Railway vehicle traction system anomaly detection method, device and railway vehicle

Abstract

An abnormality detection method and device of a rail vehicle traction system and a rail vehicle, the method comprising: obtaining current data of a parameter related to the rail vehicle traction system; obtaining k sample data nearest to the current data from a pre-constructed training set, wherein the training set is constructed based on historical normal data of the parameter; constructing a class based on the k sample data, determining whether the current data belongs to the class, and when the current data does not belong to the class, outputting early warning information indicating an abnormality. The method, device and rail vehicle construct a training set based on historical normal data of a parameter related to the rail vehicle traction system, and determine whether the current operation of the rail vehicle traction system is abnormal according to whether a sample point similar to the current data can be found in the training set, thereby solving the problem of low accuracy of a constructed early warning model due to less historical fault data, and being very suitable for scenarios such as a rubber-tyred tramcar traction system with less historical fault data.

Description

Technical Field

[0001] This application relates to the field of rail transit technology, and more specifically to a method, device, and rail vehicle for detecting anomalies in a rail vehicle traction system. Background Technology

[0002] In the field of rail transit, the tolerance for faults in the traction system of rail vehicles is low. Therefore, the detection of anomalies (faults) in the traction system of rail vehicles is very important.

[0003] Currently, common methods for anomaly detection and early warning in rail vehicle traction systems include: setting standard values ​​and early warning thresholds for individual parameters in advance, and issuing an early warning when the difference between the operating parameter and the standard value exceeds the threshold; or performing deviation analysis on parameters of different traction systems in the same vehicle and assessing the traction system status based on the deviation analysis results; or analyzing historical fault data and pre-fault data of the traction system to construct a fault diagnosis and early warning model, and applying it to the traction system operating parameters for fault diagnosis.

[0004] The aforementioned method uses a warning model constructed from sample data corresponding to the fault states, minor fault states, and performance degradation states before a fault occurs in key components of the traction system. This model is not suitable for anomaly detection and early warning in rubber-tired trams. This is because there is limited effective fault data for the traction system of rubber-tired trams, and warning models built based on a small amount of fault sample data have low accuracy. Summary of the Invention

[0005] This application is proposed to address the aforementioned problems. According to one aspect of this application, an anomaly detection method for a rail vehicle traction system is provided. The method includes: acquiring current data of parameters related to the rail vehicle traction system; acquiring k nearest neighbor sample data from a pre-constructed training set, wherein the training set is constructed based on historical normal data of the parameters; constructing a class based on the k sample data; determining whether the current data belongs to the class; and outputting an anomaly warning message when the current data does not belong to the class.

[0006] In one embodiment of this application, the parameters related to the rail vehicle traction system include n parameters, the training set includes N sample data, each sample data includes the value of each of the n parameters at a normal operating time, the value of each of the n parameters at the N normal operating times is used as a feature component of the training set, and the current data includes the value of each of the n parameters at the current operating time.

[0007] In one embodiment of this application, the method further includes: after obtaining the current data, performing standardization processing on the current data to obtain standardized current data; and the k sample data are the k nearest neighbor sample data to the standardized current data obtained from the training set, wherein the training set is a training set that has undergone standardization processing.

[0008] In one embodiment of this application, the standardization process performed on the training set includes performing the following operations for each feature component: calculating the mean and standard deviation of the feature component; calculating the difference between the feature component and the mean; calculating the ratio between the difference and the standard deviation, and using the ratio as the standardized feature component, wherein the mean of the standardized feature component is 0 and the standard deviation is 1; wherein the standardization process performed on the current data is based on the mean and the standard deviation.

[0009] In one embodiment of this application, obtaining the k nearest neighbor samples from a pre-constructed training set includes: obtaining k sample data from the training set that satisfy the following conditions: constructing a k nearest neighbor sample set based on the k sample data, wherein the distance from each sample data in the k nearest neighbor sample set to the current data is used as a first distance, the distance from each sample data in the complement of the k nearest neighbor sample set to the current data is used as a second distance, and the first distance is less than or equal to the second distance.

[0010] In one embodiment of this application, the step of constructing a class based on the k sample data includes: constructing a minimum hypersphere that can contain the k sample data by using a planning problem solving method based on the k sample data; the step of determining whether the current data belongs to the class includes: determining whether the current data falls within the hypersphere.

[0011] In one embodiment of this application, determining whether the current data falls within the hypersphere includes: determining whether the distance between the current data and the center of the hypersphere is less than or equal to the radius of the hypersphere.

