Rail surface damage diagnosis method, system and equipment based on dependence metric

Abstract

This invention provides a method, system, and device for diagnosing rail surface damage based on dependence metrics. The method includes: preprocessing standard profile data and profile data of various types of damage; reconstructing the phase space of both to obtain their respective vector groups, which are considered as two-dimensional random variable samples; calculating the martingale divergence matrix of the two variables in the samples; calculating all eigenvalues ​​of the MDDM and normalizing them; calculating the dependence index DI in the form of standardized Shannon entropy; calculating the distance covariance and variance of X and Y; calculating the dependence metric DM in the form of cross-sample entropy; constructing a correlation plane with DM as the abscissa and DI as the ordinate; obtaining the planar distribution area corresponding to each type of damage for known damage profile data; and determining the fault type of the data to be detected based on its distribution position on the correlation plane. This invention can accurately diagnose the fault types of various damage types in rail profile data.

Description

Technical Field

[0001] This invention relates to the field of intelligent operation and maintenance of heavy-haul railways, and in particular to a method, system and equipment for diagnosing rail surface damage based on dependency measurement. Background Technology

[0002] To date, there are two main methods for geometric measurement of rail profile damage: contact measurement and non-contact measurement. Contact measurement primarily involves mechanical contact, requiring the probe of the inspection equipment to contact the rail. This fault detection method has been widely used in the Chinese railway system, but it suffers from low measurement accuracy and high labor intensity, failing to meet the current needs of domestic railway development. A newer fault detection method is the non-contact rail profile inspection system equipped with a laser profile sensor. This system collects rail surface data using a laser sensor and constructs the rail profile by establishing and transforming the coordinates of the data points. This new inspection technology boasts high automation, high accuracy, and fast inspection speed. Using this system, the collected and processed rail profile data can be used for feature extraction and fault identification of surface damage on in-service rails.

[0003] Essentially, rail surface damage detection based on profile data is a process of extracting correlation features. For identifying rail surface damage types, the relationships between the analyzed objects are sometimes deterministic and sometimes indeterminate. Features used to describe the correlation between things can be broadly categorized as deterministic and random. Therefore, features used to describe the relationships between analyzed objects can be broadly classified as deterministic features and random features. Correspondingly, when the characteristics or attributes of the research objects are represented in the form of variables, the relationships between them can also be divided into two categories: functional relationships (deterministic relationships) and statistical relationships. However, in reality, the relationships between most research objects cannot be simply expressed by a function, but need to be represented by statistical relationships. To measure the degree of correlation between analyzed objects and to characterize this relationship with appropriate statistical indicators, researchers have proposed correlation indicators such as Pearson correlation coefficient, Spearman correlation coefficient, Kendall coefficient, mutual information, and maximum information coefficient. However, existing methods have some limitations: Pearson coefficients can only measure linear relationships; Spearman coefficients can only measure monotonic functional relationships of sequential variables; Kendall coefficients have limited capabilities in data acquisition, and obtaining the density function of variables is also relatively difficult for mutual information methods.

[0004] In 2007, distance correlation (DC) was proposed and used to describe the joint dependence between random vector variables of arbitrary dimensions. Distance correlation has attracted widespread attention in statistical theory, time series analysis, and various other fields. It is not only an easy-to-calculate measure of correlation, but it can also effectively detect nonlinear and non-monotonic relationships between variables.

[0005] The Martingale difference distance correlation (MDDC) theory was proposed in 2014 by Professor X. Shao's team, a data scientist in the United States. It is a natural extension and refinement of distance correlation theory, primarily studying the measurement of the conditional mean correlation between response and predictor variables. Subsequently, in 2017 and 2020, the research team extended this result to the multivariate case, proposing the Martingale Difference Divergence Matrix (MDDM) and Volatility Martingale Difference Divergence Matrix (VMDDM) methods for studying the correlation of multivariate high-dimensional random variables. Due to the simplicity, efficiency, and stability of this statistical algorithm, and its applicability to measuring the degree of correlation between different high-dimensional signals, it immediately attracted significant attention. Currently, it has important and profound applications in signal feature extraction, anomaly detection, information fusion, medical information detection, machine learning, and economics.

[0006] In some cases, α∈R p There exists a linear combination of variables Y such that E(α) T Y|X)=E(α T Even though Y is not strictly independent of X in terms of conditional mean, in such cases, a linear transformation can separate the part that is independent of the conditional mean of X, reducing the effective dimension of E(Y|X). Therefore, modeling the conditional mean of Y as a function of X can be simplified. Thus, the martingale difference matrix (MDDM) theory can be used to measure the independence of conditional means between the detected objects.

