High-speed train tread tiny defect geometric description method, system, equipment and medium

Abstract

The invention belongs to the technical field of rail traffic safety monitoring, and discloses a high-speed train tread tiny defect geometric description method, system and device and a medium, and the method comprises the following steps: collecting morphological characteristic data of high-speed train wheel tread tiny defects, grouping and extracting key characteristics, and constructing an original defect geometric database; establishing a transient finite element model of wheel-rail rolling contact, simulating mechanical responses of defects under different working conditions, and outputting a stress nephogram and a stress intensity factor; defining a working condition parameter matrix, calculating a defect expansion rate, extracting defect expansion characteristics, and generating a defect instability expansion form characteristic set by adopting a depth auto-encoder and a conditional generative adversarial network; constructing a mapping relation model and an inversion model of the steel rail strain signal and the geometric parameter vector by adopting a convolutional neural network; and acquiring real-time steel rail strain signals, inputting the real-time steel rail strain signals into an inversion model to predict geometric parameter vectors, and accurately generating geometric description of tiny defects.

Description

Technical Field

[0001] The present invention relates to the technical field of rail transit safety monitoring and intelligent diagnosis, and particularly to a geometric description method, system, device and medium for minute defects on the tread of high-speed trains. Background Art

[0002] High-speed trains are an important pillar of the modern transportation system, and their operation safety and reliability are of great significance for ensuring the safety of passengers' lives and improving transportation efficiency. As the core component in direct contact with the rail, the condition of the wheelset directly determines the running stability and safety of the train. Currently, the maintenance of high-speed train wheelsets mainly adopts a planned maintenance system, that is, the maintenance cycle is set according to the service mileage or service time, and there are five corresponding maintenance levels. This maintenance system aims to ensure the safe operation of the train through regular maintenance, but it shows certain limitations in actual applications. Specifically for the wheel tread, as the key area of wheel-rail contact, its condition monitoring is particularly important and is usually listed as a daily inspection item. After the high-speed train enters the depot, the staff evaluates the damage condition of the wheel tread by visual inspection. However, it is found in the follow-up tests that the treads of high-speed train wheels are often damaged due to minute defects such as indentations, especially the wheels of the leading car and the trailing car. Due to the complex forces, if these early minute defects are not detected and processed in time, they may gradually expand to the instability stage (such as tread peeling), thus threatening the service safety of the wheel-rail system.

[0003] The existing technologies have the following significant defects and deficiencies in the detection and treatment of minute defects on the treads of high-speed train wheels: 1. Limitations of visual inspection and risk of undetected defects: The current daily inspection of the wheel tread condition mainly relies on the visual inspection of the staff. This method is limited by visual blind spots, and minute defects are easily overlooked especially under conditions of insufficient light or complex environments. In addition, visual inspection cannot scientifically evaluate the morphological characteristics of the defects and their expansion trends, resulting in potential safety hazards not being detected in time and increasing the driving risk.

[0004] 2. Arbitrariness and low efficiency of temporary turning decision-making: When minute defects are found by visual inspection, whether to perform temporary turning and the depth and number of turnings often depend on the personal experience of the staff, lacking a scientific decision-making basis. Since the existing technologies cannot accurately grasp the depth and internal morphological characteristics of the defects, the temporary turning operation shows strong arbitrariness, not only with low efficiency, but also the turning quality is difficult to guarantee, further affecting the service performance of the wheels. Summary of the Invention

[0005] The object of the present invention is to provide a geometric description method, system, device and medium for micro defects on the tread of high-speed trains. Through parametric modeling and transient simulation technology, accurate geometric description and mechanical response analysis of micro defects on the wheel tread are realized, so as to improve the scientificity and accuracy of defect identification.

[0006] In the first aspect, the present invention provides a geometric description method for micro defects on the tread of high-speed trains, including the following steps: S1: Collect the morphological feature data of micro defects on the tread of the wheels of high-speed trains, construct an original defect feature data set, and group the defects in the original defect feature data set by aspect ratio; extract key morphological features for each group of defects to construct a feature parameter matrix of the micro defects on the wheel tread; use piecewise polynomial fitting and cubic spline interpolation method to construct a geometric function of the micro defects on the wheel tread; construct a standardized original defect geometric database matrix according to the feature parameter matrix and the geometric parameter vector in the geometric function; S2: Establish a transient finite element model of the rolling contact between the wheels and rails of high-speed trains, import the geometric function into the transient finite element model, and define the wheel-rail contact relationship; use the penalty function method to define the wheel-rail contact constraint, and set the tangential friction in combination with the Coulomb friction model, and adopt adaptive meshing in the defect area; perform mechanical response simulation of the transient finite element model under different operating conditions, output the Mises stress in the defect area, and calculate the stress intensity factor by the J integral method; S3: Define a working condition parameter matrix, and the working condition parameter matrix includes the geometric parameter vector; according to the stress intensity factor, approximately calculate the defect growth rate in combination with the Paris law; for each operating condition, extract the maximum Mises stress, the maximum stress intensity factor and the average defect growth rate to construct the defect growth characteristics; based on the defect growth characteristics, use a deep autoencoder and a conditional generative adversarial network to generate a defect instability growth morphological feature set; S4: Collect the rail strain signal through multiple FBG sensors; extract time-domain, frequency-domain and time-frequency-domain features from the rail strain signal to generate a first standardized feature matrix; use a convolutional neural network to construct a mapping relationship model between the first standardized feature matrix and the geometric parameter vector, and develop a convolutional neural network inversion model; S5: Obtain the real-time rail strain signal and convert it into a second standardized feature matrix; input the second standardized feature matrix into the trained convolutional neural network inversion model to obtain a predicted geometric parameter vector and perform weighted correction; based on the weighted corrected predicted geometric parameter vector, obtain a three-dimensional predicted geometric description according to the geometric function.

