High-gas extra-long tunnel construction ventilation method, system, equipment and medium

Abstract

The invention relates to a high-gas extra-long tunnel construction ventilation method, system and device and a medium, and belongs to the technical field of tunnel construction.The method comprises the steps that a to-be-monitored tunnel is divided into a plurality of monitoring intervals based on three-dimensional geographic information of the to-be-monitored tunnel; acquiring gas concentration data of each monitoring interval in real time and preprocessing the data; extracting space-time correlation characteristics according to the preprocessed gas concentration data, and constructing a space-time characteristic matrix of each monitoring interval; inputting the spatial-temporal characteristic matrix into a pre-trained risk prediction model, and outputting the risk probability of each monitoring interval; determining the monitoring interval of which the risk probability is greater than a preset threshold value as a risk interval, and determining the risk level of each risk interval based on a preset risk level mapping table; generating a ventilation strategy corresponding to each risk interval based on the risk level; and performing simulation optimization on the ventilation strategy corresponding to each risk interval and sending the ventilation strategy to the ventilation control terminal. The construction safety and the ventilation efficiency can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of tunnel construction, and in particular to a ventilation method, system, equipment and medium for the construction of a high-gas extra-long tunnel. Background Art

[0002] With the continuous development of underground engineering construction, especially the construction of high-gas tunnels and extra-long tunnels, environmental management in tunnel construction faces increasingly severe challenges. Tunnel construction is a high-risk working environment. Especially in high-gas areas, the change of gas concentration during construction directly affects the life safety and health of construction workers.

[0003] In the traditional tunnel construction process, due to the narrow and complex internal space of the tunnel, there are significant non-uniformities in air flow and gas exchange, resulting in large differences in gas concentration at different positions. Moreover, the construction section of a high-gas tunnel is often affected by multiple factors, including the geographical location, geological conditions, construction progress, construction method, etc. The common ventilation control system cannot respond in real time to the dynamic environmental changes at different positions in the tunnel, especially the fluctuations of gas concentration, which easily leads to excessive local gas concentration, making the ventilation system often lag behind the actual needs, difficult to cope with the sudden gas risks in local areas, and reducing the construction safety and ventilation efficiency. Summary of the Invention

[0004] In order to improve construction safety and ventilation efficiency, the present application provides a ventilation method, system, equipment and medium for the construction of a high-gas extra-long tunnel.

[0005] In the first aspect, the present application provides a ventilation method for the construction of a high-gas extra-long tunnel, adopting the following technical solution: A ventilation method for the construction of a high-gas extra-long tunnel, the method comprising: Dividing the tunnel to be monitored into multiple monitoring intervals based on the three-dimensional geographical information of the tunnel to be monitored; Obtaining the gas concentration data of each monitoring interval in real time and performing data preprocessing; Extracting spatio-temporal correlation features according to the preprocessed gas concentration data, and constructing a spatio-temporal feature matrix for each monitoring interval; Inputting the spatio-temporal feature matrix into a pre-trained risk prediction model, and outputting the risk probability of each monitoring interval; Determining the monitoring intervals with the risk probability greater than a preset threshold as risk intervals, and determining the risk level of each risk interval based on a preset risk level mapping table; Generating a ventilation strategy corresponding to each risk interval based on the risk level; Simulating and optimizing the ventilation strategy corresponding to each risk interval and sending it to the ventilation control terminal.

[0006] By adopting the above technical solutions, the intelligent monitoring and control of ventilation safety during the construction of high-gas extra-long tunnels is realized. Through accurate tunnel section division, spatio-temporal data feature extraction, risk assessment based on risk prediction models, and flexible generation and execution of ventilation strategies, it is ensured that the gas concentration in each monitored section during the construction process can be effectively controlled, the risk of local gas concentration exceeding the standard is reduced, and the safety and ventilation efficiency of tunnel construction are improved.

[0007] Optionally, the steps of extracting spatio-temporal correlation features from the preprocessed gas concentration data and constructing a spatio-temporal feature matrix for each monitored section include: Sort the gas concentration data of each monitored section after preprocessing according to the time stamp; Based on the time series analysis method, model the gas concentration data of each monitored section and extract the time series feature set of each monitored section; Perform spatial interpolation on the gas concentration data of each monitored section, and calculate the spatial correlation in combination with the gas concentration data of adjacent monitored sections to extract the spatial correlation feature set; Combine the time series feature set and the spatial correlation feature set to construct a spatio-temporal feature matrix for each monitored section.

[0008] By adopting the above technical solutions, the time series and spatial features of the gas concentration are combined to construct a spatio-temporal feature matrix, which can comprehensively reflect the variation law of the gas concentration with time and space. Through time sorting, time series modeling, spatial interpolation and spatial correlation analysis of the gas concentration data, rich spatio-temporal features are extracted. These features not only help to better understand the dynamic changes of the gas concentration, but also provide accurate support for subsequent work such as risk assessment and ventilation optimization.

[0009] Optionally, it further includes the training steps of the risk prediction model, and the training steps include: Obtain historical sample data and perform preprocessing; the historical sample data includes historical gas concentration data and pre-labeled risk labels; Extract spatio-temporal correlation features from the preprocessed historical gas concentration data to obtain a sample spatio-temporal feature set; Divide the sample spatio-temporal feature set and the pre-labeled risk labels into training samples and test samples; Input the sample spatio-temporal feature set of the training samples into a pre-constructed support vector machine model to obtain a training label result; Compare the training label result with the risk labels in the training samples to obtain a training comparison result; Iterate the support vector machine model based on the training comparison result to obtain the trained risk prediction model.

