A method for decoupling and predicting wheel-rail contact loads, electronic equipment, and storage medium

Abstract

This invention provides a method for decoupling and predicting wheel-rail contact loads, an electronic device, and a storage medium. The prediction method includes: establishing a three-dimensional simulation model of wheel-rail contact and determining the strain acquisition radius and acquisition points on the wheel; setting the boundary conditions, operating speed, and traction torque of the three-dimensional simulation model of wheel-rail contact; solving the three-dimensional simulation model of wheel-rail contact to obtain the strain values ​​and contact forces at each strain acquisition point of the wheel; inputting the obtained strain values ​​and contact forces at each strain acquisition point of the wheel into a neural network model for training to establish a wheel-rail contact neural network model of strain-contact force relationship; inputting unknown strain data into the wheel-rail contact neural network model of strain-contact force relationship to obtain the wheel-rail contact force; this invention can accurately identify and predict wheel-rail contact force.

Description

Technical Field

[0001] This invention relates to the field of rail vehicle technology, and in particular to a method for predicting wheel-rail contact load decoupling, an electronic device, and a storage medium. Background Technology

[0002] Currently, research on the identification of wheel-rail contact forces in trains primarily employs experimental testing methods, such as strain gauge-based on-board testing, optical and acoustic-based wheel-rail contact force testing, and multi-technology integrated wheel-rail contact force testing. These experimental testing methods require high levels of equipment integration and data acquisition accuracy, and are limited by the finite number of measurement points, making it difficult to achieve large-scale, high-precision acquisition and real-time processing of large amounts of direct data. Furthermore, given the current maturity of artificial intelligence theory and technology, there is a lack of research on combining these technologies to achieve accurate identification of wheel-rail contact forces in heavy-haul trains.

[0003] Therefore, there is an urgent need for a wheel-rail contact load decoupling prediction method, electronic equipment, and storage medium that, combined with artificial intelligence methods, can accurately identify and predict wheel-rail contact forces. Summary of the Invention

[0004] The purpose of this invention is to provide a method for decoupling and predicting wheel-rail contact load, an electronic device, and a storage medium, aiming to solve the technical problems of traditional train wheel-rail contact force testing methods, such as high requirements for equipment integration and acquisition accuracy, and difficulty in obtaining a large amount of high-precision data.

[0005] To achieve the above objectives, in a first aspect, the present invention provides a method for decoupling and predicting wheel-rail contact loads, comprising:

[0006] S1. Establish a three-dimensional simulation model of wheel-rail contact and determine the strain acquisition radius and acquisition points on the wheel;

[0007] S2. Set the boundary conditions, running speed, and traction torque of the wheel-rail contact three-dimensional simulation model;

[0008] S3. Solve the three-dimensional simulation model of wheel-rail contact to obtain the strain value and contact force at each strain acquisition point of the wheel;

[0009] S4. Using a large number of strain values ​​and contact forces from various strain acquisition points of the wheel as sample data, input them into the neural network model for training to establish a wheel-rail contact neural network model of strain-contact force relationship.

[0010] S5. Input the unknown strain data into the wheel-rail contact neural network model of the strain-contact force relationship to obtain the corresponding wheel-rail contact force.

[0011] As a further improvement to the above scheme, the steps for determining the strain acquisition radius and acquisition point on the wheel-rail in step S1 are as follows:

[0012] S11. First, obtain the strain values ​​of several nodes of the wheel and rail within a specified radius range under specified lateral load and specified vertical load, and conduct comparative analysis to obtain the relationship between the strain on the outer side of the wheel spoke and the radius of the acquisition node.

[0013] S12. Based on the relationship diagram between the strain on the outer side of the wheel spoke and the radius of the acquisition node, determine the non-contact strain acquisition radius C2 for lateral force decoupling identification and the non-contact strain acquisition radius C4 for vertical force decoupling identification, where C2 < C4.

