the period growth rate as the critical growth rate.
[0125] The ratio of the concentration value that still needs to increase to the critical growth rate is calculated to obtain the number of calculation windows. The number of calculation windows is then multiplied by the interval between the two collection time points to obtain the warning time.
[0126] It needs to be explained that for a phased key growth trend, the essence is to match the core characteristics of phased key growth, which is that the growth is discontinuous and the real risk only exists within the effective growth period. Using the fastest growth rate to calculate the warning time, we can obtain the earliest time when the concentration reaches the alarm value (i.e., the most conservative prediction result), ensuring that intervention is not delayed due to underestimating the growth rate. Furthermore, phased key growth may have a situation where the growth rate gradually accelerates (such as an increase in leakage). Taking the maximum growth rate during the period can cover this acceleration risk. Even if the subsequent growth rate does not reach the maximum value, the predicted warning time can still leave enough room for intervention.
[0127] By predicting the specific time when the corrected concentration reaches the alarm value (early warning time), a clear time reference is provided for subsequent interventions, including but not limited to:
[0128] If there is sufficient warning time (such as 1 hour later), personnel can be arranged to investigate the source on-site and start the ventilation system in advance for pre-adjustment;
[0129] If the warning time is urgent (e.g., within 10 minutes), an emergency response can be triggered immediately, such as cutting off the power supply to the risk area and guiding personnel to evacuate.
[0130] If the current level has been exceeded, immediate intervention will be initiated (such as full ventilation and audible and visual alarms).
[0131] The gas signal alarm system in this embodiment consists of four main modules. The model library construction module collects historical concentration data of various types of gases at different vertical heights, combines this data with environmental parameters such as temperature and wind speed, and uses Pearson correlation coefficient analysis and K-means clustering to construct a vertical concentration gradient model library coupled with "gas type-vertical height-operating condition category." The corrected concentration output module collects data in real time and matches it to the model, outputting corrected concentration values for the entire height and extracting the height interval sequence. The time series analysis module performs trend analysis on the concentration sequences at key heights, identifying continuous growth and phased key growth trends, and extracting high-risk vertical heights requiring warning. The warning time prediction module, based on different growth trends, calculates the time when the concentration reaches the alarm value using concentration difference and growth intensity, providing a basis for emergency intervention.
[0132] By upgrading tunnel gas monitoring from a single-point, fixed-height model output to multi-vertical-height monitoring, it can capture the explosion risk of light gas leaks at the top and quickly identify the poisoning hazards of heavy gas deposits at the bottom, reducing blind spots in safety protection. Through differentiated early warning time prediction logic, it provides clear time basis for emergency intervention and can flexibly initiate investigation, evacuation and other measures according to the urgency of the early warning, including but not limited to using other telecommunications services such as mobile voice service and mobile data communication service for early warning notification, reducing resource mismatch. In addition, the structured model library design and environmental adaptability ensure that the system can operate stably under different tunnel conditions, and is suitable for gas safety management in long tunnels and complex operating scenarios, effectively protecting the safety of personnel and equipment.
[0133] Example 2
[0134] Based on the same inventive concept as the signal alarm system for gas inside a tunnel in the foregoing embodiments, such as Figure 2 As shown, this application provides a signal alarm method for gas in a tunnel, which includes the following steps:
[0135] Step S10: By acquiring historical concentration data of various types of gases in the tunnel, synchronously linking working condition information of different vertical heights and different environmental parameters, and based on data correlation analysis, constructing a vertical concentration gradient model library that couples gas type, vertical height, and environmental parameters.
[0136] Step S20: Collect gas concentration data and environmental parameters in real time, match the corresponding vertical concentration gradient model, output the corrected concentration values at different heights, and extract the corrected concentration values within the preset height range to obtain the corrected concentration value sequence.
[0137] Step S30: Obtain the sequence of corrected concentration values at different collection time points, perform time-series trend analysis on the corrected concentration values corresponding to the same vertical height, and extract the vertical height values corresponding to the increasing trend as the vertical height to be warned.
[0138] Step S40: Based on the growth pattern of the vertical height to be warned, predict the time point when the corrected concentration at the vertical height to be warned reaches the alarm value through trend fitting and alarm value deduction, so as to provide a time basis for subsequent early warning intervention.
[0139] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.
Claims
1. A signal warning system for gases in a tunnel, characterized by: The system specifically comprises: A model library construction module: by acquiring historical concentration data of multiple types of gases in the tunnel, synchronously associating working condition information of different vertical heights and different environmental parameters, and based on data association analysis, a vertical concentration gradient model library coupled with gas type, vertical height and environmental parameter is constructed; A corrected concentration output module: real-time acquisition of concentration data and environmental parameters of the gas, matching of the vertical concentration gradient model, output of corrected concentration values at different heights, and interception of corrected concentration values in a preset height interval to obtain a corrected concentration value sequence; A time series analysis module: acquisition of corrected concentration value sequences at different acquisition time points, time series trend analysis of corrected concentration values corresponding to the same vertical height, extraction of vertical height values corresponding to high-risk growth trends as vertical heights to be warned; A warning time prediction module: according to the growth change law of the vertical height to be warned, through trend fitting and alarm value deduction, the warning time for the corrected concentration to reach the alarm value at the vertical height to be warned is predicted, providing a time basis for subsequent early warning intervention.