[0007] As an alternative implementation of the first aspect of the present application, in step S1, the morphological feature data includes the depth, area, width, length, curvature, and edge gradient of the defect.

[0008] As an alternative implementation of the first aspect of the present application, in step S1, the geometric function is obtained by performing piecewise polynomial fitting on the cross-sectional profile of the defect and using cubic spline interpolation in the edge transition region.

[0009] As an alternative implementation of the first aspect of the present application, in step S2, the transient finite element model is established using ABAQUS software, and the dynamic update of the defect geometry is realized through a Fortran subroutine.

[0010] As an alternative implementation of the first aspect of the present application, in step S3, the deep autoencoder is used to extract latent features from the defect expansion characteristics, the conditional generative adversarial network uses the latent features and operating condition parameters to generate extended feature samples, and then the instability expansion mode is identified through a clustering algorithm to form the defect instability expansion morphological feature set.

[0011] As an alternative implementation of the first aspect of the present application, in step S4, the time-domain features include peak strain, root mean square strain, and kurtosis; the frequency-domain features include peak frequency and spectral energy; the time-frequency domain features are used to extract wavelet energy through wavelet transform.

[0012] As an alternative implementation of the first aspect of the present application, in step S5, the weighted correction is to perform weighted fusion on the geometric parameter vector predicted by the convolutional neural network inversion model and the clustering center in the defect instability expansion morphological feature set.

[0013] In a second aspect, an embodiment of the present application provides a geometric description system for minute defects on the tread of a high-speed train, including: A data acquisition and preprocessing module, configured to collect morphological feature data of minute defects on the tread of the wheels of a high-speed train, construct an original defect feature data set, group the defects in the original defect feature data set using the aspect ratio; extract key morphological features for each group of defects to construct a feature parameter matrix for the minute defects on the tread of the wheels; construct a geometric function for the minute defects on the tread of the wheels using piecewise polynomial fitting and cubic spline interpolation; construct a standardized original defect geometry database matrix according to the feature parameter matrix and the geometric parameter vector in the geometric function; A finite element simulation module, configured to establish a transient finite element model of the wheel-rail rolling contact of a high-speed train, import the geometric function into the transient finite element model, and define the wheel-rail contact relationship; define the wheel-rail contact constraint by the penalty function method, and set the tangential friction in combination with the Coulomb friction model, and adopt an adaptive mesh in the defect area; perform a mechanical response simulation of the transient finite element model under different operating conditions, output the Mises stress in the defect area, and calculate the stress intensity factor by the J integral method; A defect propagation characteristic analysis and feature set generation module, configured to define a working condition parameter matrix, where the working condition parameter matrix includes the geometric parameter vector; approximately calculate the defect propagation rate in combination with the Paris law according to the stress intensity factor; for each operating condition, extract the maximum Mises stress, the maximum stress intensity factor, and the average defect propagation rate to construct the defect propagation characteristics; based on the defect propagation characteristics, use a deep autoencoder and a conditional generative adversarial network to generate a defect instability propagation morphology feature set; A signal acquisition and inversion modeling module, configured to collect rail strain signals through multiple FBG sensors; extract time-domain, frequency-domain, and time-frequency domain features from the rail strain signals to generate a first standardized feature matrix; use a convolutional neural network to construct a mapping relationship model between the first standardized feature matrix and the geometric parameter vector, and develop a convolutional neural network inversion model; A real-time geometric description module, configured to obtain real-time rail strain signals and convert them into a second standardized feature matrix; input the second standardized feature matrix into the trained convolutional neural network inversion model to obtain a predicted geometric parameter vector and perform weighted correction; based on the weighted corrected predicted geometric parameter vector, obtain a three-dimensional predicted geometric description according to the geometric function.

[0014] In a third aspect, an embodiment of the present application provides an electronic device, which includes a processor, a memory, and a program or instruction stored on the memory and executable on the processor. When the program or instruction is executed by the processor, the steps of the method described in the first aspect are implemented.

[0015] In a fourth aspect, an embodiment of the present application provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by the processor, the steps of the method described in the first aspect are implemented.

[0016] Compared with the prior art, the present invention first collects the morphological feature data of the micro-defects on the wheel tread, groups the defects by using the aspect ratio, and constructs a geometric function by combining piecewise polynomial fitting and cubic spline interpolation method, so as to realize the accurate parametric modeling and three-dimensional geometric description of the micro-defects on the wheel tread, which directly solves the problem that visual inspection cannot scientifically evaluate the morphological features of the defects. Secondly, the present invention establishes a transient finite element model of the wheel-rail rolling contact of high-speed trains, imports the constructed geometric function for mechanical response simulation, defines the contact constraint by using the penalty function method, sets the tangential friction by using the Coulomb friction model, and adopts adaptive meshing in the defect area to output the Mises stress and calculate the stress intensity factor, so as to quantitatively analyze the mechanical behavior and damage degree of the defect area under different operating conditions, thus making up for the deficiency of the prior art in not being able to deeply master the defect depth and internal morphological features. Further, the present invention approximately calculates the defect propagation rate based on the stress intensity factor in combination with the Paris law, and uses the deep autoencoder and conditional generative adversarial network to generate the set of morphological features of the defect unstable propagation, realizing the scientific prediction of the future propagation trend of the defect and the assessment of the instability risk, providing a quantitative scientific basis for optimizing the temporary turning decision and avoiding the arbitrariness of the decision. Finally, the present invention collects the rail strain signals in real time through multiple FBG sensors, constructs a mapping relationship model and its inversion model between the rail strain signal and the defect geometric parameter vector by using a convolutional neural network, and then obtains and predicts the three-dimensional geometric description of the defect in real time. This innovative method completely overcomes the defects of visual inspection being limited by the visual blind area, prone to missed detection and low efficiency, and realizes non-contact and high-efficiency real-time detection and evaluation of defects. Generally speaking, the multi-level technical means such as parametric modeling, transient simulation, model prediction and real-time monitoring adopted by the present invention solve the core problems of unscientific identification of micro-defects on the wheel tread of high-speed trains, high risk of missed detection, and lack of basis for maintenance decisions in a linked manner, significantly improving the scientificity, accuracy and real-time performance of defect identification, reducing the driving safety risk caused by the expansion of early micro-defects, and effectively improving the service safety and maintenance efficiency of the train wheelset. Brief Description of the Drawings

