Railway vehicle interior pressure adjusting method and device and machine learning model construction method

Abstract

The invention discloses a method and device for adjusting the pressure in a railway vehicle and a machine learning model construction method, and the method comprises the following steps: obtaining the intersection starting moment and the intersection ending moment of the railway vehicle and a target section through a machine learning algorithm according to the operation data and line data of the railway vehicle at the current moment when the railway vehicle runs; according to the intersection starting moment of the rail vehicle and the target section, the moment of sending an opening instruction to the pressure wave protection device in advance is determined, and according to the intersection ending moment of the rail vehicle and the target section, the moment of delaying sending of a closing instruction to the pressure wave protection device is determined. Compared with the prior art, the machine learning algorithm is utilized to predict the intersection starting moment and the intersection ending moment of the rail vehicle and the target section, so that a large amount of hardware cost is saved, the opening and closing time of the pressure wave protection device is adaptively changed according to the actual situation, the action is more accurate, and the accuracy of the pressure wave protection device is improved. And untimely opening or advanced closing caused by fixed opening and closing time is avoided.

Description

Technical Field

[0001] This invention relates to the field of rail vehicle technology, specifically to a method and device for regulating the pressure inside a rail vehicle, and a method for constructing a machine learning model. Background Technology

[0002] When rail vehicles intersect on open tracks or pass through tunnel entrances / ventilation shafts or other locations with varying cross-sections, external pressure fluctuates dramatically and is transmitted into the vehicle through the air conditioning and ventilation system, causing discomfort such as ear pain and tinnitus in passengers. Pressure wave protection devices are designed to prevent this pressure wave transmission into the vehicle. However, existing pressure wave protection devices can only detect and activate pressure wave protection after the train has entered the tunnel, based on measured pressure differences. At the moment the train enters the tunnel, pressure waves cannot be completely prevented from entering the vehicle, still causing adverse effects on passengers and resulting in a poor passenger experience.

[0003] To address this issue, Chinese invention patent application CN110745148A discloses a train pressure wave protection control method, device, equipment, and readable storage medium. This patent involves installing signal generators at tunnel entrances and exits, and determining whether the train is entering or exiting a tunnel by identifying the signals emitted by these generators, thereby activating the pressure wave protection device before the train enters the tunnel. Its disadvantages are: the need to install signal generators at all tunnel entrances and exits along the operating line, resulting in high hardware costs; and the inability to resolve the problem of pressure waves generated when rail vehicles meet on open tracks causing adverse effects on passengers. Summary of the Invention

[0004] This invention provides a method and device for regulating the pressure inside a rail vehicle, as well as a method for constructing a machine learning model, to solve the technical problem that existing technologies require the installation of signal generating devices at all tunnel entrances and exits on the operating line, resulting in high costs.

[0005] To achieve the above objectives, the present invention adopts the following technical solution.

[0006] On the one hand, a method for regulating the pressure inside a rail vehicle is provided, comprising the following steps:

[0007] When the rail vehicle is running, the machine learning algorithm is used to obtain the start time and end time of the intersection between the rail vehicle and the target section based on the current rail vehicle operation data and line data.

[0008] The timing for sending the activation command to the pressure wave protection device is determined based on the start time of the intersection between the track vehicle and the target section, and the timing for sending the deactivation command to the pressure wave protection device is determined based on the end time of the intersection between the track vehicle and the target section. The track vehicle operation data includes the current position, speed, and acceleration of the track vehicle; the line data includes the line gradient, the start and end points of the target section; and the target section includes tunnels and / or ventilation shafts on the operating line.

[0009] This invention, by collecting real-time track vehicle operation data and track data, and utilizing machine learning algorithms to predict the start and end times of the intersection between the track vehicle and the target section, not only saves significant hardware costs but also allows the opening and closing timing of the pressure wave protection device to adapt to actual conditions, resulting in more precise operation. Furthermore, when using machine learning algorithms to predict the start and end times of the intersection between the track vehicle and the target section, parameters such as distance and speed are considered, along with real-time acceleration and track gradient. This makes the opening and closing timing of the pressure wave protection device more accurate, avoiding untimely opening or premature closing caused by fixed opening and closing times.

[0010] 1) Construction and training of intelligent prediction models

[0011] a. Data Acquisition and Preprocessing. Historical operational data acquisition: This includes rail vehicle operation data and track data. Rail vehicle operation data includes train position, track gradient, speed curve, and acceleration. Track data includes tunnel entrance / ventilation shaft location and length. Data annotation: For each target segment, the actual times of the start and end of the intersection between the front and rear of the rail vehicle and the target segment are annotated. Data preprocessing: Through normalization and noise reduction, a training dataset is constructed.

[0012] b. Model Architecture Design: An LSTM neural network architecture is adopted, leveraging its ability to process long-term dependencies in time-series data to capture dynamic features during train operation. The model input layer receives multi-dimensional feature vectors (rail vehicle operation data and track data), the hidden layer learns the temporal patterns in the data through multiple layers of LSTM units, and the output layer predicts the start and end times of the intersection between the rail vehicle and the target segment.

[0013] c. Model Training and Optimization: The Adam optimization algorithm is used with mean squared error (MSE) as the loss function. The model is iteratively trained on historical datasets until the prediction error converges (target error ≤ 0.5 seconds). Regularization techniques (such as dropout) are introduced to avoid overfitting and improve the model's generalization ability.

[0014] 2) Real-time control process based on intelligent predictive model

[0015] a. Target Distance Segment Identification and Data Acquisition: Real-time acquisition of rail vehicle operation data and track data. Based on real-time data, the prediction model determines whether a target distance segment (such as a tunnel entrance) exists in the current direction of operation, and extracts the coordinates of the first end of the target distance segment (such as the starting point of the tunnel entrance).

