A method for rapid prediction and risk assessment of temporal and spatial evolution of hydrogen concentration on tunnel ceiling

Abstract

The application relates to a tunnel ceiling hydrogen concentration space-time evolution rapid prediction and risk assessment method. The method comprises the following steps: obtaining target monitoring point position coordinates and ventilation wind speed and normalizing; inputting the normalized features into a two-stage hydrogen concentration space-time coupling prediction model, which is composed of a diffusion stage submodel and a ventilation stage submodel in cascade, to respectively predict hydrogen concentration time series in the free diffusion stage and the mechanical ventilation stage; splicing and inverse transforming the two-stage sequences to obtain a complete hydrogen concentration time curve; and automatically extracting risk indexes such as dangerous peak concentration, peak arrival time, overall exposure level, ventilation efficiency and safety recovery time based on the curve. The application adopts a machine learning method to replace computational fluid dynamics numerical simulation, realizes millisecond-level prediction of hydrogen concentration evolution at any point of the tunnel ceiling, and can quantitatively evaluate the risk differences under different ventilation strategies, thereby providing forward-looking decision support for emergency ventilation scheduling and personnel evacuation.

Description

Technical Field

[0001] This application relates to the field of hydrogen energy transportation safety monitoring and risk assessment technology, and in particular to a method for rapid prediction and risk assessment of the spatiotemporal evolution of hydrogen concentration in tunnel ceilings. Background Technology

[0002] With the increasing application of hydrogen-powered trains in the rail transit sector, the safety risks of hydrogen leakage in confined spaces such as tunnels are receiving growing attention. Hydrogen has characteristics such as low ignition energy, wide combustion range, and rapid diffusion rate. Once it accumulates in the tunnel ceiling area, it can easily cause a combustion and explosion accident.

[0003] Currently, the monitoring and risk assessment of hydrogen leaks in tunnels mainly employ the following two technical approaches: One approach is a real-time monitoring method based on a fixed sensor network. This scheme involves arranging an array of hydrogen concentration sensors at equal intervals along the longitudinal direction of the tunnel ceiling, and placing sensors at the axis of the exhaust vents. The monitoring data is transmitted to the control center in real time, triggering ventilation or an alarm based on preset logic. However, this scheme has inherent limitations: due to the cost and installation conditions of sensor deployment, it can only cover a limited number of discrete points in practical applications, failing to achieve continuous spatial concentration field perception across the entire tunnel, and blind spots are unavoidable. If a leak occurs at a non-monitoring point, or if the diffusion path after the leak deviates from the sensor deployment area, the system will be unable to obtain the true concentration distribution in a timely manner.

[0004] Secondly, numerical simulation methods based on computational fluid dynamics (CFD) are employed. This approach establishes a refined tunnel geometry and physical field model to solve the fluid control equations, simulating the diffusion and transport processes following a hydrogen leak. While numerical simulation can reveal the continuous evolution of hydrogen concentration in time and space, the computation is extremely time-consuming—a single transient simulation typically takes several hours or even days, failing to meet the real-time requirements of emergency response. Furthermore, numerical simulation demands significant computational resources, making it difficult to deploy on-site or run in real-time.

[0005] In summary, existing technologies have the following shortcomings in practical applications: monitoring methods are limited to discrete points and cannot achieve dynamic perception of the concentration field across the entire tunnel; numerical simulation based on physical models is inefficient and cannot provide rapid predictions at the second or millisecond level; and there is a lack of quantitative risk assessment tools that can provide forward-looking data support for key decisions such as optimized scheduling of ventilation systems and emergency evacuation of personnel.

[0006] Therefore, there is an urgent need for a method in related technologies that can quickly and accurately predict the spatiotemporal evolution of hydrogen concentration in tunnel ceilings and conduct quantitative risk assessments. Summary of the Invention

[0007] Therefore, it is necessary to provide a method for rapid prediction and risk assessment of the spatiotemporal evolution of hydrogen concentration in tunnel ceilings to address the aforementioned technical problems.

