A Method and System for Monitoring the Degradation State of Axle Box Bearings Based on Morphological Feature Extraction

Abstract

The application discloses a kind of based on morphological feature extraction axle box bearing degradation state monitoring method and system, the method includes the following steps: S1, acquisition axle box bearing vibration signal x (n);S2, construct morphological composite open operator O and composite closed operator C, construct signal feature extraction operator D;Iterative use signal feature extraction operator D, extract 5 feature sequences of axle box bearing vibration signal x (n);S3, corresponding multivariate normal distribution is constructed as normal model f of vibration signal x (n);S4, establish benchmark normal model f B And real-time normal model f R ;S5, calculate the deviation between real-time normal model f R And benchmark normal model f B , as axle box bearing degradation state evaluation value V, measure the degradation degree of axle box bearing, send early warning information.Adopting the technical scheme, the deviation between real-time normal model and benchmark normal model is obtained, the anti-noise ability is strong, and the degradation state of axle box bearing can be effectively evaluated.

Description

Technical Field

[0001] This invention belongs to the technical field of axle box bearing monitoring equipment, and relates to a method and system for monitoring the degradation state of axle box bearings based on morphological feature extraction. Background Technology

[0002] The bogie is a crucial component of railway trains. It mainly consists of a frame, wheelsets, axle boxes, and braking systems, serving primarily to provide support and steering. The axle box, which fits onto the journals of the wheelsets and connects them to the bogie frame, is responsible for transferring the weight and load of the train body to the wheelsets. Through the axle box, the train's weight and various loads are effectively transferred to the wheelsets, ensuring the train's stability and safety during operation.

[0003] Railway transportation plays an irreplaceable and vital role in national economic and social development. With the rapid development of railway transportation, the axle box bearings of train bogies, as key components in train operation, directly affect the safety and efficiency of train operation. Under complex operating environments and conditions, axle box bearings frequently face the effects of fatigue damage, overload, and corrosion, all of which lead to a gradual decline in their performance, i.e., degradation, thus directly impacting the safe operation of trains.

[0004] Monitoring the health and degradation status of bogie axle box bearings has always been a key focus and challenge in railway operation and maintenance. Traditional methods for monitoring the health and degradation status of axle box bearings often rely on manual experience and professional knowledge, requiring the calculation of a series of complex indicators. This cumbersome and inefficient diagnostic process fails to meet the demands of modern train operation and maintenance. Therefore, developing an efficient and accurate method for monitoring the degradation status of axle box bearings is of great significance for ensuring train safety and improving train operating efficiency. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for monitoring the degradation state of axle box bearings based on morphological feature extraction, which can automatically assess the degradation state of axle box bearings and improve work efficiency.

[0006] To achieve the above objectives, the basic solution of this invention is: a method for monitoring the degradation state of axle box bearings based on morphological feature extraction, comprising the following steps:

[0007] S1, Vibration sensors are placed on the axle box bearings of the train bogie to collect the vibration signal x(n) of the axle box bearings;

[0008] S2, constructing the morphological composite opening operator O and the composite closing operator C;

[0009] Based on the morphological composite opening operator O and composite closing operator C, a signal feature extraction operator D is constructed.

[0010] The signal feature extraction operator D is used iteratively to extract multiple feature sequences of the bearing vibration signal x(n).

[0011] S3. Based on multiple feature sequences of the vibration signal x(n) of the axle box bearing, a corresponding multivariate normal distribution is constructed as the normal model f of the vibration signal x(n);

[0012] S4. Collect vibration signal data of the bearing box under normal operating conditions. Following steps S2 and S3, establish a multivariate normal distribution as the reference normal model f. B ;

[0013] Collect real-time vibration signals of the axle box bearing, and establish a real-time normal model f according to steps S2 and S3. R ;

[0014] S5, Calculate the real-time normal model f R Compared with the baseline normal model f B The deviation between the two values ​​is used as the assessment value V for the degradation state of the axle box bearing, and the degree of degradation of the axle box bearing is measured to issue a pre-alarm message.