[0016] b. Dynamic prediction of encounter time: Input the current track vehicle operation data and track data into the trained LSTM model to calculate the first encounter time (T) from the current moment until the train head meets the target end. predict The model simultaneously considers the effects of train acceleration and deceleration, as well as the track gradient on speed. For example, it automatically extends the predicted meeting time on uphill sections to ensure prediction accuracy under dynamic conditions.

[0017] c. Dynamic Sending and Feedback Correction of Shutdown Command: A shutdown command is sent to the pressure wave protection device in advance according to a preset time, ensuring the device is in a closed state when the vehicle head meets the target end. After the pressure wave protection device activates, the actual shutdown time is fed back. The predicted and actual values ​​are fused using Kalman filtering to correct the model parameters in real time and optimize the prediction accuracy for the next cycle.

[0018] 3) Model iteration and self-optimization mechanism

[0019] a. Online learning and parameter updates: During train operation, the actual encounter time and pressure wave protection device action data are continuously collected. The model is fine-tuned periodically (e.g., after each trip) to update the weight parameters of the LSTM network and adapt to the dynamic characteristics of different lines and train types.

[0020] To more accurately determine the timing of sending the shutdown command in advance, in some embodiments, the method for determining the timing of sending the start command to the pressure wave protection device in advance includes: setting the timing of sending the start command to the pressure wave protection device in advance as the starting time of the intersection between the rail vehicle and the target section minus the preset action time and response delay time of the pressure wave protection device; wherein, the response delay time is obtained by a Kalman filter estimation algorithm.

[0021] Specifically, the preset operating time of the pressure wave protection device is manually set to ensure that the device activates before reaching the tunnel entrance. This is combined with the preset operating time (T) of the pressure wave protection device. action ) and real-time response delay (estimated in real time via Kalman filtering), calculate the lead time for sending the pressure wave protection device's shutdown command:

[0022] T advance =T predict -T action -ΔT delay

[0023] Where, ΔT delayThis refers to the response delay (such as valve mechanical delay, signal transmission delay) dynamically compensated by Kalman filtering. The response delay time ΔT delay The time difference between sending an opening command and the complete operation (e.g., the valve closing completely) of the pressure wave protection device (fresh air valve / exhaust valve) is mainly composed of the following factors:

[0024] a. Mechanical delay: The time for the valve motor to start and the transmission mechanism to move (accounting for 60-70%);

[0025] b. Signal delay: The communication transmission time of instructions from the edge computing unit to the actuator (accounting for 20-30%);

[0026] c. Environmental interference delay: Additional delays caused by environmental factors such as low temperature and vibration (accounting for less than 10%).

[0027] This delay is "dynamic" (it varies with operating conditions and equipment aging) and "observable" (it can be measured through device action feedback data), making it suitable for real-time estimation using Kalman filtering.

[0028] 2) Kalman filter model construction

[0029] a. State definition: Let the true response delay at time k be ΔT. k As a state variable in Kalman filtering, its physical meaning is the actual time difference between the sending of the command to the device and its complete operation at that moment.

[0030] b. State equations (prediction phase)

[0031] Based on the dynamic characteristics of response delay (gradual and approximately uniform change over a short period of time), the state equation is defined as:

[0032] ΔT k =ΔT k-1 +ω k-1 ;

[0033] In the formula, ΔT k Let ΔT be the true response delay at time k. k-1 Let ω be the true response delay at time k-1. k-1 The process noise follows a Gaussian distribution with a mean of 0 and a variance of Q (Q = 0.01, calibrated using historical data to reflect the random fluctuations in delay). This equation indicates that the current delay value is obtained by superimposing the previous delay value with random disturbances, which is consistent with the actual characteristic of slow delay changes.

[0034] c. Observation Equations (Update Phase)

[0035] The response delay ΔT at time k is measured using the action feedback signal of the pressure wave protection device. oβs,k(The difference between the actual action time and the command transmission time), the observation equation is defined as:

[0036] ΔT oβs,k =ΔT k +v k ;

[0037] In the formula, ΔT oβs,k The response delay observation at time k (calculated using the device's built-in position sensor and command timestamp), v k To observe the noise, it follows a Gaussian distribution with a mean of 0 and a variance of R (R = 0.02, which reflects the measurement error through sensor accuracy calibration).

[0038] 3) Kalman filter real-time compensation process

[0039] Kalman filtering achieves real-time estimation and compensation of response delay through the following 5 steps, executed once every 50ms (synchronized with the data acquisition frequency):

[0040] a. Initialization

[0041] Initial state estimation Based on the historical average delay value (e.g., 0.5s) of the same model of device;

[0042] Initial covariance matrix P0: reflects the uncertainty of the initial estimate (P0 = 0.1).

[0043] b. Prediction Steps

[0044] Predict the delay estimate and covariance at time k based on the state equation:

[0045]

[0046] In the formula, This is the prior estimate at time k (without incorporating observations). For the prior covariance (reflecting the uncertainty of the prediction), P k-1 Let be the posterior covariance matrix at time k-1.

[0047] b. Calculate the Kalman gain

[0048] Calculate the filter gain K based on the prior covariance and observation noise. k (Weighing the credibility of predictions versus observations):

[0049]

[0050] c. Update state estimation

[0051] Combined with the observed value ΔT oβs,k Update to obtain the posterior estimate at time k. (Optimal estimate):

[0052]

[0053] d. Update the covariance matrix

[0054]

[0055] 4) Compensation Effects and Applications

[0056] a. Real-time performance: Outputs the optimal delay estimate every 50ms. Ensure that the control cycle (100ms) matches that of the pressure wave protection device;

[0057] b. Accuracy: The compensated delay estimation error is ≤0.1s (60% lower than the uncompensated version), making the calculation error of the turn-off command advance ≤0.3s;

[0058] c. Application Scenarios: [The following text appears to be incomplete and requires further context:] By directly substituting the closing lead time formula, the command sending time is dynamically adjusted to ensure that the device is fully closed when the locomotive reaches the first end of the target distance segment.

