Point switch fault diagnosis method based on case similarity matching

Abstract

The invention provides a switch machine fault diagnosis method based on case similarity matching. The method comprises the following steps: S1, establishing a case action curve database based on a switch machine action curve of K types of cases; s2, sampling the action curve of the point switch of the ith type of case to generate an action curve sequence of the ith type of case; sampling a to-be-analyzed action curve of the switch machine to generate a to-be-analyzed action curve sequence; s3, calculating a dynamic time warping distance between the to-be-analyzed action curve sequence and the ith case action curve sequence; s4, calculating a segmentation time distance between the to-be-analyzed action curve sequence and the ith case action curve sequence; s5, based on the dynamic time warping distance and the segmentation time distance, calculating the similarity between the to-be-analyzed action curve sequence and the ith case action curve sequence; and S6, based on a fault category corresponding to the to-be-analyzed action curve sequence. According to the invention, the accuracy of fault diagnosis of the switch machine is improved, and the dependence on manpower is reduced.

Description

Technical Field

[0001] The present invention relates to the field of rail transit technology, and in particular to a switch machine fault diagnosis method based on case similarity matching. Background Art

[0002] As an important railway signal trackside equipment, the switch machine is the direct execution unit of the turnout. Since the switch machine is affected by factors such as its own installation, external environment and wear and tear, it may cause the equipment to operate poorly or even fail to operate, posing a hidden danger to driving safety and operational efficiency. Therefore, staff are required to perform fault detection and diagnosis and maintenance on the switch machine in a timely manner, which relies heavily on the professional level of the personnel and has a heavy workload. Therefore, the automated switch machine fault diagnosis method is an important guarantee for the safe and efficient operation of trains, and is also one of the research hotspots in the field of rail transit.

[0003] At present, the switch machine fault diagnosis method mainly analyzes the action curve during the rotation process to determine whether the switch machine has a fault and the fault type. According to the technical principle, the switch machine fault diagnosis method can be divided into model-driven method and data-driven method.

[0004] The shortcomings of the model-driven approach include: ① Due to the variety of switch types and complex internal structures, it is difficult to establish a high-precision mechanism model. ② The cost is too high, and complex diagnostic logic needs to be written through manual experience and mechanism. ③ Due to the influence of different equipment operating environments and service times, the action characteristics of similar switch machines will be different, and the same set of diagnostic logic is difficult to promote.

[0005] The data-driven method uses big data analysis and artificial intelligence technology to perform reasoning diagnosis by training network models, which is highly applicable and economical. However, the data-driven method mainly relies on machine learning, deep learning, etc., which requires a large amount of sample data, and different diagnostic models need to be trained for different regions or stations. In addition, the switch machines in different regions have different equipment models, different suppliers, and different service times. The massive historical action curves of switch machines have complicated specifications and cannot be directly used to train network models without professional manual labeling.

[0006] How to improve the universality of switch machine fault diagnosis methods while reducing the economic cost and reliance on manual experience in the diagnosis process is an urgent problem that needs to be solved. Summary of the invention

[0007] The purpose of the present invention is to provide a switch machine fault diagnosis method based on case similarity matching, which diagnoses switch machine faults by calculating the dynamic time regularization distance and segmented time length distance between the action curve sequence to be analyzed and the case action curve sequence, thereby greatly improving the diagnostic accuracy, having good universality and reducing dependence on manual labor.

[0008] In order to achieve the above object, the present invention provides a switch machine fault diagnosis method based on case similarity matching, comprising the steps of:

[0009] S1. Sampling the switch machine action curves of the first to K-th cases to obtain the action curve sequences of the first to K-th cases; establishing a case action curve database based on the action curve sequences of the first to K-th cases;

[0010] S2, sampling the switch machine action curve to be analyzed, and generating a corresponding action curve sequence to be analyzed;

[0011] S3, calculating the dynamic time warping distance between the action curve sequence to be analyzed and the action curve sequence of the i-th case; i∈[1, K];

[0012] S4, calculating the segmented time distance between the action curve sequence to be analyzed and the action curve sequence of the i-th case;

[0013] S5. Calculate the similarity between the action curve sequence to be analyzed and the action curve sequence of the i-th case based on the dynamic time warping distance and the segment duration distance;

[0014] S6. Determine the fault category corresponding to the action curve sequence to be analyzed based on the similarity.

