Real-time monitoring method and system for the wheel-rail matching state and wheel wear of rail vehicles

Abstract

The present invention discloses a method and system for real-time monitoring of the wheel-rail matching state and wheel wear of a rail vehicle. The method comprises the following steps: establishing and performing wear simulation processing on a multi-body dynamics model to obtain vehicle wheel wear at different mileages; collecting axle box vibration accelerations of wheels at different wheel wear levels and performing feature analysis on the axle box vibration accelerations, and extracting axle box lateral vibration accelerations from the axle box vibration accelerations as a training data set; calculating that there is a positive relationship between the wheel wear level and the equivalent taper, and using a correlation model to verify that there is a mapping relationship between wheel wear, equivalent taper, and axle box lateral vibration acceleration; constructing a pre-trained spectrum model, and inputting the training data set into the pre-trained spectrum model for feature training to obtain a spectrum model; collecting a set of axle box lateral vibration accelerations as a test data set and inputting the test data set into the spectrum model to obtain the vehicle wheel-rail matching state and wheel wear level under the axle box lateral vibration accelerations.

Description

Technical Field

[0001] The present invention relates to the technical field of high-speed train wheels and rails, and in particular to a method and system for real-time monitoring of the wheel-rail matching state and wheel wear of a rail vehicle. Background Art

[0002] In traditional wheel-rail profile matching design, low contact stress between wheels and rails is an important indicator. Low contact stress can effectively reduce wheel-rail wear and rolling contact fatigue. With the popularization of high-speed trains, wheel-rail profile matching design is more concerned with reasonable equivalent taper, which can give wheels a higher critical speed and better curve negotiating ability. Scholars agree that parameters such as wheel tread profile and rail profile directly change the wheel-rail contact geometry, thereby affecting the dynamic performance of high-speed trains. Therefore, it can be seen that the running quality of high-speed trains is directly related to wheel profile matching. By monitoring changes in wheel-rail profile and maintaining good wheel-rail profile matching performance, the safe operation of high-speed trains can be guaranteed.

[0003] At present, railway management departments usually control the equivalent wheel-rail taper by wheel boring and rail grinding. Wheel boring and rail grinding are the responsibilities of the vehicle depot and the engineering section respectively. The vehicle depot plans to boring the wheels within a predetermined period according to the standard wheel-rail profile, while the engineering section uses small mechanical equipment to perform regular rail grinding. Although this operation method has improved the wheel-rail matching relationship and improved the train running quality to a certain extent, due to the difference between the wear profile of the in-service wheel and rail and the standard wheel-rail profile, the wheel-rail matching performance at this time is subject to great uncertainty; sometimes it even deteriorates rapidly, leading to frequent maintenance, which greatly reduces the service life of the wheel and rail and causes a huge waste of resources.

[0004] By monitoring the vehicle dynamic response induced by wheel-rail excitation in real time and establishing a mapping relationship between the two, a scientific reasoning algorithm is constructed to invert the wheel-rail service status. Based on this, an opportunistic maintenance strategy for maintaining the wheel-rail profile matching performance is proposed. This not only provides important guarantees for the safe operation of high-speed trains, but also realizes intelligent operation and maintenance based on the status of the wheel-rail system, thereby significantly reducing maintenance costs. Summary of the Invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, the present invention proposes a method and system for real-time monitoring of the wheel-rail matching status and wheel wear of a railway vehicle.

[0006] According to a first embodiment of the present invention, a method for real-time monitoring of wheel-rail matching status and wheel wear of a railway vehicle comprises:

[0007] Establishing a vehicle multi-body dynamics model, and performing wear simulation on the multi-body dynamics model to obtain vehicle wheel wear at different mileages;

[0008] collecting axle box vibration accelerations of wheels at different wheel wear levels, performing feature analysis on the axle box vibration accelerations, determining that the axle box lateral vibration acceleration has the greatest impact on vehicle wear changes, and extracting the axle box lateral vibration accelerations from the axle box vibration accelerations as a training data set;

[0009] According to the calculation, there is a positive relationship between the wheel wear degree and the equivalent taper, and a correlation model is used to verify that there is a mapping relationship between the wheel wear, the equivalent taper and the lateral vibration acceleration of the axle box;

[0010] Constructing a pre-trained spectrum model, inputting the training data set into the pre-trained spectrum model for feature training to obtain a spectrum model;

[0011] A set of axle box lateral vibration accelerations is collected as a test data set and input into the spectrum model to obtain the vehicle wheel-rail matching state and wheel wear degree under the axle box lateral vibration acceleration.

