A method and device for predicting wind pressure of an automatic air brake system

Abstract

The present invention provides a method and device for predicting wind pressure in an automatic air brake system. The automatic air brake system includes a locomotive brake system corresponding to a train locomotive and a vehicle foundation brake system corresponding to each vehicle section of the train. The locomotive brake system is used to transmit air volume corresponding to different operating signals to each vehicle foundation brake system through a train pipe to control the operation of each vehicle. The vehicle foundation brake system includes an auxiliary air cylinder, a brake cylinder, and a brake shoe brake device connected in sequence. The method includes: constructing a model architecture for the automatic air brake system; collecting actual pressure change data of the train pipe, auxiliary air cylinder, and brake cylinder in the automatic air brake system as training data; and training the model architecture based on the training data to obtain a simulation model for predicting pressure changes in the train pipe, auxiliary air cylinder, and brake cylinder. The method for predicting wind pressure in an automatic air brake system of the present invention can construct a simulation model that accurately predicts wind pressure changes in the train pipe, auxiliary air cylinder, and brake cylinder of a train.

Description

Technical Field

[0001] The embodiments of the present invention relate to the technical field of air brake technology, and in particular to a method and device for predicting wind pressure of an automatic air brake system. Background Art

[0002] As heavy-haul railways develop toward larger axle loads, longer train lengths, and higher transport volumes, heavy-haul technology is increasingly becoming intelligent, interconnected, and automated. The increase in train lengths results in longer air travel time in train ducts, and the asynchrony of individual train movements creates longitudinal impacts between trains. Simulating the braking system of heavy-haul trains is crucial for their longitudinal dynamics, and accurate simulation of the air brake system is crucial for resolving the brake excitation issues inherent in these longitudinal dynamics.

[0003] Research on air brake system simulation began in the 1970s, followed by the development of models for train and brake pipes in the United States, Japan, South Korea, Poland, Italy, and India. Current air brake model architectures are primarily categorized into empirical models, fluid dynamics models, and empirical fluid dynamics models. Empirical models offer simplicity and high computational efficiency, but their applicability is limited. Fluid dynamics models primarily employ differential methods, operator splitting, the finite element method (FEA), and the method of characteristics (MoC). These models are more sophisticated but computationally inefficient. Empirical fluid dynamics models are a combination of the former two, offering an intermediate computational efficiency. One method uses parallel computing to increase simulation speed, resulting in consistent results with actual results, with a maximum discrepancy of 15%. The method also compared a 2D model of a 20,000-ton heavy-load train with a traditional longitudinal dynamics model under different braking conditions, including coupler force and locomotive speed. However, the air brake model utilizes a lookup table approach, limiting simulation to air brake system pressure changes under specific braking conditions for specific train configurations.

[0004] In recent years, the rise of machine learning intelligent algorithms has also seen significant development in the rail transit industry. For example, neural networks have achieved promising results in predicting subway rail corrugation, wheel and rail wear, and lateral wheel-rail contact force. Machine learning has also been applied in the field of longitudinal vehicle dynamics. For example, some methods have introduced physical topology into neural networks, effectively improving model performance. Other methods have built autonomous vehicle trajectory planning systems based on machine learning and compared them with traditional computational methods, concluding that they are over 9,000 times faster than traditional methods for finding optimal trajectories.

[0005] Currently, there are three main approaches to describing the behavior of vehicle braking systems when they receive braking and release commands. The first is synchronous control, where all vehicles use control signals synchronized with the locomotive, regardless of signal attenuation. The second employs random motion within a controllable range, and the third uses fixed propagation velocities for braking and release waves. In reality, none of these approaches accurately describes the actual train operation. The nonlinear fluid dynamics of the air in the braking system and the delicate valve structure make the simulation of automatic air brake systems very difficult. The identification and determination of model parameters such as brake pipe friction, valve sliding friction, and orifice friction are highly nonlinear and vary from vehicle to vehicle, making them difficult to measure. Summary of the Invention

[0006] The technical problems that the present invention intends to solve include providing a method and device for predicting the wind pressure of an automatic air brake system, so as to construct a simulation model that can accurately predict the wind pressure changes of the train pipe, auxiliary air cylinder and brake cylinder of the train, and realize the wind pressure prediction of the automatic air brake system based on the simulation model.

[0007] To address the above-mentioned problems, the present invention provides a method for predicting wind pressure in an automatic air brake system. The automatic air brake system includes a locomotive brake system corresponding to a train locomotive and a vehicle foundation brake system corresponding to each vehicle section of the train. The locomotive brake system is used to transmit air volume corresponding to different operating signals to each vehicle foundation brake system through a train pipe, thereby controlling the operation of each vehicle. The vehicle foundation brake system includes an auxiliary air cylinder, a brake cylinder, and a brake shoe brake device connected in sequence. The method includes:

[0008] Constructing a model architecture of the automatic air brake system;

[0009] Collecting actual pressure change data of the train pipe, auxiliary air cylinder and brake cylinder in the automatic air brake system as training data;

[0010] Training the model architecture based on the training data to obtain a simulation model for predicting pressure changes in the train pipe, auxiliary air cylinder, and brake cylinder;

[0011] The wind pressure prediction of the automatic air brake system is realized based on the simulation model.

[0012] As an optional embodiment, the step of constructing the model architecture of the automatic air brake system includes:

[0013] A train pipe model, an auxiliary air cylinder sub-model and a brake cylinder sub-model are constructed, and the train pipe model, the auxiliary air cylinder sub-model and the brake cylinder sub-model cooperate to form the model architecture of the automatic air brake system.

[0014] As an optional embodiment, the step of constructing the train tube model includes:

[0015] The train tube model is constructed based on a deep neural network having a first number of hidden layers, the first number comprising twenty.

[0016] As an optional embodiment, the constructing of the auxiliary air cylinder sub-model includes:

[0017] The auxiliary air cylinder sub-model is constructed based on a deep neural network having a second number of hidden layers, wherein the second number includes ten.

[0018] As an optional embodiment, the step of constructing the brake cylinder sub-model includes:

[0019] The brake cylinder sub-model is constructed based on a deep neural network having a third number of hidden layers, the third number including fifteen.