[0015] Optionally, step S2 includes:

[0016] S21, sampling the switch machine action curve to be analyzed at a set time interval to obtain the action curve sequence to be analyzed Wherein, n represents the length of the action curve sequence to be analyzed; x 1 ~x n It indicates the results of the first to nth sampling of the switch machine action curve to be analyzed;

[0017] S22, sampling the action curve sequence of the i-th case according to the time interval to obtain the action curve sequence of the i-th case Among them, m i is the length of the action curve sequence of the i-th case; Indicates the first to mth operations on the action curve sequence of case i i The result of subsampling.

[0018] Optionally, step S3 includes:

[0019] S31. Constructing distance matrix

[0020]

[0021] make Represents the distance matrix The element in row a and column b in x a is the result of the ath sampling of the switch machine action curve to be analyzed, is the action curve sequence Y for case i i The result of the b-th sampling; aa∈[1,n], b∈[1,m];

[0022] S32, let the distance matrix middle arrive Path Q is a natural number, max(n,m)≤Q<n+m-1;

[0023] is the path W i The qth step in the distance matrix Middle a q Row b q Elements of a column in,

[0024] S33, Order Action curve sequence X and action curve sequence Y of the i-th case i The dynamic time warping distance between

[0025]

[0026] Optional, The paths W i The qth step and the q+1th step in the equation satisfy: 0≤a q+1 -a q ≤1 and 0≤b q+1 -b q ≤1.

[0027] Optionally, step S4 includes:

[0028] S41, calculate the segment point position index of the action curve sequence X to be analyzed, and obtain the change point index sequence BD X =[t 1 , t 2 …, t U ];

[0029] U is a natural number, t u+1 ∈BD X , 1≤t u <t u+1 ≤n;

[0030] S42, dividing the action curve sequence X to be analyzed into U subsequences in chronological order, wherein the u-th subsequence includes the elements in the action curve sequence X to be analyzed

[0031] where t 1 =1,t U+1 -1=n; generate time interval sequence in, express and The sampling time interval between them; u∈[1,U];

[0032] S43, for the action curve sequence Y of the i-th case i Calculate the segment point position index to obtain the change point index sequence BD i =[T 1 , …, T V ];

[0033] V is a natural number, T v+1 ∈BD i , 1≤T v <T v+1 ≤m;

[0034] S44, the action curve sequence Y of the i-th case i It is divided into V subsequences in chronological order, where the vth subsequence includes the action curve sequence Y of the i-th case i Elements in Where T 1 =1,T V+1 -1=m i ;

[0035] Generate time interval sequence in express and The sampling time interval between them, v∈[1,V];

[0036] S45. If U=V, calculate the time interval sequence T X With TY i Euclidean distance Will As the action curve sequence X to be analyzed and the action curve sequence Y of the i-th case i The time distance between the segments,

[0037]

[0038] Optionally, step S41 includes:

[0039] S411, mapping the action curve sequence X to be analyzed to the reproducing kernel Hilbert space by using the mapping function φ(·) Get the mapped sequence

[0040] S412, by minimizing the objective function V(·), the segment point position index calculation is performed on the sequence XH to obtain the change point index sequence BD X =[t 1 ,…,t U ],satisfy:

[0041]

[0042]

[0043] in, for The mean of represents the reproducing kernel Hilbert space 2-norm operation in .

[0044] Optionally, step S43 includes:

[0045] S431, transform the action curve sequence Y of the i-th case into i Mapping to the reproducing kernel Hilbert space Get the mapped sequence

[0046] S432, by minimizing the objective function V(·) i Calculate the segment point position index and obtain the change point index sequence [T 1 ,…,T V ],satisfy:

[0047]

[0048]

[0049] in, for The mean of .

[0050] Optional, φ(x r )=xh r ∈XH, r∈[1,n], x r ∈X,xh r For x r In the reproducing kernel Hilbert space The mapping value in ;

[0051] The reproducing kernel Hilbert space Inner product operation in is the square of the 2-norm operation,

[0052]

[0053] Among them, k(·,·) is the kernel function.