[0007] For the effective extraction of correlation features from profile data, existing correlation measurement methods have the following shortcomings:

[0008] Existing methods cannot characterize the correlation of certain complex non-stationary data (such as profile data). For example, even when the two analyzed objects are completely independent, the detection results of some correlation measures tend to approach 1 as the dimensionality of the analyzed objects increases, indicating the strongest correlation.

[0009] Existing methods are often only effective in characterizing linear and monotonic correlations between detected objects, but cannot successfully extract more complex nonlinear and non-monotonic correlation features.

[0010] For the data to be analyzed, existing methods lack necessary preprocessing work such as phase space reconstruction, which results in the inability to effectively extract the high-dimensional information hidden in the data, sensitivity to noise, and low fault identification rate. Summary of the Invention

[0011] The purpose of this invention is to provide a method, system, and device for diagnosing rail surface damage based on dependence measurement, which can automatically and accurately extract various damage features based on rail profile data.

[0012] To achieve the above objectives, the present invention provides the following solution:

[0013] A method for diagnosing rail surface damage based on dependence metrics, comprising:

[0014] A. Based on standard profile data and various injury profile data, generate standard profile data vector groups and various injury profile data vector groups; the standard profile data vector groups and various injury profile data vector groups are two-dimensional random variable samples;

[0015] B. Calculate the martingale divergence matrix of the two-dimensional random variable sample to obtain all the eigenvalues ​​of the martingale divergence matrix;

[0016] C. Normalize all eigenvalues ​​to generate normalized eigenvalues;

[0017] D. Calculate the dependency index in the form of standardized Shannon entropy based on the normalized eigenvalues;

[0018] E. Calculate the distance covariance between the standard profile data and the various types of damage profile data, the distance variance of the standard profile data, and the distance variance of the various types of damage profile data;

[0019] F. Calculate the dependency measure in the form of cross-sample entropy based on the distance covariance, the distance variance of the standard profile data, and the distance variance of the various types of damage profile data;

[0020] G. Construct a correlation plane with the dependence measure as the horizontal axis and the dependence index as the vertical axis;

[0021] H. For known damage profile data, implement steps A to G to determine the planar distribution area corresponding to each type of damage;

[0022] I. For the profile data to be detected, implement steps A to G, based on the planar distribution area corresponding to the various types of damage, determine the fault type to which the profile data to be detected belongs according to the distribution position of the profile data to be detected on the correlation plane.

[0023] Optionally, based on standard profile data and various damage profile data, generate standard profile data vector sets and various damage profile data vector sets, specifically including:

[0024] The difference between the various types of damage profile data and the standard profile data is measured using Euclidean distance. Abnormal data in the various types of damage profile data are removed to generate preprocessed damage profile data.

[0025] Phase space reconstruction is used to capture the dynamic features in the preprocessed damage profile data and the standard profile data, generating standard profile data vector sets and various damage profile data vector sets.

[0026] Optionally, the martingale divergence matrix is ​​a real symmetric positive semi-definite matrix;

[0027] The martingale divergence matrix is:

[0028]

[0029] Among them, MDDM n (Y|X) is the martingale divergence matrix; X is the vector group after phase space reconstruction of the standard profile data sequence; Y is the vector group after phase space reconstruction of various damage profile data sequences; n is the number of vectors after phase space reconstruction; h and l are subscripts and are positive integers; Y h Let Y be the h-th vector in the vector group Y; l Let l be the l-th vector in the vector group Y; X is the mean of all vectors in the vector group Y; h Let X be the h-th vector in the vector group X; l Let q be the l-th vector in the vector group X; q is the dimension of the vector group X.

[0030] Optionally, the normalized feature values ​​are:

[0031]

[0032] Among them, Λ i λ represents the normalized eigenvalues. i Let λ be the i-th eigenvalue;j Let be the j-th eigenvalue; r is the total number of eigenvalues.

[0033] Optionally, the dependency index ranges from [0, 1], and the value of the dependency index is proportional to the correlation between the standard profile data and the various types of damage profile data;

[0034] The dependency index is:

[0035]

[0036] Wherein, DI(Y|X) is the dependence index between X and Y.

[0037] Optionally, the distance covariance is:

[0038]

[0039] Where σ(X,Y) is the distance covariance between the standard profile data and the various damage profile data; n is the number of vectors after phase space reconstruction; A ij For a standard profile data matrix, B ij This is a matrix of damage profile data for various types of injuries.

[0040] Optionally, the value of the dependency measure is inversely proportional to the correlation between the standard profile data and the various types of damage profile data;

[0041] The dependency measure is:

[0042]

[0043] Among them, DM (m) (X,Y) is the dependency measure; m is the vector dimension of the phase space reconstruction.