[0014] S13. Based on the symmetrical four-group bridging method, at least eight strain acquisition points are selected on the circle corresponding to each radius.

[0015] As a further improvement to the above scheme, in step S12, based on the obtained strain acquisition radii C2 and C4, at least three more non-contact strain acquisition radii C1, C3 and C5 are selected, and C1 < C2 < C3 < C4 < C5.

[0016] As a further improvement to the above scheme, in step S2, a fixed constraint is applied to the two tracks, an angular velocity within a preset range is applied to the wheel axle, and a traction torque within a preset range is applied to the wheel axle;

[0017] The angular velocities within the preset range are stored as an angular velocity load array N, and the traction torques within the preset range are stored as a traction torque load array M.

[0018] As a further improvement to the above scheme, in step S2, when setting the boundary conditions, running speed, and traction torque of the wheel-rail contact three-dimensional simulation model,

[0019] Batch modification of model boundary conditions, running speed, and traction torque, and generation of corresponding inp files.

[0020] As a further improvement to the above scheme, in step S3, when solving the three-dimensional simulation model of wheel-rail contact,

[0021] Apply each angular velocity value in the angular velocity load array, traverse all traction torque values ​​in the traction torque load array, and obtain the strain value and contact force value of each strain acquisition point of the wheel under the corresponding load;

[0022] Apply each traction torque value in the traction torque load array, traverse all angular velocity values ​​in the angular velocity load array, and obtain the strain value and contact force value of each strain acquisition point of the wheel under the corresponding load;

[0023] The strain values ​​and contact force values ​​of each strain acquisition point of N*M groups of wheels were obtained.

[0024] As a further improvement to the above scheme, the strain value includes strain data in three directions: normal strain in the x-direction, normal strain in the y-direction, and xy shear strain.

[0025] As a further improvement to the above scheme, in step S3, after all load conditions have been calculated, the script is used to extract and store the results of all odb result files for subsequent result analysis and neural network training.

[0026] As a further improvement to the above scheme, in step S4, the coefficient of determination (R²) is used. 2 The model performance of the constructed wheel-rail contact neural network model of the strain-contact force relationship is evaluated using this method.

[0027] The formula for calculating the coefficient of determination is as follows:

[0028]

[0029] Where N is the sample size. It is the predicted output value. It is the actual output value. and These are the true value and the predicted value of the i-th sample, respectively.

[0030] As a further improvement to the above scheme, in step S4, the neural network model adopts, but is not limited to, a feedforward neural network model.

[0031] In a second aspect, the present invention also provides an apparatus comprising a memory for storing computer program instructions and a processor for executing the computer program instructions, wherein when the computer program instructions are executed by the processor, the apparatus is triggered to execute a wheel-rail contact load decoupling prediction method provided in the first aspect.

[0032] Thirdly, the present invention also provides a storage medium storing a computer program thereon, wherein, when the program is running, it controls the device where the storage medium is located to execute the wheel-rail contact load decoupling prediction method provided in the first aspect.

[0033] Because the present invention adopts the above technical solutions, the beneficial effects of this application are as follows:

[0034] This invention provides a method for decoupling and predicting wheel-rail contact loads. First, a three-dimensional simulation model of wheel-rail contact is established, and the strain acquisition radius and acquisition points on the wheel are determined. Then, boundary conditions, operating speed, and traction torque are set for the three-dimensional simulation model. Next, based on the boundary conditions, different operating speeds, and different traction forces, a large amount of strain values ​​and contact force data at various strain acquisition points on the wheel are obtained. This data is used as sample data and input into a neural network model for training, establishing a wheel-rail contact neural network model with a strain-contact force relationship. Finally, unknown strain data is input into the strain-contact force relationship wheel-rail contact neural network model. A neural network model is used to obtain wheel-rail contact force. In this invention, a large amount of sample data is obtained through simulation, and this sample data is then input into a neural network model for training, thereby obtaining a wheel-rail contact neural network model of the strain-contact force relationship. This model can then be used to predict wheel-rail contact force, meaning that this invention achieves decoupling and identification of wheel-rail contact force through a neural network. This invention overcomes the difficulties of traditional experimental testing methods, such as high requirements for equipment integration and acquisition accuracy, and the limitation of experimental measurement points, making it difficult to achieve large-scale, high-precision acquisition and real-time processing of large amounts of direct data. The wheel-rail contact force prediction method proposed in this invention significantly reduces the requirements for manpower, equipment, and environment, and has significant advantages in the extraction and processing of required data, improving data utilization and work efficiency. Furthermore, by combining artificial intelligence technology, the neural network model training achieves high prediction accuracy for the contact force at the contact points on both sides of the wheel and rail. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0036] Figure 1 This is a flowchart illustrating a wheel-rail contact load decoupling prediction method disclosed in this invention.