[0054] Optionally, the kernel function is any one of a linear kernel function, a Gaussian kernel function, and a cosine kernel function.

[0055] Optionally, step S5 includes:

[0056] S51, yes Normalize it and get

[0057]

[0058]

[0059] S52. For the DC switch machine, the action curve sequence X to be analyzed and the action curve sequence Y of the i-th case are calculated. i The similarity between i ;

[0060]

[0061] S53. For AC switch machines:

[0062] The switch machine action curve includes: A phase current curve, B phase current curve, C phase current curve, power curve;

[0063] The action curve sequence to be analyzed includes: sampling the A-phase current curve, B-phase current curve, C-phase current curve, and power curve to be analyzed, to obtain the A-phase current curve sequence, B-phase current curve sequence, C-phase current curve sequence, and power curve sequence to be analyzed;

[0064] The action curve sequence of the i-th case includes: sampling the A-phase current curve, the B-phase current curve, the C-phase current curve, and the power curve of the i-th case to obtain the A-phase current curve sequence, the B-phase current curve sequence, the C-phase current curve sequence, and the power curve sequence of the i-th case;

[0065] include are respectively the dynamic time warping distances between the A-phase current curve sequence, the B-phase current curve sequence, the C-phase current curve sequence, and the power curve sequence to be analyzed and the A-phase current curve sequence, the B-phase current curve sequence, the C-phase current curve sequence, and the power curve sequence for the i-th case;

[0066] make Respectively express The result after normalization:

[0067]

[0068]

[0069]

[0070]

[0071] include are the segmented time length distances between the A-phase current curve sequence, B-phase current curve sequence, C-phase current curve sequence, and power curve sequence to be analyzed and the A-phase current curve sequence, B-phase current curve sequence, C-phase current curve sequence, and power curve sequence for the i-th case;

[0072] make Respectively express The result after normalization:

[0073]

[0074]

[0075]

[0076]

[0077] Order i，A 、s i，B 、s i，C 、s i，P Respectively represent the similarity between the A-phase current curve sequence, B-phase current curve sequence, C-phase current curve sequence, and power curve sequence to be analyzed and the A-phase current curve sequence, B-phase current curve sequence, C-phase current curve sequence, and power curve sequence of the i-th case;

[0078]

[0079]

[0080]

[0081]

[0082] The action curve sequence X to be analyzed and the action curve sequence Y of the i-th case i The similarity between i ,

[0083] Optionally, step S6 includes:

[0084] S61. Calculate the action curve sequence X to be analyzed and the action curve sequence Y of the i-th case i The probability of matching is p i ,

[0085] S62. Calculate p max with i max ,

[0086] p max =max(P), P = [p 1 , p 2 , ..., p K ];

[0087] i max = argmax i＝1，2，…，K (P), P = [p 1 , p 2 ,…,p K ];

[0088] Let θ represent the preset probability threshold,

[0089] If p max >θ, then determine whether the action curve sequence X to be analyzed is consistent with the i-th max Case-like action curve sequence The fault category is the same;

[0090] If p max <θ, it is determined that the action curve sequence X to be analyzed does not belong to the fault category in the case action curve database; the action curve sequence X to be analyzed is manually analyzed, and if the analysis result is a fault, the switch machine action curve sequence to be analyzed is added to the case action curve database as a new type of fault case.

[0091] Optionally, the following steps are also included:

[0092] S7. Send the diagnostic results of the action curve sequence to be analyzed to the intelligent operation and maintenance platform:

[0093] If the diagnosis is normal, the diagnosis result "normal" is returned;

[0094] If the diagnosis is a fault, the fault type, cause analysis and treatment method will be sent, and an alarm will be displayed.

[0095] Optionally, the switch machine action curve includes a switch machine action three-phase current curve, a switch machine action DC current curve and a switch machine action power curve.