[0044] A rail surface damage diagnosis system based on dependency metric, comprising:

[0045] The vector group generation module is used to generate standard profile data vector groups and various damage profile data vector groups based on standard profile data and various damage profile data; the standard profile data vector groups and various damage profile data vector groups are two-dimensional random variable samples;

[0046] The martingale divergence matrix calculation module is used to calculate the martingale divergence matrix of the two-dimensional random variable sample and obtain all the eigenvalues ​​of the martingale divergence matrix.

[0047] The eigenvalue normalization module is used to normalize all eigenvalues ​​and generate normalized eigenvalues.

[0048] The dependency index calculation module is used to calculate the dependency index in the form of standardized Shannon entropy based on the normalized eigenvalues.

[0049] The distance covariance and distance variance calculation module is used to calculate the distance covariance between the standard profile data and the various types of damage profile data, the distance variance of the standard profile data, and the distance variance of the various types of damage profile data;

[0050] The dependency measurement calculation module is used to calculate the dependency measure in the form of cross-sample entropy based on the distance covariance, the distance variance of the standard profile data, and the distance variance of the various types of damage profile data.

[0051] The correlation plane construction module is used to construct a correlation plane with the dependence metric as the horizontal axis and the dependence index as the vertical axis.

[0052] The module for determining the planar distribution area corresponding to various types of damage is used to determine the planar distribution area corresponding to various types of damage given known damage profile data.

[0053] The fault type determination module is used to determine the fault type of the profile data to be detected based on the planar distribution area corresponding to the various types of damage, according to the distribution position of the profile data to be detected on the correlation plane.

[0054] An electronic device includes a memory and a processor, the memory storing a computer program, and the processor running the computer program to cause the electronic device to perform the rail surface damage diagnosis method based on dependence metric as described above.

[0055] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the rail surface damage diagnosis method based on dependency metrics as described above.

[0056] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects: The present invention provides a method, system, and device for diagnosing rail surface damage based on dependence measurement. By constructing a correlation plane, the distribution area of ​​various damage profiles on the correlation plane is determined, and the correlation relationship of the profile data to be detected is more intuitively displayed on a two-dimensional plane, achieving the effect of clustering and characterizing the nonlinear and non-monotonic correlation information in the profile data; moreover, it is not sensitive to the influence of noise, improves the fault identification rate, and can automatically and accurately extract various damage features based on rail profile data. Attached Figure Description

[0057] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0058] Figure 1 The flowchart of the rail surface damage diagnosis method based on dependence measurement provided by the present invention is shown below.

[0059] Figure 2 The cross-sectional shape diagram of the 60kg / m rail provided by this invention;

[0060] Figure 3 The rail cross-sectional shape diagram drawn in MATLAB is provided by this invention;

[0061] Figure 4 The rail profile diagram obtained from on-site sampling of the rail is drawn in MATLAB, as provided in this invention.

[0062] Figure 5 The rail profile diagram after coordinate transformation is drawn in MATLAB, as provided in this invention.

[0063] Figure 6 The present invention provides Figure 5 Data chart of the top section of the middle rail;

[0064] Figure 7 This invention provides a sample image of data with large differences that is removed using Euclidean distance.

[0065] Figure 8 This is a diagram showing the distribution of undamaged profiles and various damaged profiles on the correlation plane. Detailed Implementation

[0066] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0067] The purpose of this invention is to provide a method, system, and device for diagnosing rail surface damage based on dependence measurement, which can automatically and accurately extract various damage features based on rail profile data.

[0068] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0069] Example 1

[0070] like Figure 1 As shown, this invention provides a method for diagnosing rail surface damage based on dependence metrics, comprising:

[0071] Step A: Perform data preprocessing on standard profile data X and profile data Y of various types of damage, and reconstruct the phase space of both.

[0072] In practical applications, after prolonged use, rails will gradually wear down on their load-bearing surfaces. To assess rail wear, it is necessary to measure the rail cross-section and evaluate its difference from the profile of a standard rail to obtain the rail wear value.

[0073] Railway rail damage mainly includes several forms such as side wear damage, stripe damage, peeling and fish scale damage, accounting for more than 80% of all rail damage.

[0074] For the collection of rail profile data, the rail inspection vehicle used is equipped with three data acquisition hosts and has four seats. The driver is responsible for driving safety and controlling the vehicle speed at different times each time. The speed control is performed at low speed twice, medium speed twice, and high speed twice. Figure 2 A cross-sectional shape diagram of the rail, such as... Figure 2 As shown, the surface profile is described using coordinates as shown in Table 1. Table 1 is a coordinate table of the cross-sectional shape of a 60kg / m rail. Its shape is drawn as follows: Figure 3 As shown.

