Railway vehicle fault early warning method and storage medium

Abstract

The invention discloses a rail vehicle fault early warning method and a storage medium, and the method comprises the steps: obtaining operation data of a rail vehicle, and determining first operation data corresponding to a first marshalling type and second operation data corresponding to a second marshalling type; updating the first operation data by using the first variable, updating the second operation data by using the second variable, and inputting the updated first operation data and the updated second operation data into an early warning model; the early warning model is pre-trained to identify a first variable and a second variable; and determining rail vehicle fault early warning information according to the output of the early warning model. According to the method, the variables are configured for the corresponding operation data according to the marshalling types, the accuracy and applicability of the data are improved, the early warning model can cope with rail vehicles of different marshalling types, and reusability and universality are high.

Description

Technical Field

[0001] The present invention relates to the technical field of urban rail vehicle braking systems, and in particular to a rail vehicle fault early warning method and a storage medium. Background Art

[0002] With the widespread application of Internet of Things technology in the field of rail transit, rail transit systems have realized the real-time collection and feedback of train operation data. These data provide a solid foundation for building an accurate vehicle fault warning model, enabling the rail transit system to identify potential operating risks in advance and effectively avoid the huge human and financial losses caused by train suspension. However, in the actual project implementation process, the existing warning model construction technology faces many challenges.

[0003] In order to cope with complex situations such as mixed train formations, existing technologies usually use a method of adjusting the internal logic of the early warning model one by one according to the train formation and project-specific variables. This method requires backend developers to have a deep understanding of the specific conditions of each project, including differences in train formations and differences in collected data, to ensure that the early warning model can accurately reflect the actual operating conditions. Once the collected variables change, developers need to re-evaluate and modify all early warning models involving the variable to ensure the accuracy and effectiveness of the model.

[0004] This method of building early warning models based on project-specific variables one by one has obvious disadvantages. First, a tight coupling relationship is formed between the early warning model and the collected variables, which greatly reduces the reusability and versatility of the model. Once the collected variables change, a large number of models need to be tediously modified, which not only increases the development cost but also prolongs the project delivery cycle. Secondly, different projects use independent variable naming rules, which makes it difficult for developers to reach a unified consensus, and the readability and maintainability of the code are also reduced accordingly. Summary of the invention

[0005] One of the purposes of the present invention is to provide a rail vehicle fault warning method to solve the technical problems in the prior art that mixed operation data with different marshaling types do not match the input data required by the warning model, resulting in poor reusability and versatility of the prediction model, and low prediction accuracy.

[0006] One of the objects of the present invention is to provide a computer storage medium.

[0007] To achieve one of the above-mentioned invention purposes, the present invention provides a method for early warning of faults in rail vehicles, including: obtaining operation data of the rail vehicle, determining first operation data corresponding to a first formation type and second operation data corresponding to a second formation type; updating the first operation data using a first variable, and updating the second operation data using a second variable, and inputting the updated first operation data and the updated second operation data into an early warning model; the early warning model is pre-trained to be able to identify the first variable and the second variable; determining the early warning information of the rail vehicle fault according to the output of the early warning model.

[0008] As a further improvement of an embodiment of the present invention, the updating the first operation data using the first variable includes: identifying a first data value with the same attribute and other second data values without the same attribute in the first operation data, and updating the first data value and the second data value respectively using the first variable; there is a mapping relationship between the variable and the data value in the updated first operation data.

[0009] As a further improvement of an embodiment of the present invention, the updating the first data value and the second data value respectively using the first variable includes: performing an update operation on each data value in the first operation data respectively using the first variable to obtain first intermediate operation data; there is a mapping relationship between the variable and the data value in the first intermediate operation data; identifying and abstracting the variables with the same attribute in the first intermediate operation data to determine second intermediate operation data; combining the first intermediate operation data and the second intermediate operation data to obtain the updated first operation data.

[0010] As a further improvement of an embodiment of the present invention, the identifying and abstracting the variables with the same attribute in the first intermediate operation data to determine the second intermediate operation data includes: identifying the variables with the same prefix in the first intermediate operation data, summarizing them into a unified abstract variable, and determining the corresponding second intermediate operation data based on the abstract variable.