[0008] Firstly, this application provides a method for rapid prediction and risk assessment of the spatiotemporal evolution of hydrogen concentration in tunnel ceilings. The method includes: Obtain the location coordinates of the target monitoring point inside the tunnel and the ventilation speed of the tunnel ventilation system; The location coordinates and the ventilation speed are normalized to obtain normalized location coordinates and normalized ventilation speed. The normalized location coordinates and the normalized ventilation wind speed are input into a pre-trained two-stage spatiotemporal coupling prediction model for hydrogen concentration. The two-stage spatiotemporal coupling prediction model for hydrogen concentration includes a diffusion stage sub-model and a ventilation stage sub-model. The diffusion stage sub-model is used to output the time series of hydrogen concentration during the diffusion stage from the start of hydrogen leakage to the start of the ventilation system. The ventilation stage sub-model is used to output the time series of hydrogen concentration during the ventilation stage from the start of the ventilation system to the end of the prediction. The hydrogen concentration time series during the diffusion stage and the hydrogen concentration time series during the ventilation stage are spliced ​​along the time axis to obtain an absolute logarithmic concentration prediction series. An exponential inverse transformation is performed on the absolute logarithmic concentration prediction series to obtain the hydrogen concentration time curve of the target monitoring point from the start of the leak to the end of the prediction. Risk indicators are extracted based on the hydrogen concentration-time curve. These risk indicators include the dangerous peak concentration, the time to reach the dangerous peak, the overall level of dangerous exposure, ventilation efficiency, and the time required to restore safety.

[0009] Optionally, in one embodiment of this application, the training process of the two-stage spatiotemporal coupling prediction model for hydrogen concentration includes: Spatiotemporal evolution data of hydrogen concentration under different operating conditions were generated through computational fluid dynamics numerical simulation. The spatiotemporal evolution data of hydrogen concentration are truncated by a threshold and then logarithmized to obtain the absolute logarithmic concentration. The complete time series is divided into diffusion stage label sequences and ventilation stage label sequences according to the start-up time of the ventilation system. The normalized coordinates of the monitoring points and the ventilation wind speed are used as input features to train the diffusion stage sub-model and the ventilation stage sub-model, respectively.

[0010] Optionally, in one embodiment of this application, the diffusion stage sub-model is a feedforward fully connected neural network, whose input layer receives the normalized location coordinates and the normalized ventilation velocity, and whose output layer outputs the absolute logarithmic concentration at each time point of the diffusion stage.

[0011] Optionally, in one embodiment of this application, the ventilation stage sub-model is a feedforward fully connected neural network, whose input layer receives the normalized position coordinates, the normalized ventilation wind speed, and the concentration value at the end of the diffusion stage predicted by the diffusion stage sub-model, and whose output layer outputs the absolute logarithmic concentration at each time point of the ventilation stage.

[0012] Optionally, in one embodiment of this application, the method for extracting the dangerous peak concentration and its arrival time is as follows: determining the maximum concentration value and the time point corresponding to the maximum concentration value on the hydrogen concentration-time curve; The method for extracting the overall hazardous exposure level is as follows: calculate the average concentration of the hydrogen concentration time curve over a specified time period; The ventilation efficiency is extracted by calculating the ratio of the difference between the concentration at the end of the diffusion phase and the concentration at the end of ventilation to the concentration at the end of the diffusion phase. The method for extracting the time required to restore safety is as follows: during the ventilation phase, the time point corresponding to the hydrogen concentration dropping to a preset safety threshold is calculated by interpolation.

[0013] Optionally, in one embodiment of this application, when training the two-stage hydrogen concentration spatiotemporal coupling prediction model, the mean squared error is used as the loss function, the AdamW optimizer is used for optimization, and dynamic scheduling of the learning rate, gradient pruning, Dropout regularization and early stopping strategy are introduced.

[0014] Secondly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the steps of the methods described in the various embodiments above.