[0096] Compared with the prior art, the present invention has the following beneficial effects:

[0097] The switch machine fault diagnosis method based on case similarity matching of the present invention diagnoses the switch machine fault by calculating the dynamic time warping distance and the segmented time length distance between the action curve sequence to be analyzed and the case action curve sequence, overcoming the singularity problem caused by the continuous monotonic characteristics of the dynamic time warping algorithm (i.e., unnatural distortion or warping occurs when the sequence is aligned, resulting in reduced accuracy of the similarity measurement), greatly reducing the probability of mismatching with the case switch machine action curve in the switch machine fault diagnosis, effectively improving the diagnostic accuracy and reducing dependence on manual labor.

[0098] The present invention is applicable to complex time series, does not require data preprocessing and large amounts of training data modeling, and has good universality.

[0099] The diagnostic results of the present invention can be applied to the intelligent operation and maintenance system and generate corresponding alarm information, so as to facilitate on-site staff to quickly locate the source of the fault, thereby handling the fault as soon as possible and ensuring the safe operation of the train and the turnout machine. BRIEF DESCRIPTION OF THE DRAWINGS

[0100] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for the description. Obviously, the drawings described below are an embodiment of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work:

[0101] Figure 1 The flowchart of the switch machine fault diagnosis method based on case similarity matching in an embodiment of the present invention. DETAILED DESCRIPTION

[0102] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0103] It should be understood that when used in this specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0104] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0105] It should be further understood that the term “and / or” used in the specification and appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0106] As used in this specification and the appended claims, the term "if" may be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrases "if it is determined" or "if [described condition or event] is detected" may be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0107] In addition, in the description of the present application, the terms "first", "second", "third", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0108] The present invention provides a switch machine fault diagnosis method based on case similarity matching, such as Figure 1 As shown, it includes the steps:

[0109] S1. Sampling the switch machine action curves of the first to K-th cases to obtain the action curve sequences of the first to K-th cases; establishing a case action curve database based on the action curve sequences of the first to K-th cases;

[0110] In this embodiment, for a DC switch, the action curve is a current curve. For an AC switch, the action curve includes an A-phase current curve, a B-phase current curve, a C-phase current curve and a power curve, that is, each type of case includes an A-phase current curve sequence, a B-phase current curve sequence, a C-phase current curve sequence, and a power curve sequence, for a total of 4 curve sequences. K types of cases correspond to K types of switch fault types.

[0111] S2, sampling the switch machine action curve to be analyzed, and generating a corresponding action curve sequence to be analyzed;

[0112] Step S2 includes:

[0113] S21, sampling the switch machine action curve to be analyzed at the set time interval to obtain the action curve sequence to be analyzed Where n represents the length of the action curve sequence to be analyzed; x 1 ~x n It indicates the results of the first to nth sampling of the switch machine action curve to be analyzed;

[0114] S22, sampling the action curve sequence of the i-th case at time intervals to obtain the action curve sequence of the i-th case Among them, m i is the length of the action curve sequence of the i-th case; Indicates the first to mth operations on the action curve sequence of case i i The result of subsampling.

[0115] S3, calculating the dynamic time warping distance between the action curve sequence to be analyzed and the action curve sequence of the i-th case;

[0116] Step S3 includes:

[0117] S31. Constructing distance matrix

[0118]

[0119] make Represents the distance matrix The element in row a and column b in x a is the result of the ath sampling of the switch machine action curve to be analyzed, is the action curve sequence Y for case i i The result of the b-th sampling is a∈[1,n] and b∈[1,m].

[0120] S32, let the distance matrix middle arrive Path The path W i The following constraints are met:

[0121] 1) Q is a natural number, max(n, m)≤Q <n+m-1。

[0122] 2) is the path W i The qth step in the distance matrix Middle a q Row b q Elements of a column

[0123] 3) The paths W i The qth step and the q+1th step in the equation satisfy: 0≤aq+1 -a q ≤1 and 0≤b q+1- b q ≤1.

[0124] S33, Order Action curve sequence X and action curve sequence Y of the i-th case i The dynamic time warping distance between

[0125]

[0126] S4, calculating the segmented time distance between the action curve sequence to be analyzed and the action curve sequence of the i-th case;

[0127] Step S4 includes:

[0128] S41, calculate the segment point position index of the action curve sequence X to be analyzed, and obtain the change point index sequence BD X =[t 1 , t 2 …, t U ];

[0129] U is a natural number, t u+1 ∈BD X , 1≤t u <t u+1 ≤n.