[0010] By adopting the above technical solution, the support vector machine model can make full use of historical data for training, and accurately predict the risk probability of each monitoring interval based on spatio-temporal features, thereby facilitating the assessment of risks during tunnel construction, providing accurate early warnings, effectively improving construction safety, and reducing the error of manual judgment.

[0011] Optionally, after the step of obtaining the trained risk prediction model, the following steps are further included: Input the sample spatio-temporal feature set in the test sample into the trained risk prediction model to obtain a test label result; Based on the risk labels in the test sample, compare the test label result to obtain a test comparison result; According to the test comparison result, evaluate the performance index of the trained risk prediction model and perform iteration until the preset performance index is met or the preset number of iterations is reached, to obtain the trained risk prediction model.

[0012] By adopting the above technical solution, continuously adjust and improve the model, improve the prediction accuracy, and finally generate a relatively ideal risk prediction model, which can accurately judge risks.

[0013] Optionally, the steps of determining the monitoring intervals with risk probability greater than the preset threshold as risk intervals and determining the risk level of each risk interval based on the preset risk level mapping table include: Respectively judge whether the risk probability of each monitoring interval is greater than the preset threshold; If so, determine the monitoring interval as a risk interval; Based on the preset mapping table, determine the corresponding probability interval according to the risk probability of the risk interval, and determine the corresponding risk level according to the probability interval; wherein, multiple groups of mapping relationships between probability intervals and risk levels are pre-configured in the preset mapping table.

[0014] By adopting the above technical solution, screen out the risk intervals that need to be focused on, ensure the monitoring focus, and then map each risk interval to the corresponding risk level, providing a scientific basis for subsequent ventilation strategies and risk prevention and control, thereby realizing automated and accurate risk classification and reducing the error of human judgment.

[0015] Optionally, the steps of simulating and optimizing the ventilation strategy corresponding to each risk interval and sending it to the ventilation control terminal include: Construct a digital twin model based on the three-dimensional geographical information of the tunnel to be monitored; Real-time obtain the gas concentration data of each monitoring interval of the tunnel to be monitored and transmit it to the digital twin model; In the digital twin model, based on the ventilation strategy corresponding to each risk interval, a hydrodynamic simulation model of the air flow in each risk interval is established to obtain the simulation results; Based on the simulation results, the ventilation strategy corresponding to each risk interval is feedback-optimized; The feedback-optimized ventilation strategy is sent to the ventilation control terminal.

[0016] By adopting the above technical solution, through the precise simulation and optimization process of the digital twin model, the dynamic ventilation strategy optimization based on risk intervals is realized in tunnel construction. Through meticulous data collection, simulation, and feedback optimization, it is ensured that the ventilation strategy can respond to the change of gas concentration in real time and is optimized for each risk interval. Finally, by sending the optimized ventilation strategy to the control terminal, the efficient and safe operation of the tunnel ventilation system is ensured. This technical solution not only improves the automation level of ventilation control, reduces the error of human intervention, but also effectively improves the safety and efficiency in the tunnel construction process through precise air flow simulation and strategy optimization.

[0017] In a second aspect, the present application provides a ventilation system for high-gas extra-long tunnel construction, adopting the following technical solution: A ventilation system for high-gas extra-long tunnel construction, the system includes: An interval division module, configured to divide the to-be-monitored tunnel into multiple monitoring intervals based on the three-dimensional geographic information of the to-be-monitored tunnel; A data processing module, configured to obtain the gas concentration data of each monitoring interval in real time and perform data preprocessing; A feature extraction module, configured to extract spatio-temporal correlation features according to the preprocessed gas concentration data and construct a spatio-temporal feature matrix for each monitoring interval; A risk prediction module, configured to input the spatio-temporal feature matrix into a pre-trained risk prediction model and output the risk probability of each monitoring interval; A risk level evaluation module, configured to determine the monitoring intervals with the risk probability greater than a preset threshold as risk intervals and determine the risk level of each risk interval based on a preset risk level mapping table; A ventilation strategy generation module, configured to generate a ventilation strategy corresponding to each risk interval based on the risk level; A strategy simulation and optimization module, configured to simulate and optimize the ventilation strategy corresponding to each risk interval and send it to the ventilation control terminal.

[0018] Optionally, the feature extraction module includes: A data sorting module, configured to sort the gas concentration data of each monitoring interval after preprocessing according to the time stamp; The time-series feature extraction module is used to model the gas concentration data in each monitoring interval based on the time-series analysis method and extract the time-series feature set of each monitoring interval; The spatial feature extraction module is used to perform spatial interpolation on the gas concentration data in each monitoring interval, calculate the spatial correlation by combining the gas concentration data in adjacent monitoring intervals, and extract the spatial correlation feature set; The feature matrix construction module is used to construct the spatio-temporal feature matrix of each monitoring interval by combining the time-series feature set and the spatial correlation feature set.

[0019] In a third aspect, the present application provides a computer device, adopting the following technical solution: A computer device includes a memory, a processor, and a computer program stored on the memory. The processor executes the computer program to implement the steps of the method described in the first aspect.

[0020] In a fourth aspect, the present application provides a computer-readable storage medium, adopting the following technical solution: A computer-readable storage medium stores a computer program that can be loaded and executed by a processor to perform any of the methods described in the first aspect.