[0015] Thirdly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the steps of the methods described in the various embodiments above.

[0016] The above-mentioned method for rapid prediction and risk assessment of the spatiotemporal evolution of hydrogen concentration in tunnel ceilings has the following advantages compared with existing technologies: First, this invention employs a two-stage cascaded neural network model to decouple the complex physical process of hydrogen diffusion into two stages: free diffusion and mechanical ventilation, which are modeled separately. The model structure is clear and can be implemented using only conventional fully connected neural network components, making it easy to deploy and apply.

[0017] Second, this invention learns the distribution patterns of computational fluid dynamics simulation data through machine learning methods, enabling it to predict the evolution of hydrogen concentration at any point on the tunnel roof under different ventilation velocities within milliseconds. The prediction speed is several orders of magnitude faster than traditional numerical simulation, truly meeting the real-time requirements of emergency response.

[0018] Third, this invention can replace densely deployed physical sensors to achieve virtual perception of the concentration field across the entire tunnel ceiling, eliminating monitoring blind spots.

[0019] Fourth, this invention can automatically extract quantitative risk indicators such as dangerous peak concentration, peak arrival time, overall exposure level, ventilation efficiency, and safe recovery time, providing accurate and forward-looking data support for the optimized scheduling of ventilation systems and emergency evacuation decisions. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating a method for rapid prediction and risk assessment of the spatiotemporal evolution of hydrogen concentration in a tunnel ceiling, as described in one embodiment. Figure 2 This is a schematic diagram of tunnel geometry modeling in one embodiment; Figure 3 This is a schematic diagram of structured mesh generation in one embodiment; Figure 4 This is a schematic diagram of mesh encryption near the leak point in one embodiment; Figure 5 This is a simulation of the hydrogen concentration at each detection point during the leakage phase in one embodiment, showing the change of hydrogen concentration over time. Figure 6 This is a cloud map showing the hydrogen concentration distribution in the radial section of a tunnel during a simulated leakage phase in one embodiment. Figure 7 This is a cloud map showing the hydrogen concentration distribution along the axial section of a tunnel during a simulated leakage phase in one embodiment. Figure 8 This is a schematic diagram of the structure of a two-stage spatiotemporal coupling prediction model for hydrogen concentration in one embodiment. Figure 9 This is a comparison between the prediction results of two monitoring points based on a two-stage spatiotemporal coupling prediction model of hydrogen concentration and the actual sequence in one embodiment; Figure 10 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0022] In one embodiment, such as Figure 1 As shown, a method for rapid prediction and risk assessment of the spatiotemporal evolution of hydrogen concentration in tunnel ceilings is provided, including the following steps: S101: Obtain the location coordinates of the target monitoring point inside the tunnel and the ventilation speed of the tunnel ventilation system.

[0023] S102: Normalize the position coordinates and the ventilation speed to obtain normalized position coordinates and normalized ventilation speed.

[0024] S103: Input the normalized location coordinates and the normalized ventilation wind speed into a pre-trained two-stage spatiotemporal coupling prediction model for hydrogen concentration. The two-stage spatiotemporal coupling prediction model for hydrogen concentration includes a diffusion stage sub-model and a ventilation stage sub-model. The diffusion stage sub-model is used to output the time series of hydrogen concentration during the diffusion stage from the start of hydrogen leakage to the start of the ventilation system. The ventilation stage sub-model is used to output the time series of hydrogen concentration during the ventilation stage from the start of the ventilation system to the end of the prediction.

[0025] S104: The hydrogen concentration time series of the diffusion stage and the hydrogen concentration time series of the ventilation stage are spliced ​​along the time axis to obtain the absolute logarithmic concentration prediction sequence. The absolute logarithmic concentration prediction sequence is subjected to an exponential inverse transformation to obtain the hydrogen concentration time curve of the target monitoring point from the start of the leak to the end of the prediction.