[0059] Traditional methods use a fixed delay value (e.g., 0.8s) for compensation, which cannot adapt to dynamic changes in delay (e.g., when the delay increases to 1.2s in a low-temperature environment, the fixed value will lead to insufficient advance). This invention, however, achieves a closed loop of "real-time measurement - dynamic estimation - precise compensation" through Kalman filtering, significantly improving the operating accuracy of the pressure wave protection device, and is especially suitable for complex operating conditions (e.g., cross-line operation, equipment aging).

[0060] In some embodiments, the method for determining the time for delaying the sending of the shutdown command to the pressure wave protection device includes: calculating the time required for the pressure difference between the inside and outside of the vehicle to decay from the peak value to a stable value after the intersection of the track vehicle and the target section is completed according to the pressure decay curve, and adding the time to the end time of the intersection of the track vehicle and the target section to obtain the time for delaying the sending of the shutdown command to the pressure wave protection device; the pressure decay curve is a curve of the relationship between the pressure difference between the inside and outside of the vehicle and time obtained by fitting method.

[0061] After the rail vehicle and the target section intersect, the time required for the pressure difference between the inside and outside of the vehicle to decrease from the peak of the fluctuation to a stable value (≤20Pa) is calculated using an exponential decay curve model, as follows:

[0062] 1) Model building principles

[0063] After the train's rear passes through tunnel entrances, ventilation shafts, and other locations with varying cross-sections, the external pressure fluctuations gradually dissipate. The internal pressure, influenced by the coupling effect of the air conditioning and ventilation system with the external environment, exhibits an exponential decay process in its pressure difference (the pressure difference between the inside and outside of the train, denoted as ΔP). This law stems from the energy dissipation characteristics of pressure fluctuations within a confined space, characterized by initial rapid decay followed by a gradual stabilization.

[0064] 2) Pressure decay curve formula

[0065] The pressure difference decay curve model over time is defined as follows:

[0066] ΔP(t)=ΔP0·e -t / τ +ΔP ∞

[0067] In the formula, ΔP(t) is the pressure difference between the inside and outside of the vehicle at time t after the rear of the vehicle passes the second end (unit: Pa), ΔP0 is the initial peak pressure difference when the rear of the vehicle passes the second end (unit: Pa, collected in real time by pressure sensors), τ is the decay time constant (unit: s), which reflects the pressure decay rate and is related to the tunnel cross-sectional area, train speed, and the airtightness of the doors / windows; ΔP ∞ The pressure difference after stabilization (unit: Pa, usually the steady-state pressure difference in the environment, approximately ±5 Pa); e is the natural constant (≈2.718).

[0068] 3) Method for determining the decay time constant τ

[0069] τ is the core parameter of the model, determined by a combination of historical data fitting and real-time correction:

[0070] Baseline value fitting: Based on historical operating data of the same route and the same vehicle type (at least 100 tunnel / ventilation shaft passage cases), the least squares method is used to fit the pressure decay curve to obtain the baseline value of τ under different scenarios.

[0071] Dynamic correction: Real-time influencing factors (such as tunnel cross-sectional area S and train speed v) are introduced to correct the base values.

[0072]

[0073] In the formula, τ0 is the base value for this scenario, and S is the current tunnel cross-sectional area (m²). 2 S0 is the reference cross-sectional area, v is the current train speed (km / h), v0 is the reference speed (e.g., 120km / h), and α and β are correction coefficients (calibrated experimentally, α = 0.3, β = 0.2, reflecting that the larger the cross-section and the higher the speed, the faster the pressure decay).

[0074] 4) Co-optimization with other models

[0075] This model works in synergy with the LSTM prediction model and Kalman filtering: the tunnel cross-section parameters and train speed output by the LSTM model are used to dynamically correct τ; the Kalman filter compensates for the measurement error of the pressure sensor in real time, improving the acquisition accuracy of ΔP0; after each pressure stabilization, the actual stabilization time is compared with the predicted value, and the coefficients of α and β are fine-tuned by the gradient descent method to continuously optimize the model accuracy (prediction error ≤ 1s).

[0076] The above model improves the prediction accuracy of pressure stabilization time by 60% compared to traditional fixed delays (such as 8 seconds), ensuring that the protection device is activated after the pressure is fully stabilized, preventing secondary pressure fluctuations from entering the vehicle and further improving passenger comfort.

[0077] In some embodiments, the target segment further includes adjacent track vehicles traveling in opposite directions; the track data also includes the speed and acceleration of the adjacent track vehicles.

[0078] Therefore, this invention can not only avoid the adverse effects of pressure waves generated when rail vehicles pass through tunnels / ventilation shafts on passengers, but also avoid the adverse effects of pressure waves generated when open tracks intersect.

[0079] In some embodiments, before the rail vehicle stops and opens its doors, it is determined whether the following conditions are met:

[0080] F door_open ≥(P in -P out )×A+F friction And |P in -P out ≤20Pa;

[0081] If so, send an open door command;

[0082] If not, adjust the opening of the air conditioning fresh air valve and air conditioning waste air valve on the rail vehicle until the above conditions are met, and then send the door opening command.

[0083] Among them, F door_open P is the opening force of the rail vehicle door. in For the pressure inside the vehicle, P out Let F be the external pressure on the vehicle, A be the area of ​​the door, and F be the area of ​​the door. friction The preset value for the mechanical friction resistance of the car door.