[0011] As a further improvement of an embodiment of the present invention, the updating the first data value and the second data value respectively using the first variable includes: obtaining a plurality of first data values with the same attribute in the first operation data; performing an abstract update process of the variable on the plurality of first data values using the first variable to obtain second intermediate operation data; there is a mapping relationship between the variable and the data value in the second intermediate operation data; performing an update operation on other second data values without the same attribute respectively using the first variable to obtain first intermediate operation data; combining the first intermediate operation data and the second intermediate operation data to obtain the updated first operation data.

[0012] As a further improvement of an embodiment of the present invention, before inputting the updated first operation data and the updated second operation data into the early warning model, the method further includes: traversing a plurality of data values corresponding to each variable in the first operation data, calculating the Z value corresponding to each data value, generating a first array according to the Z value, and the first array is used to mark the abnormal conditions of each data value; counting the number of elements continuously marked as abnormal values in the first array, and determining a corresponding second array according to the number of elements; wherein, the array lengths of the first array and the second array are equal; updating the first operation data according to the differences between the first array and the second array.

[0013] As a further improvement of an embodiment of the present invention, the generating a first array according to the Z value includes: determining whether the Z value corresponding to each data value is greater than a preset threshold; if so, marking the data value at the corresponding position as an abnormal value; if not, marking the data value at the corresponding position as a normal value; arranging the marking results of the data values according to the position distribution of the data values in the first operation data to obtain the first array.

[0014] As a further improvement of an embodiment of the present invention, the counting the number of elements continuously marked as abnormal values in the first array and determining a corresponding second array according to the number of elements includes: initializing an abnormal count flag p, and the abnormal count flag p is used to represent the number of elements continuously abnormally marked in the first array; traversing each element in the first array, and determining whether the value of the current element is an abnormal value; if so, incrementing the abnormal count flag p by 1, and using the result of the abnormal count flag p as the element value at the corresponding position in the second array; if not, setting the abnormal count flag p to 0, and using the result of the abnormal count flag p as the element value at the corresponding position in the second array.

[0015] As a further improvement of an embodiment of the present invention, the updating the first operation data according to the differences between the first array and the second array includes: shifting each element in the second array one bit to the left and setting the last element to null to generate a third array, and the number of elements in the third array is equal to that of the second array; calculating the difference between the second array and the third array to obtain a fourth array, and the number of elements in the fourth array is equal to that of the third array; determining that the data value at the corresponding position in the first operation data is an abnormal value according to the position where the element value in the fourth array is an abnormal value, and updating to determine the abnormal data value.

[0016] As a further improvement of an embodiment of the present invention, determining that the data value at the corresponding position in the first operation data is an outlier and updating to determine the outlier data value includes: determining whether the outlier data value is at the first position or the last position in the first operation data; if not, obtaining a third data value and a fourth data value adjacent to the outlier data value, and determining an estimated value of the outlier data value by using a linear filling method according to the third data value and the fourth data value, and updating the estimated value to the corresponding outlier data value.

[0017] As a further improvement of an embodiment of the present invention, after determining whether the outlier data value is at the first position or the last position in the first operation data, the method further includes: if so, when the outlier data value is at the first position in the first operation data, filling the outlier data value with the data value at the next position; when the outlier data value is at the last position in the first operation data, filling the outlier data value with the data value at the position before the last position.

[0018] To achieve one of the above-mentioned invention purposes, the present invention also provides a computer storage medium, in which a computer program is stored, and when the computer program runs, it causes the device where the computer storage medium is located to execute the steps of any one of the above-mentioned rail vehicle fault warning methods.

[0019] Compared with the prior art, the embodiment of the present invention has at least one of the following beneficial effects:

[0020] The present invention adopts a rail vehicle fault warning method, which determines the first operation data corresponding to the first formation type and the second operation data corresponding to the second formation type by analyzing the operation data of the rail vehicle, and updates these operation data by using corresponding variables to ensure that the data input into the warning model meets the model requirements. In this way, the warning model can more accurately warn of the faults of the rail vehicle according to the updated operation data, avoiding warning errors caused by chaotic and mismatched data, not only improving the accuracy and applicability of the data, but also enabling the warning model to handle rail vehicles of different formation types without separately constructing a warning model for each formation type, with strong reusability and generality. Description of the Drawings

[0021] Figure 1 is a schematic diagram of the steps of the rail vehicle fault warning method in an embodiment of the present invention.