[0017] Figure 1 is a flowchart of a method for geometric description of micro-defects on the tread of a high-speed train according to an embodiment of the present invention; Figure 2 is a structural diagram of a high-resolution image demoireing model based on pyramid feature extraction and attention feature fusion; Figure 3 is a structural diagram of an encoder / decoder according to an embodiment of the present invention; Figure 4 is a structural diagram of a dilated residual dense block (DRDB) according to an embodiment of the present invention; Figure 5It is a structural diagram of an Enhanced Attention Gate (EAG) according to an embodiment of the present invention; Figure 6 It is a schematic structural diagram of a geometric description system for minute defects on the tread of a high - speed train provided by an embodiment of the present invention. Detailed implementation manners

[0018] Next, the technical solutions in the embodiments of the present application will be clearly and completely described with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0019] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, rather than to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present application can be implemented in an order different from those illustrated or described herein. In addition, "and / or" in the specification and claims means at least one of the connected objects, and the character " / " generally represents an "or" relationship between the associated objects before and after. In the description of the present invention, "a plurality of" means two or more, unless otherwise specifically defined.

[0020] Embodiment 1 Please refer to Figure 1 , which is a flowchart of a geometric description method for minute defects on the tread of a high - speed train provided by an embodiment of the present invention. The method includes five stages: the original defect data acquisition and modeling stage, the wheel - rail contact simulation and mechanical response analysis stage, the defect expansion characteristic generation and morphological feature set construction stage, the strain signal processing and recognition model construction stage, and the real - time recognition and geometric description application stage. The specific implementation process is as follows.

[0021] Stage 1: Original defect data acquisition and modeling. This stage can be divided into the following steps: Ⅰ. Collect the morphological feature data of minute defects on the tread of the wheels of a high - speed train, construct the original defect feature dataset, and group the defects in the original defect feature dataset by the aspect ratio.

[0022] During the service life cycle of the wheels of a high - speed train, the morphological feature data of minute defects on the tread of the wheels are collected by regularly using a variety of detection devices (such as a wheel profile detector, a wheel out - of - roundness detector, a metal surface minute defect detector, a laser ultrasonic damage detector, and a steel ruler, etc.). Let t represent the service time (unit: days), and m represent the service mileage (unit: kilometers). For each defect , the morphological features at time t are represented by the feature vector : ; The definitions of each component are as follows: : The depth of defect i at time t (unit: mm); : The area of defect i at time t (unit: mm2); : The width of defect i at time t (unit: mm); : The length of defect i at time t (unit: mm); : The curvature of defect i at time t (unit: mm -1 ), reflecting the degree of bending of the defect surface; : The edge gradient of defect i at time t (dimensionless), indicating the steepness of the edge.

[0023] Furthermore, the original defect feature dataset is constructed: , where: is the total number of defects, is the total number of measurement time points.

[0024] To group the defects, the aspect ratio is defined. Based on this, the defects are divided into groups ( is the total number of groups), and each group corresponds to an aspect ratio range . The set of defects in the k-th group is represented as: .

[0025] II. Extract the key morphological features for each group of defects to construct the feature parameter matrix of the micro-defects on the wheel tread.

[0026] For each group of defects , extract its key morphological feature parameters and construct the feature parameter matrix . Suppose the k-th group contains feature vectors, and the matrix is defined as: , where, is a three-dimensional matrix, and each row corresponds to a feature vector , ; is the total number of feature vectors of the k-th group of defects; is the time point corresponding to the n-th feature vector.

[0027] : is the eigenvalue of the n-th defect in the k-th group at time with the same definition as .

[0028] Analyze the variation law of the characteristic parameters with the service mileage m. For each parameter , use a polynomial regression model to describe: , where , , …, are the regression coefficients; n is the polynomial order; is the random error term. Define the vector form as: coefficient vector: ; independent variable vector: . The prediction model is: , Use the least squares method to solve . The observation data set is , construct the matrix: ; Then the solution of the regression coefficients can be obtained as: .

[0029] III. Adopt piecewise polynomial fitting and cubic spline interpolation method to construct the geometric function of the micro-defects on the wheel tread.

[0030] To describe the geometric shape of the defect, use piecewise polynomial fitting and cubic spline interpolation method. Assume that the defect cross-sectional profile consists of S segments, and each segment is represented by a polynomial: where: s = 1, 2, …, S represents the total number of segments, represents the position of the s-th node; P is the polynomial order; is the p-th order coefficient of the s-th segment, represents the abscissa of the profile to the power, used to construct a polynomial function to describe the profile characteristics; To ensure smoothness, adjacent segments need to satisfy the continuity condition: ; In the edge transition region, use cubic spline interpolation. Assume the interpolation nodes are , and the cubic spline function satisfies: , where is the abscissa of the interpolation node, is the corresponding depth value, represents the number of interpolation nodes.

[0031] The defect geometry model is a parametric function: , where: , representing the total parameter vector composed of the coefficient vectors of all segments, with a dimension of ; is the coefficient vector of the s-th segment polynomial.

[0032] is expressed as the contour function of the s-th segment.

[0033] IV. Construct a standardized original defect geometry database matrix based on the characteristic parameter matrix and the geometric parameter vector in the geometric function.

[0034] Based on the above geometric function , for each defect , (where represents the -th abscissa interval of the s-th segment, , satisfying ; represents the -th order coefficient of the -th segment polynomial of defect i; represents the geometric parameter vector of defect i.