[0012] According to a method for real-time monitoring of the wheel-rail matching state and wheel wear of a railway vehicle of the present invention, first, a vehicle multi-body dynamics model is constructed by using dynamics simulation software, and the vehicle multi-body dynamics model is simulated using a wear model to obtain the vehicle wheel wear under different mileages; the axle box vibration acceleration of the wheel under different wheel wear is collected, and the characteristics of the axle box vibration acceleration are analyzed to obtain that the axle box lateral vibration acceleration has the greatest impact on the vehicle wear change; the axle box vibration acceleration is processed to extract the axle box lateral vibration acceleration from the axle box vibration acceleration, and the axle box vertical vibration acceleration is eliminated, and the extracted axle box lateral vibration acceleration is used as a training data set; according to the calculation, it is obtained that there is a positive relationship between the wheel wear degree and the equivalent taper, and the relevant model is used to calculate the relationship between the wheel wear degree and the equivalent taper. Verify that there is a mapping relationship between wheel wear, equivalent conicity and the axle box lateral vibration acceleration. The greater the wheel wear, the greater the equivalent conicity, and the greater the equivalent conicity, the poorer the wheel-rail matching state, so it is concluded that the greater the vehicle wear, the worse the wheel-rail matching state; construct a pre-trained spectrum model, input the training data set into the pre-trained spectrum model for feature training to obtain a spectrum model; the spectrum model is the final model for real-time acquisition of vehicle wheel-rail matching state and wheel wear degree, by regularly collecting a set of axle box lateral vibration acceleration as a test data set and inputting it into the spectrum model, the vehicle wheel-rail matching state and wheel wear degree under the axle box lateral vibration acceleration are obtained, so that the vehicle wheel-rail matching state and wheel wear degree during operation can be collected in real time.

[0013] According to some embodiments of the present invention, the wear simulation processing is specifically performed by processing the multi-body dynamics model using the Archard wear model;

[0014] In the Archard wear model, the wear volume at a certain contact point of the wheel can be expressed as:

[0015]

[0016] Among them, V wear is the material wear volume; k w is the wear coefficient, N is the wheel-rail normal force, d is the related slip

[0017] Dynamic displacement, H is the hardness of the wheel material;

[0018] The contact spot is discretized into countless strip rectangles, and the area of each rectangle is ΔS. The wear depth calculation within the rectangle can be expressed as:

[0019]

[0020]

[0021] Where, ΔV wear (x, y) is the wear volume of the particle; k w is the wear coefficient, ΔN(x, y) is the normal force of the particle, Δd is the sliding displacement in the rectangle where the contact point is located, s x is the sliding velocity of the particle in the x direction, s y is the sliding velocity of the particle in the y direction, V C is its passing speed within the contact patch, and H is the hardness of the wheel material.

[0022] According to some embodiments of the present invention, the characteristic analysis of the axle box vibration acceleration is performed, specifically,

[0023] An analysis of the vertical acceleration and lateral acceleration collected on the axle box shows that the average value of the vertical acceleration is significantly smaller than the lateral acceleration;

[0024] Analysis of vertical acceleration at different mileages under the same track conditions revealed that the degree of wear varied at different mileages, but the vertical acceleration amplitude did not change significantly. An FFT analysis of the vertical acceleration yielded a vertical amplitude spectrum, which showed no characteristic peaks.

[0025] Analysis of lateral acceleration at different mileages under the same track conditions revealed that the degree of wear at different mileages varied significantly, with the lateral acceleration amplitude varying significantly. As wear increased, the lateral extreme velocity amplitude also increased significantly. An FFT analysis of the lateral acceleration yielded a lateral amplitude spectrum, revealing characteristic peaks.

[0026] It can be concluded that the axle box lateral vibration acceleration has the greatest impact on vehicle wear changes.

[0027] According to some embodiments of the present invention, the equivalent taper is calculated as follows:

[0028] The motion of a freewheel pair without inertia running on a track is expressed by the following formula:

[0029]

[0030] in, is the lateral acceleration of the wheelset; Δr is the difference in rolling circle radius between the left and right wheels; V is the running speed; e is the distance between the left and right contact points of the wheel and rail; r0 is the nominal radius of the wheel;

[0031] Assuming the velocity is constant, the velocity can be expressed as,

[0032]

[0033]

[0034] Where x is the longitudinal displacement; t is the running time;

[0035] The rolling radius between the left and right wheels of the vehicle tread with a conical profile angle γ can be expressed as:

[0036] Δr=2ytanγ,

[0037] Then the differential equation can be obtained as:

[0038]

[0039] Solve the differential equation whose solution is a sine wave with wavelength λ,

[0040]

[0041] According to the above calculation method, the equivalent taper at the wheelset lateral displacement of 3mm is taken as the nominal equivalent taper, and it is found that there is a positive correlation between the degree of wear and the equivalent taper.

[0042] According to some embodiments of the present invention, in verifying the mapping relationship between wheel wear, equivalent taper, and the axle box lateral vibration acceleration using a correlation model, the correlation model is a nonlinear SVM model; in order to evaluate the modeling accuracy of the SVM method, the root mean square error and correlation coefficient are used to evaluate the modeling results;

[0043] The formula for calculating the root mean square error is as follows:

[0044]

[0045] Among them, yi is the real data, is the predicted value;

[0046] The correlation coefficient is used to measure the degree of linearity between the predicted results and the actual results. Its calculation formula is:

[0047]

[0048] The mean square error and correlation coefficient in the above formula vary with the parameters. It is clear that as the parameters increase, the mean square error of the model increases significantly, while the correlation coefficient decreases. This indicates that using smaller parameters can effectively improve the accuracy of the modeling between vibration signals and track equivalent taper and wear.