[0021] In summary, the present application includes at least one of the following beneficial technical effects: Through precise monitoring and data processing, combined with spatio-temporal feature extraction and a risk prediction model, it is possible to achieve real-time risk assessment and dynamic ventilation strategy optimization for each monitoring interval in the tunnel. By converting the gas concentration data into a spatio-temporal feature matrix and inputting it into the risk prediction model, the risk probability of each monitoring interval can be accurately judged, and corresponding ventilation strategies can be generated based on this. This technical solution can not only timely identify high-risk areas, avoiding the lag of traditional ventilation methods, but also improve ventilation efficiency through simulation optimization, effectively reducing safety risks such as gas explosions, significantly enhancing the safety and work efficiency during tunnel construction, and having important practical significance. Description of the Drawings

[0022] Figure 1 is the first process schematic diagram of the construction ventilation method for a high-gas long tunnel in one embodiment of the present application.

[0023] Figure 2 is the second process schematic diagram of the construction ventilation method for a high-gas long tunnel in one embodiment of the present application.

[0024] Figure 3 is the third process schematic diagram of a construction ventilation method for a high-gas long tunnel in one embodiment of the present application.

[0025] Figure 4It is the schematic diagram of the fourth process of a high-gas extra-long tunnel construction ventilation method according to one embodiment of the present application.

[0026] Figure 5 It is the schematic diagram of the fifth process of a high-gas extra-long tunnel construction ventilation method according to one embodiment of the present application.

[0027] Figure 6 It is the schematic diagram of the sixth process of a high-gas extra-long tunnel construction ventilation method according to one embodiment of the present application. Detailed implementation manners

[0028] In order to make the purpose, technical solutions and advantages of the present application clearer, the following further details the present application in conjunction with the appended Figures 1-6 drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0029] An embodiment of the present application discloses a high-gas extra-long tunnel construction ventilation method.

[0030] Referring to Figure 1 , a high-gas extra-long tunnel construction ventilation method includes: Step S101, based on the three-dimensional geographical information of the tunnel to be monitored, divide the tunnel to be monitored into multiple monitoring intervals; Among them, the tunnel is spatially divided by using three-dimensional geographical information to determine different monitoring intervals. The three-dimensional geographical information can comprehensively describe the shape, length, depth, structure, geology and other information of the tunnel, and multiple monitoring intervals can be divided according to the actual situation of the tunnel.

[0031] In some embodiments, it can be divided based on the tunnel structure characteristics. High-gas extra-long tunnels usually consist of different structural segments, such as straight segments, curved segments, gradient change segments, different lining type segments, etc. With the help of the tunnel three-dimensional geographical model, accurately identify the starting and ending positions and lengths of each structural segment of the tunnel. For example, for straight segments and curved segments, since gas is more likely to accumulate in curved segments, they can be separately divided into monitoring intervals for key monitoring.

[0032] In addition, it can also be divided based on ventilation zones. The ventilation system forms different ventilation zones in the tunnel, and the ventilation effect and gas dilution ability of each zone are different. Analyze the layout of ventilation ducts, the positions of ventilation openings and ventilation flow directions through the three-dimensional geographical model, divide the tunnel into different ventilation zones according to the ventilation effect, and use each ventilation zone as a monitoring interval. For example, divide the well-ventilated area and the area with poor ventilation and prone to gas accumulation into different monitoring intervals.

[0033] Step S102, obtain the gas concentration data of each monitoring interval in real time and perform data preprocessing; Among them, sensors are installed in each monitoring section in the tunnel to obtain gas concentration data in real time. The sensors transmit the collected data to the central data processing system through wireless or wired networks. Gas concentration, as one of the most critical parameters in monitoring the tunnel construction process, directly affects construction safety.

[0034] However, since the collected original data is usually affected by environmental interferences such as temperature and humidity fluctuations, electrical noise, etc., it is necessary to preprocess the gas concentration data. Common data preprocessing methods include denoising, interpolation to fill in missing data, outlier detection, etc. For example, median filtering is used to remove instantaneous fluctuation data, and interpolation method is used to fill in the data during sensor failure. The preprocessed data can effectively eliminate noise interference and fill in data gaps, improving the accuracy of subsequent analysis.

[0035] Step S103, extract spatio-temporal correlation features based on the preprocessed gas concentration data, and construct a spatio-temporal feature matrix for each monitoring section; Among them, since the change of gas concentration is not only related to time but also closely related to spatial location, it is necessary to consider spatio-temporal correlation features. For example, if the gas concentration in a certain section changes in a short period of time, it may be due to poor ventilation or local airflow problems caused by construction activities, or due to changes in the geological characteristics of this section.

[0036] In some embodiments, spatio-temporal feature extraction algorithms can be adopted, such as time series analysis methods (such as autoregressive model ARIMA) and spatial interpolation techniques (such as Kriging method), so as to help construct a feature matrix describing the change of gas concentration over time and space, facilitating subsequent risk assessment and ventilation strategy optimization.

[0037] It can be understood that by extracting spatio-temporal correlation features, the change law of gas concentration in time and space can be comprehensively described, providing more accurate input data for subsequent risk prediction. This step helps to discover the dynamic changes and spatial distribution characteristics of gas concentration, thus improving the accuracy of risk prediction.

[0038] Step S104, input the spatio-temporal feature matrix into a pre-trained risk prediction model, and output the risk probability of each monitoring section; Among them, taking the spatio-temporal feature matrix as the input, a machine learning model is used for risk prediction. The pre-trained risk prediction model is usually constructed based on a large amount of historical data and trained through neural network algorithms. In this model, the input is the spatio-temporal feature matrix of each monitoring section, and the model analyzes these features and outputs the risk probability of each section. Through the application of the machine learning model, the risk probability of each monitoring section can be automatically and accurately predicted, reducing the error of manual judgment and improving the safety during the tunnel construction process.

[0039] Exemplarily, if the gas concentration in a certain interval increases significantly and is accompanied by a decrease in oxygen concentration, the model will evaluate the risk probability of that interval as high, indicating a high risk of gas explosion or poisoning in that interval.