[0035] And combined with its characteristic parameter matrix (see Part II of Phase I for details), construct the original defect geometry database matrix: , where represents the characteristic parameter matrix of defect i, with a dimension of M×6, containing all the eigenvectors of the k-th group of defects at the M-th time point; represents the transpose of the geometric parameter vector of defect i; represents the original defect geometry database matrix, with a dimension of , integrating the geometric parameters and characteristic parameters of all defects.

[0036] For easy analysis, the characteristic parameter matrix is standardized, and the original defect geometry database matrix after standardization is: .

[0037] Phase II: Wheel-rail contact simulation and mechanical response analysis. This phase can be divided into the following steps: Ⅰ. Establish a transient finite element model for the wheel-rail rolling contact of high-speed trains, import the geometric function into the transient finite element model, and define the wheel-rail contact relationship.

[0038] Use ABAQUS to establish a transient finite element model for the wheel-rail rolling contact of high-speed trains to simulate the wheel-rail contact behavior and the defect propagation process. Create the geometric models of the wheel and the rail in ABAQUS, and the rail cross-section is based on the actual standard (UIC60). After defining the wheel-rail contact relationship, generate the initial finite element model and export the.inp file; according to the on-site tracking test data, define the initial defect of the wheel tread in the.inp file through the *IMPERFECTION keyword. Among them, the defect geometry is represented by a parametric function, and the defect geometry function constructed in Phase I is imported and the standardized original defect geometry database matrix , as follows: Extract the characteristic parameters of each defect from , and define the parameter vector of typical defects: , where respectively represent the depth, area, width, length, curvature, and edge gradient of the i-th defect.

[0039] The defect surface height function is: , where is the polynomial coefficient of the s-th segment, which is called from the database in Phase I.

[0040] In ABAQUS, the dynamic update of the defect geometry is realized through a Fortran subroutine (USDFLD), input and , and adjust the coordinates of the tread nodes: , where is the original node coordinate and n is the normal vector.

[0041] Ⅱ. Define the wheel-rail contact constraint using the penalty function method, set the tangential friction in combination with the Coulomb friction model, and use an adaptive mesh in the defect area.

[0042] The boundary conditions of wheel-rail rolling contact need to consider the influence of the geometry of the stage-I defect. The penalty function method is used to define the contact constraint: , where is the normal penalty function stiffness, is the contact gap, , are the surface coordinates of the wheel and the rail respectively, and the gap is directly affected by ; The tangential friction is set based on the Coulomb friction model: , where μ = 0.3 is the friction coefficient, is the tangential penalty function stiffness, is the tangential relative velocity. An adaptive mesh is adopted in the defect area, and the element size is controlled by the defect width of stage I: , where represents the minimum mesh size in the defect area, which is determined by of stage I, is the mesh scaling factor to ensure the accurate capture of the local stress field of the defect.

[0043] III. Perform the mechanical response simulation of the transient finite element model under different operating conditions and output the Mises stress in the defect area.

[0044] The load is applied to the wheel center, combined with the defect position of stage I (extracted from ), to simulate the dynamic response when the train passes through the defect. Perform the transient finite element simulation in ABAQUS / Explicit, and use the defect parameter vector , the geometric description function and the normalization matrix provided by stage I to simulate the mechanical response of the micro-defects on the wheel tread under different operating conditions. The output results are the Mises stress contour map of the wheel (as shown in Figure 2 ) and the stress intensity factor (as shown in Figure 3 ) to evaluate the defect propagation trend. Specifically as follows: Extract from the normalized defect database , the defect instance corresponding to a specific operating condition, and input into the Fortran subroutine of ABAQUS. Update the coordinates of the mesh nodes in the defect area by calculating .

[0045] Convert to the surface height distribution in the defect area through the Fortran subroutine , for updating the wheel tread geometry.

[0046] Extract each defect from the characteristic parameters , for defining the operating conditions. According to the service mileage, select the corresponding standardized characteristic vector .

[0047] Denormalize to physical units and input it together with into the simulation model to determine the geometric dimensions and boundary conditions of the defect. By looping through the N defect instances and M time points of

[0048] Furthermore, use the explicit dynamics method to solve the transient process, and the control equation is: , where is the density of the wheel material; is the displacement field; is the stress tensor; is the external force in vitro.

[0049] The time discretization adopts the central difference method, and the time step satisfies the stability condition: , where represents the minimum grid size of the defect area; c represents the longitudinal wave velocity of the material; = 210 GPa, which is the Young's modulus of the wheel material.

[0050] Calculate the Von Mises stress distribution of the wheel tread and the defect area as the output of the stress nephogram: , represents the Von Mises stress (unit: MPa) at the position (three-dimensional space coordinates, unit: millimeter, mm) and time , which represents the equivalent stress of the material and is used to evaluate whether the yield limit is reached.

[0051] where: is the deviatoric stress tensor; defined as: , In the formula, is the stress tensor, including the components , through the constitutive relationship Calculation; C is the elastic stiffness tensor, is the total strain, is the plastic strain, represents the unit tensor ( identity matrix); is the volumetric stress tensor (unit: MPa), a symmetric tensor, representing the volumetric component of the stress tensor (i.e., the isotropic pressure or tensile part), defined as: ; In ABAQUS, the Mises stress nephogram is generated through the post-processing module and stored as a spatial distribution matrix: ; where, is the coordinate of the wheel tread mesh node; is the simulation time step; N is the total number of nodes; M is the total number of measurement time points.

[0052] IV. The stress intensity factor is calculated using the J-integral method.

[0053] For the defects defined in Stage 1, the stress intensity factor K is calculated to evaluate the crack growth driving force. It is calculated using the J-integral method: ; where: is the strain energy density; is the surface traction; is the integral path around the defect tip; is the path element.