[0049] According to some embodiments of the present invention, the pre-trained spectral model includes a Transformer module, and the Transformer module includes four submodules: a multi-head attention mechanism, a feedforward neural network, layer normalization, and a residual connection;

[0050] The multi-head attention mechanism uses a matrix to achieve parallel computation in the model; first, the dot product of the query and all keys is calculated, and then multiplied by the scaling factor 1 / d k To prevent the product from being too large; send the above calculation results to the Softmax function to obtain the weight corresponding to the Value; according to such weights, the Value vector can be configured to obtain the final output; the process of calculating the attention relationship output can be simply expressed as:

[0051]

[0052] Among them, d k Refers to the initial dimension of the Key vector;

[0053] The feedforward neural network consists of a linear layer and a linear activation function GeLU, which realizes the output of nonlinear transformation of input data and is expressed by the following formula:

[0054] FNN(x)=(GeLU(x)W1+b1)W2+b2,

[0055] Among them, W1, W2 represent weight parameters; b1, b2 represent bias parameters;

[0056] The layer normalization can improve the training speed of the model and make the model more robust. In layer normalization, statistics are calculated based on the values of each single sample in all dimensions, that is, the values x(1), x(2), ..., x(n) in all dimensions of the sample are counted, thereby realizing the normalization operation. Assuming that the input of the lth layer is x, its mean and variance can be expressed as,

[0057]

[0058]

[0059] Among them, H represents the number of hidden nodes in this layer;

[0060] The mathematical definition expression of the layer normalization is as follows,

[0061]

[0062] Among them, g and b represent the parameter vectors of scaling and translation, respectively. They are a set of parameters learned by backpropagation in layer normalization and are small positive numbers used to ensure that the denominator is greater than zero;

[0063] The residual connection can prevent the problem of gradient disappearance or explosion caused by the multi-layer iterative structure, thereby transmitting information deeper to enhance the fitting ability of the model;

[0064] The mathematical definition expression of the residual connection is as follows:

[0065] h(x)=f(x)+x,

[0066] Among them, x is the input of the residual layer, f(x) is the residual part to be learned by the middle layer, and h(x) represents the output of the residual layer.

[0067] According to a second aspect of the present invention, a real-time monitoring system for the wheel-rail matching state and wheel wear of a railway vehicle is provided, wherein:

[0068] a first modeling module, configured to establish a vehicle multi-body dynamics model and a wheel wear model, and send the vehicle multi-body dynamics model and the wheel wear model to a first processing module;

[0069] a first processing module, receiving the vehicle multi-body dynamics model and the wheel wear model, performing wear simulation processing on the vehicle multi-body dynamics model using the wheel wear model to obtain vehicle wheel wear at different mileages, and sending the vehicle wheel wear to a first acquisition module;

[0070] a first acquisition module, receiving and acquiring corresponding wheel axle box vibration acceleration according to different wheel wear of the vehicle, and sending the wheel axle box vibration acceleration to a first analysis module;

[0071] a first analysis module receiving the wheel axle box vibration acceleration and performing a characteristic analysis on the wheel axle box vibration acceleration according to a preset algorithm program to determine that the axle box lateral vibration acceleration has the greatest impact on vehicle wear change, taking the axle box lateral vibration acceleration among the axle box vibration acceleration as a training data set, and sending the training data set to the second modeling module;

[0072] a second modeling module, configured to construct a pre-trained spectrum model, and train the pre-trained spectrum model according to the received training data set to obtain a spectrum model;

[0073] The first input module is used to input test data set parameters;

[0074] The first output module is used to output the vehicle wheel-rail matching status and wheel wear degree corresponding to the test data set.

[0075] According to a rail vehicle wheel-rail matching state and wheel wear real-time monitoring system of the present invention, firstly, a vehicle multi-body dynamics model and a wheel wear model are established by a first modeling module, and the vehicle multi-body dynamics model and the wheel wear model are sent to a first processing module; the first processing module receives the vehicle multi-body dynamics model and the wheel wear model, uses the wheel wear model to perform wear simulation processing on the vehicle multi-body dynamics model, obtains the vehicle wheel wear under different mileage, and sends the vehicle wheel wear to a first acquisition module; the first acquisition module receives the vehicle wheel wear, and according to different vehicle wheel wear, obtains the corresponding wheel axle box vibration acceleration, and sends the wheel axle box vibration acceleration to a first analysis module; the first analysis module receives the wheel axle The wheel axle box vibration acceleration is characterized by analyzing the axle box vibration acceleration according to a preset algorithm program, and it is obtained that the axle box lateral vibration acceleration has the greatest impact on the vehicle wear change. The axle box lateral vibration acceleration in the axle box vibration acceleration is taken as a training data set, and the training data set is sent to the second modeling module; the second modeling module constructs a pre-trained spectrum model, and trains the pre-trained spectrum model according to the received training data set to obtain a trained spectrum model; a group of axle box lateral vibration acceleration is collected as a test data set and input into the system through the first input module, and after being processed by the spectrum model in the system, the vehicle wheel-rail matching status and wheel wear degree corresponding to the test data set are output from the first output module, thereby realizing real-time monitoring of the wheel-rail matching status and wheel wear of the rail vehicle.