[0037] Figure 2 Figure 2(a) shows the 3D modeling of the rail and the wheel-rail CAD model established according to the standard of this invention, where the RF2 wheelset and the JM3 tread dimensions are shown. Figure 2 (b) shows the dimensions of a 75kg rail; Figure 2(c) shows the wheel-rail contact model built in Solidworks based on standard dimensions;

[0038] Figure 3This is a diagram showing the strain variation curve of the outer side of the wheel spoke as a function of the node radius, along with the sampling radius and node distribution. Figure 3 (a) is the curve of strain on the outer side of the wheel spoke as a function of the nodal radius; Figure 3 (b) is a schematic diagram of the wheel sampling radius and node distribution;

[0039] Figure 4 The neural network training set, validation set, and total set result curves for the lateral and vertical forces at the contact points of the left and right wheels on both sides.

[0040] The realization of the objective, functional characteristics and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0041] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0042] It should be noted that all directional indicators (such as up, down, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicator will also change accordingly.

[0043] Furthermore, in this invention, descriptions involving "first," "second," etc., are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature.

[0044] Furthermore, the technical solutions of the various embodiments of the present invention can be combined with each other, but only if they are based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0045] Example 1

[0046] See Figures 1-4 This invention provides a method for decoupling and predicting wheel-rail contact loads, comprising:

[0047] S1. Establish a three-dimensional simulation model of wheel-rail contact and determine the strain acquisition radius and acquisition points on the wheel; specifically, when establishing the three-dimensional simulation model of wheel-rail contact, the multi-point wheel-rail contact system of heavy-haul trains on existing lines is adopted, and a wheel-rail model containing complete wheelsets and rails is established based on literature reports and national standard specifications, such as... Figure 2 As shown. Specifically, according to TB1010-2016 "Railway Vehicle Wheelsets and Bearing Types and Basic Dimensions", this invention selects the RF2 wheelset, consisting of HFS type wheels and RF2 axles. The wheel tread type is JM3 type, specifically for heavy-haul trains, and the dimensions are determined according to the national standard TB / T449-2003 "Roller Wheel Flange Tread Shape". Furthermore, according to GB / T2582-2021 "Hot-rolled Rails for Railways", the rail type is determined to be 75kg rail for heavy-haul train lines. Finally, after determining the model type and specifications of each component in the numerical simulation, the 3D modeling software Solidworks is used to build and assemble the model. The specific dimensional parameters in the model are given by the respective standard documents. Detailed dimensional data of axles and wheels not marked in some figures are given in Tables 1 and 2.

[0048]

[0049]

[0050] S2. Set the boundary conditions, running speed, and traction torque of the wheel-rail contact three-dimensional simulation model;

[0051] In this embodiment, the vertical load value remains constant for the boundary conditions. A central coupling reference point is added on both sides of the wheel axle, with a value of 250.0 kN, the axle load of the existing heavy-haul railway line. According to the "Design Code for Heavy-Haul Railways," the maximum traction force of a heavy-haul train is Fmax = 570.0 kN. The interval from 120.0 kN to the maximum traction force is divided into six equal parts. The traction torque is applied through the coupling reference point at the wheel axle center. The converted traction torque intervals are listed in Table 3. The train speed interval is divided into five equal parts, with an upper limit of 80.0 km / h and a lower limit of 16.0 km / h for the existing straight-line speed. This speed is converted into wheel rotational angular velocity and applied to the wheels through the central coupling reference point. The specific values ​​are listed in Table 4. A total of 30 load conditions are formed by changing the traction force and travel speed.