[0012] In one embodiment of this application, the construction of the minimum hypersphere capable of containing the k sample data is performed based on the penalty factors of the k sample data respectively, wherein the penalty factor of each sample data in the training set is pre-constructed.

[0013] In one embodiment of this application, the construction of the penalty factor for each sample data in the training set includes: calculating the median of each feature component in the training set, wherein the medians of all feature components constitute the median vector of the training set; for each sample data, calculating the distance between the sample data and the median vector, and calculating the penalty factor of the sample data based on the distance.

[0014] In one embodiment of this application, the method further includes: when the current data does not fall into the hypersphere, calculating and outputting the distance from each parameter in the current data to the center of the hypersphere.

[0015] In one embodiment of this application, the parameters include at least one of the following: radiator parameters, braking force parameters, reducer parameters, motor parameters, inverter parameters, transistor parameters, and speed parameters.

[0016] In one embodiment of this application, the radiator parameters include radiator temperature; the braking force parameters include at least one of the following: electric braking force capability value, electric braking actuation force value, and vehicle-required braking force; the reducer parameters include at least one of the following: reducer oil sensor level and reducer oil temperature sensor temperature; the motor parameters include at least one of the following: motor speed, motor actual actuation torque, motor oil temperature, and armature winding temperature; the inverter parameters include at least one of the following: inverter DC bus voltage, inverter output U-phase effective current, and inverter output W-phase effective current; the transistor parameters include the maximum temperature of the insulated gate bipolar transistor; and the speed parameters include vehicle speed.

[0017] In one embodiment of this application, the warning information includes: the name of the vehicle under warning, the carriage number under warning, and the warning time.

[0018] In one embodiment of this application, the warning information further includes at least one of the following: possible consequences, each possible cause, the probability of each possible cause, and guidance measures.

[0019] In one embodiment of this application, the presentation of the warning information includes at least one of the following: SMS reminder, warning details and icon prompts on a personal computer page, and automatic generation and dispatch of application software repair work orders.

[0020] In one embodiment of this application, the method can be implemented using an R script.

[0021] In one embodiment of this application, the method can be executed at preset frequencies.

[0022] In one embodiment of this application, the training set can be updated at a preset update frequency.

[0023] In one embodiment of this application, the rail vehicle includes a rubber-tired tram.

[0024] According to another aspect of this application, an anomaly detection device for a rail vehicle traction system is provided. The device includes a memory and a processor, wherein the memory stores a computer-executable program that is executed by the processor. When the computer-executable program is executed by the processor, it causes the processor to perform the above-described anomaly detection method for the rail vehicle traction system.

[0025] According to another aspect of this application, a rail vehicle is provided that includes the anomaly detection device for the rail vehicle traction system described above.

[0026] The present application discloses an anomaly detection method, device, and rail vehicle for rail vehicle traction systems. Based on historical normal data of parameters related to the rail vehicle traction system, a training set is constructed. The method determines whether there is an anomaly in the current operation of the rail vehicle traction system by finding sample points in the training set that are similar to the current data. This method can solve the problem of low accuracy of the early warning model due to the scarcity of historical fault data. It is very suitable for scenarios where there is limited historical fault data for traction systems, such as rubber-tired trams. Attached Figure Description

[0027] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0028] Figure 1 A schematic flowchart illustrating an anomaly detection method for a rail vehicle traction system according to an embodiment of this application is shown.

[0029] Figure 2 An example diagram showing the range of parameters used to construct the training set in the anomaly detection method for a rail vehicle traction system according to an embodiment of this application is provided.

[0030] Figure 3 An exemplary flowchart is shown when the anomaly detection method of the rail vehicle traction system according to an embodiment of this application is applied to a rubber-tired tram.

[0031] Figure 4 A schematic structural block diagram of an anomaly detection device for a rail vehicle traction system according to an embodiment of this application is shown. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this application more apparent, exemplary embodiments according to this application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are merely a part of the embodiments of this application, and not all of the embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein. Based on the embodiments of this application described herein, all other embodiments obtained by those skilled in the art without inventive effort should fall within the protection scope of this application.

[0033] First, refer to Figure 1 This application describes an anomaly detection method for a rail vehicle traction system according to an embodiment of the present application. Figure 1 A schematic flowchart of an anomaly detection method 100 for a rail vehicle traction system according to an embodiment of this application is shown. Figure 1 As shown, the anomaly detection method 100 for a rail vehicle traction system according to an embodiment of this application may include the following steps:

[0034] In step S110, current data of parameters related to the rail vehicle traction system are obtained.

[0035] In step S120, the k nearest neighbor samples to the current data are obtained from the pre-constructed training set, wherein the training set is constructed based on historical normal data of the parameters.