[0076] Additional aspects and advantages of the present invention will be set forth in part in the description which follows and, in part, will be obvious from the description which follows, or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0078] Figure 1This is a flow chart of a method for real-time monitoring of wheel-rail matching status and wheel wear of a railway vehicle according to an embodiment of the present invention;

[0079] Figure 2 Schematic diagram of a discrete coordinate system of contact spots in a method for real-time monitoring of wheel-rail matching status and wheel wear of a railway vehicle according to an embodiment of the present invention;

[0080] Figure 3 Schematic diagram of vibration signal collection for a method for real-time monitoring of wheel-rail matching status and wheel wear of a railway vehicle according to an embodiment of the present invention;

[0081] Figure 4 A diagram showing a vertical acceleration vibration signal of a method for real-time monitoring of wheel-rail matching status and wheel wear of a railway vehicle according to an embodiment of the present invention;

[0082] Figure 5 This is a diagram of a lateral acceleration vibration signal of a method for real-time monitoring of wheel-rail matching status and wheel wear of a railway vehicle according to an embodiment of the present invention;

[0083] Figure 6 A diagram of the vertical vibration of the axle box in a method for real-time monitoring of the wheel-rail matching status and wheel wear of a railway vehicle according to an embodiment of the present invention;

[0084] Figure 7 This is a diagram of the lateral vibration of the axle box in a method for real-time monitoring of the wheel-rail matching status and wheel wear of a railway vehicle according to an embodiment of the present invention;

[0085] Figure 8 This is a graph showing the amplitude spectrum of the axle box lateral vibration in a method for real-time monitoring of the wheel-rail matching status and wheel wear of a railway vehicle according to an embodiment of the present invention;

[0086] Figure 9 An equivalent conicity curve of a wheel profile in a method for real-time monitoring of a rail vehicle wheel-rail matching state and wheel wear according to an embodiment of the present invention;

[0087] Figure 10 This is a relationship diagram between the wear amount and the equivalent taper in a method for real-time monitoring of the wheel-rail matching status and wheel wear of a railway vehicle according to an embodiment of the present invention. DETAILED DESCRIPTION

[0088] The embodiments of the present invention are described in detail below. The embodiments described with reference to the accompanying drawings are exemplary. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0089] It should be noted that when an element is referred to as being “fixed to” another element, it may be directly on the other element or there may be an intermediate element. When an element is referred to as being “connected to” another element, it may be directly connected to the other element or there may be an intermediate element.

[0090] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this invention pertains. The terms used herein in the specification of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0091] Example 1

[0092] See Figure 1 As shown, a method for real-time monitoring of the wheel-rail matching state and wheel wear of a railway vehicle comprises the following steps:

[0093] Step S100: establishing a vehicle multi-body dynamics model, and performing wear simulation on the multi-body dynamics model to obtain vehicle wheel wear at different mileages;

[0094] Furthermore, the wear simulation is specifically processed by using the Archard wear model to process the multi-body dynamics model;

[0095] In the Archard wear model, the wear volume at a certain contact point of the wheel can be expressed as:

[0096]

[0097] Among them, V wear is the material wear volume; k w is the wear coefficient, N is the wheel-rail normal force, d is the related slip

[0098] Dynamic displacement, H is the hardness of the wheel material;

[0099] See Figure 2 As shown, the elliptical contact spot is discretized into countless strip rectangles, and the area of each rectangle is ΔS. The wear depth calculation within the rectangle can be expressed as:

[0100]

[0101]

[0102] Where, ΔV wear (x, y) is the wear volume of the particle; k w is the wear coefficient, ΔN(x, y) is the normal force of the particle, Δd is the sliding displacement in the rectangle where the contact point is located, s x is the sliding velocity of the particle in the x direction, s y is the sliding velocity of the particle in the y direction, V C is its passing speed within the contact patch, H is the hardness of the wheel material;

[0103] When the wheel rolls forward one circle, wear will occur within the contact area with the rail. The length of the contact patch is the distance of one roll, and the amount of wheel wear is equal to the amount of wear of the rolling profile in the contact area. To facilitate the determination of the wear of the rolling profile, the default contact patch size, wheel-rail creep force, and contact pressure parameters remain unchanged. To calculate the amount of wear after the wheel rolls forward one circle, it is only necessary to accumulate the wear in the contact area along the direction of vehicle movement. In this process, it is necessary to determine the contact position between the wheel and the rail, then discretize the wheel coordinates into a grid, and accumulate the wear of each wheel along the direction of wheel rolling to obtain the cumulative wear damage depth of the wheel rolling forward one circle at a certain point.

[0104] Step S200: collecting the axle box vibration acceleration of the wheels at different wheel wear levels and performing characteristic analysis on the axle box vibration acceleration;

[0105] Furthermore, the axle box vibration acceleration of the wheels under different wheel wear degrees is collected as follows:

[0106] Through simulation, the axle box vibration acceleration of the train running under different degrees of wear is obtained, and the track is set to be uneven to simulate the running state on the real route. The simulation is set to 300km / h and the sampling frequency is 200 in UM. The multi-body dynamics model is used to match the LMA wheels with wheel wear and the CHN60 tracks to collect the vertical and lateral vibration accelerations above the axle box. The vibration acceleration above the axle box is collected because the axle box and the wheelset are rigidly connected. In the vehicle system dynamics, the axle box has only one nodding degree of freedom relative to the wheelset. The interaction vibration information between the wheel and the rail can be better transmitted through the vibration of the axle box. The vibration information of the interaction between the wheel and the rail included in the axle box vibration acceleration can better reflect the wheel-rail matching relationship. Under actual railway operation conditions, it is also convenient to install acceleration sensors above the axle box. The schematic diagram of vibration acceleration collection is shown as follows. Figure 3 As shown. The vertical and lateral vibration accelerations collected are as follows Figure 4 and Figure 5 As shown;

[0107] Furthermore, the characteristics of the axle box vibration acceleration are analyzed as follows:

[0108] Analysis of the vertical and lateral vibrations collected on the axle box shows that the mean value of the vertical acceleration is smaller than the lateral vibration. This is because the lateral vibration amplitude between the wheel and rail caused by the snaking motion of the wheelset is greater than the vertical vibration amplitude.