[0075] Table 1

[0076] X coordinate 27.6704 27.5755 27.4807 27.3858 27.2909 27.1961 27.1012 27.0063 Y coordinate 21.2766 21.3082 21.3398 21.3714 21.4031 21.4347 21.4663 21.4979

[0077] Table 2 shows the coordinates of the rail profile obtained by sampling the rail cross-section using a two-dimensional sensor on site. As shown in Table 2, the shape is plotted as follows: Figure 4 As shown.

[0078] Table 2

[0079]

[0080] Because the coordinate system of the measurement data is determined by the sensor's position and orientation at the time of each frame's acquisition, each frame of data needs to be rotated and translated before it can be compared with the standard profile to confirm the amount of rail wear. Furthermore, since the sensor position is affected by vehicle movement and vibration, the rotation and translation parameters for each frame are different. The next step is to find the rotation and translation parameters that best match the standard profile and use these parameters for coordinate transformation.

[0081] Table 3 shows the converted orbital profile coordinates. As shown in Table 3, draw the shape as follows: Figure 5 As shown, observe the data of its top section as follows: Figure 6 As shown.

[0082] Table 3

[0083]

[0084] The original rail profile data may contain some unusable anomalies, typically caused by surface contamination encountered during the initial laser scanning of the rail surface. To address this, a simple Euclidean distance metric can be used to measure the difference between the profile data to be analyzed and the standard profile data to eliminate data samples with significant discrepancies. For example... Figure 7 The data in the four intervals between 2900 and 3600 on the horizontal axis shown are outliers. After removing them, the preprocessed original data to be detected is denoted as [missing information]. Standard profile data is Where t is a subscript and is a positive integer.

[0085] Phase space reconstruction is used to capture dynamic features in the data. m represents the length of the vectors extracted from U and V, and N-m+1 = n vectors are extracted from the two original time series respectively.

[0086] X m (i)=[u i ,u i+1 ,...,u i+(m-1) ], 1≤i≤N-m+1;

[0087] Y m (j)=[v j ,v j+1 ,...,v j+(m-1) ], 1≤j≤N-m+1.

[0088] Among them, X m (i) is the i-th vector of sequence U; u i Y is the first element in the vector, where i is the index and is a positive integer; m (j) is the j-th vector of sequence V; v j Let j be the first element in the vector, and j be the index, which is a positive integer; N be the length of the original sequence, and n be the number of vectors generated by the phase space reconstruction.

[0089] In step A, phase space reconstruction is used to map both one-dimensional sequences U and V to a high-dimensional space, obtaining their respective vector group forms, in order to reveal the intrinsic features of the sequences.

[0090] Step B: Treat the two vector sets obtained as two-dimensional random samples (X) k ,Y k )n k =1. Calculate MDDM using the above sample:

[0091]

[0092] Among them, MDDM n (Y|X) is the martingale divergence matrix; X is the vector group after phase space reconstruction of the standard profile data sequence; Y is the vector group after phase space reconstruction of various damage profile data sequences; n is the number of vectors after phase space reconstruction; h and l are subscripts and are positive integers; Y h Let Y be the h-th vector in the vector group Y; l Let l be the l-th vector in the vector group Y; X is the mean of all vectors in the vector group Y; h Let X be the h-th vector in the vector group X; l Let q be the l-th vector in the vector group X; q is the dimension of the vector group X.

[0093] Clearly, the MDDM theory originates from the martingale difference correlation (MDC) theory, which in turn is a generalization of distance correlation (DC). Distance correlation has attracted widespread attention in many fields, and numerous scholars have published a wealth of research findings based on it. Here, we consider a p-dimensional real space R... p A random vector X and a q-dimensional real space R q Given a random vector Y (p>0, q>0), define f X (t) and f Y (s) are the characteristic functions of X and Y, respectively, f X,Y (t,s) are their joint characteristic functions. Distance covariance V 2 The definition of (X,Y;w) is as follows:

[0094]

[0095] Here, w(t,s) is a positive weight function, therefore the integral in the above equation exists. GJSzékely and MLRizzo proved that in the weight function w(t,s) and When the weights are inversely proportional, their expression is unique. Thus, the weight function is expressed as: in:

[0096]

[0097] c q c is the first intermediate variable. q As the second intermediate variable, For p-norm, Let Γ be the q-norm, p and q be the dimensions of the random vectors Y and X respectively, and Γ be a positive integer, where Γ(·) is the gamma function.