[0084] In some embodiments, before the rail vehicle stops and the doors close, it is determined whether the following conditions are met:

[0085] F door_close ≥(P out -P in )×A+F friction And |Pout -P in ≤20Pa;

[0086] If so, send a door-closing command;

[0087] If not, adjust the opening of the air conditioning fresh air valve and air conditioning waste air valve on the rail vehicle until the above conditions are met, and then send a door closing command.

[0088] Among them, F door_close The closing force of the vehicle door.

[0089] This invention, when a train stops / starts at a platform in a tunnel, monitors the opening of the air conditioning fresh air valve and air conditioning exhaust valve on the rail vehicle in real time based on feedback of the door force and pressure difference. This achieves dynamic matching between the pressure difference and the door force, effectively regulating the pressure difference between the inside and outside of the train. It ensures that the pressure difference between the inside and outside of the train matches the door opening / closing force, thus solving the problem of mismatch between the door opening / closing force and the pressure difference between the inside and outside of the train when the train stops at a platform in a tunnel, which leads to the door anti-pinch function being accidentally activated and thus the door being unable to open and close normally.

[0090] In some embodiments, the machine learning algorithm is a long short-term memory network algorithm.

[0091] In some embodiments, a PID algorithm or a fuzzy control algorithm is used to adjust the opening degree of the vehicle's air conditioning fresh air valve and air conditioning exhaust valve.

[0092] On the other hand, a method for building machine learning models is provided, including the following steps:

[0093] Collect rail vehicle operation data and track data; among which, rail vehicle operation data includes train position, track gradient, speed curve and acceleration, and track data includes the location and length of tunnel entrances / ventilation shafts;

[0094] Mark the actual time when the front of the rail vehicle intersects with the target segment and the actual time when the rear of the rail vehicle intersects with the target segment for each target segment;

[0095] The above data was preprocessed to construct a training dataset;

[0096] Using preprocessed rail vehicle operation data and track data as input, and the actual times when the front of the preprocessed rail vehicle intersects with the target segment and when the rear of the rail vehicle intersects with the target segment as output, a trained machine learning model is obtained.

[0097] In another aspect, a pressure regulating device for the interior of a rail vehicle is provided, comprising a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above method.

[0098] This invention has at least the following technical effects or advantages:

[0099] 1. This invention, by collecting real-time track vehicle operation data and track data, and utilizing machine learning algorithms to predict the start and end times of the intersection between the track vehicle and the target section, not only saves significant hardware costs but also allows the opening and closing timing of the pressure wave protection device to adapt to actual conditions, resulting in more precise operation. Furthermore, when using machine learning algorithms to predict the start and end times of the intersection between the track vehicle and the target section, parameters such as distance and speed are considered, along with real-time acceleration and track gradient, making the opening and closing timing of the pressure wave protection device more accurate and avoiding untimely opening or premature closing due to fixed opening and closing times.

[0100] 2. Traditional methods use a fixed delay value (e.g., 0.8s) for compensation, which cannot adapt to dynamic changes in delay (e.g., when the delay increases to 1.2s in a low-temperature environment, the fixed value will lead to insufficient advance). This invention, however, achieves a closed loop of "real-time measurement - dynamic estimation - precise compensation" through Kalman filtering, significantly improving the operating accuracy of the pressure wave protection device, and is especially suitable for complex operating conditions (e.g., cross-line operation, equipment aging).

[0101] 3. This invention can not only avoid the adverse effects of pressure waves generated when rail vehicles pass through tunnels / ventilation shafts on passengers, but also avoid the adverse effects of pressure waves generated when open tracks intersect.

[0102] 4. When a train stops / starts at a platform in a tunnel, this invention monitors the opening of the air conditioning fresh air valve and air conditioning exhaust valve on the rail vehicle in real time based on the feedback of the door force and the pressure difference. This achieves dynamic matching between the pressure difference and the door force, effectively controlling the pressure difference between the inside and outside of the train. It ensures that the pressure difference between the inside and outside of the train matches the door opening / closing force, thus solving the problem that the door anti-pinch function is accidentally activated due to the mismatch between the door opening / closing force and the pressure difference between the inside and outside of the train when the train stops at a platform in a tunnel, which in turn prevents the door from opening and closing normally. Attached Figure Description

[0103] Figure 1 This is a schematic flowchart of a method for regulating the pressure inside a rail vehicle according to an embodiment of the present invention (when the rail vehicle is running);

[0104] Figure 2 This is a flowchart illustrating a method for regulating the pressure inside a rail vehicle according to an embodiment of the present invention (when the rail vehicle is stopped at a station). Detailed Implementation

[0105] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0106] Example 1

[0107] See Figure 1 A method for regulating the pressure inside a rail vehicle includes the following steps:

[0108] When the rail vehicle is running, the machine learning algorithm is used to obtain the start time and end time of the intersection between the rail vehicle and the target section based on the current rail vehicle operation data and line data.

[0109] The timing for sending the start command to the pressure wave protection device is determined based on the start time of the intersection between the rail vehicle and the target section, and the timing for sending the stop command to the pressure wave protection device is determined based on the end time of the intersection between the rail vehicle and the target section.

[0110] The rail vehicle operation data includes the current position, speed, and acceleration of the rail vehicle; the line data includes the line gradient, the start and end points of the target section; and the target section includes tunnels and / or ventilation shafts on the operating line.

[0111] Specifically, the steps for building and training an intelligent prediction model include:

[0112] 1. Data Acquisition and Preprocessing. Historical operational data acquisition: This includes rail vehicle operation data and track data. Rail vehicle operation data includes train position, track gradient, speed curve, and acceleration. Track data includes tunnel entrance / ventilation shaft location and length. Data annotation: For each target segment, the actual times of the start and end of the intersection between the front and rear of the rail vehicle and the target segment are annotated. Data preprocessing: Through normalization and noise reduction, a training dataset is constructed.