[0015] The working principle and beneficial effects of this basic scheme are as follows: This technical scheme acquires the vibration signal of the axle box bearing, constructs a composite operator based on the basic transformations of mathematical morphology, and builds a signal feature extraction operator to extract the feature sequence of the signal in an iterative manner. A normal model of the vibration signal is then established from the feature sequence. Next, the vibration signal of the axle box bearing under normal operating conditions is acquired to establish a reference normal model, and the real-time vibration signal of the axle box bearing is acquired to establish a real-time normal model. Finally, the deviation between the real-time normal model and the reference normal model is calculated to assess the degradation state of the axle box bearing and output pre-alarm information. This method is simple, efficient, and has strong noise resistance, effectively assessing the degradation state of the axle box bearing.

[0016] Furthermore, in step S2, based on the fundamental morphological transformations, a morphological composite opening operator O is constructed as follows:

[0017]

[0018] Where x is the signal sequence; g1 is the first structuring element sequence; and g2 is the second structuring element sequence. The basic morphological transformation, opening, is defined as follows:

[0019]

[0020] Where x is the signal sequence participating in the opening operation; s is the structuring element sequence participating in the opening operation; "⊙" represents the basic morphological transformation: erosion; This represents the basic morphological transformation: dilation;

[0021] The basic morphological transformations, corrosion and dilation, are defined as follows:

[0022]

[0023] Where x(n) is the signal sequence participating in the transformation, n is the index of the point in the signal sequence x(n), and N is the total number of points in the signal sequence x(n); s(j) is the sequence of structuring elements participating in the transformation, j is the index of the point in the structuring element sequence s(j), and J is the total number of points in the structuring element sequence s(j); [](n) represents the value of the nth point of the transformation result;

[0024] Furthermore, after the signal sequence x(n) undergoes erosion and dilation transformations by the structuring element s(j), missing data is filled with 0s to maintain the length of the signal sequence x(n).

[0025]

[0026] Based on the fundamental transformations of mathematical morphology, a morphological composite opening operator is constructed, which is simple to operate.

[0027] Furthermore, in step S2, the composite closing operator C is constructed as follows:

[0028] C[x,g1,g2]=(x·g1)·g2

[0029] C[x,g1,g2](n)=[(x·g1)·g2](n)={[x(n)·g1(m)]·g2(h)}

[0030] Where x is the signal sequence; g1 is the first structuring element sequence; g2 is the second structuring element sequence; · represents the basic morphological transformation: the closing operation, which is defined as follows:

[0031]

[0032] Where x is the signal sequence participating in the closing operation; w is the structuring element sequence participating in the closing operation;

[0033] The first structuring element sequence g1(m) is a triangle with height H1 and width L1, and its expression is:

[0034]

[0035] The second structuring element sequence g2(h) is sinusoidal, with a height of H2 and a width of L2, and its expression is:

[0036]

[0037] The calculation is simple and easy to operate.

[0038] Furthermore, the method for constructing a signal feature extraction operator D to extract the five feature sequences of the axle box bearing vibration signal x(n) is as follows:

[0039] The input to the feature extraction operator D is the signal sequence to be processed. The feature extraction operator D includes two parallel lines, each of which is a combination of a composite open operator O and a composite closed operator C. Finally, the results of the two lines are averaged to obtain the output, which is the feature sequence of the signal.

[0040] After feature extraction by feature extraction operator D, the signal sequence x(n) is used to obtain the feature sequence y1(n), and the corresponding residual signal r1(n) is calculated:

[0041] r1(n) = x(n) - y1(n)

[0042] The feature extraction operator D is used to extract the features of r1(n) to obtain the feature sequence y2(n), and the corresponding residual signal r2(n) is calculated:

[0043] r2(n) = r1(n) - y2(n)

[0044] The feature extraction operator D is used iteratively until five feature sequences are extracted: y1(n), y2(n), ..., y5(n).

[0045] Five feature sequences were extracted for later use.

[0046] Furthermore, in step S3, the method for constructing the corresponding multivariate normal distribution as the normal model f of the vibration signal x(n) based on the five feature sequences of the axle box bearing vibration signal x(n) is as follows:

[0047] Five feature sequences y1(n), y2(n), ..., y5(n) are extracted from the vibration signal x(n). The mean values ​​μ1, μ2, ..., μ5 and the covariance matrix Σ of each feature sequence are calculated.