[0130] Step S41 includes:

[0131] S411, mapping the action curve sequence X to be analyzed to the reproducing kernel Hilbert space through the mapping function φ(·) Get the mapped sequence

[0132] φ(x r )=xh r ∈XH, r∈[1,n], x r ∈X,xh r For x r In the reproducing kernel Hilbert space The mapping value in . That is

[0133] They represent the reproducing kernel Hilbert space The 2-norm operation and inner product operation in

[0134] Since the inner product operation in high-dimensional space is complicated, the kernel method is used to transform the inner product operation in high-dimensional space into the kernel function operation in low-dimensional space:

[0135] Among them, k(·,·) is the kernel function.

[0136] The kernel function is any one of a linear kernel function, a Gaussian kernel function, and a cosine kernel function.

[0137] (1) Linear kernel function: k(α, β) = α T β.

[0138] (2) Gaussian kernel function: Where σ>0 is a user-settable parameter, and the default value is 1.

[0139] (3) Cosine kernel function: k(α, β) = (α T β) / (||α||·||β||).

[0140] S412, by minimizing the objective function V(·), the segment point position index calculation is performed on the sequence XH to obtain the change point index sequence BD X =[t 1 ,…,t U ],satisfy:

[0141]

[0142]

[0143] in, for The mean of .

[0144] S42, dividing the action curve sequence X to be analyzed into U subsequences in chronological order, wherein the u-th subsequence includes the elements in the action curve sequence X to be analyzed

[0145] where t 1 =1,t U+1 -1=n; generate time interval sequence in, express and The sampling time interval between them; u∈[1,U];

[0146] S43, for the action curve sequence Y of the i-th case i Calculate the segment point position index to obtain the change point index sequence BD i =[T 1 , …, T V ].

[0147] V is a natural number, T v+1 ∈BD i, 1≤T v <T v+1 ≤m.

[0148] Step S43 includes:

[0149] S431, transform the action curve sequence Y of the i-th case into i Mapping to the reproducing kernel Hilbert space Get the mapped sequence

[0150] S432, by minimizing the objective function V(·) i Calculate the segment point position index and obtain the change point index sequence [T 1 ,…,T V ],satisfy:

[0151]

[0152]

[0153] in, for The mean of .

[0154] S44, the action curve sequence Y of the i-th case i It is divided into V subsequences in chronological order, where the vth subsequence includes the action curve sequence Y of the i-th case i Elements in Where T 1 =1,T V+1 -1=m i .

[0155] Generate time interval sequence in express and The sampling time interval between them is v∈[1,V].

[0156] S45. If U=V, calculate the time interval sequence T X and Euclidean distance Will As the action curve sequence X to be analyzed and the action curve sequence Y of the i-th case i The time distance between the segments,

[0157]

[0158] S5. Calculate the similarity between the action curve sequence to be analyzed and the action curve sequence of the i-th case based on the dynamic time warping distance and the segment duration distance;

[0159] Step S5 includes:

[0160] S51, yes Normalize it and get

[0161]

[0162]

[0163] S52. For the DC switch machine, the action curve sequence X to be analyzed and the action curve sequence Y of the i-th case are calculated. i The similarity between i ;

[0164]

[0165] S53. For AC switch machines:

[0166] The switch machine action curve includes: A phase current curve, B phase current curve, C phase current curve, power curve;

[0167] The action curve sequence to be analyzed includes: sampling the A-phase current curve, B-phase current curve, C-phase current curve, and power curve to be analyzed, to obtain the A-phase current curve sequence, B-phase current curve sequence, C-phase current curve sequence, and power curve sequence to be analyzed;

[0168] The action curve sequence of the i-th case includes: sampling the A-phase current curve, the B-phase current curve, the C-phase current curve, and the power curve of the i-th case to obtain the A-phase current curve sequence, the B-phase current curve sequence, the C-phase current curve sequence, and the power curve sequence of the i-th case;