[0040] Step S105: Determine the monitoring intervals with risk probability greater than the preset threshold as risk intervals, and determine the risk level of each risk interval based on the preset risk level mapping table; Among them, a preset threshold (such as 0.4) is set. If the risk probability of a certain monitoring interval is greater than this threshold, then this interval is determined as a risk interval. According to different risk probabilities, the corresponding risk levels can be assigned to each risk interval through the risk level mapping table (for example, 0.8 - 1.0 is the red high-risk area, 0.6 - 0.8 is the orange medium-risk area, 0.4 - 0.6 is the green low-risk area).

[0041] By setting the threshold and the risk level mapping table, through automated risk assessment and level division, the errors and subjectivity brought by human judgment are avoided, and the scientificity and effectiveness of risk management are improved.

[0042] Step S106: Generate the ventilation strategy corresponding to each risk interval based on the risk level; Among them, based on the determined risk level, specific ventilation control strategies are generated. For different risk levels, different ventilation schemes should be adopted. For example, in the red high-risk area, the air volume needs to be increased immediately or the standby fan needs to be started. In the orange medium-risk area, only the rotation speed of the existing fan needs to be adjusted. In the green low-risk area, normal ventilation can be maintained.

[0043] Step S107: Simulate and optimize the ventilation strategy corresponding to each risk interval and send it to the ventilation control terminal.

[0044] Among them, the simulation and optimization of the ventilation strategy need to combine real-time data such as gas concentration and oxygen content, as well as the actual ventilation equipment situation of the tunnel. Through the real-time data synchronization between the physical entity and the virtual model, accurate simulation and prediction are carried out, so as to carry out simulation and optimization adjustment in the virtual environment, and then transmit the ventilation strategy to the ventilation control terminal through the automated system to command the ventilation equipment to adjust in real time.

[0045] In the above embodiments, the intelligent monitoring and control of ventilation safety during the construction of high-gas extra-long tunnels are realized. Through accurate tunnel interval division, spatio-temporal data feature extraction, risk assessment based on the risk prediction model, and flexible ventilation strategy generation and execution, it is ensured that the gas concentration in each monitoring interval during the construction process can be effectively controlled, the risk of local gas concentration exceeding the standard is reduced, and the safety and ventilation efficiency of tunnel construction are improved.

[0046] Refer toFigure 2 , as an implementation of step S103, the steps of extracting spatio-temporal correlation features based on the preprocessed gas concentration data and constructing a spatio-temporal feature matrix for each monitoring interval include: Step S201, sorting the gas concentration data of each monitoring interval after preprocessing according to the time stamp; Among them, the gas concentration data changes with time. Sorting by time stamp can ensure that subsequent data analysis and modeling are based on the correct time series.

[0047] Step S202, modeling the gas concentration data of each monitoring interval based on time series analysis method, and extracting a time series feature set for each monitoring interval; Among them, the trend and fluctuation law of gas concentration changing with time are important bases for understanding the dynamic change of gas concentration. The time series analysis method, such as the autoregressive integrated moving average (ARIMA) model, is used to capture the trend, seasonality and volatility hidden in the time series. Through modeling, in-depth analysis can be carried out on the gas concentration data of each monitoring interval, and important features such as trend, periodicity and volatility can be extracted.

[0048] Specifically, the ARIMA model can identify the autocorrelation and trend of the data by fitting the historical data. For data with trend, the trend is eliminated through differencing processing. The parameter estimation method of the ARIMA model is used to fit the data and extract the time series features of each monitoring interval, such as predicted values, residuals, seasonal changes, etc.

[0049] Step S203, performing spatial interpolation on the gas concentration data of each monitoring interval, and calculating the spatial correlation in combination with the gas concentration data of adjacent monitoring intervals, and extracting a spatial correlation feature set; Among them, the gas concentration is not only affected by time factors, but also closely related to the spatial position. Through spatial interpolation methods (such as Kriging method), the gas concentration at unmeasured positions can be estimated, and at the same time, the spatial correlation of gas concentration between different monitoring intervals can be calculated to help analyze the spatial distribution law of gas concentration. Spatial correlation analysis helps to identify the spatial change trend of gas concentration and discover potential abnormal gas concentration areas.

[0050] Specifically, use spatial interpolation methods (such as Kriging method) to predict the gas concentration of some uncollected monitoring intervals, combine the gas concentration of adjacent intervals, calculate the spatial autocorrelation (such as using Moran's I index), evaluate the aggregation or distribution characteristics of gas concentration in space, and extract the spatial correlation feature set through spatial interpolation and autocorrelation analysis to understand the change law of gas concentration in space.

[0051] Step S204: Combine the time - series feature set and the spatial correlation feature set to construct a spatio - temporal feature matrix for each monitoring interval.

[0052] Among them, the time - series feature and the spatial feature are two key dimensions of the change in gas concentration. Combining these two types of features to construct a spatio - temporal feature matrix can comprehensively describe the change law of gas concentration in time and space, providing an important basis for subsequent risk assessment and ventilation optimization.

[0053] Specifically, each row of the spatio - temporal feature matrix represents the position of a monitoring interval, each column represents the gas concentration feature at a time point, and each element value includes the gas concentration feature of the monitoring interval at a specific time point and its spatial correlation feature with adjacent intervals.

[0054] In the above - mentioned embodiment, the time - series and spatial features of gas concentration are combined to construct a spatio - temporal feature matrix, which can comprehensively reflect the change law of gas concentration over time and space. By sorting gas concentration data in time, performing time - series modeling, spatial interpolation, and spatial correlation analysis, rich spatio - temporal features are extracted. These features not only help to better understand the dynamic changes of gas concentration but also provide accurate support for subsequent work such as risk assessment and ventilation optimization.