[0020] As an optional embodiment, the collecting actual pressure change data of the train pipe, the auxiliary air cylinder, and the brake cylinder in the automatic air brake system as training data includes:

[0021] collecting actual pressure change data of a train pipe in the automatic air brake system to form first training data for training the train pipe model;

[0022] Collecting actual pressure change data of the auxiliary air cylinder in the automatic air brake system to form second training data for training the auxiliary air cylinder sub-model;

[0023] Actual pressure change data of a brake cylinder in the automatic air brake system is collected to form third training data for training the brake cylinder sub-model.

[0024] As an optional embodiment, collecting actual pressure change data of the train pipe in the automatic air brake system to form first training data for training the train pipe model includes:

[0025] collecting data on the pressure changes over time in the train pipes corresponding to the locomotive and each vehicle in the automatic air brake system and the distance between each vehicle and the locomotive as the first training data;

[0026] The train pipe model trained based on the first training data can predict the train pipe pressure changes corresponding to each vehicle respectively.

[0027] As an optional embodiment, the collecting actual pressure change data of the auxiliary cylinder in the automatic air brake system to form second training data for training the auxiliary cylinder sub-model includes:

[0028] The train pipe pressure change data corresponding to each vehicle output by the train pipe model and the auxiliary air cylinder pressure change data generated by the auxiliary air cylinder of each vehicle in response to the train pipe pressure change are used as the second training data, wherein the auxiliary air cylinder pressure change of each vehicle is positively correlated with the corresponding train pipe pressure change;

[0029] Among them, the auxiliary air cylinder sub-model trained based on the second training data can respectively predict the auxiliary air cylinder pressure changes corresponding to each vehicle.

[0030] As an optional embodiment, the auxiliary air cylinder is used to adjust the air pressure change of the brake cylinder of the corresponding vehicle;

[0031] The collecting actual pressure change data of the brake cylinder in the automatic air brake system to form third training data for training the brake cylinder sub-model includes:

[0032] The auxiliary cylinder pressure change data corresponding to each vehicle output by the auxiliary cylinder sub-model and the brake cylinder pressure change data generated by the brake cylinder of each vehicle in response to the auxiliary cylinder pressure change are used as the third training data;

[0033] The brake cylinder sub-model trained based on the third training data can respectively predict the brake cylinder pressure changes corresponding to each vehicle.

[0034] Another embodiment of the present invention also provides a wind pressure prediction device for an automatic air brake system. The automatic air brake system includes a locomotive brake system corresponding to a train locomotive and a vehicle foundation brake system corresponding to each vehicle section of the train. The locomotive brake system is used to transmit air volume corresponding to different operating signals to each vehicle foundation brake system through a train pipe, thereby controlling the operation of each vehicle. The vehicle foundation brake system includes an auxiliary air cylinder, a brake cylinder, and a brake shoe brake device connected in sequence. The device includes:

[0035] A building module, configured to build a model architecture of the automatic air brake system;

[0036] an acquisition module, for acquiring actual pressure change data of the train pipe, auxiliary air cylinder and brake cylinder in the automatic air brake system as training data;

[0037] A training module, configured to train the model architecture according to the training data to obtain a simulation model for predicting pressure changes in the train pipe, auxiliary air cylinder, and brake cylinder;

[0038] A prediction module is used to predict the wind pressure of the automatic air brake system through the simulation model.

[0039] Based on the disclosure of the above embodiments, it can be known that the beneficial effects of the embodiments of the present invention include the need to determine complex time-varying and nonlinear parameters such as the train's brake pipe friction, valve sliding friction and hole friction, and only consider the interaction of the air cylinders in the braking system, that is, only collecting the actual pressure changes of the train pipe, auxiliary air cylinder and brake cylinder of the train to construct a simulation model of the automatic air brake system, which is used to quickly and accurately predict the wind pressure changes of the train pipe, auxiliary air cylinder and brake cylinder, that is, to realize the wind pressure prediction of the automatic air brake system, reducing the difficulty of constructing existing simulation models of trains, and at the same time improving the calculation efficiency and accuracy of the model.

[0040] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purposes and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings.

[0041] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0043] Figure 1 2 is a structural diagram of an automatic air brake system in an embodiment of the present invention.

[0044] Figure 2 This is a flow chart of a method for predicting wind pressure in an automatic air brake system according to an embodiment of the present invention.

[0045] Figure 3 2 is a diagram illustrating the application process of the wind pressure prediction method for the automatic air brake system in an embodiment of the present invention.

[0046] Figure 4 Schematic diagram of the train pipe pressure change when the normal braking pressure is reduced by 50kPa.

[0047] Figure 5 This is a diagram showing the relationship between the time when pressure starts to drop and the vehicle serial number.

[0048] Figure 6 Schematic diagram of the relationship between the pressure drop completion time and the vehicle serial number.

[0049] Figure 7 This is a schematic diagram of the relationship between the decompression process completion time and the vehicle serial number.

[0050] Figure 8 Schematic diagram of the relationship between the maximum pressure change and the vehicle serial number.

[0051] Figure 9 This is a diagram of the train brake decompression process in an embodiment of the present invention.

[0052] Figure 10 This is a diagram of the train pipe relief and pressure boosting process in an embodiment of the present invention.

[0053] Figure 11 Schematic diagram of the multi-particle dynamics model.

[0054] Figure 12 Schematic diagram of the locomotive train pipe pressure change: (a) braking stage; (b) relief stage.

[0055] Figure 13 Figure 2 shows the simulation results of the braking stage: (a) locomotive speed; (b) maximum hook pressing force.

[0056] Figure 14 Simulation results for the relief phase: (a) locomotive speed; (b) maximum hook pressing force.

[0057] Figure 15 4 is a structural block diagram of a wind pressure prediction device for an automatic air brake system in an embodiment of the present invention. DETAILED DESCRIPTION

[0058] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings, but are not intended to limit the present invention.

[0059] It should be understood that various modifications may be made to the embodiments disclosed herein. Therefore, the following description should not be considered as limiting, but merely as an example of an embodiment. Other modifications within the scope and spirit of the present disclosure will occur to those skilled in the art.

[0060] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the present disclosure and, together with the general description of the present disclosure given above and the detailed description of the embodiments given below, serve to explain the principles of the present disclosure.

[0061] These and other characteristics of the invention will become apparent from the following description of a preferred form of embodiment given as a non-limiting example with reference to the accompanying drawings.