[0169] include are respectively the dynamic time warping distances between the A-phase current curve sequence, the B-phase current curve sequence, the C-phase current curve sequence, and the power curve sequence to be analyzed and the A-phase current curve sequence, the B-phase current curve sequence, the C-phase current curve sequence, and the power curve sequence for the i-th case;

[0170] make Respectively express The result after normalization:

[0171]

[0172]

[0173]

[0174]

[0175] include are the segmented time length distances between the A-phase current curve sequence, B-phase current curve sequence, C-phase current curve sequence, and power curve sequence to be analyzed and the A-phase current curve sequence, B-phase current curve sequence, C-phase current curve sequence, and power curve sequence for the i-th case;

[0176] make Respectively express The result after normalization:

[0177]

[0178]

[0179]

[0180]

[0181] Order i，A 、s i，B 、s i，C 、s i，P Respectively represent the similarity between the A-phase current curve sequence, B-phase current curve sequence, C-phase current curve sequence, and power curve sequence to be analyzed and the A-phase current curve sequence, B-phase current curve sequence, C-phase current curve sequence, and power curve sequence of the i-th case;

[0182]

[0183]

[0184]

[0185]

[0186] The action curve sequence X to be analyzed and the action curve sequence Y of the i-th case i The similarity between i ,

[0187] S6. Determine the fault category corresponding to the action curve sequence to be analyzed based on the similarity.

[0188] Step S6 includes:

[0189] S61. Calculate the action curve sequence X to be analyzed and the action curve sequence Y of the i-th case i The probability of matching is p i :

[0190]

[0191] S62. Calculate p max with i max :

[0192] p max =max(P), P = [p 1 , p 2 , ..., p K ];

[0193]

[0194] Let θ represent a preset probability threshold, in this embodiment, θ=0.3 (this is only used as an example).

[0195] If p max >θ, then determine whether the action curve sequence X to be analyzed is consistent with the i-th max Case-like action curve sequence The fault category is the same.

[0196] If p max <θ, it is judged that the action curve sequence X to be analyzed does not belong to the fault category in the case action curve database; the action curve sequence X to be analyzed is manually analyzed. If the analysis result is that the switch machine has a fault, the characteristic curve of the switch machine to be analyzed is added to the case action curve database as a new type of fault case.

[0197] In this embodiment, Figure 1 As shown, the switch machine fault diagnosis method based on case similarity matching also includes the steps of:

[0198] S7. Send the diagnosis result of the action curve sequence to be analyzed to the intelligent operation and maintenance platform.

[0199] (1) If the diagnosis is normal, the diagnosis result "normal" is returned.

[0200] (2) If the fault is diagnosed, the fault type, cause analysis and treatment method will be sent, and an alarm will be displayed.

[0201] The switch machine fault diagnosis method based on case similarity matching of the present invention diagnoses the switch machine fault by calculating the dynamic time warping distance and the segmented time length distance between the action curve sequence to be analyzed and the case action curve sequence, thereby overcoming the singularity problem caused by the continuous monotonic characteristics of the dynamic time warping algorithm (i.e., unnatural distortion or warping occurs when the sequence is aligned, resulting in reduced accuracy of the similarity measurement). The present invention greatly reduces the probability of mismatching with the case switch machine action curve in the switch machine fault diagnosis, effectively improves the diagnostic accuracy and reduces dependence on manual labor. The diagnostic results of the present invention can be applied to the intelligent operation and maintenance system, and generate corresponding alarm information, which is convenient for on-site staff to quickly locate the source of the fault, thereby handling the fault as soon as possible and ensuring the safe operation of the train and the turnout switch machine. The present invention is applicable to complex time series, does not require data preprocessing and large-scale training data modeling, and has good universality.

[0202] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0203] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A switch machine fault diagnosis method based on case similarity matching, characterized in that: Contains steps: S1. Sampling the switch machine action curves of the first to K-th cases to obtain the action curve sequences of the first to K-th cases; establishing a case action curve database based on the action curve sequences of the first to K-th cases; S2, sampling the switch machine action curve to be analyzed, and generating a corresponding action curve sequence to be analyzed; S3, calculating the dynamic time warping distance between the action curve sequence to be analyzed and the action curve sequence of the i-th case; i∈[1,K]; S4, calculating the segmented time distance between the action curve sequence to be analyzed and the action curve sequence of the i-th case; S5. Calculate the similarity between the action curve sequence to be analyzed and the action curve sequence of the i-th case based on the dynamic time warping distance and the segment duration distance; S6. Determine the fault category corresponding to the action curve sequence to be analyzed based on the similarity.