[0098] Similar to

[0099] Therefore, the distance covariance between X and Y is:

[0100]

[0101] Thus, the distance correlation R between X and Y 2 (X,Y) is defined as:

[0102]

[0103] Among them, V 2 (X) is the variance of the distance to X, V 2 (Y) represents the variance of the distance to Y.

[0104] Martingale divergence (MDD) is a natural generalization of DC, used to quantify the conditional mean independence of one variable given a vector variable. Users can modify the MDC method according to different application scenarios to filter out variables that do not contribute to a specific aspect of the conditional distribution of the response, given a finite number of covariates, such as for conditional quantile screening. Similar to the definition of distance covariance mentioned earlier, martingale divergence (MDD) is defined as follows:

[0105]

[0106] Among them, g Y,X (s)=E(Y e i＜s,X＞ ), g Y,X (s) is the joint characteristic function of X and Y, E(Y e i＜s,X＞ ) for Y e i ＜s,X＞ Expectations; g Y (s)=E(Y), g Y (s) is the characteristic function of Y, and E(Y) is the expected value of Y; g X (s)=E(e i＜s,X＞ ), g X (s) is the characteristic function of X, E(e) i＜s,X＞ ) for e i＜s,X＞ The expected value. When X is given as a non-negative value, the MDC of Y is:

[0107]

[0108] Where MDD(Y|X) is the martingale divergence between X and Y, and var(X) is the difference between X and Y. 2 Let be the variance of X, and var(Y) 2 Let Y be the variance.

[0109] In practical applications, these concepts need to be applied to profile-based detection data. This applies to independent and identically distributed observations from a joint distribution (X, Y). The non-negative sample form of MDD is:

[0110]

[0111] in, a kl =Y k Y l , akin, b kl =|X k -X l | q , Where k, l = 1, ..., n, n is the number of vectors after phase space reconstruction, that is, the number of vectors after phase space reconstruction of the original profile sequence.

[0112] The non-negative sample form of MDC is:

[0113]

[0114] in It is the sample variance. Let V be the sample distance variance.

[0115] In some cases, there may exist a linear combination of Y such that E(α) T Y|X)=E(α T Y)(α∈R p Even if Y is not necessarily independent of X, the effective dimension of E(Y|X) can be reduced, thus simplifying the modeling of the conditional mean of Y given X. To this end, the martingale divergence matrix MDDM is introduced, which can be viewed as a generalization of MDD.

[0116] Given a vector Y = (Y1, ..., Y2) p ) T ∈R p ,X∈R q ,

[0117]

[0118] Where G(s) = cov(Y,e)i＜s,X＞ )=(G1(s),···,G p (s)) T , s∈R q G j (s)=ccov(Y j ,e i ＜s,X＞ If satisfied. Then MDDM can be simplified to:

[0119] MDDM(Y|X)=-E[(YE(Y))(Y'-E(Y')) T |X-X'| q ].

[0120] (Y',X') are independent and identically distributed (Y,X). MDDM(Y|X) is a real symmetric positive semi-definite matrix and also a Hermitian matrix with positive real eigenvalues. It is also worth noting that the trace of MDDM satisfies tr(MDDM(Y|X))=MDD(Y|X). 2 Furthermore, the rank of the MDDM is closely related to the number of linear combinations of Y that are conditionally independent of X.

[0121] Step C: Calculate all eigenvalues ​​λ1, λ2, ..., λ of MDDM. r And by normalizing, we get:

[0122]

[0123] Among them, Λ i λ represents the normalized eigenvalues. i Let λ be the i-th eigenvalue; j Let be the j-th eigenvalue; r is the total number of eigenvalues.

[0124] In step C, all the feature values ​​of MDDM are calculated to describe the frequency information of each major component in the detected data.

[0125] Step D: To ensure that the dependency index takes values ​​within [0, 1], consider the standardized Shannon entropy. The final expression for DI, DI(Y|X), is:

[0126]

[0127] The DI statistic ranges from 0 to 1. When two different objects being analyzed have strong dependence or synchronicity, their DI values ​​are relatively large. On the other hand, when the object being analyzed is an individual and itself with a time delay, i.e., X = (x1, x2, ..., x...), the DI statistic is relatively large. N-Δt ) and Y = X' = (x 1+Δt ,x2+Δt ,...,x N The value of DI is directly proportional to the degree of autocorrelation of the system. DI is defined here as:

[0128]

[0129] Where Δt is the time delay.

[0130] The dependency index (DI) obtained in step D ranges from [0, 1], and the magnitude of the DI value is proportional to the correlation between the two objects being detected.