[0022] Figure 2 is a schematic diagram of the process before step S1 in an embodiment of the present invention.

[0023] Figure 3(a) is a schematic diagram of the specific steps of step S2 in an embodiment of the present invention.

[0024] Figure 3(b) is a schematic diagram of the specific steps of step S2 in another embodiment of the present invention.

[0025] Figure 4 is a schematic diagram of the steps before step S3 in an embodiment of the present invention.

[0026] Figure 5 is a schematic diagram of the specific steps of step P32 in a specific embodiment of an embodiment of the present invention.

[0027] Figure 6 is a schematic diagram of the specific steps of step P32 in a specific embodiment of an embodiment of the present invention.

[0028] Figure 7(a) is a schematic diagram of the process of processing abnormal data values in an embodiment of the present invention.

[0029] Figure 7(b) is a schematic diagram of the calculation process of processing abnormal data values in an embodiment of the present invention.

[0030] Figure 8(a) is a schematic diagram of the process of a method for fault warning of rail vehicles in an embodiment of the present invention.

[0031] Figure 8(b) is a schematic diagram of the principle of a method for fault warning of rail vehicles in an embodiment of the present invention. Detailed implementation manners

[0032] The present invention will be described in detail below in conjunction with the specific implementation manners shown in the accompanying drawings. However, these implementation manners do not limit the present invention, and any structural, method, or functional transformation made by those of ordinary skill in the art based on these implementation manners is included in the protection scope of the present invention.

[0033] It should be noted that the term "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article, or device including a series of elements not only includes those elements but also includes other elements not explicitly listed, or further includes elements inherent to such process, method, article, or device. In addition, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0034] As Figure 1 shown, an embodiment of the present invention provides a method for fault warning of rail vehicles.

[0035] Step S1, obtain the operation data of the rail vehicle, and determine the first operation data corresponding to the first formation type and the second operation data corresponding to the second formation type;

[0036] Step S2, update the first operation data using the first variable, and update the second operation data using the second variable, and input the updated first operation data and the updated second operation data into the warning model; the warning model is pre-trained to recognize the first variable and the second variable;

[0037] Step S3, determine the fault warning information of the rail vehicle according to the output of the warning model.

[0038] In this way, by obtaining the operation data of the rail vehicle, the first operation data corresponding to the first formation type and the second operation data corresponding to the second formation type are determined, and these data are updated using the corresponding variables to ensure that the data input into the warning model is accurate and meets the model requirements. In this way, the warning model can more accurately warn of the faults of the rail vehicle based on the updated operation data, avoiding warning errors caused by chaotic and mismatched data. This not only improves the accuracy and applicability of the data, but also enables the warning model to handle rail vehicles of different formation types without the need to separately construct a warning model for each formation type, with strong reusability and generality.

[0039] In step S1, the operation data refers to various data generated during the actual operation of the rail vehicle, including but not limited to speed, position, acceleration, braking state, braking components, traction force, etc. The operation data can reflect the operation state and performance of the vehicle and is the basis for subsequent fault warning analysis.

[0040] The existence status of the first operation data and the second operation data can reflect whether the current operation data of the rail vehicle belongs to single-formation operation data or mixed-formation operation data. It can be directly determined by the mixed-formation flag bit Flag in the operation data. For example, True means mixed formation, and Flase means non-mixed formation; it can also be determined by the train number in the operation data, and the corresponding formation is determined, and based on the preset mapping relationship, it is determined whether the formation is a single formation or a mixed formation.

[0041] In one embodiment, when the first operation data is empty and the second operation data is not empty, it indicates that the currently operating rail vehicle corresponds to the first formation type and is single-formation operation data. Similarly, when the first operation data is not empty and the second operation data is empty, it indicates that the currently operating rail vehicle corresponds to the second formation type and is also single-formation operation data.