[0036] In step S130, a class is constructed based on k sample data, and it is determined whether the current data belongs to the class. When the current data does not belong to the class, an early warning message indicating an anomaly is output.

[0037] In the embodiments of this application, a training set is constructed based on historical normal data (i.e., normal operation data, non-fault data) of parameters related to the rail vehicle traction system. At the current detection time, the current data of parameters related to the rail vehicle traction system is obtained, and the k nearest neighbor sample data (k-nearest neighbor algorithm, k is a natural number) is obtained from the training set. If the current data belongs to the class constructed by the k nearest neighbor sample data, that is, a sample point similar to the current data can be found in the training set (also known as the historical normal database), the current data can be determined to be normal operation data, that is, the current operation of the traction system is normal. Conversely, if the current data does not belong to the class constructed by the k nearest neighbor sample data, that is, a sample point similar to the current data cannot be found in the training set, the current data can be determined to be abnormal operation data, that is, the current operation of the traction system is abnormal.

[0038] Therefore, the anomaly detection method for rail vehicle traction systems according to the embodiments of this application constructs a training set based on historical normal data of parameters related to the rail vehicle traction system. It determines whether there is an anomaly in the current operation of the rail vehicle traction system by finding sample points in the training set that are similar to the current data. This method can solve the problem of low accuracy of the early warning model due to the scarcity of historical fault data (because it does not need to construct an early warning model based on historical fault data, but rather constructs a training set based on a large amount of historical normal data). It is very suitable for scenarios where there is little historical fault data for traction systems, such as rubber-tired trams.

[0039] In embodiments of this application, the parameters related to the rail vehicle traction system used to construct the training set may include at least one of the following: radiator parameters, braking force parameters, reducer parameters, motor parameters, inverter parameters, transistor parameters, and speed parameters. That is, the data in the training set can be multi-dimensional (each dimension corresponding to a parameter), making the diagnosis of whether the current data is abnormal based on the training set more comprehensive and accurate.

[0040] For example, the above-mentioned radiator parameters may include radiator temperature (°C); the above-mentioned braking force parameters may include at least one of the following: electric braking force capacity (kN), electric braking force (kN), and vehicle required braking force (kN); the above-mentioned reducer parameters may include at least one of the following: reducer oil sensor level (mm), and reducer oil temperature sensor temperature (°C); the above-mentioned motor parameters may include at least one of the following: motor speed (rpm), motor actual torque (Nm), motor oil temperature (°C), and armature winding temperature (°C); the above-mentioned inverter parameters may include at least one of the following: inverter DC bus voltage (V), inverter output U-phase effective current (A), and inverter output W-phase effective current (A); the above-mentioned transistor parameters may include the maximum temperature of the insulated gate bipolar transistor (IGBT) (°C); and the above-mentioned speed parameters may include vehicle speed (km / h).

[0041] In one example, the above 15 parameters are used as 15-dimensional features of the training set to construct the training set. Figure 2 An example diagram showing the range of parameters used to construct the training set in the anomaly detection method for a rail vehicle traction system according to an embodiment of this application is shown. Figure 2 The normal range of the above 15 parameters is shown in the figure (these 15 parameters are also related to the traction system of rubber-tired trams). Only when all parameters are within the normal range at a certain moment will the record be included as a sample in the training set.

[0042] In general, in the embodiments of this application, the parameters related to the rail vehicle traction system may include n parameters (n is a natural number, in the above example, n = 15), the training set includes N sample data (N is a natural number, which can represent N times when the sample is acquired, and each time a sample data including n parameters is acquired), each sample data includes the value of each of the n parameters at a normal operating time, the value of each of the n parameters at the N normal operating times is used as a feature component of the training set, and the current data includes the value of each of the n parameters at the current operating time.

[0043] Therefore, in the embodiments of this application, when the training set is represented as X, the training set X can be represented in two ways, one of which is:

[0044]

[0045] Where x i The training samples, i = 1, 2, ..., N, are shown below:

[0046]

[0047] In this representation of the training set, the training set includes N sample data, each sample data includes the values ​​of n parameters at a normal operating time (n=15 in the example of this paper).

[0048] Another way to represent the training set X is:

[0049] X = (X1, X2, ..., X...) 15 (3)

[0050] Where X j =(x 1j ,x 2j ,...,x Nj ),j=1,2,...,15 represents the j-th dimension of the training set.

[0051] In this representation of the training set, the training set includes N sample data, each sample data includes the values ​​of n parameters at a normal operating time (n=15 in the example of this paper), and the value of each of the n parameters at the N normal operating times is used as a feature component (i.e., a dimension) of the training set.