[0062] It should also be understood that although the invention has been described with reference to certain specific examples, those skilled in the art will be able to realize many other equivalent forms of the invention that have the characteristics recited in the claims and are therefore within the scope of protection defined thereby.

[0063] The above and other aspects, features and advantages of the present disclosure will become more apparent from the following detailed description when taken in conjunction with the accompanying drawings.

[0064] Specific embodiments of the present disclosure will be described hereinafter with reference to the accompanying drawings; however, it should be understood that the disclosed embodiments are merely examples of the present disclosure, which may be implemented in a variety of ways. Well-known and / or repetitive functions and structures are not described in detail to avoid obscuring the present disclosure with unnecessary or redundant detail. Therefore, the specific structural and functional details disclosed herein are not intended to be limiting, but rather serve merely as a basis and representative basis for the claims to teach those skilled in the art to variously employ the present disclosure with substantially any suitable detailed structure.

[0065] This description may use the phrases "in one embodiment," "in another embodiment," "in a further embodiment," or "in other embodiments," each of which may refer to one or more of the same or different embodiments according to the present disclosure.

[0066] Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

[0067] The embodiment of the present invention provides a method for predicting wind pressure of an automatic air brake system. Taking a train as an example, the structure of the automatic air brake system can be referred to Figure 1 As shown, it includes a locomotive braking system corresponding to the train locomotive and a vehicle basic braking system corresponding to each section of the train vehicle, a locomotive braking system air compressor 1, a main air cylinder 2 and a brake valve 3, wherein the brake valve 3 has three gears of braking, pressure holding and relief, which are used to control the braking state of the train. The vehicle basic braking system includes a distribution valve 4, an auxiliary air cylinder 6, a brake cylinder 5 and a brake shoe braking device 8 connected in sequence. The train pipe 7 in the figure is an air circulation channel connecting the locomotive and the vehicle, that is, the locomotive braking system transmits the air volume corresponding to different operating signals to each vehicle basic braking system through the train pipe, thereby controlling the operation of each vehicle. By studying the above structure and the braking principle and actual braking situation of the train, this embodiment determines that the train pipe, the auxiliary air cylinder and the brake cylinder are the three key components of the automatic air brake system. By predicting the wind pressure of these three components, the overall action process and state of the automatic air brake system when receiving the braking and relief instructions can be understood, and the high-precision simulation of the system can be truly achieved. This invention is different from any existing simulation method for the automatic air brake system. Accordingly, Figure 2 As shown, the wind pressure prediction method for the automatic air brake system provided in this embodiment includes:

[0068] S101: Build a model architecture for the automatic air brake system;

[0069] S102: Collecting actual pressure change data of the train pipe, auxiliary air cylinder, and brake cylinder in the automatic air brake system as training data;

[0070] S103: Training a model architecture based on the training data to obtain a simulation model for predicting pressure changes in the train pipe, auxiliary air cylinder, and brake cylinder;

[0071] S104: Predicting the wind pressure of the automatic air brake system based on the simulation model.

[0072] Based on the disclosure of the above embodiments, it can be known that the beneficial effects of this embodiment include the need to eliminate the need to determine complex time-varying and nonlinear parameters such as the train's brake pipe friction, valve sliding friction, and hole friction. Only the interaction of the air cylinders in the brake system is considered, that is, only the actual pressure changes of the train pipe, auxiliary air cylinder, and brake cylinder of the train are collected to construct a simulation model of the automatic air brake system, which is used to quickly and accurately predict the wind pressure changes of the train pipe, auxiliary air cylinder, and brake cylinder, that is, to achieve wind pressure prediction of the automatic air brake system. This method effectively reduces the difficulty of constructing existing simulation models for trains, while improving the computational efficiency and accuracy of the model. In addition, the wind pressure prediction method for the automatic air brake system in this embodiment can be applied to the simulation of the brake system of a 10,000-ton heavy-load train, and can also be used for the simulation of automatic air brake systems of 20,000 tons and above, and has a wide range of applications.

[0073] Furthermore, by conducting experimental comparisons on simulation models prepared by various model architectures, such as fine tree, fine dead vector machine, and ascending tree, based on the comparison results, the automatic air brake system of this embodiment is preferably constructed using a deep neural network. The structure of the neural network includes an input layer, a hidden layer, and an output layer. The input layer consists of p layers composed of x i The output layer consists of q neurons, represented by y j Denote j = 1, 2, 3, ..., q, where q is the number of output variables. Neural networks operate by forward propagating signals and backward propagating errors. This involves comparing the actual output with the target output. Using the backpropagation learning rule, the error is distributed across each node in each layer, and the weights of each connection are adjusted until the objective function is minimized. The relationship between input parameters and output values ​​can be expressed as follows: y = f(∑wx + b).

[0074] Where w and b represent the weight matrix and the bias vector from the adjacent layer, respectively, and f represents the activation function. In this embodiment, the activation functions of the hidden layer and the output layer are both Sigmoid functions, which are of course not unique and other types of activation functions can also be selected.

[0075] Specifically, if Figure 3 As shown in the figure, the model architecture of the automatic air brake system is constructed, including:

[0076] S105: Constructing a train pipe model, an auxiliary air cylinder sub-model, and a brake cylinder sub-model. The train pipe model, the auxiliary air cylinder sub-model, and the brake cylinder sub-model together constitute a model architecture of the automatic air brake system.

[0077] As previously mentioned, research and analysis revealed that the three key components of the air brake system are the train pipe, the auxiliary air cylinder, and the brake cylinder. To achieve high-precision simulation of the automatic air brake system while avoiding complex, time-varying, and nonlinear parameters such as brake pipe friction, valve sliding friction, and orifice friction, which are difficult to measure, this embodiment proposes that the air pressure conditions of these three components must be studied and predicted separately to effectively improve the overall computational efficiency and accuracy of the model. Therefore, the model architecture of the automatic air brake system in this embodiment, which can also be considered a simulation model, includes a train pipe sub-model, an auxiliary air cylinder sub-model, and a brake cylinder sub-model.

[0078] Specifically, the train tube model is constructed, including:

[0079] S106: Construct a train tube model based on a deep neural network having a first number of hidden layers, the first number including twenty.