2. The switch machine fault diagnosis method based on case similarity matching according to claim 1 is characterized in that: Step S2 includes: S21, sampling the switch machine action curve to be analyzed at a set time interval to obtain the action curve sequence to be analyzed X = [x1, x2, ..., x n ], Wherein, n represents the length of the action curve sequence to be analyzed; x1~x n It indicates the results of the first to nth sampling of the switch machine action curve to be analyzed; S22, sampling the action curve sequence of the i-th case according to the time interval to obtain the action curve sequence of the i-th case Among them, m i is the length of the action curve sequence of the i-th case; Indicates the first to mth operations on the action curve sequence of case i i The result of subsampling.

3. The switch machine fault diagnosis method based on case similarity matching according to claim 2 is characterized in that: Step S3 includes: S31. Constructing distance matrix make Represents the distance matrix The element in row a and column b in x a is the result of the ath sampling of the switch machine action curve to be analyzed, is the action curve sequence Y for case i i The result of the b-th sampling; a∈[1,n], b∈[1,m]; S32, let the distance matrix middle arrive Path Q is a natural number, max(n,m)≤Q <n+m-1; is the path W i The qth step in the distance matrix Middle a q Row b q Elements of a column q∈[1,Q]; where S33, Order Action curve sequence X and action curve sequence Y of the i-th case i The dynamic time warping distance between 4. The switch machine fault diagnosis method based on case similarity matching according to claim 3 is characterized in that: The paths W i The qth step and the q+1th step in the equation satisfy: 0≤a q+1 -a q ≤1 and 0≤b q+1 -b q ≤1.

5. The switch machine fault diagnosis method based on case similarity matching according to claim 4 is characterized in that: Step S4 includes: S41, calculate the segment point position index of the action curve sequence X to be analyzed, and obtain the change point index sequence BD X =[t1, t2…, t U ]; U is a natural number, 1≤t u <t u+1 ≤n; S42, dividing the action curve sequence X to be analyzed into U subsequences in chronological order, wherein the u-th subsequence includes the elements in the action curve sequence X to be analyzed Where t1 = 1, t U+1 -1=n; generate time interval sequence in, express and The sampling time interval between them; u∈[1,U]; S43, for the action curve sequence Y of the i-th case i Calculate the segment point position index to obtain the change point index sequence BD i =[T1,…,T V ]; V is a natural number, 1≤T v <T v+1 ≤m; S44, the action curve sequence Y of the i-th case i It is divided into V subsequences in chronological order, where the vth subsequence includes the action curve sequence Y of the i-th case i Elements in Where T1 = 1, T V+1 -1=m i ; Generate time interval sequence in express and The sampling time interval between them, v∈[1,V]; S45. If U=V, calculate the time interval sequence T X and Euclidean distance Will As the action curve sequence X to be analyzed and the action curve sequence Y of the i-th case i The segment duration distance between 6. The switch machine fault diagnosis method based on case similarity matching according to claim 5, characterized in that: Step S41 includes: S411, mapping the action curve sequence X to be analyzed to the reproducing kernel Hilbert space by using the mapping function φ(·) Get the mapped sequence S412, by minimizing the objective function V(·), the segment point position index calculation is performed on the sequence XH to obtain the change point index sequence BD X =[t1, ..., t u ],satisfy: in, for The mean of represents the reproducing kernel Hilbert space 2-norm operation in .

7. The switch machine fault diagnosis method based on case similarity matching according to claim 6 is characterized in that: Step S43 includes: S431, transform the action curve sequence Y of the i-th case into i Mapping to the reproducing kernel Hilbert space Get the mapped sequence S432, by minimizing the objective function V(·) i Calculate the segment point position index and obtain the change point index sequence [T1, ..., T V ],satisfy: in, for The mean of .