[0055] Refer to Figure 3 , as an embodiment of the risk prediction model, the training steps include: Step S301: Obtain historical sample data and perform pre - processing; the historical sample data includes historical gas concentration data and pre - labeled risk labels; Among them, the historical gas concentration data contains gas concentration information of different monitoring intervals at different time points, and these data are the basis for model training. Each piece of historical data is attached with a risk label, which is usually judged according to historical actual situations and represents the risk level of the monitoring interval at that time point (such as low risk, medium risk, high risk).

[0056] In one embodiment of the present application, historical gas concentration and other relevant data (such as oxygen concentration, temperature, etc.) can be collected from the tunnel monitoring system, and it is ensured that each piece of data contains a corresponding risk label. The risk label is usually based on historical experience or expert judgment and represents different risk levels (such as low risk, medium risk, high risk).

[0057] In addition, the data pre - processing steps include removing outliers, filling missing values, performing standardization or normalization processing, etc., to ensure that the data can be effectively input into the machine - learning model. At the same time, standardization can ensure that the feature data has a unified dimension and avoid some features having too much influence on the model.

[0058] Step S302: Extract spatiotemporal correlation features from the preprocessed historical gas concentration data to obtain a sample spatiotemporal feature set; Specifically, time series analysis methods (such as ARIMA or autoregressive moving average) can be used to model the gas concentration data in each monitoring interval to extract time features such as trends, seasonal fluctuations, and periodic changes. At the same time, spatial interpolation techniques (such as Kriging method) are used to perform spatial interpolation on the gas concentration data, and combined with the gas concentration data in adjacent monitoring intervals to calculate spatial correlation. Then, the time series features and spatial features are combined to form the spatiotemporal feature vector of each sample, constituting the sample spatiotemporal feature set. Each sample includes the time-dependent features and spatial distribution features of the gas concentration.

[0059] Step S303: Divide the sample spatiotemporal feature set and the pre-annotated risk labels into training samples and test samples; Among them, using common training set and test set division methods, the historical data set is divided into training samples and test samples according to a certain ratio (for example, 80% training set, 20% test set), ensuring that the sample distributions of each data set are as similar as possible to improve the generalization ability of the model. When dividing, ensure that the risk label distributions in each data set are similar to avoid biases caused by data division.

[0060] Step S304: Input the sample spatiotemporal feature set of the training samples into the pre-constructed support vector machine model to obtain the training label results; Among them, input the spatiotemporal feature set of the training samples into the support vector machine (SVM) model. SVM is a model based on maximum margin classification, which is good at dealing with high-dimensional data and is particularly suitable for classification problems of such spatiotemporal features.

[0061] Specifically, regarding the model structure of the support vector machine, for the non-linear problems in the embodiments of the present application, the radial basis function (RBF) can be selected as the kernel function to help the model find the optimal decision boundary; during the training process, an optimal hyperplane can be found to maximize the interval between data points, and at the same time, the model complexity is controlled by regularization to avoid overfitting.

[0062] In one embodiment of the present application, the support vector machine, as a powerful classifier, maps the original input to a high-dimensional feature space through a kernel function to find an optimal decision hyperplane to maximize the classification margin. This hyperplane effectively distinguishes different risk levels (such as low risk, medium risk, and high risk). During the training process, the model adjusts the weights and biases by optimizing the loss function (such as the hinge loss) to minimize classification errors as much as possible, and through iterative optimization, enables the model to make accurate predictions on new data. The choice of the kernel function (such as the radial basis function RBF) is crucial for handling non-linear relationships and helps the model find a suitable decision boundary in the complex spatio-temporal data space. The trained model can perform risk classification on the real-time input spatio-temporal features, automatically identify potential high-risk areas, and thus provide accurate early warnings for safety management in environments such as tunnel construction.

[0063] Step S305: Compare the training label result with the risk label in the training sample to obtain a training comparison result. Among them, compare the training label output by the model with the actual risk label in the training sample, and calculate evaluation metrics such as accuracy, recall rate, and F1 score. According to the comparison result, analyze the prediction error of the model, identify which specific samples have inaccurate prediction results, and may need to adjust the model or add more features. By comparing with the actual label, evaluate the training effect of the model, help discover the weak links of the model, and provide a basis for subsequent model adjustment, so as to continuously optimize the model performance.

[0064] Step S306: Iterate the support vector machine model based on the training comparison result to obtain a trained risk prediction model.

[0065] Among them, according to the comparison result, use gradient descent or other optimization algorithms to iterate the SVM model, optimize the model parameters (such as kernel function parameters, regularization coefficients, etc.). If the model performs poorly, the type of kernel function can be adjusted, the feature set can be reselected, or the input data can be optimized through feature selection methods to further improve the prediction accuracy of the model.

[0066] In the above embodiment, the support vector machine model can make full use of historical data for training, and accurately predict the risk probability of each monitoring interval based on spatio-temporal features, so as to facilitate the assessment of risks during tunnel construction, provide accurate early warnings, effectively improve construction safety, and reduce the error of manual judgment.

[0067] Refer to Figure 4 , as a further embodiment of the risk prediction model, after the step of obtaining the trained risk prediction model, it further includes: Step S401: Input the sample spatio-temporal feature set in the test sample into the trained risk prediction model to obtain the test label result; Among them, input the spatio-temporal feature set of the test sample into the trained SVM model to obtain the predicted label result (i.e., the predicted risk level) of each sample. The model will output the corresponding risk label result according to the features in the test sample for further performance evaluation.