[0054] For the plane stress state, the relationship between the stress intensity factor and the J-integral is ; In ABAQUS, the is extracted through the *CONTOUR INTEGRAL function for each defect instance in Stage 1 to generate a time series vector:

[0055] In summary, the Mises stress nephogram is output in the form for visual analysis of stress concentration in the defect area. The stress intensity factor is output in the vector form, characterizing the crack growth driving force.

[0056] Stage 3: Generation of crack growth characteristics and construction of morphological feature sets. This stage can be divided into the following steps: Ⅰ. Define the operating condition parameter matrix, where the operating condition parameter matrix includes the geometric parameter vector.

[0057] To comprehensively simulate the defect growth characteristics under different operating conditions, define the operating condition parameter matrix, covering key variables such as train speed, load, and rail surface conditions , where, is the defect geometric parameter vector corresponding to the k-th operating condition, extracted from the standardized defect geometry database in Phase I and describes the contour shape of the defect, is the total number of segments of the defect contour.

[0058] is the train speed (unit: m / s, ); is the axle load (unit: N, ); is the height of rail surface irregularity; : the total number of operating conditions; Ⅱ. Based on the stress intensity factor, approximately calculate the defect growth rate in combination with the Paris law.

[0059] Relying on the ABAQUS simulation software, based on the transient finite element framework in Phase II, batch-simulate the defect growth characteristics under different operating conditions: for each operating condition k, input into the dynamic load time series in Phase II: , where A is the amplitude coefficient, is the wavelength of rail surface irregularity, and t is the simulation time. The defect geometry is defined by . Automatically execute N simulations through Python scripts in ABAQUS, and output the Mises stress nephogram and stress intensity factor for each calculation.

[0060] Approximately calculate the defect growth rate based on the Paris law: , where, is the Mode I stress intensity factor at time step t for the k-th operating condition, , are material constants. The simulation results are stored as the growth characteristic matrix: , .

[0061] Ⅲ. For each operating condition, extract the maximum Mises stress, the maximum stress intensity factor, and the average defect growth rate to construct the defect growth characteristics.

[0062] For each condition k, extract the maximum Mises stress (as Figure 4 shown), the maximum stress intensity factor, and the average growth rate: , where ; ; .

[0063] Ⅳ. Based on the defect growth characteristics, use a deep autoencoder and a conditional generative adversarial network to generate a set of morphological features of defect instability growth.

[0064] Integrate the characteristics of all conditions as . Further, based on , use a method combining a deep autoencoder (DAE) and a conditional generative adversarial network (CGAN) to generate a set of morphological features of the instability growth of micro-defects. Define the DAE to include an encoder and a decoder: Encoder: ; Decoder: ; where is the input feature vector; is the latent representation; is the weight matrix; is the bias vector; is the activation function, minimizing the reconstruction error: , where is the regularization parameter; obtain the latent feature matrix through the Adam optimizer; Subsequently, the CGAN uses Z and the condition parameter to generate extended feature samples, and the generator is defined as: ; the discriminator is: ; In the formula, , are the neural network parameters of the generator and the discriminator; The loss function uses the conditional GAN adversarial loss as: , denotes the probability distribution of the random variable according to its probability distribution Take the expectation; is the probability distribution of the true features, characterizing the statistical characteristics of the feature vectors in

[0065] denotes the random variable take the joint expectation according to its probability distribution ; is the probability distribution of the potential noise, usually a standard normal distribution, used to provide random input for the generator.

[0066] where the objective is set to minimize and maximize ; By alternately optimizing the generator and the discriminator (learning rate , iterate ), generate the feature set: , M is the number of generated samples, merge the true features and the generated features ; use the DBSCAN algorithm for density clustering (neighborhood radius , minimum number of samples MinPts = 5), identify K unstable expansion patterns, and the cluster centers are where , where is the k-th cluster center, is the total number of clusters.

[0067] The final unstable expansion form feature set is defined as: .

[0068] where is the comprehensive feature set, including true features, generated features, and cluster centers, used to characterize the morphological pattern of defect instability.

[0069] Stage 4: Strain signal processing and recognition model construction. This stage can be divided into the following steps: Ⅰ. Collect the rail strain signals through multiple FBG sensors.

[0070] To achieve high-precision and real-time rail strain monitoring, design and build an FBG-based signal acquisition system, deploy FBG sensors along the wheel-rail contact path at the bottom of the rail, and represent their positions with a matrix: , where represents the j-th sensor ( The coordinate vector of ) is aligned with the defect center in Stage 2 where are the lateral and vertical coordinates along the rail respectively, and the sensor spacing is ; represents the length of the wheel defect at time t, which is extracted from the defect parameter vector in the standardized defect geometry database of Stage 1 .

[0071] The relationship between the wavelength shift of the FBG sensor and strain and temperature is: where In the formula: is the wavelength shift of the j-th sensor, is the initial wavelength of the j-th sensor, =0.22 is the photoelastic coefficient, is the local strain, is the coefficient of thermal expansion, is the temperature change; The local strain of the j-th sensor is simplified to: where In the data acquisition part, the acquisition frequency is set , and the strain signal is recorded within the time interval [0, T] to generate a time series matrix: ; where is the k-th sampling time point, , and the complete strain signal matrix is: where In the formula, is the strain vector transpose at time .

[0072] II. Extract the time domain, frequency domain, and time-frequency domain features of the rail strain signal to generate the first standardized feature matrix.

[0073] Extract and quantify the sensitivity features caused by micro-defects from , and use time domain, frequency domain, and time-frequency domain analysis methods to comprehensively characterize the defect response. To capture the strain anomaly caused by defect expansion, calculate the time domain features of the j-th sensor, including peak strain, root mean square strain, and kurtosis, where the peak strain, root mean square strain, and kurtosis are respectively: , , , where , representing the average strain of the j-th sensor; ​is the strain value of the j-th sensor at time ; In frequency-domain analysis, perform a fast Fourier transform (FFT) on to obtain the frequency spectrum . Extract the peak frequency and spectral energy. The peak frequency of the j-th sensor is: , which is responsible for reflecting the main frequency component of the signal; The spectral energy of the j-th sensor is: , which is responsible for reflecting the frequency distribution intensity of the signal.