[0026] S105: Extract risk indicators based on the hydrogen concentration-time curve. The risk indicators include the dangerous peak concentration, the time to reach the dangerous peak, the overall dangerous exposure level, the ventilation efficiency, and the time required to restore safety.

[0027] In this embodiment, a rectangular underground tunnel of an urban rail transit system is used as the application scenario, with a total length of 140 meters. A high-pressure hydrogen leak occurs in the hydrogen storage tank of a hydrogen-powered Type A train within the tunnel, with a leak pressure of 35 MPa. The leak stops after 30 seconds, and the tunnel ventilation system activates longitudinal ventilation 180 seconds after the leak occurs. The purpose of this invention is to rapidly and accurately predict the evolution of hydrogen volume concentration over time at any location within the tunnel in this scenario, from the start of the leak to 60 seconds after ventilation (a total of 240 seconds), and to conduct a risk assessment.

[0028] In one embodiment of this application, the training process of the two-stage spatiotemporal coupling prediction model for hydrogen concentration includes: Spatiotemporal evolution data of hydrogen concentration under different operating conditions were generated through computational fluid dynamics numerical simulation. The spatiotemporal evolution data of hydrogen concentration are truncated by a threshold and then logarithmized to obtain the absolute logarithmic concentration. The complete time series is divided into diffusion stage label sequences and ventilation stage label sequences according to the start-up time of the ventilation system. The normalized coordinates of the monitoring points and the ventilation wind speed are used as input features to train the diffusion stage sub-model and the ventilation stage sub-model, respectively.

[0029] In one embodiment of this application, training data is first generated using computational fluid dynamics numerical simulation. The professional fluid simulation software Ansys Fluent is used, based on the Realizable k-ε turbulence model and component transport model, to perform refined modeling and transient simulation of the rectangular tunnel scenario. Specifically, as shown... Figure 2 , 3 As shown in Figure 4, the geometric model of the hydrogen-powered train tunnel was completed, and a high-quality structured hexahedral mesh was generated. The mesh near the leak point was refined to ensure that the output dataset in subsequent simulations was regularly arranged. This avoids the interpolation errors introduced by unstructured meshes, thus significantly reducing the problems caused by accuracy loss, increased computational cost, and increased model complexity. It provides a standardized, high-quality data foundation for the training and generalization of machine learning models. The simulation covers six operating conditions: a leak pressure of 35 MPa and ventilation velocities of 2 m / s, 3 m / s, 4 m / s, 5 m / s, 6 m / s, and 7 m / s. Under each operating condition, multiple fixed monitoring points were arranged along the tunnel axis. The total simulation time was 240 seconds, and the output time step was 0.05 seconds, thereby generating an original time-hydrogen concentration sequence containing 4800 time points for each monitoring point and each operating condition. ,like Figure 5 As shown, this illustrates the change in hydrogen concentration over time at various detection points during the simulated leakage phase. Figure 6 As shown, this is a simulated cloud map of hydrogen concentration distribution in the radial section of the tunnel during the leakage phase. Figure 7 The image shows a cloud map of hydrogen concentration distribution along the axial section of the tunnel during the simulated leakage phase.

[0030] Thresholding and logarithmic transformation were performed on the original simulation data. To prevent numerical calculation problems caused by concentration values ​​of zero or extremely low values, a lower limit threshold of 1 × 10⁻⁶ was set. -5 Any concentration below this threshold is set to this value. Then, the absolute logarithmic concentration is obtained by taking the logarithm to base 10 of the threshold-processed concentration data. This process compresses the concentration range to a logarithmic scale, which is beneficial for the training stability of neural network models and the prediction accuracy in low-concentration regions. The specific formula is as follows:

[0031] in The absolute logarithmic concentration; In order to be in Hydrogen concentration over time; This is a preset lower limit threshold for concentration. This operation is to avoid numerical problems caused by minimum values ​​in subsequent logarithmic calculations.