[0052] In the embodiments of this application, when the training set includes multiple feature components (i.e., multiple parameters), each sample data in the training set can be standardized, and the standardized sample data can be saved. This can avoid the impact of different parameter magnitudes on subsequent processing. For example, in the example shown above, Figure 2The 15 feature parameters shown are of different magnitudes. If modeling is performed without processing, the results will only affect the parameters with higher magnitudes, while the parameters with lower magnitudes will be ignored. Therefore, the training set X = (X1, X2, ..., X...) can be processed. 15 Standardize the process.

[0053] In embodiments of this application, the standardization process performed on the training set may include: performing the following operations for each feature component in the training set: calculating the mean and standard deviation of the feature component; calculating the difference between the feature component and the mean; calculating the ratio between the difference and the standard deviation, and using this ratio as the standardized feature component, wherein the mean of the standardized feature component is 0 and the standard deviation is 1. This will be explained more clearly below using symbols and formulas. In this embodiment, for the training set X = (X1, X2, ..., X...), ... 15 Standardization is performed, that is, the standardization process is performed on each feature component X of the training set. j mean(X) j ) and standard deviation sd(X j ), and through equation (4), each feature component of the training set is transformed into a value with a mean of 0 and a standard deviation of 1.

[0054]

[0055] Where j = 1, 2, ..., 15. In the embodiments of this application, the mean(X) of each dimension can be... j ) and sd(X j The data is saved for standardization of the current operating parameters of the traction system. That is, in this embodiment, after obtaining the current data of the parameters related to the rail vehicle traction system in step S110, the current data can be standardized based on the mean and standard deviation of each feature component in the training set (using equation (4) above). Then, in step S120, the k nearest neighbor samples of the standardized current data are obtained from the training set. Finally, in step S130, it is determined whether the standardized current data belongs to the class constructed by the k samples, so as to determine whether the current operating condition of the traction system is abnormal. As mentioned above, standardizing the sample data in the training set and the current data can avoid the influence of different parameter magnitudes on subsequent processing.

[0056] In the embodiments of this application, after standardizing the sample data in the training set, a penalty factor can be constructed for each sample data for use in the class construction step S130 (described below). Specifically, the construction of the penalty factor for each sample data in the training set may include: calculating the median of each feature component in the training set, where the medians of all feature components constitute the median vector of the training set; for each sample data, calculating the distance between the sample data and the median vector, and calculating the penalty factor of the sample data based on the distance. The following description still uses examples.

[0057] For the training set X = (X1, X2, ..., X...), 15 The j-th dimension X j =(x 1j ,x 2j ,...,x Nj The order statistic is expressed as: x j (1)≤x j (2)≤...≤x j If (N), then the median of the j-th dimension in the training set is:

[0058]

[0059] Where j = 1, 2, ..., 15, then the median vector of the training set is μ = (μ1, μ2, ..., μ...). 15 ).

[0060] For each training sample x i =(x i1 ,x i2 ,...,x i15 ), calculate the distance d from the median vector. i Here, Euclidean distance is used:

[0061]

[0062] Where i = 1, 2, ..., N, then the penalty factor c i for:

[0063]

[0064] Therefore, the farther a sample point is from the center of the training set, the greater the probability that it is an outlier, and the smaller the corresponding penalty factor; conversely, the closer a sample point is to the center of the training set, the less likely it is to be an outlier, and the greater the corresponding penalty factor.

[0065] In an embodiment of this application, step S120, which involves obtaining the k nearest neighbor samples from a pre-constructed training set, may include: obtaining k samples from the training set that satisfy the following conditions: constructing a k-nearest neighbor sample set based on the k samples, wherein the distance from each sample in the k-nearest neighbor sample set to the current data is used as a first distance, and the distance from each sample in the complement of the k-nearest neighbor sample set to the current data is used as a second distance, and the first distance is less than or equal to the second distance. This will be illustrated below with examples.

[0066] In one example, k = 5, meaning the k nearest neighbors (k = 5) are obtained from the training set. For the current data (current operating parameters) x of the rail vehicle traction system... t From the training set Obtain the k-nearest neighbor sample set Make the k-nearest neighbor sample set From each sample point to x t The distance from each sample point in the complement of the training set X to x is less than or equal to X'. t The distance satisfies:

[0067] d(x t ,x′ i )≤d(x t ,y) (8)

[0068] in d(x,y) can be a Euclidean distance metric.

[0069] In the embodiments of this application, the step S130 of constructing a class based on k sample data may include: constructing a minimum hypersphere that can contain k sample data using a planning problem solving method based on k sample data; determining whether the current data belongs to a class may include: determining whether the current data falls within the hypersphere. The following description is based on the examples shown above.