[0109] See Figure 6As shown in the figure, the vertical acceleration at different mileages under the same track line conditions is analyzed, and it is found that the degree of wear at different mileages is different, but the vertical acceleration amplitude does not change significantly. The vertical amplitude spectrum obtained by FFT of the vertical acceleration does not show any characteristic peaks.

[0110] Transverse vibration is different from vertical vibration. For transverse vibration at different mileages, refer to Figure 7 As shown in the figure, the degree of wear has a significant impact on the lateral vibration. As the wear increases, the lateral vibration amplitude also increases significantly. Performing FFT on the lateral vibration to obtain the amplitude spectrum. The amplitude spectrum shows the distribution of the signal in the frequency domain, where the horizontal axis represents the frequency and the vertical axis represents the amplitude. The amplitude diagram helps to more easily observe the characteristics of the signal in the frequency domain, especially for those cases where the frequency components are relatively small. The presence of an amplitude peak in the amplitude spectrum indicates that the frequency component is strong in the signal. The peak usually corresponds to the main vibration or periodic change in the signal. Figure 8 As shown, it can be seen that the axle box lateral vibration amplitude spectrum at different mileages has peaks, and it can be found that the peaks are concentrated in 0-30Hz. Moreover, as the mileage increases, the wear increases, the peak characteristics become more obvious, and the peak also increases accordingly.

[0111] Step S300: obtaining the axle box lateral vibration acceleration that has the greatest impact on vehicle wear change, and extracting the axle box lateral vibration acceleration from the axle box vibration acceleration as a training data set;

[0112] Step S400: Based on the calculated positive relationship between the wheel wear degree and the equivalent taper, a correlation model is used to verify that a mapping relationship exists between the wheel wear, the equivalent taper, and the axle box lateral vibration acceleration;

[0113] Furthermore, the motion of the freewheel pair without inertia running on the track is expressed by the following formula,

[0114]

[0115] in, is the lateral acceleration of the wheelset; Δr is the difference in rolling circle radius between the left and right wheels; V is the running speed; e is the distance between the left and right contact points of the wheel and rail; r0 is the nominal radius of the wheel;

[0116] Assuming the velocity is constant, the velocity can be expressed as,

[0117]

[0118]

[0119] Where x is the longitudinal displacement; t is the running time;

[0120] The rolling radius between the left and right wheels of the vehicle tread with a conical profile angle γ can be expressed as:

[0121] Δr=2ytanγ

[0122] Then the differential equation can be obtained as:

[0123]

[0124] Solve the differential equation whose solution is a sine wave with wavelength λ,

[0125]

[0126] According to the above calculation method, the equivalent taper at the wheelset lateral displacement of 3mm is taken as the nominal equivalent taper, and it is found that there is a positive correlation between the wear degree and the equivalent taper;

[0127] Select LMA wear profiles with different wear degrees and match them with CHN60 rails to calculate the equivalent taper. Figure 9 As shown in the figure, the equivalent taper of the wheelset lateral displacement of 1mm-12mm under different wear degrees; when the wear is small, the equivalent taper curve is gentle. As the wear increases, the equivalent taper of the wheelset lateral displacement at 1mm-1.5mm shows a positive slope growth, and then shows a negative growth with a smaller slope. The equivalent taper slowly decreases until the wheelset lateral displacement is 9mm-12mm, at which point the equivalent taper curve rises sharply.

[0128] According to the above calculation method, the equivalent taper at the wheelset lateral displacement of 3mm is the nominal equivalent taper. There is a positive correlation between the wear degree and the equivalent taper. Figure 10 As shown, the equivalent taper increases at a higher rate in the initial stage of surface wear, increases steadily from 50,000 to 150,000 kilometers of operation, and the growth rate of the equivalent taper increases after 150,000 kilometers;

[0129] Furthermore, the related model is a nonlinear SVM model; in order to evaluate the modeling accuracy of the SVM method, the root mean square error and correlation coefficient are used to evaluate the modeling results;

[0130] The formula for calculating the root mean square error is as follows:

[0131]

[0132] Among them, y i For real data, is the predicted value;

[0133] The correlation coefficient is used to measure the degree of linearity between the predicted results and the actual results. Its calculation formula is:

[0134]

[0135] The mean square error and correlation coefficient in the above formula vary with the parameters. It is clear that as the parameters increase, the mean square error of the model increases significantly, while the correlation coefficient decreases. This indicates that using smaller parameters can effectively improve the accuracy of the modeling between vibration signals and track equivalent taper and wear.