[0032] Based on the physical processes of leakage and ventilation, the complete 240-second sequence was divided into two phases: the diffusion phase, from 0 to 180 seconds, corresponds to the occurrence, development, and free diffusion of the leak, comprising 3600 time points. This phase is labeled with the absolute logarithmic concentration sequence from 0 to 180 seconds. The ventilation phase lasts from 180 to 240 seconds, corresponding to the forced exhaust process after the ventilation system is activated, comprising 1200 time points. This phase is labeled with the absolute logarithmic concentration sequence from 180 to 240 seconds. .

[0033] Input features are normalized to the [0,1] interval to accelerate model convergence. Monitoring point location coordinates are normalized to... Ventilation velocity normalized to ,in, The total length of the tunnel. Minimum wind speed, This represents the maximum wind speed.

[0034] Ultimately, each training sample contains: a normalized input feature vector. The corresponding diffusion stage label sequence (3600 dimensions), ventilation stage label sequence (1200 dimensions), and the end value of the diffusion stage. .

[0035] In one embodiment of this application, the diffusion stage sub-model is a feedforward fully connected neural network, whose input layer receives the normalized location coordinates and the normalized ventilation velocity, and whose output layer outputs the absolute logarithmic concentration at each time point of the diffusion stage.

[0036] In one embodiment of this application, the ventilation stage sub-model is a feedforward fully connected neural network. Its input layer receives the normalized position coordinates, the normalized ventilation wind speed, and the concentration value at the end of the diffusion stage predicted by the diffusion stage sub-model. Its output layer outputs the absolute logarithmic concentration at each time point of the ventilation stage.

[0037] In one embodiment of this application, two independent but cascaded feedforward fully connected neural networks are constructed, with the specific structure as follows: Figure 8 As shown. The diffusion stage neural network model: The input layer receives 2D input features (normalized position coordinates). and normalized ventilation velocity The model consists of five fully connected hidden layers with 128, 256, 512, 1024, and 2048 neurons respectively. Each hidden layer is followed by a ReLU activation function and a dropout operation with a dropout rate of 0.05. The output layer is a linear layer with a dimension of 3600, directly predicting the absolute logarithmic concentration at each time point during the diffusion phase. The ventilation phase neural network model receives 3D input features (normalized position coordinates). Normalized ventilation velocity The absolute logarithmic concentration at the end of the diffusion stage, predicted by the diffusion stage model. The model consists of four fully connected hidden layers with 128, 256, 512, and 1024 neurons respectively, using ReLU activation and Dropout with a dropout rate of 0.05. The output layer is a linear layer with a dimension of 1200, predicting the absolute logarithmic concentration at each time point during the ventilation phase. The weight matrices of all fully connected layers in both phases are initialized using the Kaiming normal distribution method, and the bias vectors are initialized to zero.

[0038] In one embodiment of this application, when training the two-stage spatiotemporal coupling prediction model for hydrogen concentration, the mean squared error is used as the loss function, the AdamW optimizer is used for optimization, and dynamic scheduling of the learning rate, gradient pruning, Dropout regularization, and early stopping strategies are introduced.

[0039] In one embodiment of this application, model training employs the PyTorch machine learning framework and is performed on a GPU-equipped computing device. All samples generated by CFD simulation are randomly shuffled and divided into training, validation, and test sets in a 6:3:1 ratio. Mean squared error is used as the loss function, the AdamW optimizer is selected, and the initial learning rate is set to 5×10⁻⁶. - ³, the weight decay coefficient is set to 4×10 -4 The training strategy includes: setting the batch size to 8; using the ReduceLROnPlateau scheduler to monitor the validation set loss, and multiplying the learning rate by a factor of 0.85 to decay the validation loss if it does not decrease for 100 consecutive training epochs; pruning the gradient norm after each backpropagation, limiting the maximum value to no more than 1.0; and terminating training early and rolling back to the model parameters with the lowest validation loss if the validation set loss does not decrease for 300 consecutive training epochs. The model converges after approximately 3000 epochs during the diffusion phase and approximately 2000 epochs during the ventilation phase. Figure 9 As shown, the results of the prediction based on the two-stage spatiotemporal coupling prediction model of hydrogen concentration at two monitoring points are compared with the actual sequence. The results show that the error between the result of the two-stage spatiotemporal coupling prediction model of hydrogen concentration and the actual value is very small, and the prediction is accurate.