[0048]

[0049] Where, σ ij Represents the sequence y i (n) and y j The covariance between (n) is calculated using the following formula:

[0050]

[0051] in, Representing the feature sequence y respectively i (n), y j The average value of (n); R is the total number of points in the feature sequence y, and y(t) represents the value of point t in the feature sequence y;

[0052] Mean vector composed of the average values ​​of each feature sequence And the covariance matrix Σ, construct a multivariate normal distribution f as the normal model of the vibration signal x(n), and the probability density function of the multivariate normal distribution f. for:

[0053]

[0054] in, Let be the independent variable of the probability density function, which is a vector with respect to the mean. Column vectors of the same length; Let Σ be the mean vector of the multivariate normal distribution f, Σ be the covariance matrix of the multivariate normal distribution f, exp[] denote the exponent of the natural constant e; k is the mean vector. The length of the vector is T, where T represents finding the transpose of the vector.

[0055] Five feature sequences extracted from the vibration signal x(n) are used to construct a corresponding multivariate normal distribution as the normal model f of the vibration signal x(n).

[0056] Furthermore, in step S5, the real-time normal model f is calculated. R Compared with the baseline normal model f B The deviation between them is determined by the following steps:

[0057] Calculate the normal model f of real-time vibration signal R Compared with the baseline normal model f B The JS divergence between JSD(f) R ||f B ), as the deviation between the two models:

[0058]

[0059] Where KL represents the KL divergence; f M It is f R with f B The mixed average distribution satisfies:

[0060]

[0061] in, The distribution is a multivariate normal distribution f M The mean vector, Σ M The distribution is a multivariate normal distribution f M The covariance matrix; The distribution is a multivariate normal distribution f R The mean vector, The distribution is a multivariate normal distribution f B The mean vector; Σ R The distribution is a multivariate normal distribution f R The covariance matrix, Σ B The distribution is a multivariate normal distribution f B The covariance matrix;

[0062] The KL divergence is:

[0063]

[0064] Among them, f p with f q For the two multivariate normal distributions involved in the KL divergence calculation, The distribution is a multivariate normal distribution f p f q The mean vector, Σ p Σ q The distribution is a multivariate normal distribution f p f q The covariance matrix is ​​given by f, det() calculates the determinant of the matrix, Tr() calculates the trace of the matrix, T calculates the transpose of the vector, and d represents the multivariate normal distribution f. p with f q mean vector The dimension of the mean vector is the length of the mean vector.

[0065] Calculate the real-time normal model f R Compared with the baseline normal model f B The deviation between them is used to assess the degradation condition of the axle box bearings.

[0066] Furthermore, using the normal model f of the real-time vibration signal of the axle box bearing... R Compared with the baseline normal model f B The JS divergence between the two is used as the degradation state assessment value V. The larger V is, the more severe the degradation state of the bearing box is.

[0067] When 0 ≤ V < 0.20, the axle box bearing is in normal condition;

[0068] When 0.20≤V<0.40, the axle box bearing is in a state of slight degradation;

[0069] When 0.40≤V<0.60, the axle box bearing is in a state of moderate degradation;

[0070] When V≥0.60, the axle box bearing is in a severely degraded state;

[0071] A warning message is issued when the bearing in the axle box is in a state of moderate degradation.

[0072] By dividing the state into multiple state ranges and using the degradation state evaluation value V, the degradation state of the axle box bearing can be quickly determined, resulting in higher efficiency.

[0073] The present invention also provides a system for monitoring the degradation status of axle box bearings based on morphological feature extraction, including a data acquisition module and a processing module;

[0074] The data acquisition module includes a vibration sensor arranged on the axle box bearing of the train bogie, and the data acquisition module is used to acquire the vibration signal x(n) of the axle box bearing;

[0075] The input terminal of the processing module is connected to the output terminal of the data acquisition module. The processing module executes the method described in this invention to monitor the degradation state of the axle box bearing.

[0076] Using this system to efficiently and accurately monitor the degradation status of axle box bearings is of great significance for ensuring train safety and improving train operation efficiency. Attached Figure Description

[0077] Figure 1 This is a flowchart illustrating the method for monitoring the degradation state of axle box bearings based on morphological feature extraction according to the present invention.