[0131] Step E: Calculate the distance covariance σ(X,Y) and the respective distance variances σ(X) and σ(Y) of X and Y to obtain the distance correlation statistic DC of X and Y.

[0132] For a joint data sample (X, Y), the distance covariance statistic σ(X, Y) is expressed as:

[0133]

[0134] Where n is the number of vectors after the phase space is reconstructed.

[0135] Matrix A={A ij The definition of} is as follows:

[0136]

[0137] Among them, A ij For a standard profile data matrix, B ij For various damage profile data matrices, a ij =|X i -X j | represents the Euclidean distance between any two vectors in the vector set generated from the standard profile data sample. The row arithmetic mean of each vector in the vector set generated for the standard profile data sample. The column arithmetic mean of each vector in the vector set generated for the standard profile data sample. The arithmetic mean of each vector in the vector group generated for the standard profile data sample, where i and j are subscripts and are positive integers.

[0138] Similarly, matrix B = {B ij} is defined as:

[0139]

[0140] Among them, b ij The Euclidean distance between every two vectors in the vector set generated for various damage profile data samples. The arithmetic mean of each vector in the vector set generated for various damage profile data samples. The column arithmetic mean of each vector in the vector set generated for various damage profile data samples. The arithmetic mean of each vector in the vector group generated for various damage profile data samples.

[0141] The distance variance statistic for a single data sample X is shown below:

[0142]

[0143] For sample Y:

[0144]

[0145] The DC statistic is defined as follows:

[0146]

[0147] The calculation of the distance covariance σ(X,Y) and the respective distance variances σ(X) and σ(Y) of X and Y in step E only involves the calculation of the first moment and the second moment.

[0148] Step F: Calculate the dependency statistic (DM) in the form of cross-sample entropy:

[0149] Among them, DM (m) (X,Y) is the dependency measure; m is the vector dimension of the phase space reconstruction.

[0150] Generally, for two different analytical objects, if they are highly correlated or have higher synchronicity, their DM values ​​are smaller.

[0151] In step F, the dependency metric (DM) is calculated in the form of cross-sample entropy, which measures the correlation between the two objects being tested: the magnitude of the DM value is inversely proportional to the correlation between the two objects.

[0152] Step G: Construct a correlation plane with DM as the x-axis and DI as the y-axis to more intuitively display the correlation relationship of the diagnosed profile data on a two-dimensional plane, so as to achieve a clear clustering effect: individuals represented by points that are close to each other on the plane have similar characteristics, while individuals represented by points that are far apart have greater differences.

[0153] Step H: For known damage profile data of various types, implement steps A to G to construct a correlation plane, which can obtain the distribution area of ​​various types of damage on the plane.

[0154] Step I: For the profile data to be detected, perform steps A to G to determine the fault type based on the distribution of the detected data on the correlation plane.

[0155] The experiments collected various rail surface damage data, including three main categories: side wear damage, stripe damage, and spalling / fish scale damage. For each of these three categories, 20 data samples were randomly selected from the collected dataset for each damage profile (standard undamaged profile, side wear damage, stripe damage, spalling, and fish scale damage). For each experiment, 20 samples were randomly selected, analyzed, replaced, and repeated 10 times, and the average DM and DI values ​​were calculated. Figure 8 The distribution of undamaged profiles and profiles of various types of damage on the correlation plane is shown in the experimental results. Figure 8 As shown in the figure, the distribution ranges of various damaged profiles and undamaged profiles are different, indicating a significant clustering effect.

[0156] Figure 8 The distribution areas of various damage profiles on the correlation plane are given. Therefore, by calculating the correlation plane based on the data to be measured, the specific type of damage can be determined by its distribution location.

[0157] Example 2

[0158] In order to implement the method corresponding to Embodiment 1 above and achieve the corresponding functions and technical effects, a rail surface damage diagnosis system based on dependency measurement is provided below.

[0159] A rail surface damage diagnosis system based on dependency metric, comprising:

[0160] The vector group generation module is used to generate standard profile data vector groups and various damage profile data vector groups based on standard profile data and various damage profile data; the standard profile data vector groups and various damage profile data vector groups are two-dimensional random variable samples.

[0161] The martingale divergence matrix calculation module is used to calculate the martingale divergence matrix of the two-dimensional random variable sample and obtain all the eigenvalues ​​of the martingale divergence matrix.

[0162] The eigenvalue normalization module is used to normalize all eigenvalues ​​and generate normalized eigenvalues.

[0163] The dependency index calculation module is used to calculate the dependency index in the form of standardized Shannon entropy based on the normalized eigenvalues.

[0164] The distance covariance and distance variance calculation module is used to calculate the distance covariance between the standard profile data and the various types of damage profile data, the distance variance of the standard profile data, and the distance variance of the various types of damage profile data.