[0052]

[0053]

[0054] S3. Solve the three-dimensional simulation model of wheel-rail contact to obtain the strain value and contact force at each strain acquisition point of the wheel;

[0055] S4. Using a large number of strain values ​​and contact forces from various strain acquisition points of the wheel as sample data, input them into the neural network model for training to establish a wheel-rail contact neural network model of strain-contact force relationship.

[0056] S5. Input the unknown strain data into the wheel-rail contact neural network model of the strain-contact force relationship to obtain the corresponding wheel-rail contact force;

[0057] In this invention, a large amount of sample data is obtained through simulation, and this sample data is then input into a neural network model for training, thereby obtaining a wheel-rail contact neural network model of strain-contact force relationship. This model can then be used to predict wheel-rail contact force, meaning that this invention achieves decoupling and identification of wheel-rail contact force through neural networks. This invention overcomes the difficulties of traditional experimental testing methods, such as high requirements for equipment integration and acquisition accuracy, and the limitation of experimental measurement points, making it difficult to achieve large-scale, high-precision acquisition and real-time processing of large amounts of direct data. The wheel-rail contact force prediction method proposed in this invention significantly reduces the requirements for manpower, equipment, and environment, and has significant advantages in the extraction and processing of required data, improving data utilization and work efficiency. Furthermore, by combining artificial intelligence technology, the invention achieves high prediction accuracy for the contact force at both sides of the wheel-rail contact point based on neural network model training.

[0058] As a preferred embodiment, considering the complex distribution of strain at different locations on the wheel circumference, selecting appropriate strain acquisition points is crucial. This invention determines a reasonable strain acquisition radius and acquisition points through numerical simulation calculations. Specifically, in step S1, the method for determining the strain acquisition radius and acquisition points on the wheel-rail is as follows:

[0059] S11. First, obtain the strain values ​​of several nodes on the wheel-rail system within a specified radius range under specified lateral and vertical loads, and perform comparative analysis to obtain a graph showing the relationship between the strain on the outer side of the wheel spoke and the radius of the sampling nodes. Specifically, considering that under normal circumstances, the strain value of the radial node element where the external load is applied is larger than the strain value of other node elements at the same radius, this invention selects nodes on the radial side of the external load application point with a radius of 155.0 mm-407.5 mm. Through numerical calculation, the strain distribution of these nodes under a 10.0 kN lateral load and a 10.0 kN vertical load is calculated, and a comparative analysis is performed to obtain a graph showing the relationship between the strain on the outer side of the wheel spoke and the radius of the sampling nodes. Figure 3 As shown in (a);

[0060] S12. Based on the relationship diagram between the strain on the outer side of the wheel spoke and the radius of the acquisition node, determine the non-contact strain acquisition radius C2 for lateral force decoupling identification and the non-contact strain acquisition radius C4 for vertical force decoupling identification, where C2 < C4.

[0061] Based on the obtained strain acquisition radii C2 and C4, at least three more non-contact strain acquisition radii C1, C3 and C5 are selected, where C1 < C2 < C3 < C4 < C5;

[0062] S13. Based on the symmetrical four-group bridge arrangement method, at least eight strain acquisition points are selected on the circle corresponding to each radius.