[0042] In one embodiment, when both the first operation data and the second operation data are not empty, it indicates that the currently operating rail vehicle is a vehicle mixed with two formation types. Of course, it can also include operation data of more than two different formation types, which is not specifically limited here.

[0043] Such as Figure 2As shown, in a specific embodiment, a data file corresponding to the Multifunction Vehicle Bus and / or a data file collected by a data logger is obtained; the train operation data in the data file is parsed and extracted, and the train operation data is preprocessed, such as removing duplicate data, processing missing values and outliers, etc., to obtain the processed operation data.

[0044] The data file of the Multifunction Vehicle Bus (MVB) is used to store and transmit data based on the MVB protocol. MVB refers to a communication protocol and data transmission method, which is a bus protocol for serial data communication between interconnected devices with interoperability and interchangeability requirements.

[0045] A data logger, also known as a Datalogger, is a technology used to collect, store and share real-time data. In the field of data storage, a Datalogger is used to record the real-time data of various sensors, instruments or systems, allowing specific variable values to be written into a data log file in CSV format, and operations such as management, viewing, overwriting or creating new ones can be performed.

[0046] In step S1, the formation type refers to the composition method and car body configuration of rail vehicles. Different formation types may have different numbers of vehicles, car body types (such as motor car bodies, trailer car bodies), power configurations, etc. For example, one formation type may consist of 4 motor car bodies and 4 trailer car bodies, while another formation type may consist of 6 motor car bodies and 2 trailer car bodies.

[0047] In a specific embodiment, the operation data is parsed to determine the car body number, and based on a preset mapping relationship, the formation type corresponding to the car body number is determined. For example, if the operation data obtained is for car body No. 3 and car body No. 6, based on the preset mapping relationship, the formation type corresponding to car body No. 3 is determined to be 4 - formation, and the formation type corresponding to car body No. 6 is determined to be 8 - formation.

[0048] In one embodiment, the use of the first variable to update the first operation data in step S2 includes: step S2’, identifying the first data values with the same attributes and other second data values without the same attributes in the first operation data, and using the first variable to update the first data values and the second data values respectively; there is a mapping relationship between the variables and the data values in the updated first operation data.

[0049] In step S2’, the same - attribute values are data values with common characteristics or identifiers, and these characteristics or identifiers enable these data values to be grouped or classified into the same category.

[0050] For example, assume that the first operation data corresponding to a 4 - formation (the corresponding train number is 3, and it is determined to belong to a 4 - formation train based on a preset mapping relationship) is "3_15_16, 3_17_18, 3_17_20, 3_17_21, 3_17_22, 3_17_23", corresponding to the first running speed of the vehicle, the first timestamp, and the axle speeds of the four axles of the vehicle respectively. Among this first operation data, the data values with the same attribute are the axle speeds of the last four axles, and the remaining data values (i.e., the first running speed and the first timestamp) do not have the same attribute.

[0051] Use the first variables: "TB_Ref_Speed", "Time_Stamp", "TB_Alex_Spe ed1", "TB_Alex_Speed2", "TB_Alex_Speed3", and "TB_Alex_Speed4" to update it, and obtain the updated first operation data. The updated first operation data is:

[0052] signals_TB = {"TB_Ref_Speed": 3_15_16, "Time_Stamp": 3_17_18, "TB_Alex_Speed_1": 3_17_20, "TB_Alex_Speed_2": 3_17_21, "TB_Alex_Speed_3": 3_17_22, "TB_Alex_Speed_4": 3_17_23}.

[0053] When some variables in the updated first operation data have the same or similar attributes, use the abstract variables to update the first operation data again, and obtain the first operation data after the second update.

[0054] Similarly, in step S2", identify the fifth data values with the same attribute and the sixth data values with different attributes in the second operation data, and use the second variables to update the fifth data values and the sixth data values respectively; there is a mapping relationship between the variables and the data values in the updated second operation data.

[0055] For example, assume that the second operation data corresponding to an 8 - formation (the corresponding train number is 6, and it is determined to belong to an 8 - formation train based on a preset mapping relationship) is "6_15_16, 6_17_18, 6_17_20, 6_17_21, 6_17_22, 6_17_23", corresponding to the second running speed of the vehicle, the second timestamp, and the axle speeds of the four axles of the vehicle respectively. Among this second operation data, the data values with the same attribute are the axle speeds of the last four axles, and the remaining data values (i.e., the second running speed and the second timestamp) do not have the same attribute.