[0136] Step S500: constructing a pre-trained spectrum model, inputting the training data set into the pre-trained spectrum model for feature training to obtain a spectrum model;

[0137] Furthermore, the pre-trained spectral model includes a Transformer module, which includes four submodules: a multi-head attention mechanism, a feedforward neural network, layer normalization, and a residual connection;

[0138] The multi-head attention mechanism uses a matrix to achieve parallel computation in the model; first, the dot product of the query and all keys is calculated, and then multiplied by the scaling factor 1 / d k To prevent the product from being too large; send the above calculation results to the Softmax function to obtain the weight corresponding to the Value; according to such weights, the Value vector can be configured to obtain the final output; the process of calculating the attention relationship output can be simply expressed as:

[0139]

[0140] Among them, d k Refers to the initial dimension of the Key vector;

[0141] The feedforward neural network consists of a linear layer and a linear activation function GeLU, which implements the nonlinear transformation of the input data and outputs it, which is expressed by the following formula:

[0142] FNN(x)=(GeLU(x)W1+b1)W2+b2,

[0143] Among them, W1, W2 represent weight parameters; b1, b2 represent bias parameters;

[0144] Layer normalization can improve the training speed of the model and make the model more robust. In layer normalization, statistics are calculated based on the values of each single sample in all dimensions, that is, statistics are taken on the values x(1), x(2), ..., x(n) in all dimensions of the sample, thereby achieving normalization operation. Assuming that the input of the lth layer is x, its mean and variance can be expressed as,

[0145]

[0146]

[0147] Among them, H represents the number of hidden nodes in this layer;

[0148] The mathematical definition expression of the layer normalization is as follows,

[0149]

[0150] Among them, g and b represent the parameter vectors of scaling and translation, respectively. They are a set of parameters learned by backpropagation in layer normalization and are small positive numbers used to ensure that the denominator is greater than zero;

[0151] Residual connections can prevent the problem of gradient vanishing or exploding caused by multi-layer iterative structures, thereby transmitting information deeper to enhance the model's fitting ability;

[0152] The mathematical definition expression of the residual connection is as follows:

[0153] h(x)=f(x)++x

[0154] Among them, x is the input of the residual layer, f(x) is the residual part to be learned by the middle layer, and h(x) represents the output of the residual layer.

[0155] Step S600: collecting a set of axle box lateral vibration accelerations as a test data set and inputting it into the spectrum model to obtain the vehicle wheel-rail matching state and wheel wear degree under the axle box lateral vibration acceleration.

[0156] According to a method for real-time monitoring of the wheel-rail matching state and wheel wear of a railway vehicle of the present embodiment, first, a vehicle multi-body dynamics model is constructed by using dynamics simulation software, and the vehicle multi-body dynamics model is simulated using a wear model to obtain the vehicle wheel wear under different mileages; the axle box vibration acceleration of the wheel under different wheel wear is collected, and the characteristics of the axle box vibration acceleration are analyzed to obtain that the axle box lateral vibration acceleration has the greatest impact on the vehicle wear change; the axle box vibration acceleration is processed to extract the axle box lateral vibration acceleration from the axle box vibration acceleration, and the axle box vertical vibration acceleration is eliminated, and the extracted axle box lateral vibration acceleration is used as a training data set; according to the calculation, it is found that there is a positive relationship between the wheel wear degree and the equivalent taper, and the relevant model is used to obtain the vehicle wheel wear under different mileages; The model verifies that there is a mapping relationship between wheel wear, equivalent conicity and axle box lateral vibration acceleration. The greater the wheel wear, the greater the equivalent conicity, and the greater the equivalent conicity, the poorer the wheel-rail matching state. Therefore, it is concluded that the greater the vehicle wear, the worse the wheel-rail matching state; a pre-trained spectrum model is constructed, and the training data set is input into the pre-trained spectrum model for feature training to obtain the spectrum model; the spectrum model is the final model for obtaining the vehicle wheel-rail matching state and wheel wear degree in real time. By regularly collecting a set of axle box lateral vibration acceleration as a test data set and inputting it into the spectrum model, the vehicle wheel-rail matching state and wheel wear degree under the axle box lateral vibration acceleration are obtained, so that the vehicle wheel-rail matching state and wheel wear degree during operation can be collected in real time.

[0157] Example 2

[0158] A rail vehicle wheel-rail matching status and wheel wear real-time monitoring system, comprising:

[0159] a first modeling module, configured to establish a vehicle multi-body dynamics model and a wheel wear model, and send the vehicle multi-body dynamics model and the wheel wear model to a first processing module;

[0160] A first processing module receives a vehicle multi-body dynamics model and a wheel wear model, performs wear simulation on the vehicle multi-body dynamics model using the wheel wear model, obtains vehicle wheel wear at different mileages, and sends the vehicle wheel wear to a first acquisition module;

[0161] A first acquisition module receives and acquires corresponding wheel axle box vibration acceleration according to different vehicle wheel wear, and sends the wheel axle box vibration acceleration to a first analysis module;

[0162] A first analysis module receives the wheel axle box vibration acceleration and performs characteristic analysis on the wheel axle box vibration acceleration according to a preset algorithm program to determine that the axle box lateral vibration acceleration has the greatest impact on vehicle wear changes. The axle box lateral vibration acceleration of the axle box vibration acceleration is used as a training data set, and the training data set is sent to the second modeling module;

[0163] The second modeling module is used to construct a pre-trained spectrum model and train the pre-trained spectrum model according to the received training data set to obtain a spectrum model;

[0164] The first input module is used to input test data set parameters;

[0165] The first output module is used to output the vehicle wheel-rail matching status and wheel wear degree corresponding to the test data set.