[0080] Construct the auxiliary air cylinder sub-model, including:

[0081] S107: Constructing an auxiliary air cylinder sub-model based on a deep neural network having a second number of hidden layers, where the second number includes ten.

[0082] Construct the brake cylinder sub-model, including:

[0083] S108: Constructing a brake cylinder sub-model based on a deep neural network having a third number of hidden layers, where the third number includes fifteen.

[0084] The values ​​of the first quantity, the second quantity and the third quantity are not fixed and may be other values.

[0085] When collecting actual pressure change data of the train pipe, auxiliary air cylinder and brake cylinder in the automatic air brake system as training data, it includes:

[0086] S109: collecting actual pressure change data of the train pipe in the automatic air brake system to form first training data for training the train pipe model;

[0087] S110: Collecting actual pressure change data of the auxiliary cylinder in the automatic air brake system to form second training data for training the auxiliary cylinder sub-model;

[0088] S111: Collecting actual pressure change data of a brake cylinder in an automatic air brake system to form third training data for training a brake cylinder sub-model.

[0089] For example, a domestic factory conducted braking and braking tests on a 108-car train with a capacity of 10,000 tons on a 150-car braking test platform. The train consisted of one HXD1 car and 108 C80 cars. Emergency braking test data was collected at a 50Hz sampling frequency for train pipe pressure, auxiliary air cylinder pressure, and brake cylinder pressure. The sampling time was 53.8 seconds, and the brake decompression was 50kPa. To ensure effective testing, test points were arranged on the leading and trailing cars, with one test point added every 10 cars, for a total of 10 test points located on cars 1, 10, 20, 30, 40, 50, 58, 68, 78, 88, 98, and 108.

[0090] The train pipe pressure start drop time, train pipe pressure drop completion time and train pipe pressure maximum change are important indicators in the train pipe braking and relief process. Figure 4 Projecting the surface onto the time-vehicle number plane shows that the decompression time for the first car starts at 0.66 seconds, and for the 108th car at 2.04 seconds. The difference in decompression start time between the first and last cars is 1.38 seconds. The start and completion times of decompression increase with increasing car number. This is due to the propagation of the air brake wave from the locomotive to the following car, causing the air pressure in the car closest to the locomotive to begin decreasing earlier than that in the following car. The maximum change in train pipe pressure decreases gradually with car number. The maximum pressure drop in the train pipe for the first car reaches 43.7151 kPa, while that for the 108th car reaches 40.8283 kPa. This indicates that the car closest to the locomotive is more fully decompressed, while the rear train pipe, due to its distance from the locomotive, has not been fully decompressed to 50 kPa. During the decompression process, a small spike in train pipe pressure appears at the rear of the train, due to backflow caused by air impacting the rear wall of the train pipe.

[0091] Assume that the vehicle number is x, the time when the train pipe pressure starts to drop is y1, the time when the train pipe pressure ends to drop is y2, the time of the decompression stage is y3, and the maximum decompression amount of the train pipe is y4. Then the time when the train pipe pressure starts to drop and the vehicle number satisfy the linear change relationship of y1=0.63+0.013x, and the regression coefficient R is 0.997. The time when the train pipe ends to drop and the vehicle number satisfy the linear change relationship of y2=7.5+0.01x, and the regression coefficient R is 0.999. The decompression stage time and the vehicle number show a linear change relationship of y3=6.88-0.003x, and the regression coefficient R is 0.972. The maximum decompression amount of the train pipe is related to the cube of the vehicle number, and the fitting curve is y4=44+0.1x-0.001x. 2 +2.67·e -6 ·x 3 , the regression coefficient R is 0.997.

[0092] The vehicles are numbered sequentially from the leading locomotive to the rear, and the vehicle number reflects the distance between the vehicle and the locomotive. Figure 5 、 6 It can be seen that the vehicle number (distance from the exhaust port) is linearly correlated with the train pipe pressure start and completion time of each vehicle, and the slope is not much different. The locomotive train pipe pressure start and completion time is 0.63s. For each additional vehicle, the train pipe pressure start and completion time increases by 0.013s. The locomotive train pipe pressure completion time is 7.5s. For each additional vehicle, the train pipe pressure completion time increases by 0.01s. Figure 7 It can be seen that the time it takes for each vehicle to complete decompression is around 6.88, with a slope of -0.003 and a range of 0.3s. This indicates that as the vehicle number increases, the time it takes to complete decompression decreases, but the magnitude of the decrease is very low. The time it takes for each vehicle to complete the decompression process is not much different, except for the braking time. Figure 8 It can be seen that the distance from the locomotive (car number) affects the braking effect of each car. The front car exhausts more completely, but cannot fully reduce the pressure to the desired 50kPa. The actual pressure reduction of the train pipe is 44.1366kPa, with a pressure reduction efficiency of 88.27%. The actual pressure reduction of the first car is 43.7151kPa, with a pressure reduction efficiency of 87.43%. The actual pressure reduction of the 58th car is 40.6089kPa, with a pressure reduction efficiency of 81.22%. The actual pressure reduction of the 108th car is 40.8283kPa, with a pressure reduction efficiency of 81.66%. It can be seen that the exhaust efficiency of the front car is the highest, the exhaust efficiency of the middle car is the lowest, and the exhaust efficiency of the rear car is slightly higher than that of the middle car but lower than that of the front car due to the presence of the tail.

[0093] Therefore, based on the above test results, this embodiment collects actual pressure change data of the train pipe in the automatic air brake system to form first training data for training the train pipe model, including:

[0094] S112: collecting data on the pressure changes over time in the train pipes corresponding to the locomotive and each vehicle in the automatic air brake system and the distance between each vehicle and the locomotive as first training data;

[0095] Among them, the train pipe model trained based on the first training data can respectively predict the train pipe pressure changes corresponding to each vehicle.

[0096] For example, the locomotive train pipe pressure change data relative to time and the vehicle serial number (representing the distance between the vehicle and the locomotive) are used as the features of the train pipe model training, that is, the input features of the model, and the train pipe pressure change data of each vehicle behind the locomotive are used as the response, that is, the output features of the model for training. The specific training algorithm can be selected from but not limited to Levenberg-Marquardt (an estimation method for least squares estimation of regression parameters in nonlinear regression). In this embodiment, a total of 27,448 test samples are involved in the training. Figure 9 and Figure 10 It can be seen that the ball representing the predicted value falls near the surface. The mean square error (MSE) of the training set is 0.228, and the regression coefficient (R) is 0.999. The mean square error (MSE) of the test set is 0.202, and the regression coefficient (R) is 0.999, both close to 1. This indicates that the train pipe model based on the deep neural network can accurately simulate the changes in train pipe pressure during braking and easing. In other words, after training, the train pipe model can accurately predict the pressure changes in the train pipes of each vehicle.