[0056] Update it using the second variables: "MB_Ref_Speed", "Time_Stamp1", "TB_Alex_Speed1", "TB_Alex_Speed2", "TB_Alex_Speed3", "TB_Alex_Speed4", "TB_Alex_Speed5", "TB_Alex_Speed6", "TB_Alex_Speed7", and "TB_Alex_Speed8" to obtain the updated second operation data. The updated second operation data is as follows:

[0057] signals_MB = {"MB_Ref_Speed": 6_15_16, "Time_Stamp1: 6_17_18", "TB_Alex_Speed_1": 6_17_20, "TB_Alex_Speed_2": 6_17_21, "TB_Alex_Speed_3": 6_17_22, "TB_Alex_Speed_4": 6_17_23, "TB_Alex_Speed_5": 6_17_24, "TB_Alex_Speed_6": 6_17_25, "TB_Alex_Speed_7": 6_17_26, "TB_Alex_Speed_8": 6_17_27}.

[0058] When there are some variables with the same or similar attributes in the updated second operation data, use the abstract unified variables to update the second operation data again to obtain the second operation data updated again.

[0059] It should be noted that in the data initialization stage, different values are assigned to the same variable name (such as TB_AlexSpeedX) according to the formation type of the rail vehicle. This means that although the variable names are the same, the specific numerical values or meanings represented by this variable are different under different formation types. This initialization process ensures that the same variable name can represent different information in different contexts. For example, in a 4-car train, TB_AlexSpeedX represents the axle speed of four axles, while in an 8-car train, TB_AlexSpeedX represents the axle speed of eight axles.

[0060] Once the variables are correctly assigned during initialization, during the operation phase of the warning model, the variable values represented by the same variable names (such as TB_AlexSpeedX) initialized under different grouping types are input into the warning model. Inside the model, since the correctness of the variables has been ensured through the initialization process, the model processes and issues warnings based on the current values of these variables, without the need to distinguish their different sources or names during initialization. Therefore, in the warning model, these variables (although they may represent different meanings during initialization) are uniformly processed with the same variable name.

[0061] Of course, the first variable and the second variable can also use different variable names. Specifically, during the data initialization phase, these two variables are assigned different values to distinguish different grouping types; during the operation phase of the warning model, these two variables are passed to the model as independent inputs, and the model processes and issues warnings based on their current values.

[0062] As shown in Figure 3(a), in a specific embodiment, the step of using the first variable to update the first data value and the second data value respectively in step S2' includes the following steps.

[0063] Step S211, use the first variable to update each data value in the first operation data respectively to obtain the first intermediate operation data; there is a mapping relationship between the variables and the data values in the first intermediate operation data;

[0064] Step S212, identify and abstract the variables with the same attributes in the first intermediate operation data to determine the second intermediate operation data;

[0065] Step S213, combine the first intermediate operation data and the second intermediate operation data to obtain the updated first operation data.

[0066] In this way, by performing a one-by-one update operation on each data value in the first operation data, configuring a corresponding variable for each data value, so that each data value obtains an appropriate variable identifier, ensuring the comprehensive update of the data, and then abstracting and classifying the attributes of the comprehensively updated data, making the data processing flow clearer.

[0067] In this embodiment, a data processing method of "updating first and then abstracting" is adopted, which pays more attention to the comprehensive update of the data and the retention of all details during the data update process. Specifically, through the method of updating first and then abstracting, it is ensured that each data value in the operation data is properly processed, and then classified and simplified.

[0068] In a specific embodiment, in step S212, the second intermediate operation data can also be determined through the following steps. That is: identify the variables with the same prefix in the first intermediate operation data, classify them into a unified abstract variable, and determine the corresponding second intermediate operation data based on the abstract variable.

[0069] For example, the variables are respectively: "TB_Alex_Speed1", "TB_Alex_Speed2", "TB_Alex_Speed3" and "TB_Alex_Speed4", which have the same prefix TB_Alex_Speed, and they are classified into the same abstract train variable TB_AlexSpeedX, where X = [1, 2, 3, 4].