[0166] According to a real-time monitoring system for the wheel-rail matching state and wheel wear of a railway vehicle of this embodiment, a vehicle multi-body dynamics model and a wheel wear model are first established by a first modeling module, and the vehicle multi-body dynamics model and the wheel wear model are sent to a first processing module; the first processing module receives the vehicle multi-body dynamics model and the wheel wear model, uses the wheel wear model to perform wear simulation processing on the vehicle multi-body dynamics model, obtains the vehicle wheel wear under different mileages, and sends the vehicle wheel wear to a first acquisition module; the first acquisition module receives the vehicle wheel wear, and according to different vehicle wheel wear, obtains the corresponding wheel axle box vibration acceleration, and sends the wheel axle box vibration acceleration to a first analysis module; the first analysis module receives the wheel axle box vibration acceleration According to a preset algorithm program, the wheel axle box vibration acceleration is characterized and analyzed, and it is found that the axle box lateral vibration acceleration has the greatest impact on the vehicle wear change. The axle box lateral vibration acceleration in the axle box vibration acceleration is taken as a training data set, and the training data set is sent to the second modeling module; the second modeling module constructs a pre-trained spectrum model, and trains the pre-trained spectrum model according to the received training data set to obtain a trained spectrum model; a group of axle box lateral vibration acceleration is collected as a test data set and input into the system through the first input module. After being processed by the spectrum model in the system, the vehicle wheel-rail matching status and wheel wear degree corresponding to the test data set are output from the first output module, thereby realizing real-time monitoring of the wheel-rail matching status and wheel wear of the rail vehicle.

[0167] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential" and the like to indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore should not be understood as a limitation to the invention.

[0168] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "illustrative embodiments," "example," "specific example," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with the embodiment or example is included in at least one embodiment or example of the present invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example.

[0169] Obviously, the described embodiments are only some of the embodiments of the present application, rather than all of the embodiments. Mentioning "embodiment" in this article means that the specific features, structures or characteristics described in conjunction with the embodiment may be included in at least one embodiment of the present embodiment application. The appearance of this phrase in various positions in the specification does not necessarily mean that they are all the same embodiments, nor are they independent or alternative embodiments that are mutually exclusive with other embodiments. It can be understood explicitly and implicitly by those skilled in the art that the embodiments described herein can be combined with other embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without making creative work are within the scope of protection of this application.

[0170] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the claims and their equivalents.

Claims

1. A method for real-time monitoring of wheel-rail matching status and wheel wear of a railway vehicle, characterized in that: The method comprises, Establishing a vehicle multi-body dynamics model, and performing wear simulation on the multi-body dynamics model to obtain vehicle wheel wear at different mileages; collecting axle box vibration accelerations of wheels at different wheel wear levels, performing feature analysis on the axle box vibration accelerations, determining that the axle box lateral vibration acceleration has the greatest impact on vehicle wear changes, and extracting the axle box lateral vibration accelerations from the axle box vibration accelerations as a training data set; According to the calculation, there is a positive correlation between the wheel wear degree and the equivalent taper, and a correlation model is used to verify that there is a mapping relationship between the wheel wear, the equivalent taper and the lateral vibration acceleration of the axle box; Constructing a pre-trained spectrum model, inputting the training data set into the pre-trained spectrum model for feature training to obtain a spectrum model; A set of axle box lateral vibration accelerations is collected as a test data set and input into the spectrum model to obtain the vehicle wheel-rail matching state and wheel wear degree under the axle box lateral vibration acceleration.

2. A method for real-time monitoring of wheel-rail matching status and wheel wear of a railway vehicle according to claim 1, characterized in that: The wear simulation process is specifically performed by processing the multi-body dynamics model using the Archard wear model; In the Archard wear model, the wear volume at a certain contact point of the wheel can be expressed as: ， Among them, V wear is the material wear volume; k w is the wear coefficient, N is the wheel-rail normal force, d is the related sliding displacement, and H is the hardness of the wheel material; The contact spot is discretized into countless strip rectangles, and the area of each rectangle is ∆S. The wear depth calculation within the rectangle can be expressed as: ， in, is the wear volume of the particle; k w is the wear coefficient, is the normal force of the particle, ∆d is the sliding displacement in the rectangle where the contact point is located, s x is the sliding velocity of the particle in the x direction, s y is the sliding velocity of the particle in the y direction, V C is its passing speed in the contact patch, and H is the hardness of the wheel material.

3. A method for real-time monitoring of wheel-rail matching status and wheel wear of a railway vehicle according to claim 1, characterized in that: The characteristic analysis of the axle box vibration acceleration is carried out, specifically, An analysis of the vertical acceleration and lateral acceleration collected on the axle box shows that the average value of the vertical acceleration is significantly smaller than the lateral acceleration; Analysis of vertical acceleration at different mileages under the same track conditions revealed that the degree of wear varied at different mileages, but the vertical acceleration amplitude did not change significantly. An FFT of the vertical acceleration yielded a vertical amplitude spectrum, which showed no characteristic peaks. Analysis of lateral acceleration at different mileages under the same track conditions revealed that the degree of wear varied at different mileages, with significant changes in lateral acceleration amplitude. As wear increased, the lateral extreme velocity amplitude also increased significantly. An FFT of the lateral acceleration yielded a lateral amplitude spectrum, revealing characteristic peaks. It can be concluded that the axle box lateral vibration acceleration has the greatest impact on vehicle wear changes.