[0097] Furthermore, actual pressure change data of the auxiliary cylinder in the automatic air brake system is collected to form second training data for training the auxiliary cylinder sub-model, including:

[0098] S113: Using the train pipe pressure change data corresponding to each vehicle output by the train pipe model and the auxiliary air cylinder pressure change data generated by the auxiliary air cylinder of each vehicle in response to the train pipe pressure change as second training data, the vehicle auxiliary air cylinder pressure change is positively correlated with the corresponding train pipe pressure change;

[0099] Among them, the auxiliary air cylinder sub-model trained based on the second training data can respectively predict the auxiliary air cylinder pressure changes corresponding to each vehicle.

[0100] For example, since the changes in the auxiliary air cylinder pressure of each vehicle are positively correlated with the changes in the train pipe pressure of each vehicle, the train pipe pressure and time of each vehicle are used as input variables of the auxiliary air cylinder sub-model, and the changes in the auxiliary air cylinder pressure of each vehicle are output. That is, the data on the changes in the train pipe pressure over time are used as the input features of the model, and the pressure change data of the auxiliary air cylinder of each vehicle are used as the output features of the model to train the auxiliary air cylinder sub-model. Based on experimental data, it is known that the pressure of the auxiliary air cylinder begins to drop almost at the same time as the train pipe. The descent time of the first car is the same. As the train number increases, the auxiliary air cylinder air pressure drop time will slightly lag behind the train pipe descent time, with a range of 0.1s. The completion time of the auxiliary air cylinder descent has little to do with the vehicle number, but is related to the basic braking system of each vehicle. The basic braking device of vehicles with less fatigue damage has high sensitivity and short braking decompression time. The basic braking device of vehicles with more fatigue damage has a longer braking time. All vehicles completed the rapid braking decompression process of the auxiliary air cylinder within 4.52s-6.94s, with a range of 2.42s. There are obvious differences in the auxiliary air cylinder decompression time of each vehicle. The auxiliary air cylinder decompression time of the middle vehicle is generally lower than that of the vehicles at both ends, indicating that the basic braking device of the middle vehicle is more sensitive than that of the two ends and has a lower degree of fatigue damage. The predicted changes in the auxiliary cylinder pressure are consistent with the actual pressure changes. The predicted mean absolute error is 0.825 kPa, the mean square error (MSE) of the training set is 0.802, and the regression coefficient (R) is 0.997. The mean square error (MSE) of the test set is 0.793, and the regression coefficient (R) is 0.997, both close to 1. This indicates that the auxiliary cylinder sub-model based on the deep neural network can accurately simulate the changes in the auxiliary cylinder pressure of vehicles during the common braking and relief processes (reducing pressure by 50 kPa) of a 10,000-ton heavy-load combination train. In other words, after training, the auxiliary cylinder sub-model can accurately predict the pressure changes in the auxiliary cylinder of each vehicle.

[0101] Furthermore, the auxiliary air cylinder in this embodiment is used to adjust the wind pressure change of the brake cylinder of the corresponding vehicle, so the pressure change of the auxiliary air cylinder can directly reflect the wind pressure change (pressure change) of the brake cylinder;

[0102] Therefore, when collecting the actual pressure change data of the brake cylinder in the automatic air brake system to form the third training data for training the brake cylinder sub-model, the following steps are included:

[0103] S114: using the auxiliary cylinder pressure change data corresponding to each vehicle output by the auxiliary cylinder sub-model and the brake cylinder pressure change data generated by the brake cylinder of each vehicle in response to the auxiliary cylinder pressure change as third training data;

[0104] The brake cylinder sub-model trained based on the third training data can respectively predict the brake cylinder pressure changes corresponding to each vehicle.

[0105] For example, the changes and time series of each vehicle's auxiliary cylinder pressure—that is, the data on the change of auxiliary cylinder pressure over time—are used as feature inputs for the brake cylinder sub-model, and the changes in vehicle brake cylinder pressure are used as responses, namely, the model's output features, for training. Experiments show that the machine learning-based brake cylinder sub-model predicts that the trends in vehicle brake cylinder pressure changes almost coincide with actual brake cylinder pressure changes. The mean square error (MSE) of the training set is 0.223, and the regression coefficient (R) is 0.999. The mean square error (MSE) of the test set is 0.222, and the regression coefficient (R) is 0.999, both close to 1. This indicates that the deep neural network brake cylinder sub-model is effective in predicting changes in vehicle brake cylinder pressure during normal braking and relief (50 kPa pressure reduction) of 10,000-ton heavy-load combination trains.

[0106] In another embodiment, in order to verify the accuracy of the deep neural network braking simulation system, based on the basic theories of longitudinal dynamics and multi-particle dynamics, a longitudinal dynamics simulation model with a neural network braking system as the braking excitation was established. At the same time, a longitudinal dynamics simulation model with a traditional braking wave model as the braking excitation was established. The locomotive running speeds calculated by the two models were compared with the test data under the commonly used braking conditions of 50kPa for 10,000-ton heavy-load trains, so as to verify the reliability and accuracy of the braking model.

[0107] Among them, the multi-mass longitudinal dynamics treats each vehicle of the train as a mass point, which includes its own weight, velocity, acceleration and displacement properties. The coupler buffer is simplified to a nonlinear spring. The model diagram is shown in Figure 11 As shown, assuming that the heavy-load train consists of n cars, the force analysis of each car is performed using Newton's second law, and the dynamic equation is shown as follows:

[0108]

[0109] Among them, m i represents the vehicle mass, v i Fq represents the speed of the i-th vehicle, and t represents time. i and Fd i are the locomotive's traction force and electric braking force, respectively, as shown in the following formula.

[0110] Fq i (v i ,t)=fq(v i )×rang q (t),i=j1,j2,...,j n ,

[0111] Fd i (v i ,t)=fd(v i )×rang d(t),i=j1,j2,...,j n ,

[0112] Fb i represents the air braking force, which is obtained by solving the neural network braking system simulation model, Fw i is the running resistance. The locomotive running resistance and the freight car running resistance are shown in the following formula.