[0063] Specifically, in this embodiment, the principle for selecting the optimal strain acquisition point on the wheel is: to maximize self-disturbances while minimizing crosstalk. Figure 3 As shown in (a), among the locations where strain can be collected on the wheelset surface, the strain at the node with a radius of 297.0 mm is most affected by the lateral force, and this location can be used as the non-contact strain acquisition radius for lateral force decoupling identification. Similarly, the strain at the node with a radius of 324.0 mm is most affected by the vertical force, so this radius can be used as the non-contact strain acquisition radius for vertical force decoupling identification. Furthermore, considering the optimization of the strain data samples, this invention additionally selects three radii adjacent to the optimal strain acquisition radius for strain data acquisition: ultimately, strain acquisition points are placed on five radii: 284.0 mm, 297.0 mm, 310.0 mm, 324.0 mm, and 337.0 mm. Regarding the selection of strain gauge locations within the acquisition radius, this invention employs a superior placement method: a symmetrical four-group bridge arrangement, where the strain gauge angles are (0.0°, 36.0°, 60.0°, 96.0°). Considering symmetry, at least eight strain acquisition points are selected on each radius. Given that numerical simulation is a virtual experimental testing method, offering advantages in point selection and data extraction compared to traditional experimental testing, and also to expand the data sample for more comprehensive subsequent neural network training, this invention selects 30 uniformly spaced strain acquisition points for each data acquisition radius, including the aforementioned angles. Therefore, strain data in three directions—E11, E22, and E12—are ultimately extracted from 150 acquisition points on each wheel, representing the normal strain in the x and y directions and the xy shear strain, respectively, for subsequent neural network training. The strain acquisition point distribution is as follows: Figure 3 As shown in (b).

[0064] In a preferred embodiment, in step S2, a fixed constraint is applied to the two tracks, an angular velocity within a preset range is applied to the wheel axle, and a traction torque within a preset range is applied to the wheel axle;

[0065] The angular velocity within the preset range is stored as an angular velocity load array N, and the traction torque within the preset range is stored as a traction torque load array M;

[0066] Specifically, in this embodiment, the angular velocity load array N includes five sets of angular velocity values, namely 3.0 rad / s, 9.3 rad / s, 15.7 rad / s, 22.0 rad / s, and 28.3 rad / s; the traction torque load array M includes six sets of traction torque values, namely 9.7E+6 N·mm, 1.2E+7 N·mm, 1.5E+7 N·mm, 1.7E+7 N·mm, 1.9E+7 N·mm, and 2.2E+7 N·mm. In this invention, the above different angular velocity values ​​and different traction torque values ​​are stored in corresponding arrays, and the three-dimensional model is loaded in batches during loading, making the calculation and solution of the model convenient and fast. Specifically, when setting the boundary conditions, running speed, and traction torque of the wheel-rail contact three-dimensional simulation model, the model boundary conditions, running speed, and traction torque are modified in batches and corresponding inp files are generated; then, the program is used to submit, calculate, and monitor the inp files. After all load conditions are calculated, the script is used to extract and store the results of all odb result files for subsequent result analysis and neural network training. In the entire calculation process, the program script is used to load files and process result files in batches, which can reduce manual intervention and easily obtain a large amount of sample data. Compared with traditional experimental measurement, it saves time and effort.

[0067] In a preferred embodiment, during step S3, when solving the three-dimensional simulation model of the wheel-rail contact...

[0068] Apply each angular velocity value in the angular velocity load array, traverse all traction torque values ​​in the traction torque load array, and obtain the strain value and contact force value of each strain acquisition point of the wheel under the corresponding load;

[0069] Apply each traction torque value in the traction torque load array, traverse all angular velocity values ​​in the angular velocity load array, and obtain the strain value and contact force value of each strain acquisition point of the wheel under the corresponding load;

[0070] A total of N*M groups of wheels were obtained, including strain values ​​and contact force values ​​at each strain acquisition point.