[0040] In one embodiment of this application, the dangerous peak concentration and its arrival time are extracted by determining the maximum concentration value and the time point corresponding to the maximum concentration value on the hydrogen concentration-time curve. The method for extracting the overall hazardous exposure level is as follows: calculate the average concentration of the hydrogen concentration time curve over a specified time period; The ventilation efficiency is extracted by calculating the ratio of the difference between the concentration at the end of the diffusion phase and the concentration at the end of ventilation to the concentration at the end of the diffusion phase. The method for extracting the time required to restore safety is as follows: during the ventilation phase, the time point corresponding to the hydrogen concentration dropping to a preset safety threshold is calculated by interpolation.

[0041] In one embodiment of this application, the user specifies any location within the tunnel to be predicted (coordinate range 0 to 140 meters) and the ventilation speed (range 2 to 7 meters per second). The location to be predicted and the ventilation speed are then subjected to the same normalization process as in the training phase.

[0042] The normalized location and wind speed characteristics are input into a pre-trained neural network model for the diffusion phase. The model outputs an absolute logarithmic concentration prediction sequence for the diffusion phase (0-180 seconds). The last element of the sequence was extracted as the concentration state at the end of the diffusion phase. The final concentration state will be determined. With normalized position Wind speed The vectors are concatenated into a 3D vector and input into a neural network model for the ventilation phase. The model outputs a predicted absolute logarithmic concentration sequence for the ventilation phase (180-240 seconds). .

[0043] The predicted sequences from the two stages are concatenated sequentially along the time dimension to obtain a complete absolute logarithmic concentration prediction sequence at 4800 time points. The sequence is then subjected to an exponential operation (base 10) to convert it back to the actual hydrogen volume concentration value, yielding the final hydrogen concentration-time curve. .

[0044] Risk indicators are extracted based on the prediction curve: the maximum concentration value in the curve is calculated. and its arrival time The formula for assessing the most dangerous moment and its severity is as follows:

[0045]

[0046] in, This marks the end of the ventilation process.

[0047] The average concentration over a specified time period is calculated to assess the overall exposure level, as shown below:

[0048] in, The start time of the specified time period; This is the end time of the specified time period.

[0049] Concentration at the end of the extraction diffusion stage and ventilation end concentration Calculate ventilation efficiency The formula is as follows:

[0050] The time required for the concentration to drop to the threshold after ventilation is turned on is calculated using linear interpolation. This provides key time points for evacuation and recovery decisions. The calculation formula is as follows:

[0051] in, The preset safe concentration value; The last concentration value is greater than or equal to The concentration; for The corresponding time series.

[0052] By comparing the differences in the above risk indicators under different ventilation velocities, the optimal ventilation velocity that can most effectively suppress hydrogen accumulation in the ceiling and reduce risk most quickly is selected.

[0053] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0054] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 10As shown, the computer device includes a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When executed by the processor, the computer program implements a method for rapid prediction and risk assessment of the spatiotemporal evolution of hydrogen concentration in tunnel ceilings. The display screen can be an LCD screen or an e-ink display screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device casing, or an external keyboard, touchpad, or mouse.