[0113] F wi =0.001×(1.20+0.065×v+0.00279×v 2 )×M×g,

[0114] F wi =0.001×(0.92+0.048×v+0.00125×v 2 )×M×g.

[0115] Fc i Fc represents the coupling force between the i-th vehicle and the i+1-th vehicle, i-1 It represents the coupling force between the i-th car and the i-1-th car. When i=1, Fc i-1 =0, when i=n, ​​Fc i = 0. All vehicles have air braking force, while traction and electric braking force only exist in locomotives, and traction and braking force generally do not exist at the same time. fq and fd are the dynamic characteristic functions of traction and electric braking respectively. q and rang d Represent the output ratio of traction force and electric braking force respectively. j1,j2,...,j n The vehicle serial number of the truck.

[0116] Additional resistance refers to the additional resistance that a locomotive or vehicle encounters when passing through a specific section of road. For example, when a train passes through an uphill section, it will encounter slope resistance F. Wr , when passing through the curve, it is subject to the additional resistance F of the curve Wc Unlike basic resistance, additional resistance is less affected by vehicle type and is mainly related to line conditions. The value of additional resistance per unit slope is equal to the thousandth of the slope, as shown in the formula: F Wr =i.

[0117] The additional resistance of the unit curve is shown as: F Wc =0.001×(600 / R)×M×g

[0118] Where: M is the mass of the particle, t; g is the acceleration due to gravity, m / s 2 ; i is the slope in thousandths, and R is the radius of curvature.

[0119] The train starting resistance refers to the resistance that the train needs to overcome when it goes from stationary to moving. The locomotive unit starting resistance is 5.0N / t, and the vehicle unit starting resistance is 3.5N / t.

[0120] The calculation formula of coupler force is shown as follows:

[0121]

[0122] The locomotive coupler uses QKX100 elastic clay buffer, and the vehicle uses MT-2 buffer. The QKX100 elastic clay buffer has a strength of 3430kN, a coupling gap of 11.5mm, a buffer impedance of 2500kN, a maximum stroke of 82.5mm, and a capacity of 100kJ. The MT-2 buffer has an impedance of 2270kN, a maximum stroke of 83mm, and a capacity of 50kJ. The train pipe pressure changes as follows: Figure 12 As shown, the maximum decompression is 48kPa.

[0123] Depend on Figure 13 (a) It can be seen that during the braking stage of the 10,000-ton train, the speed of the test locomotive shows a trend of first increasing and then decreasing under the influence of inertia on a long slope. The step-like change in speed is due to the fact that the readings on the locomotive instrument panel can only be integers, resulting in a jump in speed. In fact, the speed should be a continuous physical change. The blue curve is the locomotive speed predicted by the longitudinal dynamics model using a deep neural network, and the black curve is the locomotive speed predicted by the longitudinal dynamics model using a traditional braking wave. It can be seen that the locomotive speed change trends predicted by the two models are the same as the actual values. The error of the locomotive speed of the longitudinal dynamics model using the traditional braking wave gradually accumulates with the increase of time. The predicted locomotive speed descent slope is slightly higher than the actual locomotive running speed. The locomotive speed of the longitudinal dynamics model using a deep neural network is interspersed between the small red balls, indicating that the locomotive speed of the model using a deep neural network is more accurate. Figure 13 (b) The maximum hook-pressing force generated by each coupler during braking using the deep neural network model is shown in the blue curve. The maximum hook-pressing force occurs on hook No. 96, reaching a maximum value of 264.8684 kN. The maximum hook-pressing force calculated using the traditional braking wave model occurs on hook No. 104, reaching a maximum value of 339.4245 kN. The maximum hook-pressing force calculated using the deep neural network longitudinal dynamics model is smaller than that calculated using the traditional braking wave model, with a range of 75.5561 kN. During braking, the hook-pulling force is below 10 kN, and the difference between the two models is minimal.

[0124] Depend on Figure 14(a) It can be seen that at the beginning of the relief stage, the sum of the braking force and resistance of the train is greater than the gravity component and inertia force of the train going downhill, and the locomotive speed shows a decreasing trend. Later, as the train is relieved, the train pipe pressure gradually increases, the braking force decreases, and the train speed gradually increases under the action of the gravity component of the long downhill slope. The locomotive speed change trends predicted by the two models are the same as the actual values. The locomotive speed predicted by the model using traditional braking waves drops faster than the actual locomotive speed at the beginning of the relief stage, and the slope of the train speed increase in the later stage of the relief is greater than the actual locomotive speed. The locomotive speed curve predicted by the model using deep neural networks passes through the point where the actual locomotive speed jumps. It can be seen that the locomotive speed predicted by the model using deep neural networks is more accurate. Figure 14 (b) As can be seen, during the release process, the maximum coupler force predicted by the deep neural network model appears to be 197.1052 kN for hook No. 2. The maximum coupler force predicted by the traditional brake wave model also appears to be 195.6643 kN for hook No. 2. The difference between the two is relatively small. The maximum coupler force predicted by the machine learning model is generally smaller than that of the traditional brake wave model, but the vibration amplitude is greater. When the step size is 0.1 ms, the traditional fluid dynamics model predicts a simulation time of 202.75 s for 30 s of braking; this simulation time is 6.75 times slower than real time. The neural network-based braking system simulation model achieves a simulation time of 0.0695 s, which is 2971 times more efficient than the traditional fluid dynamics model.

[0125] Therefore, a longitudinal dynamics model using a neural network braking system as the braking excitation can accurately predict the locomotive speed and coupler force for both normal braking and relief conditions of a 10,000-ton heavy-haul train with a 50 kPa pressure reduction. This further validates the accuracy and effectiveness of the deep neural network braking system in practical engineering applications. Furthermore, neural networks have the ability to learn autonomously. In the future, with additional training samples, the deep neural network braking system is expected to be applicable to predicting the longitudinal dynamics of heavy-haul trains with longer train lengths or emergency braking.