[0070] For the obtained k nearest neighbor sample set Based on these k nearest neighbor samples, a planning problem-solving method is used to construct the minimum hypersphere that can contain these k nearest neighbor samples. That is, the hypersphere center a and hypersphere radius R are determined so that these k nearest neighbor samples are enclosed in the hypersphere as much as possible.

[0071] For the k nearest neighbor sample set If there exists a hypersphere Ω = (a, R) that can represent all sample points x′ i If we include i = 1, 2, ..., 5, then we have:

[0072]

[0073] Where ξi ≥0, i=1,2,...,k, a is the center of the sphere, R is the radius, and C is the penalty factor (including the penalty factor c of each of the k sample points). i As calculated in combination with equation (7) above, ξ is the loss due to misjudgment.

[0074] Equation (9) can be solved using the Lagrangian multiplier method:

[0075]

[0076] Where α i γ i It is a Lagrangian multiplier and satisfies α i ≥0, γ i Equation (10) can be solved by transforming it into its dual problem:

[0077]

[0078] By differentiating the parameters in equation (10) and setting them to 0, we can obtain the expressions for each parameter and their relationships. Equation (11) can be solved using the methods commonly used in classical planning problems to calculate the center of the sphere a and the radius of the hypersphere R.

[0079] From the derivation of equations (9)-(11), it can be seen that this problem can be solved by a planning problem, where the penalty factor is a constant. If the matrix minimized in the planning problem is a positive definite matrix, then the global minimum is unique and can be solved by a quadratic programming problem; if the matrix minimized in the planning problem is a non-positive definite matrix, then the quadratic programming problem becomes a linear programming problem. Therefore, the hypersphere determined by the above k samples can be obtained by using the planning problem solving method.

[0080] In embodiments of this application, determining whether the current data falls within the hypersphere may include: determining whether the distance between the current data and the center of the hypersphere is less than or equal to the radius of the hypersphere. This will be described below with reference to examples.

[0081] In one example, determine the current runtime parameter x. t Whether x falls within the supersphere, i.e., determining x t To determine whether the distance from the center a of the hypersphere is less than or equal to the hypersphere radius R, we need to check if the following equation is satisfied:

[0082] (x t -a) T (x t -a)≤R 2 (12)

[0083] When the current operating parameter of the rail vehicle traction system is xt If equation (12) is satisfied, then x is considered to be t If it falls into the normal category, the current operating status of the rail vehicle traction system is considered to be normal; otherwise, the current operating status of the rail vehicle traction system is considered to be abnormal.

[0084] In an embodiment of this application, method 100 may further include the following step: when the current data does not fall within the hypersphere, calculating and outputting the distance from each parameter in the current data to the center of the hypersphere. In this embodiment, when the current data does not fall within the hypersphere, it is determined that the current operating state of the rail vehicle traction system is abnormal. At this time, the distance from each parameter in the current data to the center of the hypersphere can be further calculated and output, thereby clarifying the degree to which each parameter in the current data deviates from the center of the hypersphere, which helps in fault location.

[0085] Specifically, the distance of each component in the current data from the center of the sphere can be obtained through equation (13):

[0086] dcomp j =(x tj -a j ) 2 ,j=1,2,...,15 (13)

[0087] It is generally believed that the component farthest from the center of the sphere has the highest probability of causing the current operating state to be abnormal, while the component closest to the center has the lowest probability of causing the current operating state to be abnormal. This can provide some help to operation and maintenance engineers in locating problems.

[0088] In an embodiment of this application, in step S130, when it is determined that the current data does not belong to a class constructed based on the k nearest neighbor sample data, an alarm message indicating an anomaly is output. For example, the alarm message may include: the name of the vehicle under alarm, the carriage number under alarm, and the alarm time. Based on this, maintenance engineers can understand the time and location of the traction system anomaly. Further, the alarm message may also include at least one of the following: possible consequences of the anomaly, possible causes of the anomaly and their probability, guidance measures, etc. This can further provide maintenance engineers with more detailed information. In addition, the presentation method of the alarm message may include at least one of the following: SMS reminder, alarm details and icon prompts on a personal computer page, automatic generation and dispatch of maintenance work orders from application software. Based on this, maintenance engineers can perform alarm verification and corresponding maintenance based on the alarm prompts and work order content.