8. The switch machine fault diagnosis method based on case similarity matching according to claim 6 is characterized in that: φ(x r )=xh r ∈XH,r∈[1,n],x r ∈X,xh r For x r In the reproducing kernel Hilbert space The mapping value in ; The reproducing kernel Hilbert space Inner product operation in is the square of the 2-norm operation, Among them, k(·,·) is the kernel function.

9. The switch machine fault diagnosis method based on case similarity matching according to claim 8, characterized in that: The kernel function is any one of a linear kernel function, a Gaussian kernel function, and a cosine kernel function.

10. The switch machine fault diagnosis method based on case similarity matching according to claim 8, characterized in that: Step S5 includes: S51, yes Normalize it and get i=1,2,…,K; S52. For the DC switch machine, the action curve sequence X to be analyzed and the action curve sequence Y of the i-th case are calculated. i The similarity between i ; S53. For AC switch machines: The switch machine action curve includes: A phase current curve, B phase current curve, C phase current curve, power curve; The action curve sequence to be analyzed includes: sampling the A-phase current curve, B-phase current curve, C-phase current curve, and power curve to be analyzed, to obtain the A-phase current curve sequence, B-phase current curve sequence, C-phase current curve sequence, and power curve sequence to be analyzed; The action curve sequence of the i-th case includes: sampling the A-phase current curve, the B-phase current curve, the C-phase current curve, and the power curve of the i-th case to obtain the A-phase current curve sequence, the B-phase current curve sequence, the C-phase current curve sequence, and the power curve sequence of the i-th case; include are respectively the dynamic time warping distances between the A-phase current curve sequence, the B-phase current curve sequence, the C-phase current curve sequence, and the power curve sequence to be analyzed and the A-phase current curve sequence, the B-phase current curve sequence, the C-phase current curve sequence, and the power curve sequence for the i-th case; make Respectively express The result after normalization: include are the segmented time length distances between the A-phase current curve sequence, B-phase current curve sequence, C-phase current curve sequence, and power curve sequence to be analyzed and the A-phase current curve sequence, B-phase current curve sequence, C-phase current curve sequence, and power curve sequence for the i-th case; make Respectively express The result after normalization: Order i,A 、s i,B 、s i,C 、s i,P Respectively represent the similarity between the A-phase current curve sequence, B-phase current curve sequence, C-phase current curve sequence, and power curve sequence to be analyzed and the A-phase current curve sequence, B-phase current curve sequence, C-phase current curve sequence, and power curve sequence of the i-th case; The action curve sequence X to be analyzed and the action curve sequence Y of the i-th case i The similarity between i , 11. The switch machine fault diagnosis method based on case similarity matching according to claim 10, characterized in that: Step S6 includes: S61. Calculate the action curve sequence X to be analyzed and the action curve sequence Y of the i-th case i The probability of matching is p i , S62. Calculate p max with i max , p max =max(P),P=[p 1 ,p 2 ,…,p K ]; i max =argmax i＝1,2,…,K (P),P=[p 1 ,p 2 ,…,p K ]; Let θ represent the preset probability threshold, If p max >θ, then determine whether the action curve sequence X to be analyzed is consistent with the i-th max Case-like action curve sequence The fault category is the same; If p max <θ, it is determined that the action curve sequence X to be analyzed does not belong to the fault category in the case action curve database; the action curve sequence X to be analyzed is manually analyzed, and if the analysis result is a fault, the switch machine action curve sequence to be analyzed is added to the case action curve database as a new type of fault case.

12. The switch machine fault diagnosis method based on case similarity matching according to claim 1, characterized in that: Also includes the steps: S7. Send the diagnostic results of the action curve sequence to be analyzed to the intelligent operation and maintenance platform: If the diagnosis is normal, the diagnosis result "normal" is returned; If the diagnosis is a fault, the fault type, cause analysis and treatment method will be sent, and an alarm will be displayed.

13. The switch machine fault diagnosis method based on case similarity matching according to claim 1, characterized in that: The switch machine action curve includes a switch machine action three-phase current curve, a switch machine action direct current curve and a switch machine action power curve.

Images