[0126] like Figure 15 As shown, another embodiment of the present invention also provides a wind pressure prediction device 100 for an automatic air brake system. The automatic air brake system includes a locomotive brake system corresponding to a train locomotive and a vehicle foundation brake system corresponding to each vehicle section of the train. The locomotive brake system is used to transmit air volume corresponding to different operation signals to each vehicle foundation brake system through a train pipe, thereby controlling the operation of each vehicle. The vehicle foundation brake system includes an auxiliary air cylinder, a brake cylinder, and a brake shoe brake device connected in sequence, wherein the device includes:

[0127] A building module, configured to build a model architecture of the automatic air brake system;

[0128] an acquisition module, for acquiring actual pressure change data of the train pipe, auxiliary air cylinder and brake cylinder in the automatic air brake system as training data;

[0129] A training module, configured to train the model architecture according to the training data to obtain a simulation model for predicting pressure changes in the train pipe, auxiliary air cylinder, and brake cylinder;

[0130] The prediction module is used to predict the wind pressure of the automatic air brake system through the simulation model.

[0131] As an optional embodiment, the step of constructing the model architecture of the automatic air brake system includes:

[0132] A train pipe model, an auxiliary air cylinder sub-model and a brake cylinder sub-model are constructed, and the train pipe model, the auxiliary air cylinder sub-model and the brake cylinder sub-model cooperate to form the model architecture of the automatic air brake system.

[0133] As an optional embodiment, the step of constructing the train tube model includes:

[0134] The train tube model is constructed based on a deep neural network having a first number of hidden layers, the first number comprising twenty.

[0135] As an optional embodiment, the constructing of the auxiliary air cylinder sub-model includes:

[0136] The auxiliary air cylinder sub-model is constructed based on a deep neural network having a second number of hidden layers, wherein the second number includes ten.

[0137] As an optional embodiment, the step of constructing the brake cylinder sub-model includes:

[0138] The brake cylinder sub-model is constructed based on a deep neural network having a third number of hidden layers, the third number including fifteen.

[0139] As an optional embodiment, the collecting actual pressure change data of the train pipe, the auxiliary air cylinder, and the brake cylinder in the automatic air brake system as training data includes:

[0140] collecting actual pressure change data of a train pipe in the automatic air brake system to form first training data for training the train pipe model;

[0141] Collecting actual pressure change data of the auxiliary air cylinder in the automatic air brake system to form second training data for training the auxiliary air cylinder sub-model;

[0142] Actual pressure change data of a brake cylinder in the automatic air brake system is collected to form third training data for training the brake cylinder sub-model.

[0143] As an optional embodiment, collecting actual pressure change data of the train pipe in the automatic air brake system to form first training data for training the train pipe model includes:

[0144] collecting data on the pressure changes over time in the train pipes corresponding to the locomotive and each vehicle in the automatic air brake system and the distance between each vehicle and the locomotive as the first training data;

[0145] The train pipe model trained based on the first training data can predict the train pipe pressure changes corresponding to each vehicle respectively.

[0146] As an optional embodiment, the collecting actual pressure change data of the auxiliary cylinder in the automatic air brake system to form second training data for training the auxiliary cylinder sub-model includes:

[0147] The train pipe pressure change data corresponding to each vehicle output by the train pipe model and the auxiliary air cylinder pressure change data generated by the auxiliary air cylinder of each vehicle in response to the train pipe pressure change are used as the second training data, wherein the auxiliary air cylinder pressure change of each vehicle is positively correlated with the corresponding train pipe pressure change;

[0148] Among them, the auxiliary air cylinder sub-model trained based on the second training data can respectively predict the auxiliary air cylinder pressure changes corresponding to each vehicle.

[0149] As an optional embodiment, the auxiliary air cylinder is used to adjust the air pressure change of the brake cylinder of the corresponding vehicle;

[0150] The collecting actual pressure change data of the brake cylinder in the automatic air brake system to form third training data for training the brake cylinder sub-model includes:

[0151] The auxiliary cylinder pressure change data corresponding to each vehicle output by the auxiliary cylinder sub-model and the brake cylinder pressure change data generated by the brake cylinder of each vehicle in response to the auxiliary cylinder pressure change are used as the third training data;

[0152] The brake cylinder sub-model trained based on the third training data can respectively predict the brake cylinder pressure changes corresponding to each vehicle.

[0153] Another embodiment of the present invention further provides an electronic device, including:

[0154] at least one processor; and,

[0155] a memory communicatively connected to the at least one processor; wherein,

[0156] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to implement the wind pressure prediction method for the automatic air brake system as described in any one of the embodiments above.

[0157] Furthermore, an embodiment of the present invention provides a storage medium storing a computer program that, when executed by a processor, implements the aforementioned method for predicting wind pressure in an automatic air brake system. It should be understood that each solution in this embodiment has the corresponding technical effects of the aforementioned method embodiments and will not be further elaborated here.

[0158] Furthermore, an embodiment of the present invention also provides a computer program product, which is tangibly stored on a computer-readable medium and includes computer-readable instructions, which, when executed, enable at least one processor to perform a wind pressure prediction method for an automatic air brake system such as the one in the embodiment described above.

[0159] It should be noted that the computer storage medium of the present invention may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. Computer-readable media may, for example, be, but are not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems, or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection having one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage medium, a magnetic storage medium, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. In the present invention, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries computer-readable program code. This propagated data signal may take a variety of forms, including, but not limited to, electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program configured for use by or in conjunction with an instruction execution system, system, or device. Program code embodied on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, antenna, optical cable, RF, or any suitable combination thereof.

[0160] In addition, it will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage and optical storage, etc.) containing computer-usable program code.

[0161] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A system that specifies the functions of a box or boxes.

[0162] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including an instruction system that is implemented in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0163] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0164] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

[0165] The above embodiments are merely exemplary embodiments of the present invention and are not intended to limit the scope of the present invention. The scope of protection of the present invention is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present invention within the spirit and scope of protection of the present invention, and such modifications or equivalent substitutions shall also be deemed to fall within the scope of protection of the present invention.