[0068] Step S402: Compare the test label result with the risk label in the test sample to obtain the test comparison result; Among them, compare the risk label predicted by the model with the risk label of the actual test sample, calculate performance metrics such as accuracy, recall rate, and F1 score, evaluate the performance of the model based on the comparison result, and further adjust the model parameters or select other models for comparison according to the requirements. By evaluating the test sample, quantify the performance of the model, provide an objective evaluation result for the final application of the model, and ensure that it meets the preset performance requirements.

[0069] Step S403: According to the test comparison result, evaluate the performance metrics of the trained risk prediction model and perform iteration until the preset performance index is met or the preset number of iterations is reached to obtain the trained risk prediction model.

[0070] Among them, according to the test comparison result, further optimize the model, adjust hyperparameters such as the learning rate and regularization parameter, and through cross-validation and performance evaluation until the model meets the preset performance metrics (such as the accuracy reaches 95%) or reaches the preset maximum number of iterations.

[0071] In the above embodiments, continuously adjust and improve the model, improve the prediction accuracy, and finally generate a relatively ideal risk prediction model that can accurately make risk judgments.

[0072] Refer to Figure 5 , as an implementation manner of step S105, the steps of determining the monitoring interval with a risk probability greater than the preset threshold as the risk interval and determining the risk level of each risk interval based on the preset risk level mapping table include: Step S501: Determine whether the risk probability of each monitoring interval is greater than the preset threshold; if so, jump to step S502. Among them, compare the risk probability of each monitoring interval to check whether it is greater than the set preset threshold; for example, if the preset threshold is 0.4, when the risk probability of a certain monitoring interval exceeds 0.4, it indicates that there is a greater risk in this area, then this interval is determined as the "risk interval" and needs further analysis and processing.

[0073] Step S502: Determine that the monitoring interval is the risk interval; Among them, the set of intervals with a risk probability higher than the threshold is marked as a risk interval, and this mark will become the input data for subsequent risk level assessment.

[0074] Step S503: Based on a preset mapping table, determine the corresponding probability interval according to the risk probability of the risk interval, and determine the corresponding risk level according to the probability interval; among them, multiple sets of mapping relationships between probability intervals and risk levels are pre-configured in the preset mapping table.

[0075] Specifically, according to the risk probability value of each monitoring interval, referring to the preset mapping table, it is classified into a specific probability interval. For example, if the risk probability of a certain monitoring interval is 0.7, the rule in the mapping table may classify it into the "medium orange risk" interval.

[0076] It should be noted that the preset mapping table should be set in detail based on historical data, the actual construction environment, and potential risks. The design of the mapping table not only considers different risk probability intervals, but also considers the influence of different construction stages and environmental factors on the gas concentration, and adapts to the actual situation of different regions.

[0077] In the above implementation, the risk intervals that need to be focused on are screened out to ensure the monitoring focus, and then each risk interval is mapped to the corresponding risk level, providing a scientific basis for subsequent ventilation strategies and risk prevention and control, thus realizing automated and precise risk classification and reducing human judgment errors.

[0078] Refer to Figure 6 , as an implementation of step S105, the steps of simulating and optimizing the ventilation strategy corresponding to each risk interval and sending it to the ventilation control terminal include: Step S601: Construct a digital twin model based on the three-dimensional geographical information of the tunnel to be monitored; Among them, the three-dimensional geographical information of the tunnel includes the shape, structure, ventilation opening position, wall thickness, slope, etc. of the tunnel, and all this information helps to determine the path of air flow and gas diffusion.

[0079] In the embodiments of the present application, the digital twin model converts the actual physical environment of the tunnel into a virtual model through digital and three-dimensional modeling technologies. This model can comprehensively reflect the internal structural characteristics of the tunnel and provide a basis for subsequent fluid simulation and ventilation strategies. Specifically, through precise three-dimensional model construction, the air flow route inside the tunnel can be simulated, helping to analyze the gas distribution, accumulation areas, and circulation paths.

[0080] Step S602: Real-time obtain the gas concentration data of each monitoring interval of the tunnel to be monitored and transmit it to the digital twin model; Among them, the gas concentration inside the tunnel is collected by sensors and, through wireless network or wired transmission, these data are fed back to the digital twin model in real time. In the model, these data are integrated with the tunnel structure, air flow conditions, etc., so that the simulation environment in the digital twin model remains synchronized with the real world.

[0081] It can be understood that obtaining and transmitting monitoring data such as gas concentration in real time can ensure that the digital twin model reflects the latest state of the tunnel, so that subsequent simulation and optimization work can be based on the most real environmental data.

[0082] Step S603, based on the ventilation strategy corresponding to each risk interval, in the digital twin model, perform a fluid dynamics simulation modeling on the air flow in each risk interval to obtain the simulation result; Among them, according to the ventilation strategy of each risk interval, perform a fluid dynamics simulation modeling of air flow and gas concentration in the digital twin model. By using the computational fluid dynamics (CFD) method, simulate the air flow distribution, wind speed change, gas diffusion path, etc. under different ventilation conditions. The simulation results will reveal which areas have insufficient air flow or too high gas concentration, identify potential risk areas, and the deficiencies that the ventilation strategy may bring.

[0083] It can be understood that through fluid dynamics simulation, an accurate air flow distribution map can be provided for each risk interval. Such simulation results can intuitively show the impact of different fan configurations on air flow and gas dilution, providing a scientific basis for subsequent ventilation strategy optimization and avoiding manual speculation and lack of experience.