[0078] Figure 2 This is a schematic diagram of the vibration sensor used in the method for monitoring the degradation state of axle box bearings based on morphological feature extraction according to the present invention.

[0079] Figure 3 This is a schematic diagram of the signal feature extraction operator D in the axle box bearing degradation state monitoring method based on morphological feature extraction of the present invention;

[0080] Figure 4 This is a flowchart illustrating the feature sequence extraction process of the axle box bearing degradation state monitoring method based on morphological feature extraction according to the present invention. Detailed Implementation

[0081] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0082] In the description of this invention, it should be understood that the terms "longitudinal", "lateral", "up", "down", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0083] In the description of this invention, unless otherwise specified and limited, it should be noted that the terms "installation", "connection" and "linking" should be interpreted broadly. For example, they can refer to mechanical or electrical connections, or internal connections between two components. They can be direct connections or indirect connections through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms according to the specific circumstances.

[0084] This invention discloses a method for monitoring the degradation state of axle box bearings based on morphological feature extraction. This method is simple, efficient, and has strong noise resistance, effectively assessing the degradation state of axle box bearings. Figure 1 As shown, the method for monitoring the degradation condition of axle box bearings includes the following steps:

[0085] S1, as Figure 2 As shown, vibration sensors are arranged on the axle box bearings of the train bogie (one vibration sensor is arranged on each axle box of the train bogie) to collect the vibration signal x(n) of the axle box bearings;

[0086] S2, constructing the morphological composite opening operator O and the composite closing operator C;

[0087] Based on the morphological composite opening operator O and composite closing operator C, a signal feature extraction operator D is constructed.

[0088] Iteratively using the signal feature extraction operator D, multiple (e.g., 5) feature sequences of the bearing vibration signal x(n) are extracted;

[0089] S3. Based on multiple feature sequences of the vibration signal x(n) of the axle box bearing, a corresponding multivariate normal distribution is constructed as the normal model f of the vibration signal x(n);

[0090] S4. Collect vibration signal data of the bearing box under normal operating conditions. Following steps S2 and S3, establish a multivariate normal distribution as the reference normal model f. B ;

[0091] Collect real-time vibration signals of the axle box bearing, and establish a real-time normal model f according to steps S2 and S3. R ;

[0092] S5, Calculate the real-time normal model f R Compared with the baseline normal model f B The deviation between the two values ​​is used as the assessment value V for the degradation state of the axle box bearing, and the degree of degradation of the axle box bearing is measured to issue a pre-alarm message.

[0093] In a preferred embodiment of the present invention, in step S2, a morphological composite opening operator O is constructed based on the fundamental morphological transformations, as follows:

[0094]

[0095] Where x is the signal sequence; g1 is the first structuring element sequence; and g2 is the second structuring element sequence. The basic morphological transformation, opening, is defined as follows:

[0096]

[0097] Where x is the signal sequence participating in the opening operation; s is the structuring element sequence participating in the opening operation; "⊙" represents the basic morphological transformation: erosion; This represents the basic morphological transformation: dilation;

[0098] The basic morphological transformations, corrosion and dilation, are defined as follows:

[0099]

[0100] Where x(n) is the signal sequence participating in the transformation, n is the index of the point in the signal sequence x(n), and N is the total number of points in the signal sequence x(n); s(j) is the sequence of structuring elements participating in the transformation, j is the index of the point in the structuring element sequence s(j), and J is the total number of points in the structuring element sequence s(j); [](n) represents the value of the nth point of the transformation result;

[0101] Furthermore, after the signal sequence x(n) undergoes erosion and dilation transformations by the structuring element s(j), missing data is filled with 0s to maintain the length of the signal sequence x(n).

[0102]

[0103] In a preferred embodiment of the present invention, in step S2, the composite closing operator C is constructed as follows:

[0104] C[x,g1,g2]=(x·g1)·g2

[0105] C[x,g1,g2](n)=[(x·g1)·g2](n)={[x(n)·g1(m)]·g2(h)}

[0106] Where x is the signal sequence; g1 is the first structuring element sequence; g2 is the second structuring element sequence; · represents the basic morphological transformation: the closing operation, which is defined as follows:

[0107]

[0108] Where x is the signal sequence participating in the closing operation; w is the structuring element sequence participating in the closing operation;

[0109] The first structuring element sequence g1(m) is a triangle with height H1 and width L1, and its expression is:

[0110]

[0111] The second structuring element sequence g2(h) is sinusoidal, with a height of H2 and a width of L2, and its expression is:

[0112]

[0113] Based on the constructed morphological composite opening operator O and composite closing operator C, a signal feature extraction operator D is constructed, such as... Figure 3 As shown.