[0113] 2. Model Architecture Design: An LSTM neural network architecture is adopted, leveraging its ability to process long-term dependencies in time-series data to capture dynamic features during train operation. The model input layer receives multi-dimensional feature vectors (rail vehicle operation data and track data), the hidden layer learns the temporal patterns in the data through multiple layers of LSTM units, and the output layer predicts the start and end times of the intersection between the rail vehicle and the target segment.

[0114] 3. Model Training and Optimization: The Adam optimization algorithm is used with mean squared error (MSE) as the loss function. The model is iteratively trained on historical datasets until the prediction error converges (target error ≤ 0.5 seconds). Regularization techniques (such as dropout) are introduced to avoid overfitting and improve the model's generalization ability.

[0115] Real-time control processes based on intelligent predictive models include:

[0116] 1. Target Distance Segment Identification and Data Acquisition: Real-time acquisition of rail vehicle operation data and track data. Based on real-time data, the prediction model determines whether a target distance segment (such as a tunnel) exists in the current direction of operation and extracts the coordinates of the first end of the target distance segment (such as the tunnel entrance).

[0117] 2. Dynamic prediction of meeting time: Input the current track vehicle operation data and track data into the trained LSTM model to calculate the first meeting time (T) from the current time until the train head meets the target end. predict The model simultaneously considers the effects of train acceleration and deceleration, as well as the track gradient on speed. For example, it automatically extends the predicted meeting time on uphill sections to ensure prediction accuracy under dynamic conditions.

[0118] 3. Dynamic Sending and Feedback Correction of Shutdown Command: A shutdown command is sent to the pressure wave protection device in advance according to a preset time, ensuring the device is in a closed state when the vehicle head meets the target end. After the pressure wave protection device activates, the actual shutdown time is fed back. The predicted and actual values ​​are fused using Kalman filtering to correct the model parameters in real time and optimize the prediction accuracy for the next cycle.

[0119] Model iteration and self-optimization mechanisms include:

[0120] Online learning and parameter updates: During train operation, the actual encounter time and pressure wave protection device action data are continuously collected. The model is fine-tuned periodically (such as after each trip) to update the weight parameters of the LSTM network and adapt to the dynamic characteristics of different lines and train types.

[0121] As a conventional approach, the timing of sending the activation command to the pressure wave protection device in advance can be a fixed advance time set based on the start time of the intersection between the track vehicle and the target section. More preferably, the method for determining the timing of sending the activation command to the pressure wave protection device in advance includes: setting the timing of sending the activation command to the pressure wave protection device in advance as the start time of the intersection between the track vehicle and the target section minus the preset action time and response delay time of the pressure wave protection device; wherein, the response delay time is obtained through a Kalman filter estimation algorithm.

[0122] Specifically, the preset operating time of the pressure wave protection device is manually set to ensure that the device activates before reaching the tunnel entrance. This is combined with the preset operating time (T) of the pressure wave protection device. action ) and real-time response delay (estimated in real time via Kalman filtering), calculate the lead time for sending the pressure wave protection device's shutdown command:

[0123] T advance =T predict -T action -ΔT delay

[0124] Where, ΔT delay This refers to the response delay (such as valve mechanical delay, signal transmission delay) dynamically compensated by Kalman filtering. The response delay time ΔT delay The time difference between sending an opening command and the complete operation (e.g., the valve closing completely) of the pressure wave protection device (fresh air valve / exhaust valve) is mainly composed of the following factors:

[0125] a. Mechanical delay: The time for the valve motor to start and the transmission mechanism to move (accounting for 60-70%);

[0126] b. Signal delay: The communication transmission time of instructions from the edge computing unit to the actuator (accounting for 20-30%);

[0127] c. Environmental interference delay: Additional delays caused by environmental factors such as low temperature and vibration (accounting for less than 10%).

[0128] This delay is "dynamic" (it varies with operating conditions and equipment aging) and "observable" (it can be measured through device action feedback data), making it suitable for real-time estimation using Kalman filtering.

[0129] 2) Kalman filter model construction

[0130] a. State definition: Let the true response delay at time k be ΔT. k As a state variable in Kalman filtering, its physical meaning is the actual time difference between the sending of the command to the device and its complete operation at that moment.

[0131] b. State equations (prediction phase)

[0132] Based on the dynamic characteristics of response delay (gradual and approximately uniform change over a short period of time), the state equation is defined as:

[0133] ΔT k =ΔT k-1 +ω k-1 ;

[0134] In the formula, ΔT k Let ΔT be the true response delay at time k. k-1 Let ω be the true response delay at time k-1. k-1 The process noise follows a Gaussian distribution with a mean of 0 and a variance of Q (Q = 0.01, calibrated using historical data to reflect the random fluctuations in delay). This equation indicates that the current delay value is obtained by superimposing the previous delay value with random disturbances, which is consistent with the actual characteristic of slow delay changes.

[0135] c. Observation Equations (Update Phase)

[0136] The response delay ΔT at time k is measured using the action feedback signal of the pressure wave protection device. oβs,k (The difference between the actual action time and the command transmission time), the observation equation is defined as:

[0137] ΔT oβs,k =ΔT k +v k ;

[0138] In the formula, ΔT oβs,k The response delay observation at time k (calculated using the device's built-in position sensor and command timestamp), v k To observe the noise, it follows a Gaussian distribution with a mean of 0 and a variance of R (R = 0.02, which reflects the measurement error through sensor accuracy calibration).