[0165] The dependency metric calculation module is used to calculate the dependency metric in the form of cross-sample entropy based on the distance covariance, the distance variance of the standard profile data, and the distance variance of the various types of damage profile data.

[0166] The correlation plane construction module is used to construct a correlation plane with the dependence metric as the horizontal axis and the dependence index as the vertical axis.

[0167] The module for determining the planar distribution area corresponding to various types of damage is used to determine the planar distribution area corresponding to various types of damage for known damage profile data.

[0168] The fault type determination module is used to determine the fault type of the profile data to be detected based on the planar distribution area corresponding to the various types of damage, according to the distribution position of the profile data to be detected on the correlation plane.

[0169] Example 3

[0170] This invention provides an electronic device including a memory and a processor. The memory stores a computer program, and the processor runs the computer program to enable the electronic device to perform the rail surface damage diagnosis method based on dependency metric provided in Embodiment 1.

[0171] In practical applications, the aforementioned electronic devices can be servers.

[0172] In practical applications, electronic devices include: at least one processor, memory, bus, and communication interface.

[0173] The processor, communication interface, and memory communicate with each other via a communication bus.

[0174] A communication interface is used to communicate with other devices.

[0175] The processor is used to execute programs, specifically the methods described in the above embodiments.

[0176] Specifically, the program may include program code, which includes computer operation instructions.

[0177] The processor may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention. The electronic device may include one or more processors of the same type, such as one or more CPUs; or it may include processors of different types, such as one or more CPUs and one or more ASICs.

[0178] Memory is used to store programs. Memory may include high-speed RAM, and may also include non-volatile memory, such as at least one disk drive.

[0179] Based on the description of the above embodiments, this application provides a storage medium storing computer program instructions thereon, which can be executed by a processor to implement the methods described in any embodiment.

[0180] The rail surface damage diagnosis system based on dependency measurement provided in this application exists in various forms, including but not limited to:

[0181] (1) Mobile communication devices: These devices are characterized by their mobile communication capabilities and primarily aim to provide voice and data communication. These terminals include: smartphones (e.g., iPhones), multimedia phones, feature phones, and low-end phones, etc.

[0182] (2) Ultra-mobile personal computer devices: These devices fall under the category of personal computers, possessing computing and processing capabilities, and generally also have mobile internet access capabilities. These terminals include PDAs, MIDs, and UMPCs, such as the iPad.

[0183] (3) Portable entertainment devices: These devices can display and play multimedia content. This category includes: audio and video players (such as iPods), handheld game consoles, e-books, as well as smart toys and portable car navigation devices.

[0184] (4) Other electronic devices with data interaction functions.

[0185] Specific embodiments of the subject matter have now been described. Other embodiments are within the scope of the appended claims. In some cases, the actions described in the claims can be performed in a different order and still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing can be advantageous.

[0186] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0187] For ease of description, the above apparatus is described by dividing it into various functional units. Of course, in implementing this application, the functions of each unit can be implemented in one or more software and / or hardware components. Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0188] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0189] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0190] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0191] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0192] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0193] Computer-readable media, including both permanent and non-permanent, removable and non-removable media, can store information using any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, and CD-ROM.

[0194] Digital multifunction optical disc (DVD) or other optical storage, magnetic cassette tape, magnetic tape, disk storage or other magnetic storage devices

[0195] Or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transient media, such as modulated data signals and carrier waves.

[0196] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0197] This application can be described in the general context of computer-executable instructions that are executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific transactions or implement specific abstract data types. This application can also be practiced in distributed computing environments where transactions are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0198] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0199] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for diagnosing rail surface damage based on dependence metrics, characterized in that, include: A. Based on standard profile data and various injury profile data, generate standard profile data vector groups and various injury profile data vector groups; the standard profile data vector groups and various injury profile data vector groups are two-dimensional random variable samples; B. Calculate the martingale divergence matrix of the two-dimensional random variable sample to obtain all the eigenvalues ​​of the martingale divergence matrix; C. Normalize all eigenvalues ​​to generate normalized eigenvalues; D. Calculate the dependency index in the form of standardized Shannon entropy based on the normalized eigenvalues; E. Calculate the distance covariance between the standard profile data and the various types of damage profile data, the distance variance of the standard profile data, and the distance variance of the various types of damage profile data; F. Calculate the dependency measure in the form of cross-sample entropy based on the distance covariance, the distance variance of the standard profile data, and the distance variance of the various types of damage profile data; G. Construct a correlation plane with the dependence measure as the horizontal axis and the dependence index as the vertical axis; H. For known damage profile data, implement steps A to G to determine the planar distribution area corresponding to each type of damage; I. For the profile data to be detected, implement steps A to G, based on the planar distribution area corresponding to the various types of damage, determine the fault type to which the profile data to be detected belongs according to the distribution position of the profile data to be detected on the correlation plane.