[0071] In this embodiment, the angular velocity load array N includes five sets of angular velocity values, and the traction torque load array M includes six sets of traction torque values. Each acquisition point can obtain 30 sets of strain values ​​and contact force values ​​for each strain acquisition point of the wheel. Strain data in three directions are extracted from 150 acquisition points on each side of the wheel, resulting in 13,500 sample data for each side of the wheel. The data simulation method of this invention can easily and efficiently obtain a large amount of sample data, effectively overcoming the high requirements for equipment integration and acquisition accuracy of traditional experimental testing methods, and the limitation of experimental measurement supplementary points.

[0072] In a preferred embodiment, in step S4, the coefficient of determination (R²) is used. 2 The model performance of the constructed wheel-rail contact neural network model of the strain-contact force relationship is evaluated using this method.

[0073] The formula for calculating the coefficient of determination is as follows:

[0074]

[0075] Where N is the sample size. It is the predicted output value. It is the actual output value. and These are the true value and the predicted value of the i-th sample, respectively;

[0076] For details, see Figure 4 ,in Figure 4 (a) shows the fitted curves for the training set, validation set, and total set of the lateral force at the contact point of the left wheel; Figure 4 (b) is the training fitting curve for the lateral force at the contact point of the right wheel; Figure 4 (c) is the training fitting curve for the vertical force at the contact point of the left wheel; Figure 4 (d) shows the training fitting curve for the vertical force at the right wheel contact point; the regression results for each aggregate show R0. 2 The values ​​are all greater than 0.92, indicating that the neural network prediction results of the wheel-rail contact neural network model of the strain-contact force relationship can achieve the expected results.

[0077] In a preferred embodiment, in step S4, the neural network model employs, but is not limited to, a feedforward neural network model; in this application, a feedforward neural network model is used, including an input layer, a hidden layer, and an output layer. In a feedforward neural network, data is passed from the input layer to the hidden layer, and then processed by the hidden layer before being passed to the output layer, forming a unidirectional information flow. This structure enables the feedforward neural network to handle various complex nonlinear problems and learn the mapping relationship between input data and output targets through training.

[0078] Example 2:

[0079] The present invention also provides an apparatus comprising a memory for storing computer program instructions and a processor for executing the computer program instructions, wherein when the computer program instructions are executed by the processor, the apparatus is triggered to perform some or all of the steps in Embodiment 1;

[0080] A processor may include one or more processing units, such as an application processor (AP), a modem processor, a graphics processing unit (GPU), an image signal processor (ISP), a controller, a video codec, a digital signal processor (DSP), a baseband processor, and / or a neural network processing unit (NPU). These different processing units may be independent devices or integrated into one or more processors.

[0081] The controller can serve as the nerve center and command center of an electronic device. Based on the instruction opcode and timing signals, the controller generates operation control signals to control the fetching and execution of instructions.

[0082] The processor may also include memory for storing instructions and data. In some embodiments, the memory in the processor is a cache memory. This memory can store instructions or data that the processor has just used or that are used repeatedly. If the processor needs to use the instruction or data again, it can retrieve it directly from the memory. This avoids repeated accesses, reduces processor waiting time, and thus improves system efficiency.

[0083] Example 3:

[0084] The present invention also provides a storage medium having a computer program stored thereon, wherein, when the program is executed, it controls the device where the storage medium is located to perform some or all of the steps in Embodiment 1.

[0085] The storage medium may include high-speed RAM memory, and may also include nonvolatile memory, such as at least one disk storage device. It is understood that the storage medium can be any machine-readable medium capable of storing program code, such as random access memory (RAM), magnetic disk, hard disk, solid-state disk (SSD), or nonvolatile memory.

[0086] Those skilled in the art will understand that embodiments of the present invention can be provided as methods or storage media. Therefore, embodiments of the present invention can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0087] The above are merely preferred embodiments of the present invention and do not limit the patent scope of the present invention. All equivalent structural transformations made using the contents of the present invention's specification and drawings under the inventive concept of the present invention, or direct / indirect applications in other related technical fields, are included within the patent protection scope of the present invention.