[0114] In a preferred embodiment of the present invention, such as Figure 4 As shown, the method for constructing the signal feature extraction operator D and extracting the five feature sequences of the axle box bearing vibration signal x(n) is as follows:

[0115] The input to the feature extraction operator D is the signal sequence to be processed. The feature extraction operator D includes two parallel lines, each of which is a combination of a composite open operator O and a composite closed operator C. Finally, the results of the two lines are averaged to obtain the output, which is the feature sequence of the signal.

[0116] After feature extraction by feature extraction operator D, the signal sequence x(n) is used to obtain the feature sequence y1(n), and the corresponding residual signal r1(n) is calculated:

[0117] r1(n) = x(n) - y1(n)

[0118] The feature extraction operator D is used to extract the features of r1(n) to obtain the feature sequence y2(n), and the corresponding residual signal r2(n) is calculated:

[0119] r2(n) = r1(n) - y2(n)

[0120] The feature extraction operator D is used iteratively until five feature sequences are extracted: y1(n), y2(n), ..., y5(n).

[0121] In a preferred embodiment of the present invention, in step S3, the method for constructing the corresponding multivariate normal distribution as the normal model f of the vibration signal x(n) based on the five feature sequences of the axle box bearing vibration signal x(n) is as follows:

[0122] Five feature sequences y1(n), y2(n), ..., y5(n) are extracted from the vibration signal x(n). The mean values ​​μ1, μ2, ..., μ5 and the covariance matrix Σ of each feature sequence are calculated.

[0123]

[0124] Where, σ ij Represents the sequence y i (n) and y j The covariance between (n) is calculated using the following formula:

[0125]

[0126] in, Representing the feature sequence y respectively i (n), y j The average value of (n); R is the total number of points in the feature sequence y, and y(t) represents the value of point t in the feature sequence y;

[0127] Mean vector composed of the average values ​​of each feature sequence And the covariance matrix Σ, construct a multivariate normal distribution f as the normal model of the vibration signal x(n), and the probability density function of the multivariate normal distribution f. for:

[0128]

[0129] in, Let be the independent variable of the probability density function, which is a vector with respect to the mean. Column vectors of the same length; Let Σ be the mean vector of the multivariate normal distribution f, Σ be the covariance matrix of the multivariate normal distribution f, exp[] denote the exponent of the natural constant e; k is the mean vector. The length of the vector is T, where T represents finding the transpose of the vector.

[0130] In a preferred embodiment of the present invention, in step S5, the real-time normal model f is calculated. R Compared with the baseline normal model f B The deviation between them is determined by the following steps:

[0131] Calculate the normal model f of real-time vibration signal R Compared with the baseline normal model f BThe JS divergence (Jensen-Shannon Divergence, JSD) between them. R ||f B ), as the deviation between the two models:

[0132]

[0133] Where KL represents the KL divergence (KLD); f M It is f R with f B The mixed average distribution satisfies:

[0134]

[0135] in, The distribution is a multivariate normal distribution f M The mean vector, Σ M The distribution is a multivariate normal distribution f M The covariance matrix; The distribution is a multivariate normal distribution f R The mean vector, The distribution is a multivariate normal distribution f B The mean vector; Σ R The distribution is a multivariate normal distribution f R The covariance matrix, Σ B The distribution is a multivariate normal distribution f B The covariance matrix;

[0136] The KL divergence is:

[0137]

[0138] Among them, f p with f q For the two multivariate normal distributions involved in the KL divergence calculation, The distribution is a multivariate normal distribution f p f q The mean vector, Σ p Σ q The distribution is a multivariate normal distribution f p f q The covariance matrix is ​​given by f, det() calculates the determinant of the matrix, Tr() calculates the trace of the matrix, T calculates the transpose of the vector, and d represents the multivariate normal distribution f. p with f q mean vector The dimension of the mean vector is the length of the mean vector.