Claims

1. A method for predicting wind pressure in an automatic air brake system, wherein the automatic air brake system comprises a locomotive brake system corresponding to a train locomotive and a vehicle foundation brake system corresponding to each train car, wherein the locomotive brake system is used to deliver air volume corresponding to different operation signals to each vehicle foundation brake system through a train pipe, thereby controlling the operation of each vehicle, and the vehicle foundation brake system comprises an auxiliary air cylinder, a brake cylinder, and a brake shoe brake device connected in sequence, characterized in that: The method comprises: Constructing a model architecture of the automatic air brake system; Collecting actual pressure change data of the train pipe, auxiliary air cylinder and brake cylinder in the automatic air brake system as training data; Training the model architecture based on the training data to obtain a simulation model for predicting pressure changes in the train pipe, auxiliary air cylinder, and brake cylinder; Implementing wind pressure prediction of the automatic air brake system based on the simulation model; The model architecture of the automatic air brake system is constructed, including: Constructing a train pipe model, an auxiliary air cylinder sub-model, and a brake cylinder sub-model, wherein the train pipe model, the auxiliary air cylinder sub-model, and the brake cylinder sub-model cooperate to form a model architecture of the automatic air brake system; The trained train pipe model can predict the pressure changes of each vehicle's train pipes, the trained auxiliary air cylinder sub-model can predict the pressure changes of each vehicle's auxiliary air cylinders, and the trained brake cylinder sub-model can predict the pressure changes of each vehicle's brake cylinders. The train pipe pressure change data corresponding to each vehicle output by the train pipe model is used as the input variable of the auxiliary air cylinder sub-model, and the auxiliary air cylinder pressure change data corresponding to each vehicle output by the auxiliary air cylinder sub-model is used as the input variable of the brake cylinder sub-model.

2. The wind pressure prediction method for an automatic air brake system according to claim 1, characterized in that: The train tube model is constructed, comprising: The train tube model is constructed based on a deep neural network having a first number of hidden layers, the first number comprising twenty.

3. The wind pressure prediction method for an automatic air brake system according to claim 1, characterized in that: The construction of the auxiliary air cylinder sub-model includes: The auxiliary air cylinder sub-model is constructed based on a deep neural network having a second number of hidden layers, wherein the second number includes ten.

4. The wind pressure prediction method for an automatic air brake system according to claim 1, characterized in that: The construction of the brake cylinder sub-model includes: The brake cylinder sub-model is constructed based on a deep neural network having a third number of hidden layers, the third number including fifteen.

5. The wind pressure prediction method for an automatic air brake system according to claim 1, characterized in that: The collecting of actual pressure change data of the train pipe, auxiliary air cylinder and brake cylinder in the automatic air brake system as training data includes: collecting actual pressure change data of a train pipe in the automatic air brake system to form first training data for training the train pipe model; Collecting actual pressure change data of the auxiliary air cylinder in the automatic air brake system to form second training data for training the auxiliary air cylinder sub-model; Actual pressure change data of a brake cylinder in the automatic air brake system is collected to form third training data for training the brake cylinder sub-model.

6. The wind pressure prediction method for an automatic air brake system according to claim 5, characterized in that: The collecting actual pressure change data of the train pipe in the automatic air brake system to form first training data for training the train pipe model includes: collecting data on the pressure changes over time in the train pipes corresponding to the locomotive and each vehicle in the automatic air brake system and the distance between each vehicle and the locomotive as the first training data; The train pipe model trained based on the first training data can predict the train pipe pressure changes corresponding to each vehicle respectively.

7. The wind pressure prediction method for an automatic air brake system according to claim 6, characterized in that: The collecting of actual pressure change data of the auxiliary cylinder in the automatic air brake system to form second training data for training the auxiliary cylinder sub-model includes: The train pipe pressure change data corresponding to each vehicle output by the train pipe model and the auxiliary air cylinder pressure change data generated by the auxiliary air cylinder of each vehicle in response to the train pipe pressure change are used as the second training data, wherein the auxiliary air cylinder pressure change of each vehicle is positively correlated with the corresponding train pipe pressure change; Among them, the auxiliary air cylinder sub-model trained based on the second training data can respectively predict the auxiliary air cylinder pressure changes corresponding to each vehicle.

8. The wind pressure prediction method for an automatic air brake system according to claim 7, characterized in that: The auxiliary air cylinder is used to adjust the air pressure change of the brake cylinder of the corresponding vehicle; The collecting actual pressure change data of the brake cylinder in the automatic air brake system to form third training data for training the brake cylinder sub-model includes: The auxiliary cylinder pressure change data corresponding to each vehicle output by the auxiliary cylinder sub-model and the brake cylinder pressure change data generated by the brake cylinder of each vehicle in response to the auxiliary cylinder pressure change are used as the third training data; The brake cylinder sub-model trained based on the third training data can respectively predict the brake cylinder pressure changes corresponding to each vehicle.

9. A wind pressure prediction device for an automatic air brake system, the automatic air brake system comprising a locomotive brake system corresponding to a train locomotive and a vehicle foundation brake system corresponding to each vehicle section of the train, the locomotive brake system being used to deliver air volumes corresponding to different operating signals to each vehicle foundation brake system through a train pipe, thereby controlling the operation of each vehicle, the vehicle foundation brake system comprising an auxiliary air cylinder, a brake cylinder, and a brake shoe brake device connected in sequence, characterized in that: The device comprises: A building module, configured to build a model architecture of the automatic air brake system; an acquisition module, for acquiring actual pressure change data of the train pipe, auxiliary air cylinder and brake cylinder in the automatic air brake system as training data; A training module, configured to train the model architecture according to the training data to obtain a simulation model for predicting pressure changes in the train pipe, auxiliary air cylinder, and brake cylinder; a prediction module, configured to predict the wind pressure of the automatic air brake system using the simulation model; The model architecture of the automatic air brake system is constructed, including: Constructing a train pipe model, an auxiliary air cylinder sub-model, and a brake cylinder sub-model, wherein the train pipe model, the auxiliary air cylinder sub-model, and the brake cylinder sub-model cooperate to form a model architecture of the automatic air brake system; The trained train pipe model can predict the pressure changes of each vehicle's train pipes, the trained auxiliary air cylinder sub-model can predict the pressure changes of each vehicle's auxiliary air cylinders, and the trained brake cylinder sub-model can predict the pressure changes of each vehicle's brake cylinders. The train pipe pressure change data corresponding to each vehicle output by the train pipe model is used as the input variable of the auxiliary air cylinder sub-model, and the auxiliary air cylinder pressure change data corresponding to each vehicle output by the auxiliary air cylinder sub-model is used as the input variable of the brake cylinder sub-model.

Images