[0139] 3) Kalman filter real-time compensation process

[0140] Kalman filtering achieves real-time estimation and compensation of response delay through the following 5 steps, executed once every 50ms (synchronized with the data acquisition frequency):

[0141] a. Initialization

[0142] Initial state estimation Based on the historical average delay value (e.g., 0.5s) of the same model of device;

[0143] Initial covariance matrix P0: reflects the uncertainty of the initial estimate (P0 = 0.1).

[0144] b. Prediction Steps

[0145] Predict the delay estimate and covariance at time k based on the state equation:

[0146]

[0147] In the formula, This is the prior estimate at time k (without incorporating observations). For the prior covariance (reflecting the uncertainty of the prediction), P k-1 Let be the posterior covariance matrix at time k-1.

[0148] e. Calculate the Kalman gain

[0149] Calculate the filter gain K based on the prior covariance and observation noise. k (Weighing the credibility of predictions versus observations):

[0150]

[0151] f. Update state estimation

[0152] Combined with the observed value ΔT oβs,k Update to obtain the posterior estimate at time k. (Optimal estimate):

[0153]

[0154] g. Update the covariance matrix

[0155]

[0156] 4) Compensation Effects and Applications

[0157] a. Real-time performance: Outputs the optimal delay estimate every 50ms. Ensure that the control cycle (100ms) matches that of the pressure wave protection device;

[0158] b. Accuracy: The compensated delay estimation error is ≤0.1s (60% lower than the uncompensated version), making the calculation error of the turn-off command advance ≤0.3s;

[0159] c. Application Scenarios: [The following text appears to be incomplete and requires further context:] By directly substituting the closing lead time formula, the command sending time is dynamically adjusted to ensure that the device is fully closed when the locomotive reaches the first end of the target distance segment.

[0160] For example, when a train approaches a tunnel entrance at 120 km / h, the predictive model calculates an encounter time of 10 seconds and a pressure wave protection device activation time of 5 seconds. Combined with a real-time response delay compensation of 0.8 seconds, the system sends an activation command 4.2 seconds in advance, ensuring that the protection device is activated when the train arrives at the tunnel entrance. This dynamic adjustment scheme can reduce unnecessary advance activation time and improve the efficiency of fresh air supply inside the train.

[0161] As a conventional approach, the timing of delaying the sending of the shutdown command to the pressure wave protection device can be a fixed delay time set based on the start time of the intersection between the rail vehicle and the target section. More preferably, the method for determining the timing of delaying the sending of the shutdown command to the pressure wave protection device includes: calculating the time required for the pressure difference between the inside and outside of the vehicle to decay from the peak value to a stable value after the intersection of the rail vehicle and the target section is completed, based on the pressure decay curve; adding this time to the end time of the intersection between the rail vehicle and the target section to obtain the timing of delaying the sending of the shutdown command to the pressure wave protection device; the pressure decay curve is a curve showing the relationship between the pressure difference between the inside and outside of the vehicle and time obtained by a fitting method.

[0162] After the rail vehicle and the target section intersect, the time required for the pressure difference between the inside and outside of the vehicle to decrease from the peak of the fluctuation to a stable value (≤20Pa) is calculated using an exponential decay curve model, as follows:

[0163] 1) Model building principles

[0164] After the train's rear passes through tunnel entrances, ventilation shafts, and other locations with varying cross-sections, the external pressure fluctuations gradually dissipate. The internal pressure, influenced by the coupling effect of the air conditioning and ventilation system with the external environment, exhibits an exponential decay process in its pressure difference (the pressure difference between the inside and outside of the train, denoted as ΔP). This law stems from the energy dissipation characteristics of pressure fluctuations within a confined space, characterized by initial rapid decay followed by a gradual stabilization.

[0165] 2) Pressure decay curve formula

[0166] The pressure difference decay curve model over time is defined as follows:

[0167] ΔP(t)=ΔP0·e -t / τ +ΔP ∞

[0168] In the formula, ΔP(t) is the pressure difference between the inside and outside of the vehicle at time t after the rear of the vehicle passes the second end (such as the tunnel exit) (unit: Pa), ΔP0 is the initial peak pressure difference when the rear of the vehicle passes the second end (unit: Pa, collected in real time by a pressure sensor), τ is the decay time constant (unit: s), which reflects the pressure decay rate and is related to the tunnel cross-sectional area, train speed, and the airtightness of the doors / windows; ΔP ∞ The pressure difference after stabilization (unit: Pa, usually the steady-state pressure difference in the environment, approximately ±5 Pa); e is the natural constant (≈2.718).

[0169] 3) Method for determining the decay time constant τ

[0170] τ is the core parameter of the model, determined by a combination of historical data fitting and real-time correction:

[0171] Baseline value fitting: Based on historical operating data of the same route and the same vehicle type (at least 100 tunnel / ventilation shaft passage cases), the least squares method is used to fit the pressure decay curve to obtain the baseline value of τ under different scenarios.

[0172] Dynamic correction: Real-time influencing factors (such as tunnel cross-sectional area S and train speed v) are introduced to correct the base values.

[0173]

[0174] In the formula, τ0 is the base value for this scenario, and S is the current tunnel cross-sectional area (m²). 2 S0 is the reference cross-sectional area, v is the current train speed (km / h), v0 is the reference speed (e.g., 120km / h), and α and β are correction coefficients (calibrated experimentally, α = 0.3, β = 0.2, reflecting that the larger the cross-section and the higher the speed, the faster the pressure decay).

[0175] Preferably, the target section also includes adjacent rail vehicles traveling in opposite directions; the track data also includes the speed and acceleration of the adjacent rail vehicles.