[0074] To further explore the dynamic characteristics, use wavelet transform (WT) with the Morlet wavelet as the basis function to calculate the wavelet coefficient matrix : , where is the Morlet wavelet function, a is the scale, b is the time shift, is the mother wavelet, is the complex conjugate of the Morlet wavelet. And extract the wavelet energy: , Combining these features, define the feature vector of the j-th sensor ( ): ; Then the feature matrix of all sensors is ( is the feature vector of the -th sensor): , To eliminate the dimension difference, perform Z-score standardization on : , , , Generate the standardized feature matrix (i.e., the first standardized feature matrix): , which contains the standardized features of all sensors.

[0075] III. Use a convolutional neural network to construct a mapping relationship model between the first standardized feature matrix and the geometric parameter vector, and develop a convolutional neural network inversion model.

[0076] To establish the mapping relationship between the micro-defect features and the rail strain signal features, combine the instability expansion morphology feature set in stage three and the simulation results , a model is constructed using a Convolutional Neural Network (CNN) (as shown in Figure 5 ).

[0077] The training set is constructed as follows: , where: represents the standardized feature matrix of the i-th working condition, including the strain signal features of the rail under different load conditions; is the defect geometric parameter vector of the i-th working condition, describing the specific features of the micro-defects, and N represents the total number of working conditions, that is, the number of training samples.

[0078] To adapt to the input format of the CNN, is reshaped into a tensor form. The CNN model structure includes: Input layer: The input tensor , with a dimension of .

[0079] Convolutional layer: 32 convolutional kernels are used (with a size of and a stride of 1), ReLU activation, and the output has a dimension of .

[0080] Pooling layer: Max pooling (with a window of and a stride of 2), and the output has a dimension of .

[0081] Convolutional layer: 64 convolutional kernels are used (with a size of and a stride of 1), ReLU activation, and the output has a dimension of .

[0082] Pooling layer: Max pooling (with a window of and a stride of 2), and the output has a dimension of .

[0083] Fully connected layer: Flattened into a vector (with a dimension of , mapped to the output layer (with a dimension of , corresponding to . is the total number of segments of the defect profile.

[0084] The loss function for training the CNN model is: , where is the L2 regularization coefficient. The Adam optimizer is used (learning rate ), with a batch size B = 32, training for E = 5000 epochs, and monitoring the validation set error. is the square of the L2 norm of the weight matrix of the th layer; it is used to measure the difference between the predicted defect parameters and the true defect parameters.

[0085] Furthermore, develop a convolutional neural network inversion algorithm to perform feature extraction and standardization to obtain the standardized matrix .

[0086] In the inversion part of the deep learning model, input into the trained CNN to obtain the initial prediction: .

[0087] Next, optimize and correct. Combine the in Phase III to calculate the Euclidean distance from the cluster center ( is the total number of clusters): ; Select the cluster center with the minimum distance: , and perform weighted correction: , is the weighting coefficient, and is the vector of the final predicted defect parameters.

[0088] Output the confidence interval , where is estimated based on the CNN variance.

[0089] Integrate the FBG acquisition system, the feature extraction module, and the inversion algorithm on a unified platform to output in real time. The data is stored as: .

[0090] Phase Five: Real-time identification and geometric description application. This phase can be divided into the following steps: Ⅰ. Obtain the real-time rail strain signal and convert it into the second standardized feature matrix.

[0091] Ⅱ. Input the second standardized feature matrix into the trained convolutional neural network inversion model to obtain the predicted geometric parameter vector and perform weighted correction.

[0092] Obtain the real-time rail strain signal from the FBG system in Phase Four as the input of the inversion algorithm. The real-time acquisition time window is , the acquisition frequency is , and the number of sensors is The real-time strain signal is expressed as: ; Wherein, is the total number of sampling points, and the complete signal matrix is: ; To ensure that the signal is suitable for inversion, calculate the signal energy: .

[0093] Based on the convolutional neural network (CNN) inversion model trained in stage four , predict the defect parameters from the real-time signal. Generate the second normalized feature matrix (including peak strain, root mean square strain, kurtosis, peak frequency, spectral energy, and wavelet energy), and directly input it into the model to obtain the initial prediction: , Wherein, , respectively represent the predicted defect depth, width, length, edge curvature, and edge gradient. Combining the weighted fusion method in stage four, obtain the final prediction parameters .

[0094] III. Based on the weighted and corrected predicted geometric parameter vector, obtain the three-dimensional predicted geometric description according to the geometric function.

[0095] Using , based on the geometric description framework in stage one, construct the predicted geometric description of the defect. Stage one defines the defect surface function as a combination of piecewise polynomials and cubic spline interpolation. The predicted geometric description function is: .

[0096] The specific implementation is as follows. Piecewise polynomial part: Assume that the defect cross-section is divided into S segments, and the polynomial form of each segment is: , is the p-th order coefficient of the s-th segment polynomial ( , p is the polynomial order); In the edge region ( ), ( is the defect center, is the transition zone width, is the predicted defect width), use cubic spline for smooth transition: ; Coefficient is solved by the (curvature) and (gradient) constraint conditions and boundary continuity ; The three-dimensional predicted geometric description is extended to: , wherein, are the coordinates of the defect along the transverse, longitudinal, and vertical directions of the rail, is the curvature and gradient adjustment function, determined by the spline coefficients and boundary conditions.

[0097] IV. Convert the three-dimensional predicted geometric description into a predicted geometric description in the mesh form required for the finite element simulation of the defect region.