4. A method for real-time monitoring of wheel-rail matching status and wheel wear of a railway vehicle according to claim 1, characterized in that: The calculation of the equivalent taper is specifically as follows: The motion of a freewheel pair without inertia running on a track is expressed by the following formula: ， in, is the lateral acceleration of the wheelset; Δr is the difference in rolling circle radius between the left and right wheels; V is the running speed; e is the distance between the left and right contact points of the wheel and rail; r0 is the nominal radius of the wheel; Assuming the velocity is constant, the velocity can be expressed as, ， Where x is the longitudinal displacement; t is the running time; The vehicle tread has a conical profile angle γ. The rolling radius between the left and right wheels can be expressed as: ， Then the differential equation can be obtained as: ， Solve the differential equation, whose solution is a sine wave with wavelength λ, ， According to the above calculation method, the equivalent taper at the wheelset lateral displacement of 3mm is taken as the nominal equivalent taper, and it is found that there is a positive correlation between the degree of wear and the equivalent taper.

5. The method for real-time monitoring of wheel-rail matching status and wheel wear of a railway vehicle according to claim 1, characterized in that: The correlation model is used to verify that there is a mapping relationship between wheel wear, equivalent taper, and the axle box lateral vibration acceleration. The correlation model is a nonlinear SVM model. In order to evaluate the modeling accuracy of the SVM method, the root mean square error and correlation coefficient are used to evaluate the modeling results. The formula for calculating the root mean square error is as follows: ， in, yi For real data, is the predicted value; The correlation coefficient is used to measure the degree of linearity between the predicted results and the actual results. Its calculation formula is: ； Through the changes in the mean square error and correlation coefficient in the above formula with the parameters, it is obvious that as the parameters increase, the mean square error of the model increases significantly, while the correlation coefficient decreases; this indicates that for the modeling of the relationship between vibration signals and track equivalent taper and wear, using smaller parameters can effectively improve the modeling accuracy.

6. A method for real-time monitoring of wheel-rail matching status and wheel wear of a railway vehicle according to claim 1, characterized in that: The pre-trained spectral model includes a Transformer module, which includes four submodules: a multi-head attention mechanism, a feedforward neural network, layer normalization, and a residual connection; The multi-head attention mechanism is implemented in the model through a matrix to achieve parallel computation; first, the dot product of the query and all keys is calculated, and then the result is multiplied by the scaling factor 1 / d. k To prevent the product from being too large; send the above calculation results to the Softmax function to obtain the weight corresponding to the Value; according to such weights, the Value vector can be configured to obtain the final output; the process of calculating the attention relationship output can be concisely expressed as: ， Among them, d k Refers to the initial dimension of the Key vector.

7. A method for real-time monitoring of wheel-rail matching status and wheel wear of a railway vehicle according to claim 6, characterized in that: The feedforward neural network consists of a linear layer and a linear activation function GeLU, which realizes the output of nonlinear transformation of input data and is expressed by the following formula: ， Among them, W1 and W2 represent weight parameters; b1 and b2 represent bias parameters.

8. A method for real-time monitoring of wheel-rail matching status and wheel wear of a railway vehicle according to claim 6, characterized in that: The layer normalization can improve the training speed of the model and make the model more robust. In layer normalization, statistics are calculated based on the values of each single sample in all dimensions, that is, the values of all dimensions of the sample are normalized. Statistics, thereby achieving normalization operation; assuming that the input of the lth layer is x, its mean and variance can be expressed as, ， Among them, H represents the number of hidden nodes in this layer; The mathematical definition expression of the layer normalization is as follows, ， Among them, g and b represent the parameter vectors of scaling and translation, respectively. They are a set of parameters learned by backpropagation in layer normalization and are small positive numbers used to ensure that the denominator is greater than zero.

9. A method for real-time monitoring of wheel-rail matching status and wheel wear of a railway vehicle according to claim 6, characterized in that: The residual connection can prevent the problem of gradient disappearance or explosion caused by the multi-layer iterative structure, thereby transmitting information deeper to enhance the fitting ability of the model; The mathematical definition expression of the residual connection is as follows: ， Among them, x is the input of the residual layer, f(x) is the residual part to be learned by the intermediate layer, and h(x) represents the output of the residual layer.

10. A rail vehicle wheel-rail matching state and wheel wear real-time monitoring system, used to implement a rail vehicle wheel-rail matching state and wheel wear real-time monitoring method according to any one of claims 1 to 9, characterized in that: The system comprises, a first modeling module, configured to establish a vehicle multi-body dynamics model and a wheel wear model, and send the vehicle multi-body dynamics model and the wheel wear model to a first processing module; a first processing module, receiving the vehicle multi-body dynamics model and the wheel wear model, performing wear simulation processing on the vehicle multi-body dynamics model using the wheel wear model to obtain vehicle wheel wear at different mileages, and sending the vehicle wheel wear to a first acquisition module; a first acquisition module, receiving and acquiring corresponding wheel axle box vibration acceleration according to different wheel wear of the vehicle, and sending the wheel axle box vibration acceleration to a first analysis module; a first analysis module receiving the wheel axle box vibration acceleration and performing a characteristic analysis on the wheel axle box vibration acceleration according to a preset algorithm program to determine that the axle box lateral vibration acceleration has the greatest impact on vehicle wear change, taking the axle box lateral vibration acceleration among the axle box vibration acceleration as a training data set, and sending the training data set to the second modeling module; a second modeling module, configured to construct a pre-trained spectrum model, and train the pre-trained spectrum model according to the received training data set to obtain a spectrum model; The first input module is used to input test data set parameters; The first output module is used to output the vehicle wheel-rail matching status and wheel wear degree corresponding to the test data set.

Images