2. The rail surface damage diagnosis method based on dependence metric according to claim 1, characterized in that, Based on standard profile data and various injury profile data, standard profile data vector sets and various injury profile data vector sets are generated, specifically including: The difference between the various types of damage profile data and the standard profile data is measured using Euclidean distance. Abnormal data in the various types of damage profile data are removed to generate preprocessed damage profile data. Phase space reconstruction is used to capture the dynamic features in the preprocessed damage profile data and the standard profile data, generating standard profile data vector sets and various damage profile data vector sets.

3. The rail surface damage diagnosis method based on dependence metric according to claim 1, characterized in that, The martingale divergence matrix is ​​a real symmetric positive semi-definite matrix; The martingale divergence matrix is: Among them, MDDM n (Y|X) is the martingale divergence matrix; X is the vector group after phase space reconstruction of the standard profile data sequence; Y is the vector group after phase space reconstruction of various damage profile data sequences; n is the number of vectors after phase space reconstruction; h and l are subscripts and are positive integers; Y h Let Y be the h-th vector in the vector group Y; l Let l be the l-th vector in the vector group Y; X is the mean of all vectors in the vector group Y; h Let X be the h-th vector in the vector group X; l Let q be the l-th vector in the vector group X; q is the dimension of the vector group X.

4. The rail surface damage diagnosis method based on dependence metric according to claim 3, characterized in that, The normalized eigenvalues ​​are: Among them, Λ i λ represents the normalized eigenvalues. i Let λ be the i-th eigenvalue; j Let be the j-th eigenvalue; r is the total number of eigenvalues.

5. The rail surface damage diagnosis method based on dependence metric according to claim 4, characterized in that, The dependency index ranges from [0, 1], and the value of the dependency index is directly proportional to the correlation between the standard profile data and the various types of damage profile data. The dependency index is: Wherein, DI(Y|X) is the dependence index between X and Y.

6. The rail surface damage diagnosis method based on dependence metric according to claim 5, characterized in that, The distance covariance is: Where σ(X,Y) is the distance covariance between the standard profile data and the various damage profile data; n is the number of vectors after phase space reconstruction; A ij For a standard profile data matrix, B ij This is a matrix of damage profile data for various types of injuries.

7. The rail surface damage diagnosis method based on dependence metric according to claim 6, characterized in that, The value of the dependency measure is inversely proportional to the correlation between the standard profile data and the various types of damage profile data; The dependency measure is: Among them, DM (m) (X,Y) is the dependency measure; m is the vector dimension of the phase space reconstruction.

8. A rail surface damage diagnosis system based on dependence metric, characterized in that, include: The vector group generation module is used to generate standard profile data vector groups and various damage profile data vector groups based on standard profile data and various damage profile data; the standard profile data vector groups and various damage profile data vector groups are two-dimensional random variable samples; The martingale divergence matrix calculation module is used to calculate the martingale divergence matrix of the two-dimensional random variable sample and obtain all the eigenvalues ​​of the martingale divergence matrix. The eigenvalue normalization module is used to normalize all eigenvalues ​​and generate normalized eigenvalues. The dependency index calculation module is used to calculate the dependency index in the form of standardized Shannon entropy based on the normalized eigenvalues. The distance covariance and distance variance calculation module is used to calculate the distance covariance between the standard profile data and the various types of damage profile data, the distance variance of the standard profile data, and the distance variance of the various types of damage profile data; The dependency measurement calculation module is used to calculate the dependency measure in the form of cross-sample entropy based on the distance covariance, the distance variance of the standard profile data, and the distance variance of the various types of damage profile data. The correlation plane construction module is used to construct a correlation plane with the dependence metric as the horizontal axis and the dependence index as the vertical axis. The module for determining the planar distribution area corresponding to various types of damage is used to determine the planar distribution area corresponding to various types of damage given known damage profile data. The fault type determination module is used to determine the fault type of the profile data to be detected based on the planar distribution area corresponding to the various types of damage, according to the distribution position of the profile data to be detected on the correlation plane.

9. An electronic device, characterized in that, The device includes a memory and a processor, the memory being used to store a computer program, and the processor running the computer program to enable the electronic device to perform the rail surface damage diagnosis method based on dependence metric as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the rail surface damage diagnosis method based on dependence metric as described in any one of claims 1-7.

Images