[0055] Those skilled in the art will understand that Figure 10 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0056] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0057] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0058] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0059] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0060] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0061] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for rapid prediction and risk assessment of the spatiotemporal evolution of hydrogen concentration in tunnel ceilings, characterized in that, The method includes: Obtain the location coordinates of the target monitoring point inside the tunnel and the ventilation speed of the tunnel ventilation system; The location coordinates and the ventilation speed are normalized to obtain normalized location coordinates and normalized ventilation speed. The normalized location coordinates and the normalized ventilation wind speed are input into a pre-trained two-stage spatiotemporal coupling prediction model for hydrogen concentration. The two-stage spatiotemporal coupling prediction model for hydrogen concentration includes a diffusion stage sub-model and a ventilation stage sub-model. The diffusion stage sub-model is used to output the time series of hydrogen concentration during the diffusion stage from the start of hydrogen leakage to before the start of the ventilation system. The ventilation stage sub-model is used to output the time series of hydrogen concentration during the ventilation stage from the start of the ventilation system to the end of the prediction. The hydrogen concentration time series during the diffusion stage and the hydrogen concentration time series during the ventilation stage are spliced ​​along the time axis to obtain an absolute logarithmic concentration prediction series. An exponential inverse transformation is performed on the absolute logarithmic concentration prediction series to obtain the hydrogen concentration time curve of the target monitoring point from the start of the leak to the end of the prediction. Risk indicators are extracted based on the hydrogen concentration-time curve. These risk indicators include the dangerous peak concentration, the time to reach the dangerous peak, the overall level of dangerous exposure, ventilation efficiency, and the time required to restore safety.

2. The method for rapid prediction and risk assessment of the spatiotemporal evolution of hydrogen concentration in tunnel ceilings according to claim 1, characterized in that, The training process of the two-stage spatiotemporal coupling prediction model for hydrogen concentration includes: Spatiotemporal evolution data of hydrogen concentration under different operating conditions were generated through computational fluid dynamics numerical simulation. The spatiotemporal evolution data of hydrogen concentration are truncated by a threshold and then logarithmized to obtain the absolute logarithmic concentration. The complete time series is divided into diffusion stage label sequences and ventilation stage label sequences according to the start-up time of the ventilation system. The normalized coordinates of the monitoring points and the ventilation wind speed are used as input features to train the diffusion stage sub-model and the ventilation stage sub-model, respectively.

3. The method for rapid prediction and risk assessment of the spatiotemporal evolution of hydrogen concentration in tunnel ceilings according to claim 1, characterized in that, The diffusion stage sub-model is a feedforward fully connected neural network, whose input layer receives the normalized location coordinates and the normalized ventilation velocity, and whose output layer outputs the absolute logarithmic concentration at each time point of the diffusion stage.

4. The method for rapid prediction and risk assessment of the spatiotemporal evolution of hydrogen concentration in tunnel ceilings according to claim 1, characterized in that, The ventilation stage sub-model is a feedforward fully connected neural network. Its input layer receives the normalized position coordinates, the normalized ventilation wind speed, and the concentration value at the end of the diffusion stage predicted by the diffusion stage sub-model. Its output layer outputs the absolute logarithmic concentration at each time point of the ventilation stage.

5. The method for rapid prediction and risk assessment of the spatiotemporal evolution of hydrogen concentration in tunnel ceilings according to claim 1, characterized in that, The method for extracting the dangerous peak concentration and its arrival time is as follows: determine the maximum concentration value and the time point corresponding to the maximum concentration value on the hydrogen concentration-time curve; The method for extracting the overall hazardous exposure level is as follows: calculate the average concentration of the hydrogen concentration time curve over a specified time period; The ventilation efficiency is extracted by calculating the ratio of the difference between the concentration at the end of the diffusion phase and the concentration at the end of ventilation to the concentration at the end of the diffusion phase. The method for extracting the time required to restore safety is as follows: during the ventilation phase, the time point corresponding to the hydrogen concentration dropping to a preset safety threshold is calculated by interpolation.

6. The method for rapid prediction and risk assessment of the spatiotemporal evolution of hydrogen concentration in tunnel ceilings according to claim 1, characterized in that, When training the two-stage spatiotemporal coupling prediction model for hydrogen concentration, the mean squared error is used as the loss function, the AdamW optimizer is used for optimization, and dynamic learning rate scheduling, gradient pruning, Dropout regularization and early stopping strategy are introduced.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images