[0139] In a preferred embodiment of the present invention, the normal model f of the real-time vibration signal of the axle box bearing is used. R Compared with the baseline normal model f B The JS divergence between the two is used as the degradation state assessment value V. The larger V is, the more severe the degradation state of the bearing box is.

[0140] When 0 ≤ V < 0.20, the axle box bearing is in normal condition;

[0141] When 0.20≤V<0.40, the axle box bearing is in a state of slight degradation;

[0142] When 0.40≤V<0.60, the axle box bearing is in a state of moderate degradation;

[0143] When V≥0.60, the axle box bearing is in a severely degraded state;

[0144] When the axle box bearing is in a state of moderate degradation, an early warning message is issued so that the corresponding axle box bearing can be inspected and maintained in a timely manner.

[0145] This invention also provides a system for monitoring the degradation state of axle box bearings based on morphological feature extraction, including a data acquisition module and a processing module. The data acquisition module includes vibration sensors arranged on the axle box bearings of the train bogie, and is used to acquire the vibration signal x(n) of the axle box bearings.

[0146] The input terminal of the processing module is electrically connected to the output terminal of the data acquisition module. The processing module executes the method described in this invention to monitor the degradation state of the axle box bearings. Utilizing this system, the efficient and accurate monitoring of the axle box bearing degradation state is of great significance for ensuring train safety and improving train operation efficiency.

[0147] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0148] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for monitoring the degradation state of an axle box bearing based on morphological feature extraction, characterized in that, Comprising the following steps: S1, arranging a vibration sensor on an axle box bearing of a train bogie to collect a vibration signal of the axle box bearing ; S2, constructing morphological composite open operator O and composite closed operator C; Based on morphological composite open operator O and composite closed operator C, constructing signal feature extraction operator D; The signal feature extraction operator D is iteratively used to extract five feature sequences of the vibration signal of the axle box bearing ​ S3, based on the shaft bearing vibration signal of multiple feature sequences, construct the corresponding multivariate normal distribution as the normal model of the vibration signal ;​ S4, collect the vibration signal data of the axle box bearing in the normal running state, and establish a multivariate normal distribution as a reference normal model according to steps S2 and S3 ; Collecting real-time vibration signal of the axle box bearing, and establishing real-time normal model according to steps S2 and S3 ; S5, calculating real-time normal model deviation between the reference normal model , as the axle box bearing degradation state evaluation value V, and thereby measure the degradation degree of the axle box bearing, and issue a pre-warning information; A signal feature extraction operator D is constructed to extract five feature sequences of the vibration signal of the bearing of the axle box The method is as follows: The input of feature extraction operator D is the signal sequence to be processed, and the feature extraction operator D comprises two parallel lines, each of which is a combination of composite open operator O and composite closed operator C, which are respectively: sequentially connected composite open operator O, composite closed operator C, composite open operator O; sequentially connected composite closed operator C, composite open operator O, composite closed operator C; finally, the results of the two lines are averaged to obtain the output, that is, the feature sequence of the signal; signal sequence After feature extraction by the feature extraction operator D, a feature sequence is obtained, and the corresponding residual signal is calculated: ， using a feature extraction operator D , resulting in a sequence of features , and computing a corresponding residual signal : ， The feature extraction operator D is iteratively used until 5 feature sequences are extracted: ; A normal model of real-time vibration signals of a shaft box bearing JS divergence between the reference normal model is taken as a degradation state evaluation value V, the larger V is, the more serious the degradation state of the shaft box bearing is. When the shaft bearing is in normal state; When the shaft bearing is in a slightly degraded state; When the shaft bearing is in a moderate state of degradation; When the shaft bearing is in a severe degradation state; When the axle box bearing is in a moderate degradation state, a warning information is sent out.