2. A signal warning system for gases in a tunnel according to claim 1, characterized in that: The acquisition process of the vertical concentration gradient model library is as follows: Based on data association analysis of environmental parameters and gases, the strong correlation environmental parameter combination corresponding to each gas is obtained, and through clustering analysis of the strong correlation environmental parameter combination of each gas, multiple working condition categories are obtained; The historical data containing historical concentration values of all vertical heights and historical strong correlation environmental parameter values are split into independent data subsets in three dimensions of gas type, vertical height and working condition category; A random forest algorithm is used to construct a vertical concentration gradient model, and the historical concentration values at the reference height and the historical strong correlation environmental parameter values of the working condition category are used as the model input, and the concentration values corresponding to different vertical heights are used as the model output; Each data subset is divided into a training set and a validation set, the training set is used for model training, and the validation set is used for model validation; All validated models are stored in a three-level directory structure of gas type→working condition category→vertical height, and a vertical concentration gradient model library is constructed.
3. A signal warning system for gases in a tunnel according to claim 1, characterized in that: The acquisition process of the historical strong correlation environmental parameter value is as follows: The gases are divided into light gas types and heavy gas types, and the historical concentration gradient is calculated based on the types; The historical environmental parameter value sequence and the historical concentration gradient value sequence corresponding to different time stamps at the same sampling point are acquired; The Pearson correlation coefficient of the historical environmental parameter value sequence and the historical concentration gradient value sequence is calculated, and the absolute value is taken as the correlation coefficient, and the combination with an excessive correlation coefficient is extracted as the strong correlation combination; All strong correlation combinations are obtained and sorted, and for any type of gas, the environmental parameters in all corresponding strong correlation combinations are extracted to obtain the strong correlation environmental parameter combination.
4. A signal warning system for gases in a tunnel according to claim 3, characterized in that: The calculation process of the historical concentration gradient is as follows: For light gases: the difference between the historical concentration value at the highest vertical height and the historical concentration value at the reference height is calculated as the historical concentration gradient; For heavy gases: the difference between the historical concentration value at the lowest vertical height and the historical concentration value at the reference height is calculated as the historical concentration gradient.
5. A signal warning system for gases in a tunnel according to claim 1, characterized in that: The process of matching the vertical concentration gradient is as follows: Real-time concentration values and real-time environmental parameters of the gas are acquired; For any kind of gas, the corresponding real-time strong correlation environmental parameter value data is extracted and standardized; The standardized real-time strong correlation environmental parameter value is calculated, the real-time strong correlation environmental parameter value is compared with the Euclidean distance of the clustering center of all working condition categories, and the working condition of the clustering center with the smallest Euclidean distance is selected as the current real-time working condition; Based on the gas type and the current real-time working condition, the vertical concentration gradient model library is matched, and all vertical height vertical concentration gradient models are called.
6. A signal warning system for gases in a tunnel according to claim 1, characterized in that: The acquisition process of the corrected concentration value sequence is: For any vertical height, the real-time concentration value and the real-time strong correlation environmental parameter value are taken as the input of the vertical concentration gradient model, and the corresponding corrected concentration value is output; All corrected concentration values of all vertical heights are sorted from low to high according to height to form a corresponding relationship table of vertical height and corrected concentration; Based on the preset key height interval, the corrected concentration values of all heights in the preset key height interval are extracted from the corresponding relationship table of vertical height and corrected concentration to obtain the corrected concentration value sequence.
7. A signal warning system for gases in a tunnel according to claim 1, characterized in that: The acquisition process of the vertical height to be warned is: For each vertical height in the preset key height interval, the corrected concentration values of all collection time points in the collection period are obtained to obtain a time series concentration sequence; For any time series concentration sequence, the adjacent two collection time points are taken as a calculation window, the difference between the adjacent corrected concentration values in the calculation window is calculated to obtain a concentration deviation degree; The time series concentration sequence determined to have a growth trend or a stage key growth trend is extracted, the concentration deviation degree of the growth window is extracted for mean processing to obtain a growth intensity; If the growth intensity exceeds the standard, a high-risk growth trend of the corresponding vertical height is marked; All vertical heights in the preset key height interval are traversed, and the vertical height marked as a high-risk growth trend is extracted as a vertical height to be warned.
8. A signal warning system for gases in a tunnel according to claim 7, characterized in that: The determination process of the growth trend is: Extract the calculation window with a concentration deviation degree greater than zero as a growth window, count the number of growth windows, and perform ratio processing with the total number of calculation windows to obtain a growth window proportion; If the growth window proportion exceeds the standard, it is determined that the time series concentration sequence has a continuous growth trend; If the growth window proportion does not exceed the standard, it is further identified whether there is a stage key growth trend.
9. A signal warning system for gases in a tunnel according to claim 8, characterized in that: The process of whether there is a stage key growth trend is: All calculation windows are traversed, a single existing growth window is taken as an isolated growth window, and continuously adjacent growth windows are integrated as a growth period; The number of calculation windows in each growth period is taken as a continuous window number; If the continuous window number exceeds the standard, it is determined to be an effective growth period; If there is at least one effective growth period, it is determined that the time series concentration sequence has a stage key growth trend.
10. A signal warning system for gases in a tunnel according to claim 1, characterized in that: The acquisition process of the warning time is: For any time series concentration sequence corresponding to a vertical height to be warned; If the time series concentration sequence has a continuous growth trend, the last corrected concentration value of the time series concentration sequence is obtained; The alarm value is subtracted from the corrected concentration value to obtain a concentration value that still needs to be increased, and the concentration value that still needs to be increased is divided by the growth intensity to obtain a calculation window number value, which is then multiplied by the interval between the two collection time points to obtain a warning time; If the time-series concentration sequence presents a phased key growth trend, all effective growth periods are extracted from the time-series concentration sequence, for each effective growth period, the mean of the concentration deviation degrees corresponding to all calculation windows in the growth period is calculated as a period growth length, and the maximum of the period growth lengths is extracted as a key growth length; The concentration value that still needs to be increased is divided by the key growth length to obtain a calculation window number value, which is then multiplied by the interval between the two collection time points to obtain a warning time.