[0089] The above exemplarily illustrates an anomaly detection method 100 for a rail vehicle traction system according to an embodiment of this application. In the embodiments of this application, the above method 100 can be implemented using an R script. For example, the process instructions such as connecting to a database, obtaining historical normal data and current operating data of the rail vehicle traction system, calculating penalty factors, obtaining nearest neighbor samples, constructing a hypersphere, determining the current operating status, and writing the results to the database can be deployed on a server with R language installed. The R script can be executed periodically by configuring a scheduled task, making it simple, easy to operate, and highly feasible. Furthermore, the above method 100 can be executed periodically according to a preset frequency. Real-time detection can be achieved through a computer scheduled task. The scheduled task can be set with different execution frequencies according to its own real-time requirements, such as 5 seconds, 10 seconds, 30 seconds, etc. In addition, the training set used in the above method 100 can be updated according to a preset update frequency. Since the training set should contain as much data as possible about all normal operating conditions of the traction system, its sample size is relatively large. However, given that the normal operating data of the traction system in a short period of time is likely already present in the training set, the update frequency of the training set and the penalty factor can be appropriately reduced, such as once a day or once a week.

[0090] Based on the above description, the anomaly detection method for rail vehicle traction system according to the embodiments of this application constructs a training set based on historical normal data of parameters related to the rail vehicle traction system. It determines whether there is an anomaly in the current operation of the rail vehicle traction system by finding sample points in the training set that are similar to the current data. This method can solve the problem of low accuracy of the early warning model due to the scarcity of historical fault data, and is very suitable for scenarios such as rubber-tired trams where there is little historical fault data for traction systems.

[0091] Furthermore, the training set used in the anomaly detection method of the rail vehicle traction system according to the embodiments of this application can comprehensively consider 15 dimensions of features, including traction system temperature, braking force, rotational speed, torque, voltage, current, and vehicle speed. It evaluates the current operating status based on a large amount of historical normal operation data, and the model is comprehensive and has a relatively high accuracy.

[0092] Furthermore, the anomaly detection method for the rail vehicle traction system according to the embodiments of this application can not only assess the current operating status of the rail vehicle traction system, but also analyze the probability of each parameter of the traction system causing the current anomaly, which provides great convenience for operation and maintenance engineers to locate problems and is conducive to improving maintenance efficiency.

[0093] Furthermore, the anomaly detection method for the rail vehicle traction system according to the embodiments of this application can be edited in R scripts, is easy to implement, and has high feasibility.

[0094] Furthermore, the abnormal detection method of the rail vehicle traction system according to the embodiments of this application uses a computer to replace manual monitoring. The computer automatically monitors the rail vehicle operating parameters according to a pre-configured frequency, estimates the operating status of the traction system, and issues an early warning message in real time when the operating status is diagnosed as abnormal, thus providing a certain guarantee for the safe operation of the rail vehicle and greatly saving manpower costs.

[0095] Finally, the anomaly detection method for the rail vehicle traction system according to the embodiments of this application can store the state assessment results of the rail vehicle traction system and the parameter data at the time of the anomaly, which is beneficial for tracing historical anomalies of the rail vehicle traction system, trend analysis and decision-making in any time period.

[0096] The following is combined Figure 3 This describes a scenario example where the anomaly detection method for a rail vehicle traction system according to embodiments of this application is applied to a rubber-tired tram. For example... Figure 3 As shown, the anomaly detection method 300 for the traction system of a rubber-tired tram may include the following steps:

[0097] In step S301, historical normal operation data of the rubber-tired tram traction system is obtained as a training set.

[0098] In step S302, the training set is standardized, the standardized parameters are saved, and the penalty factor is calculated.

[0099] In step S303, the operating parameters of the rubber-tired tram traction system are obtained and standardized according to the parameters saved in S302.

[0100] In step S304, the k nearest neighbor sample points (k=5) are obtained from the training set.

[0101] In step S305, a hypersphere is constructed from the k nearest neighbor samples, that is, the center and radius of the hypersphere are determined by combining a variable penalty factor with a planning problem-solving method.

[0102] In step S306, it is determined whether the current operating data of the rubber-tired tram traction system falls within the hypersphere. If it falls within the hypersphere, the diagnosis ends; otherwise, the diagnosis is abnormal.

[0103] In step S307, if the current operating state of the rubber-tired tram traction system is diagnosed as abnormal, the degree of deviation of each component of the current operating parameter from the center of the sphere is calculated to obtain the probability value of each component causing the current abnormal state.

[0104] In step S308, the current abnormal information, including the probability of abnormality of each component, is displayed as an early warning.