2. The morphological feature extraction based axle box bearing degradation condition monitoring method of claim 1, wherein, In step S2, according to the basic transformation of morphology, the morphological composite open operator O is constructed, which is: ， wherein is a signal sequence; is a first structuring element sequence; is a second structuring element sequence; denotes a morphological elementary transformation: opening, which is defined as follows: ； wherein is a signal sequence participating in the opening operation; is a structuring element sequence participating in the opening operation; denotes the morphological basic transformation: erosion; denotes the morphological basic transformation: dilation; The definition of morphological basic transformation erosion and dilation is: ， ， wherein, is the signal sequence participating in the transform, n is the sequence number of the point of the signal sequence , N is the total number of points of the signal sequence ; is the structural element sequence participating in the transform, j is the sequence number of the point of the structural element sequence , J is the total number of points of the structural element sequence ; represents the value of the nth point of the transform result; In addition, the signal sequence After the erosion, dilation transform of the structure element , fill the missing data with 0 without changing the length of the signal sequence . ， 。 3. The method for monitoring the degradation condition of the axle box bearing based on morphological feature extraction according to claim 2, characterized in that, In step S2, the composite closed operator C is constructed, which is: ， ， wherein is a signal sequence; is a first structuring element sequence; is a second structuring element sequence; denotes a morphological elementary transformation: closing, which is defined as follows: ， wherein is a sequence of signals participating in the closing operation; is a sequence of structuring elements participating in the closing operation; The first sequence of structural elements The triangle is selected, the height is The width is The expression is: ， ， Second sequence of structure elements The sinusoidal shape is chosen, with a height of and a width of The expression is: ， 。 4. The morphological feature extraction based axle box bearing degradation condition monitoring method of claim 1, wherein, In step S3, based on the vibration signal of the axle box bearing Five characteristic sequences were used to construct the corresponding multivariate normal distribution as vibration signals. normal model The method is as follows: From the vibration signal 5 feature sequences are extracted The average value of each feature sequence is calculated respectively and the covariance matrix : ， wherein represents the sequence and the covariance between and, calculated as: ， wherein respectively represent the average value of the characteristic sequence , ; R is the total number of points of the characteristic sequence y, y(t) represents the value of the tth point of the characteristic sequence y. Mean vector composed of the average values ​​of each feature sequence and covariance matrix Construct a multivariate normal distribution As a vibration signal The normal model, multivariate normal distribution probability density function for: ， wherein is the length of the mean vector is a column vector of the same length as the mean vector is the mean vector of the multivariate normal distribution is the covariance matrix of the multivariate normal distribution is the covariance matrix of the multivariate normal distribution is the covariance matrix of the multivariate normal distribution denotes the exponential of the natural constant e; k is the length of the mean vector and T denotes the transposition of a vector.

5. The morphological feature extraction based axle box bearing degradation condition monitoring method as claimed in claim 1, wherein, In step S5, the deviation degree from the reference normal model is calculated with the real-time normal model The specific steps are as follows: Computing a normal model of real-time vibration signals The JS divergence between the normal model , as a measure of deviation between the two models:​ ， wherein, denotes the KL divergence; is with the mixture average distribution, satisfying: ， ， in, It is a multivariate normal distribution The mean vector, It is a multivariate normal distribution The covariance matrix; It is a multivariate normal distribution The mean vector, It is a multivariate normal distribution The mean vector; It is a multivariate normal distribution The covariance matrix, It is a multivariate normal distribution The covariance matrix; KL divergence is: ， wherein is the mean vector of the multivariate normal distribution is the covariance matrix of the multivariate normal distribution is the mean vector of the multivariate normal distribution is the covariance matrix of the multivariate normal distribution is the mean vector of the multivariate normal distribution is the covariance matrix of the multivariate normal distribution denotes the determinant of a matrix, denotes the trace of a matrix, T denotes the transpose of a vector, d is the dimension number of the multivariate normal distribution is the mean vector of the multivariate normal distribution is the mean vector of the multivariate normal distribution is the dimension number of the multivariate normal distribution, i.e. the length of the mean vector.

6. A system for monitoring the degradation state of a journal bearing based on morphological feature extraction, characterized in that Comprising a data acquisition module and a processing module; The data acquisition module comprises a vibration sensor arranged on an axle box bearing of a bogie of a train, and the data acquisition module is used for acquiring a vibration signal of the axle box bearing ; The output end of the data acquisition module is connected with the input end of the processing module, and the processing module executes the method in any one of claims 1-5 to monitor the degradation state of the axle box bearing.

Images