[0176] Preferably, such as Figure 2 As shown, before the rail vehicle stops and opens its doors, it is determined whether the following conditions are met:

[0177] F door_open ≥(P in -P out )×A+F friction And |P in -P out ≤20Pa;

[0178] If so, send an open door command;

[0179] If not, adjust the opening of the air conditioning fresh air valve and air conditioning waste air valve on the rail vehicle until the above conditions are met, and then send the door opening command;

[0180] Among them, F door_open P is the opening force of the rail vehicle door. in For the pressure inside the vehicle, P out Let F be the external pressure on the vehicle, A be the area of ​​the door, and F be the area of ​​the door. friction The preset value for the mechanical friction resistance of the car door.

[0181] Before a rail vehicle stops and its doors close, determine if the following conditions are met:

[0182] F door_close ≥(P out -P in )×A+F friction And |P out -P in ≤20Pa;

[0183] If so, send a door-closing command;

[0184] If not, adjust the opening of the air conditioning fresh air valve and air conditioning waste air valve on the rail vehicle until the above conditions are met, and then send a door closing command.

[0185] Among them, F door_close The closing force of the vehicle door.

[0186] Preferably, the opening degree of the vehicle's air conditioning fresh air valve and air conditioning exhaust valve is adjusted using a PID algorithm or a fuzzy control algorithm.

[0187] Specifically,

[0188] a. Door opening adjustment: PID control adjusts the opening of the fresh air valve, closes the waste air valve, and increases P. in ;

[0189] b. Door closing adjustment: Fuzzy control increases the opening of the exhaust valve and reduces P. in ;

[0190] c. Control parameters: PID parameters (e.g., Kp = 0.8, Ki = 0.2, Kd = 0.1), fuzzy control rule table (e.g., increase exhaust ventilation if the pressure difference is large), which is different from timed control (e.g., close the exhaust valve for 5 seconds). This step achieves dynamic parameter adjustment through algorithm.

[0191] Closed-loop feedback monitoring

[0192] a. Dual conditions: The door force meets the threshold and the pressure difference |P out -P in ≤20Pa;

[0193] b. Adjustment cycle: The pressure-force data is refreshed every 500ms to fine-tune the valve opening in real time;

[0194] c. Anti-pinch optimization: If a train passes by and causes pressure fluctuations, the system will automatically trigger emergency adjustment to prevent accidental activation of the door anti-pinch function.

[0195] In a train tunnel platform stop and door-opening scenario, the absolute pressure detection module inside and outside the train collects pressure data in real time, while the door opening force sensor provides feedback on changes in the opening force. When insufficient opening force is detected due to a pressure difference between the inside and outside of the train, the in-car pressure control module activates a PID control algorithm to dynamically adjust the opening degree of the fresh air valve and the state of the exhaust valve. For example, at a certain moment, the in-car pressure is 1312 Pa, the outside pressure is 1012 Pa, the pressure difference is -300 Pa, and the door opening force feedback indicates that the in-car pressure needs to be increased by 280 Pa. The system controls the fresh air valve opening to increase from 50% to 80%, closes the exhaust valve, and simultaneously monitors pressure changes and door opening force feedback in real time. When the pressure difference reaches -20 Pa and the door opening force meets the requirements, the adjustment stops and the door opening action is executed. This dynamic control scheme can avoid over-adjustment or under-adjustment of pressure, improving passenger comfort and door opening reliability.

[0196] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0197] Similarly, it should be understood that, in order to streamline this disclosure and aid in understanding one or more of the various aspects of the invention, in the above description of exemplary embodiments of the invention, various features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof. However, this method of disclosure should not be interpreted as reflecting an intention that the claimed invention requires more features than expressly recited in each claim. Rather, as reflected in the claims, inventive aspects lie in fewer than all features of a single foregoing disclosed embodiment. Therefore, the claims following the detailed description are hereby expressly incorporated into that detailed description, wherein each claim itself is a separate embodiment of the invention.

[0198] Those skilled in the art will understand that the modules, units, or groups of devices in the examples disclosed herein can be arranged in the device as described in this embodiment, or alternatively, can be located in one or more devices different from the device in this example. The modules in the foregoing examples can be combined into a single module or further divided into multiple sub-modules.

[0199] Those skilled in the art will understand that modules in the device of the embodiments can be adaptively changed and placed in one or more devices different from that embodiment. Modules, units, or groups in the embodiments can be combined into a single module, unit, or group, and further, they can be divided into multiple sub-modules, sub-units, or sub-groups. Except where at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or device so disclosed. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0200] Furthermore, those skilled in the art will understand that although some embodiments described herein include certain features included in other embodiments but not others, combinations of features from different embodiments are meant to be within the scope of the invention and form different embodiments.

[0201] Furthermore, some of the embodiments described herein are methods or combinations of method elements that can be implemented by a processor of a computer system or by other means of performing the functions. Therefore, a processor having the necessary instructions for implementing the methods or method elements forms means for implementing the methods or method elements. Furthermore, the elements described herein in the apparatus embodiments are examples of means for implementing the functions performed by elements for the purposes of carrying out the invention.

[0202] The various techniques described herein can be implemented in combination with hardware or software, or a combination thereof. Thus, the methods and apparatus of the present invention, or certain aspects or portions thereof, can take the form of program code (i.e., instructions) embedded in a tangible medium, such as a floppy disk, CD-ROM, hard disk, or any other machine-readable storage medium, wherein when the program is loaded into and executed by a machine such as a computer, the machine becomes an apparatus for practicing the present invention.

[0203] When the program code is executed on a programmable computer, the computing device generally includes a processor, a processor-readable storage medium (including volatile and non-volatile memory and / or storage elements), at least one input device, and at least one output device. The memory is configured to store program code; the processor is configured to execute the method of the present invention according to instructions in the program code stored in the memory.