[0105] Here, Figure 3The anomaly detection method 300 for the traction system of a rubber-tired tram, as shown, is a combination of several embodiments of the anomaly detection method 100 for the traction system of a rail vehicle described above. Those skilled in the art can understand the process of the anomaly detection method 300 for the traction system of a rubber-tired tram by referring to the preceding text; for simplicity, it will not be repeated here. Overall, the anomaly detection method 300 for the traction system of a rubber-tired tram has a high accuracy rate in detecting anomalies in the traction system, can provide problem localization, is easy to implement, saves labor costs, and is beneficial for problem tracing and statistical analysis.

[0106] The following is combined Figure 4 Describes an anomaly detection device for a rail vehicle traction system provided according to another aspect of this application. Figure 4 A schematic structural block diagram of an anomaly detection device 400 for a rail vehicle traction system according to an embodiment of this application is shown. Figure 4 As shown, the anomaly detection device 400 for a rail vehicle traction system includes a memory 410 and a processor 420. The memory 410 stores a computer-executable program that is run by the processor 420. When the computer-executable program is run by the processor 420, it causes the processor 420 to execute the anomaly detection method 100 for the rail vehicle traction system described above. Those skilled in the art can understand the structure and specific operation of each module in the anomaly detection device 400 for the rail vehicle traction system according to the embodiments of this application based on the foregoing description; for the sake of brevity, further details are omitted here.

[0107] According to another aspect of this application, a rail vehicle is also provided, which includes the anomaly detection device 400 of the rail vehicle traction system according to the embodiments of this application described above.

[0108] Furthermore, this application also provides a storage medium storing a computer program thereon, which, when run by a processor, causes the processor to execute the anomaly detection method for a rail vehicle traction system according to the embodiments of this application described above. The storage medium may, for example, include a memory card of a smartphone, a storage component of a tablet computer, a hard disk of a personal computer, a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a portable compact disc read-only memory (CD-ROM), a USB memory, or any combination of the above storage media. The computer-readable storage medium may be any combination of one or more computer-readable storage media.

[0109] Based on the above description, the anomaly detection method, device, and rail vehicle of the rail vehicle traction system according to the embodiments of this application construct a training set based on historical normal data of parameters related to the rail vehicle traction system. It can determine whether there is an anomaly in the current operation of the rail vehicle traction system by finding sample points similar to the current data in the training set. This can solve the problem of low accuracy of the early warning model due to the lack of historical fault data. It is very suitable for scenarios with limited historical fault data of traction systems, such as rubber-tired trams.

[0110] Although exemplary embodiments have been described herein with reference to the accompanying drawings, it should be understood that the above exemplary embodiments are merely illustrative and are not intended to limit the scope of this application. Various changes and modifications can be made therein by those skilled in the art without departing from the scope and spirit of this application. All such changes and modifications are intended to be included within the scope of this application as claimed in the appended claims.

[0111] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0112] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed.

[0113] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of this application may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0114] Similarly, it should be understood that, in order to streamline this application and aid in understanding one or more of the various inventive aspects, features of this application may sometimes be grouped together in a single embodiment, figure, or description thereof in the description of exemplary embodiments of this application. However, this approach should not be construed as reflecting an intention that the claimed application requires more features than are expressly recited in each claim. Rather, as reflected in the corresponding claims, its inventive point lies in solving the corresponding technical problem with features fewer than all features of a single disclosed embodiment. Therefore, the claims following the detailed description are hereby expressly incorporated into that detailed description, wherein each claim itself is a separate embodiment of this application.

[0115] Those skilled in the art will understand that, apart from the mutual exclusion of features, all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or apparatus so disclosed can be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0116] Furthermore, those skilled in the art will understand that although some embodiments described herein include certain features included in other embodiments but not others, combinations of features from different embodiments are intended to be within the scope of this application and form different embodiments. For example, in the claims, any one of the claimed embodiments can be used in any combination.

[0117] The various component embodiments of this application can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some modules according to the embodiments of this application. This application can also be implemented as a program (e.g., a computer program and computer program product) for performing part or all of the methods described herein. Such a program implementing this application can be stored on a computer-readable medium, or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0118] It should be noted that the above embodiments are illustrative of this application and not restrictive, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. This application can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several anomaly detection devices for rail vehicle traction systems, several of these anomaly detection devices for rail vehicle traction systems may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

[0119] The above description is merely a specific embodiment or illustration of the embodiments of this application. The scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. The scope of protection of this application shall be determined by the scope of the claims.