[0098] Convert into the mesh form required for finite element simulation. Define the mesh points in the defect region: , where: is the x-axis coordinate of the mesh point; (unit: millimeter, mm, ) is the y-axis coordinate of the mesh point (unit: millimeter, mm, ) represents the x-axis mesh resolution; represents the y-axis mesh resolution; , are the number of mesh points corresponding to the coordinate axes.

[0099] Calculate the height of each mesh point as ( are the coordinates of the k-th mesh point). Generate the predicted geometric description: , is the total number of mesh points, Output the predicted geometric description for subsequent use, and store it as structured data: , is the real-time strain signal matrix; is the standardized feature matrix; is the final predicted defect parameter vector.

[0100] Embodiment 2 Please refer to Figure 6 , which shows the structural schematic diagram of a geometric description system for minute defects on the tread of a high-speed train proposed in the second embodiment of this application. The system includes the following key modules: The data acquisition and preprocessing module 100 is configured to collect the morphological feature data of the micro defects on the wheel tread of a high-speed train, construct an original defect feature data set, and perform defect grouping on the original defect feature data set using the aspect ratio; extract key morphological features for each group of defects to construct the feature parameter matrix of the micro defects on the wheel tread; use piecewise polynomial fitting and cubic spline interpolation methods to construct the geometric function of the micro defects on the wheel tread; construct a standardized original defect geometric database matrix according to the feature parameter matrix and the geometric parameter vector in the geometric function; The finite element simulation module 200 is configured to establish a transient finite element model of the wheel-rail rolling contact of a high-speed train, import the geometric function into the transient finite element model, and define the wheel-rail contact relationship; use the penalty function method to define the wheel-rail contact constraint, and set the tangential friction in combination with the Coulomb friction model, and adopt adaptive meshing in the defect area; perform the mechanical response simulation of the transient finite element model under different operating conditions, output the Mises stress in the defect area, and calculate the stress intensity factor using the J integral method; The defect propagation characteristic analysis and feature set generation module 300 is configured to define a working condition parameter matrix, where the working condition parameter matrix includes the geometric parameter vector; approximately calculate the defect propagation rate according to the stress intensity factor in combination with the Paris law; for each operating condition, extract the maximum Mises stress, the maximum stress intensity factor, and the average defect propagation rate to construct the defect propagation characteristics; based on the defect propagation characteristics, use a deep autoencoder and a conditional generative adversarial network to generate a defect instability propagation morphological feature set; The signal acquisition and inversion modeling module 400 is configured to collect the rail strain signal through multiple FBG sensors; extract the time-domain, frequency-domain, and time-frequency domain features of the rail strain signal to generate a first standardized feature matrix; use a convolutional neural network to construct a mapping relationship model between the first standardized feature matrix and the geometric parameter vector, and develop a convolutional neural network inversion model; The real-time geometric description module 500 is configured to obtain the real-time rail strain signal and convert it into a second standardized feature matrix; input the second standardized feature matrix into the trained convolutional neural network inversion model to obtain a predicted geometric parameter vector and perform weighted correction; based on the weighted corrected predicted geometric parameter vector, obtain a three-dimensional predicted geometric description according to the geometric function.

[0101] A geometric description system for minute defects on the tread of a high-speed train in an embodiment of the present application may be a device, or a component, an integrated circuit, or a chip in a terminal. The device may be a mobile electronic device or a non-mobile electronic device. Exemplarily, the mobile electronic device may be a mobile phone, a tablet computer, a laptop computer, a palmtop computer, a vehicle-mounted electronic device, a wearable device, an Ultra-mobile Personal Computer (UMPC), a netbook, or a Personal Digital Assistant (PDA), etc., and the non-mobile electronic device may be a server, a Network Attached Storage (NAS), a Personal Computer (PC), etc. The embodiments of the present application do not make specific limitations.

[0102] A geometric description system for minute defects on the tread of a high-speed train in an embodiment of the present application may be a device with an operating system. The operating system may be the Android operating system, the IOS operating system, or other possible operating systems. The embodiments of the present application do not make specific limitations.

[0103] A geometric description system for minute defects on the tread of a high-speed train provided in an embodiment of the present application can implement Figure 1 each process implemented by a geometric description method for minute defects on the tread of a high-speed train in the method embodiment. To avoid repetition, it will not be elaborated here.

[0104] Optionally, an embodiment of the present application further provides an electronic device, including a processor, a memory, and a program or instruction stored on the memory and executable on the processor. When the program or instruction is executed by the processor, it implements each process of the above-mentioned geometric description method embodiment for minute defects on the tread of a high-speed train and can achieve the same technical effect. To avoid repetition, it will not be elaborated here.

[0105] An embodiment of the present application further provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, it implements each process of the above-mentioned embodiment of the geometric description method for minute defects on the tread of a high-speed train and can achieve the same technical effect. To avoid repetition, it will not be elaborated here.

[0106] Wherein, the processor is the processor in the electronic device described in the above embodiment. The readable storage medium includes a computer-readable storage medium, such as a computer Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disc, etc.

[0107] It should be noted that in this text, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising one..." does not exclude the existence of additional identical elements in the process, method, article or device comprising that element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in a reverse order according to the functions involved. For example, the described methods may be performed in an order different from that described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0108] Through the description of the above embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation. Based on this understanding, the technical solution of the present application, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions for causing a terminal (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of the present application.

[0109] The embodiments of the present application have been described above with reference to the accompanying drawings. However, the present application is not limited to the above specific embodiments. The above specific embodiments are merely illustrative and not restrictive. Under the inspiration of the present application, those of ordinary skill in the art can also make many forms without departing from the purpose of the present application and the scope protected by the claims, and all of them belong to the protection scope of the present application.