[0204] By way of example, and not limitation, computer-readable media include computer storage media and communication media. Computer storage media stores information such as computer-readable instructions, data structures, program modules, or other data. Communication media generally embodies computer-readable instructions, data structures, program modules, or other data in the form of modulated data signals such as carrier waves or other transmission mechanisms, and includes any information delivery medium. Any combination of the above is also included within the scope of computer-readable media.

[0205] As used herein, unless otherwise specified, the use of ordinal numbers such as “first,” “second,” “third,” etc., to describe ordinary objects merely indicates different instances of similar objects and is not intended to imply that the objects being described must have a given order in time, space, ordering, or any other manner.

[0206] Although the invention has been described with reference to a limited number of embodiments, those skilled in the art will understand from the foregoing description that other embodiments are conceivable within the scope of the invention described herein. Furthermore, it should be noted that the language used in this specification has been chosen primarily for readability and instructional purposes, and not for the purpose of interpreting or limiting the subject matter of the invention. Therefore, many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the appended claims. The disclosure of the invention is illustrative and not restrictive, and the scope of the invention is defined by the appended claims.

[0207] Finally, it should be noted that this invention does not explain in detail the common knowledge recognized by those skilled in the art. The above description is only a specific embodiment of this invention and is not intended to limit this invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this invention should be included within the protection scope of this invention.

Claims

1. A method for regulating the pressure in a rail vehicle, characterized in that, The method comprises the following steps: When the rail vehicle is running, the intersection start time and the intersection end time of the rail vehicle and the target section are obtained according to the rail vehicle operation data and the line data at the current time by using a machine learning algorithm; The time for sending the opening instruction to the pressure wave protection device in advance is determined according to the intersection start time of the rail vehicle and the target section, and the time for sending the closing instruction to the pressure wave protection device in delay is determined according to the intersection end time of the rail vehicle and the target section; The rail vehicle operation data comprises the current position, speed and acceleration of the rail vehicle; the line data comprises the line slope, the start point and the end point of the target section; and the target section comprises a tunnel and / or a wind shaft on the running line.

2. The rail vehicle interior pressure regulating method according to claim 1, characterized in that, The method for determining the time for sending the opening instruction to the pressure wave protection device in advance comprises: setting the time for sending the opening instruction to the pressure wave protection device in advance as the intersection start time of the rail vehicle and the target section minus the preset action time of the pressure wave protection device and the response delay time; wherein the response delay time is obtained by using a Kalman filtering estimation algorithm.

3. The rail vehicle interior pressure regulating method according to claim 2, characterized in that, The method for determining the time for sending the closing instruction to the pressure wave protection device in delay comprises: calculating the time required for the pressure difference between the inside and outside of the rail vehicle to decay from the fluctuation peak value to the stable value after the intersection end of the rail vehicle and the target section according to a pressure decay curve, and adding the time to the intersection end time of the rail vehicle and the target section to obtain the time for sending the closing instruction to the pressure wave protection device in delay; the pressure decay curve is a curve of the relationship between the pressure difference between the inside and outside of the rail vehicle and the time obtained by using a fitting method.

4. The method of pressure regulation in a rail vehicle according to any one of claims 1-3, characterized in that: The target section further comprises a rail vehicle running in the opposite direction on the adjacent line; and the line data further comprises the speed and acceleration of the rail vehicle on the adjacent line.

5. The method of pressure regulation in a rail vehicle according to any one of claims 1-3, characterized in that: Before the rail vehicle stops and opens the door, it is determined whether the following conditions are met: F door_open ≥(P in -P out )×A+F friction and |P in -P out |≤20Pa; If yes, an opening instruction is sent; If no, the opening degree of the fresh air valve and the exhaust valve of the air conditioner on the rail vehicle is adjusted until the above conditions are met, and then an opening instruction is sent; where F door_open is the opening force of the door of the rail vehicle, P in is the pressure inside the vehicle, P out is the pressure outside the vehicle, A is the area of the door, F friction is the preset value of the mechanical friction resistance of the door.

6. The rail vehicle interior pressure regulation method according to claim 5, characterized in that: Before the rail vehicle stops and closes the door, it is determined whether the following conditions are met: F door_close ≥(P out -P in )×A+F friction and |P out -P in |≤20Pa; If yes, a closing instruction is sent; If no, the opening degree of the fresh air valve and the exhaust valve of the air conditioner on the rail vehicle is adjusted until the above conditions are met, and then a closing instruction is sent; wherein F door_close is the closing force of the vehicle door.

7. The method of pressure regulation in a rail vehicle according to any one of claims 1-3, characterized in that: The machine learning algorithm is a long short-term memory network algorithm.

8. The rail vehicle interior pressure regulation method of claim 6, wherein: The opening degree of the fresh air valve and the exhaust valve of the air conditioner on the rail vehicle is adjusted by using a PID algorithm or a fuzzy control algorithm. 9.A machine learning model construction method, characterized in that, The method comprises the following steps: Rail vehicle operation data and line data are collected; wherein the rail vehicle operation data comprises the train position, the line slope, the speed curve and the acceleration, and the line data comprises the tunnel entrance / wind shaft position and its length; The actual time when the head of the rail vehicle intersects with the target section and the actual time when the tail of the rail vehicle intersects with the target section are labeled for each target section; The above data is used to construct a training data set through data preprocessing; The preprocessed rail vehicle operation data and line data are used as the input, and the actual time when the head of the rail vehicle intersects with the target section and the actual time when the tail of the rail vehicle intersects with the target section are used as the output, so as to obtain a trained machine learning model.

10. A pressure regulating device for a rail vehicle, comprising a memory, a processor and a computer program stored on the memory, characterized in that, The processor executes the computer program to implement the steps of the method of any one of claims 1-8.

Images