Claims

1. A method for detecting anomalies in a rail vehicle traction system, characterized in that, The method includes: Obtain current data on parameters related to the rail vehicle's traction system; Obtain the k nearest neighbor samples of the current data from a pre-constructed training set, wherein the training set is constructed based on historical normal data of the parameters; The class is constructed based on the k sample data, including: based on the penalty factor pre-constructed from the k sample data, using a planning problem solving method, constructing a minimum hypersphere that can contain the k sample data; The construction of the penalty factor includes: calculating the median of each feature component in the training set, wherein the medians of all feature components constitute the median vector of the training set; for each sample data, calculating the distance between the sample data and the median vector, and calculating the penalty factor of the sample data based on the distance; Determining whether the current data belongs to the class includes: determining whether the current data falls into the hypersphere; when the current data does not belong to the class, outputting an alarm message indicating an anomaly.

2. The method according to claim 1, characterized in that, The parameters related to the rail vehicle traction system include n parameters, the training set includes N sample data, each sample data includes the value of each of the n parameters at a normal operating time, the value of each of the n parameters at the N normal operating times is used as a feature component of the training set, and the current data includes the value of each of the n parameters at the current operating time.

3. The method according to claim 2, characterized in that, The method further includes: After obtaining the current data, the current data is standardized to obtain standardized current data; Furthermore, the k sample data are obtained from the training set, which are the k nearest neighbors to the standardized current data, and the training set is a standardized training set.

4. The method according to claim 3, characterized in that, The standardization process performed on the training set includes performing the following operations for each of the feature components: Calculate the mean and standard deviation of the characteristic components; Calculate the difference between the feature component and the mean; Calculate the ratio between the difference and the standard deviation, and use the ratio as the standardized feature component. The standardized feature component has a mean of 0 and a standard deviation of 1. The standardization process performed on the current data is based on the mean and the standard deviation.

5. The method according to claim 1, characterized in that, The step of obtaining the k nearest neighbor samples from the pre-constructed training set to the current data includes: Obtain k sample data from the training set that satisfy the following conditions: construct a k nearest neighbor sample set based on the k sample data, wherein the distance from each sample data in the k nearest neighbor sample set to the current data is used as a first distance, the distance from each sample data in the complement set of the k nearest neighbor sample set to the current data is used as a second distance, and the first distance is less than or equal to the second distance.

6. The method according to claim 1, characterized in that, Determining whether the current data falls within the hypersphere includes: determining whether the distance between the current data and the center of the hypersphere is less than or equal to the radius of the hypersphere.

7. The method according to claim 1, characterized in that, The method further includes: When the current data does not fall within the hypersphere, calculate and output the distance from each parameter in the current data to the center of the hypersphere.

8. The method according to claim 1, characterized in that, The parameters include at least one of the following: radiator parameters, braking force parameters, reducer parameters, motor parameters, inverter parameters, transistor parameters, and speed parameters.

9. The method according to claim 8, characterized in that, The radiator parameters include radiator temperature; The braking force parameters include at least one of the following: electric braking force capacity value, electric braking force value, and braking force required by the vehicle. The reducer parameters include at least one of the following: reducer oil level sensor and reducer oil temperature sensor temperature; The motor parameters include at least one of the following: motor speed, actual motor torque, motor oil temperature, and armature winding temperature; The inverter parameters include at least one of the following: inverter DC bus voltage, inverter output U-phase effective current, and inverter output W-phase effective current; The transistor parameters include the highest temperature of the insulated-gate bipolar transistor; The speed parameter includes vehicle speed.

10. The method according to any one of claims 1-9, characterized in that, The warning information includes: the name of the vehicle being warned, the carriage number being warned, and the warning time.

11. The method according to claim 10, characterized in that, The warning information also includes at least one of the following: possible consequences, each possible cause, the probability of each possible cause, and guidance measures.

12. The method according to any one of claims 1-7, characterized in that, The presentation of the warning information includes at least one of the following: SMS reminder, warning details and icon prompts on a personal computer page, and automatic generation and dispatch of application software repair work orders.

13. The method according to any one of claims 1-7, characterized in that, The method can be implemented using R scripts.

14. The method according to any one of claims 1-7, characterized in that, The method can be executed at preset frequencies.

15. The method according to any one of claims 1-7, characterized in that, The training set can be updated at a preset update frequency.

16. The method according to any one of claims 1-7, characterized in that, The rail vehicles include rubber-tired trams.

17. An anomaly detection device for a rail vehicle traction system, characterized in that, The device includes a memory and a processor, wherein the memory stores a computer-executable program that is executed by the processor, the computer-executable program, when executed by the processor, causes the processor to perform the anomaly detection method for a rail vehicle traction system as described in any one of claims 1-16.

18. A rail vehicle, characterized in that, The rail vehicle includes the anomaly detection device for the rail vehicle traction system as described in claim 17.

Images