[0084] Step S604, based on the simulation result, perform feedback optimization on the ventilation strategy corresponding to each risk interval; Among them, according to the simulation result, analyze the deficiencies of the existing ventilation strategy, such as insufficient air flow in some areas, mismatched fan air volume, etc. In the digital twin model, perform optimization adjustments, change the wind speed, air volume, number of fans or operation timing of the fans to achieve a better air circulation effect. Through feedback optimization, the optimization goal is to reduce the gas concentration to a safe range and ensure the uniformity of air flow, so as to ensure that the ventilation strategy can better adapt to the actual situation of the tunnel, maximize the efficiency of the ventilation system, and thus improve the safety and efficiency during tunnel construction.

[0085] Step S605, send the ventilation strategy after feedback optimization to the ventilation control terminal.

[0086] Among them, the ventilation strategy after feedback optimization includes control instructions such as the adjusted fan wind speed, air volume, fan start and stop for each risk interval.

[0087] Specifically, the optimized ventilation strategy after feedback is sent to the ventilation control terminal in the tunnel through wireless communication, network transmission, etc. After receiving the instruction, the control terminal will perform corresponding operations according to the fan type and equipment configuration, such as adjusting the fan speed, increasing the number of fans or changing the air flow direction, etc., so as to realize automatic ventilation adjustment and ensure that the ventilation system can respond to the environmental changes in the tunnel in real time.

[0088] In the above embodiments, by using the precise simulation and optimization process of the digital twin model, the dynamic ventilation strategy optimization based on risk intervals is realized in tunnel construction. Through careful data collection, simulation and feedback optimization, it is ensured that the ventilation strategy can respond to the change of gas concentration in real time and optimize each risk interval specifically. Finally, by sending the optimized ventilation strategy to the control terminal, the efficient and safe operation of the tunnel ventilation system is ensured. This technical solution not only improves the automation level of ventilation control, reduces the error of human intervention, but also effectively improves the safety and efficiency in the tunnel construction process through precise air flow simulation and strategy optimization.

[0089] The embodiment of the present application also discloses a ventilation system for high-gas long tunnel construction.

[0090] A ventilation system for high-gas long tunnel construction, the system includes: An interval division module, used to divide the tunnel to be monitored into multiple monitoring intervals based on the three-dimensional geographical information of the tunnel to be monitored; A data processing module, used to obtain the gas concentration data of each monitoring interval in real time and perform data preprocessing; A feature extraction module, used to extract spatio-temporal correlation features according to the preprocessed gas concentration data and construct a spatio-temporal feature matrix for each monitoring interval; A risk prediction module, used to input the spatio-temporal feature matrix into a pre-trained risk prediction model and output the risk probability of each monitoring interval; A risk level evaluation module, used to determine the monitoring intervals with risk probability greater than the preset threshold as risk intervals and determine the risk level of each risk interval based on the preset risk level mapping table; A ventilation strategy generation module, used to generate a ventilation strategy corresponding to each risk interval based on the risk level; A strategy simulation and optimization module, used to simulate and optimize the ventilation strategy corresponding to each risk interval and send it to the ventilation control terminal.

[0091] As an implementation manner of the feature extraction module, the feature extraction module includes: A data sorting module, used to sort the gas concentration data of each monitoring interval after preprocessing according to the time stamp; A time-series feature extraction module, which is used to model the gas concentration data of each monitoring interval based on the time-series analysis method and extract the time-series feature set of each monitoring interval; A spatial feature extraction module, which is used to perform spatial interpolation on the gas concentration data of each monitoring interval, calculate the spatial correlation by combining the gas concentration data of adjacent monitoring intervals, and extract the spatial correlation feature set; A feature matrix construction module, which is used to construct the spatio-temporal feature matrix of each monitoring interval by combining the time-series feature set and the spatial correlation feature set.

[0092] The high-gas extra-long tunnel construction ventilation system according to the embodiment of the present application can implement any one of the above ventilation methods, and the specific working processes of each module in the high-gas extra-long tunnel construction ventilation system can refer to the corresponding processes in the above method embodiments.

[0093] In several embodiments provided in the present application, it should be understood that the provided methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for example, the division of a certain module is only a logical function division, and there may be other division methods in actual implementation. For example, multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.

[0094] The embodiment of the present application also discloses a computer device.

[0095] The computer device includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements a high-gas extra-long tunnel construction ventilation method as described above.

[0096] The embodiment of the present application also discloses a computer-readable storage medium.

[0097] The computer-readable storage medium stores a computer program that can be loaded and executed by a processor to implement any one of the high-gas extra-long tunnel construction ventilation methods as described above.

[0098] Among them, the computer-readable storage medium can be any tangible medium that contains or stores a program, and this program can be used by or in combination with an instruction execution system, device, or device; the program code contained on the computer-readable medium can be transmitted by any appropriate medium, including but not limited to wireless, wire, optical fiber, RF, etc., or any suitable combination of the above.

[0099] It should be noted that in the above embodiments, the descriptions of the various embodiments have their own emphases. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0100] The above are all preferred embodiments of the present application, and do not limit the protection scope of the present application accordingly. Any feature disclosed in this specification (including the abstract and drawings), unless specifically described, can be replaced by other equivalent or similar-purpose alternative features. That is, unless specifically described, each feature is only an example of a series of equivalent or similar features.

Claims

1. A high-gas long tunnel construction ventilation method, characterized in that: The method comprises: Based on the three-dimensional geographic information of the tunnel to be monitored, the tunnel to be monitored is divided into multiple monitoring areas; Obtain gas concentration data of each monitoring interval in real time and perform data preprocessing; Extract the spatiotemporal correlation features based on the preprocessed gas concentration data and construct the spatiotemporal feature matrix of each monitoring interval; Inputting the spatiotemporal feature matrix into a pre-trained risk prediction model, and outputting the risk probability of each monitoring interval; Determine the monitoring intervals where the risk probability is greater than a preset threshold as risk intervals, and determine the risk level of each risk interval based on a preset risk level mapping table; generating a ventilation strategy corresponding to each risk interval based on the risk level; The ventilation strategy corresponding to each risk interval is simulated and optimized and sent to the ventilation control terminal.