Claims

1. A method for decoupling and predicting wheel-rail contact loads, characterized in that, include: S1. Establish a three-dimensional simulation model of wheel-rail contact and determine the strain acquisition radius and acquisition points on the wheel; Based on the relationship between the strain on the outer side of the wheel spoke and the radius of the acquisition node, the non-contact strain acquisition radius C2 for lateral force decoupling identification and the non-contact strain acquisition radius C4 for vertical force decoupling identification are determined, where C2 < C4. Based on the symmetrical four-group bridge placement method, at least eight strain acquisition points are selected on the circle corresponding to each radius; S2. Set the boundary conditions, running speed, and traction torque of the wheel-rail contact three-dimensional simulation model; specifically, apply a preset range of angular velocity and a preset range of traction torque to the wheel axle; The angular velocity within the preset range is stored as an angular velocity load array N, and the traction torque within the preset range is stored as a traction torque load array M; S3. Solve the three-dimensional simulation model of wheel-rail contact to obtain the strain values ​​and contact forces at each strain acquisition point of the wheel; specifically, apply each angular velocity value in the angular velocity load array, traverse all traction torque values ​​in the traction torque load array, and obtain the strain values ​​and contact force values ​​at each strain acquisition point of the wheel under the corresponding load; apply each traction torque value in the traction torque load array, traverse all angular velocity values ​​in the angular velocity load array, and obtain the strain values ​​and contact force values ​​at each strain acquisition point of the wheel under the corresponding load; obtain a total of N*M sets of strain values ​​and contact force values ​​at each strain acquisition point of the wheel; S4. Using a large number of strain values ​​and contact forces from various strain acquisition points of the wheel as sample data, input them into the neural network model for training to establish a wheel-rail contact neural network model of strain-contact force relationship. S5. Input the unknown strain data into the wheel-rail contact neural network model of the strain-contact force relationship to obtain the corresponding wheel-rail contact force.

2. The wheel-rail contact load decoupling prediction method according to claim 1, characterized in that, The method for obtaining the relationship between the strain on the outer side of the wheel spoke and the radius of the acquisition node in step S1 is as follows: First, obtain the strain values ​​of several nodes of the wheel and rail within a specified radius range under specified lateral and vertical loads, and then compare and analyze them to obtain the relationship between the strain on the outer side of the wheel spoke and the radius of the acquisition nodes.

3. The wheel-rail contact load decoupling prediction method according to claim 2, characterized in that, In step S12, based on the obtained strain acquisition radii C2 and C4, at least three more non-contact strain acquisition radii C1, C3 and C5 are selected, where C1 < C2 < C3 < C4 < C5.

4. A method for predicting wheel-rail contact load decoupling according to any one of claims 1-3, characterized in that, In step S2, the boundary conditions of the wheel-rail contact three-dimensional simulation model are set to apply fixed constraints to the two tracks.

5. The wheel-rail contact load decoupling prediction method according to claim 4, characterized in that, In step S2, when setting the boundary conditions, running speed, and traction torque of the wheel-rail contact three-dimensional simulation model, Batch modification of model boundary conditions, running speed, and traction torque, and generation of corresponding inp files.

6. A method for decoupling and predicting wheel-rail contact loads according to any one of claims 1-3, characterized in that, The strain values ​​include strain data in three directions: normal strain in the x-direction, normal strain in the y-direction, and xy shear strain.

7. A method for decoupling and predicting wheel-rail contact loads according to any one of claims 1-3, characterized in that, In step S3, after all load conditions have been calculated, the script is used to extract and store the results of all odb result files for subsequent result analysis and neural network training.

8. A device, characterized in that, It includes a memory for storing computer program instructions and a processor for executing the computer program instructions, wherein when the computer program instructions are executed by the processor, the device is triggered to execute a wheel-rail contact load decoupling prediction method as described in any one of claims 1-7.

9. A storage medium, characterized in that, It stores a computer program, wherein when the computer program is executed, it controls the device where the storage medium is located to perform a wheel-rail contact load decoupling prediction method as described in any one of claims 1-7.

Images