Claims

1. A geometric description method for tiny defects on the tread of a high-speed train, characterized in that, It includes the following steps: S1: Collect the morphological feature data of the micro-defects on the wheel tread of the high-speed train, construct the original defect feature dataset, and group the defects in the original defect feature dataset by the aspect ratio; extract the key morphological features for each group of defects to construct the feature parameter matrix of the micro-defects on the wheel tread; use piecewise polynomial fitting and cubic spline interpolation method to construct the geometric function of the micro-defects on the wheel tread; construct the standardized original defect geometric database matrix according to the feature parameter matrix and the geometric parameter vector in the geometric function; S2: Establish a transient finite element model of the wheel-rail rolling contact of the high-speed train, import the geometric function into the transient finite element model, and define the wheel-rail contact relationship; use the penalty function method to define the wheel-rail contact constraint, and set the tangential friction in combination with the Coulomb friction model, and adopt adaptive meshing in the defect area; Execute the mechanical response simulation of the transient finite element model under different operating conditions, output the Mises stress in the defect area, and calculate the stress intensity factor by the J integral method; S3: Define the operating condition parameter matrix, and the operating condition parameter matrix includes the geometric parameter vector; According to the stress intensity factor, approximately calculate the defect growth rate in combination with the Paris law; For each operating condition, extract the maximum Mises stress, the maximum stress intensity factor and the average defect growth rate to construct the defect growth characteristics; Based on the defect growth characteristics, use the deep autoencoder and the conditional generative adversarial network to generate the defect instability growth morphological feature set; S4: Collect the rail strain signals through multiple FBG sensors; extract the time-domain, frequency-domain and time-frequency domain features of the rail strain signals to generate the first standardized feature matrix; use the convolutional neural network to construct the mapping relationship model between the first standardized feature matrix and the geometric parameter vector, and develop the convolutional neural network inversion model; S5: Obtain the real-time rail strain signal and convert it into the second standardized feature matrix; input the second standardized feature matrix into the trained convolutional neural network inversion model to obtain the predicted geometric parameter vector and perform weighted correction; based on the weighted corrected predicted geometric parameter vector, obtain the three-dimensional predicted geometric description according to the geometric function.

2. The method according to claim 1, wherein In step S1, the morphological feature data includes the depth, area, width, length, curvature and edge gradient of the defect.

3. The method according to claim 1, characterized in that, In step S1, the geometric function is obtained by piecewise polynomial fitting of the defect cross-sectional profile and cubic spline interpolation in the edge transition region.

4. The method according to claim 1, wherein In step S2, the transient finite element model is established by using the ABAQUS software, and the dynamic update of the defect geometry is realized through the Fortran subroutine.

5. The method according to claim 1, characterized in that In step S3, the deep autoencoder is used to extract the latent features from the defect growth characteristics, and the conditional generative adversarial network uses the latent features and the operating condition parameters to generate the extended feature samples, and then identifies the instability growth mode through the clustering algorithm to form the defect instability growth morphological feature set.

6. The method according to claim 1, wherein In step S4, the time-domain features include peak strain, root mean square strain, and kurtosis; the frequency-domain features include peak frequency and spectral energy; the time-frequency domain features extract wavelet energy through wavelet transform.

7. The method according to claim 1, characterized in that, In step S5, the weighted correction is to perform weighted fusion on the geometric parameter vector predicted by the convolutional neural network inversion model and the cluster center in the defect instability propagation morphology feature set.

8. A geometric description system for minute defects on the tread surface of a high-speed train, characterized in that, It includes: A data acquisition and preprocessing module configured to collect morphological feature data of micro-defects on the wheel tread of a high-speed train, construct an original defect feature data set, and group the defects in the original defect feature data set using the aspect ratio; extract key morphological features for each group of defects to construct a feature parameter matrix of the micro-defects on the wheel tread; use piecewise polynomial fitting and cubic spline interpolation methods to construct a geometric function of the micro-defects on the wheel tread; construct a standardized original defect geometry database matrix based on the feature parameter matrix and the geometric parameter vector in the geometric function. A finite element simulation module configured to establish a transient finite element model of the rolling contact between the wheel and rail of a high-speed train, import the geometric function into the transient finite element model, and define the wheel-rail contact relationship; define the wheel-rail contact constraint using the penalty function method, and set the tangential friction in combination with the Coulomb friction model, and use adaptive meshing in the defect area. Perform a mechanical response simulation of the transient finite element model under different operating conditions, output the Mises stress in the defect area, and calculate the stress intensity factor using the J integral method. A defect propagation characteristic analysis and feature set generation module configured to define a working condition parameter matrix, where the working condition parameter matrix includes the geometric parameter vector. Approximately calculate the defect propagation rate according to the stress intensity factor in combination with the Paris law. For each operating condition, extract the maximum Mises stress, the maximum stress intensity factor, and the average defect propagation rate to construct the defect propagation characteristics. Based on the defect propagation characteristics, use a deep autoencoder and a conditional generative adversarial network to generate a defect instability propagation morphology feature set. A signal acquisition and inversion modeling module configured to collect rail strain signals through multiple FBG sensors; extract time-domain, frequency-domain, and time-frequency domain features from the rail strain signals to generate a first standardized feature matrix; use a convolutional neural network to construct a mapping relationship model between the first standardized feature matrix and the geometric parameter vector, and develop a convolutional neural network inversion model. A real-time geometry description module configured to obtain real-time rail strain signals and convert them into a second standardized feature matrix; input the second standardized feature matrix into the trained convolutional neural network inversion model to obtain a predicted geometric parameter vector and perform weighted correction; based on the predicted geometric parameter vector after weighted correction, obtain a three-dimensional predicted geometric description according to the geometric function.

9. An electronic device, characterized in that, It includes a processor, a memory, and a program or instruction stored on the memory and executable on the processor. When the program or instruction is executed by the processor, the steps of a method for geometric description of micro-defects on the tread of a high-speed train according to any one of claims 1-7 are implemented.

10. A readable storage medium, characterized in that, The program or instruction is stored on the readable storage medium, and when the program or instruction is executed by the processor, the steps of a method for geometric description of minute defects on the tread of a high-speed train as described in any one of claims 1-7 are implemented.

Images