2. A high-gas long tunnel construction ventilation method according to claim 1, characterized in that: The steps of extracting spatiotemporal correlation features based on the preprocessed gas concentration data and constructing the spatiotemporal feature matrix of each monitoring interval include: Sort the pre-processed gas concentration data of each monitoring interval by timestamp; Based on the time series analysis method, the gas concentration data of each monitoring interval is modeled and the time series feature set of each monitoring interval is extracted; The gas concentration data of each monitoring interval is spatially interpolated, and the spatial correlation is calculated by combining the gas concentration data of adjacent monitoring intervals to extract the spatial correlation feature set; By combining the time series feature set and the spatial correlation feature set, a spatiotemporal feature matrix of each monitoring interval is constructed.

3. A high-gas long tunnel construction ventilation method according to claim 1, characterized in that: The method further includes a risk prediction model training step, wherein the training step includes: Acquire historical sample data and perform preprocessing; the historical sample data includes historical gas concentration data and pre-marked risk labels; Extracting spatiotemporal correlation features from the preprocessed historical gas concentration data to obtain a sample spatiotemporal feature set; Dividing the sample spatiotemporal feature set and the pre-labeled risk labels into training samples and test samples; Inputting the sample spatiotemporal feature set of the training sample into a pre-built support vector machine model to obtain a training label result; Compare the training label result with the risk label in the training sample to obtain a training comparison result; The support vector machine model is iterated based on the training comparison result to obtain the trained risk prediction model.

4. A high-gas long tunnel construction ventilation method according to claim 3, characterized in that: After the step of obtaining the trained risk prediction model, the step further includes: Inputting the sample spatiotemporal feature set in the test sample into the trained risk prediction model to obtain a test label result; Comparing the test label result with the risk label in the test sample to obtain a test comparison result; According to the test comparison results, the performance indicators of the trained risk prediction model are evaluated and iterated until a preset performance index is met or a preset number of iterations is reached, thereby obtaining the trained risk prediction model.

5. A high-gas long tunnel construction ventilation method according to claim 1, characterized in that: The steps of determining the monitoring intervals where the risk probability is greater than a preset threshold as risk intervals, and determining the risk level of each risk interval based on a preset risk level mapping table include: Determine whether the risk probability of each monitoring interval is greater than a preset threshold; If so, determining the monitoring interval as a risk interval; Based on a preset mapping table, a corresponding probability interval is determined according to the risk probability of the risk interval, and a corresponding risk level is determined according to the probability interval; wherein the preset mapping table is pre-configured with a plurality of mapping relationships between probability intervals and risk levels.

6. A high-gas long tunnel construction ventilation method according to claim 5, characterized in that: The steps of simulating and optimizing the ventilation strategy corresponding to each risk interval and sending it to the ventilation control terminal include: Build a digital twin model based on the three-dimensional geographic information of the tunnel to be monitored; Acquire gas concentration data of each monitoring interval of the tunnel to be monitored in real time and transmit it to the digital twin model; Based on the ventilation strategy corresponding to each risk zone, in the digital twin model, fluid mechanics simulation modeling is performed on the airflow in each risk zone to obtain simulation results; Based on the simulation results, feedback optimization is performed on the ventilation strategy corresponding to each risk interval; The ventilation strategy after feedback optimization is sent to the ventilation control terminal.

7. A high-gas long tunnel construction ventilation system, characterized in that: The system comprises: An interval division module, used for dividing the tunnel to be monitored into multiple monitoring intervals based on the three-dimensional geographic information of the tunnel to be monitored; Data processing module, used to obtain gas concentration data of each monitoring interval in real time and perform data preprocessing; The feature extraction module is used to extract the spatiotemporal correlation features based on the preprocessed gas concentration data and construct the spatiotemporal feature matrix of each monitoring interval; A risk prediction module, used to input the spatiotemporal feature matrix into a pre-trained risk prediction model and output the risk probability of each monitoring interval; A risk level assessment module, used to determine the monitoring intervals where the risk probability is greater than a preset threshold as risk intervals, and determine the risk level of each risk interval based on a preset risk level mapping table; A ventilation strategy generation module, used to generate a ventilation strategy corresponding to each risk interval based on the risk level; The strategy simulation and optimization module is used to simulate and optimize the ventilation strategy corresponding to each risk interval and send it to the ventilation control terminal.

8. A high-gas long tunnel construction ventilation system according to claim 7, characterized in that: The feature extraction module comprises: A data sorting module is used to sort the pre-processed gas concentration data of each monitoring interval by timestamp; The time series feature extraction module is used to model the gas concentration data of each monitoring interval based on the time series analysis method and extract the time series feature set of each monitoring interval; The spatial feature extraction module is used to perform spatial interpolation on the gas concentration data of each monitoring interval, and calculate the spatial correlation by combining the gas concentration data of adjacent monitoring intervals to extract a spatial correlation feature set; The feature matrix construction module is used to combine the time series feature set and the spatial correlation feature set to construct a spatiotemporal feature matrix for each monitoring interval.

9. A computer device, characterized in that: The method comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 6 when executing the program.

10. A computer-readable storage medium, characterized in that: A computer program is stored which can be loaded by a processor and execute the method according to any one of claims 1 to 6.

Images