[0070] As shown in Figure 3(b), in a specific embodiment, the step of using the first variable to update the first data value and the second data value respectively in step S2' includes the following steps.

[0071] Step S221, obtain a number of first data values with the same attribute in the first operation data;

[0072] Step S222, perform an abstract update process of the variables on the number of first data values using the first variable to obtain the second intermediate operation data; there is a mapping relationship between the variables and the data values in the second intermediate operation data;

[0073] Step S223, perform an update operation on other second data values that do not have the same attribute using the first variable respectively to obtain the first intermediate operation data;

[0074] Step S224, combine the first intermediate operation data and the second intermediate operation data to obtain the updated first operation data.

[0075] In this way, by identifying the data values with the same attribute in the first operation data, and performing an abstract process on these data values using the first variable, while performing an update process on other data values. This process reduces the intermediate steps of data processing, enabling the abstract process to be carried out while identifying the data attributes, and improving the processing efficiency.

[0076] In this embodiment, a data processing method of "updating and abstracting simultaneously" is adopted to reduce the number of iterations of data processing and improve the processing speed.

[0077] In a specific embodiment, in step S222, the second intermediate operation data can also be determined through the following steps. That is: identify a number of first data values with the same attribute in the first operation data, classify them into a unified abstract variable according to the meaning of the first data value, and determine the corresponding second intermediate operation data based on the abstract variable.

[0078] In steps S211 and S222, the first variable includes specific variables for data values with different or dissimilar attributes, such as TB_Ref_Speed, and combines multiple data values with the same or similar attributes into a single abstract variable, such as TB_Alex_SpeedX. The first variable can include a single variable or multiple variables. Similarly, the second variable can include a single variable or multiple variables.

[0079] In a specific embodiment, the mapping relationship is a dictionary relationship of key-value pairs.

[0080] In steps S212 and S22, further abstract processing of the first intermediate operation data may specifically include combining multiple data values with the same or similar attributes into a single variable, or creating a common category label for these data values, such as TB_Alex_SpeedX.

[0081] For example, still taking the first operation data and the second operation data shown above as an example, the first data value may specifically refer to the axle speeds of the four axles of a vehicle, such as "3_17_20, 3_17_21, 3_17_22, 3_17_23"; the second data value may specifically refer to the running speed and time stamp of the vehicle, such as "3_15_16, 3_17_18". In the data processing method of "updating first and then abstracting" in this example, the first intermediate operation data is:

[0082] signals_TB = {"TB_Ref_Speed": 3_15_16, "Time_Stamp": 3_17_18, "TB_Alex_Speed_1": 3_17_20, "TB_Alex_Speed_2": 3_17_21, "TB_Alex_Speed_3": 3_17_22, "TB_Alex_Speed_4": 3_17_23}.

[0083] The second intermediate operation data is: model_signals_collection = {"TB_AlexSpeedX": [f f"TB_Alex_Speed_{x}" for x in X]} X = [1, 2, 3, 4]}.

[0084] Similarly, the same processing is also performed on the second operation data, and the specific processing process can refer to the relevant embodiments and specific embodiments of the first operation data, which will not be elaborated here.

[0085] It should be noted that signals_TB (the first intermediate operation data) can be a dictionary containing several key-value pairs. It can contain all the necessary information to distinguish variables in different groups and execute different calculation logics. However, directly using signals_TB to perform these operations may make the code complex and difficult to maintain. Especially when the number of variables is large and the calculation logic needs to be frequently changed according to the existence or non-existence of variables.

[0086] Based on this, model_signals_collection (the second intermediate operation data) can be extracted from signals_TB. The purposes of this operation include the following two aspects.

[0087] On the one hand, it is convenient to simply and quickly distinguish variables of the same type of components in different groups. Specifically, for the same type of component, different train formations may be in different positions or have different configurations. For example, for a 4-car train formation, the pressure valve may be located in the 2nd and 3rd cars; while for an 8-car train formation, the pressure valve may be located in the 1st, 2nd, and 3rd cars. Through the extraction operation, it is possible to clearly know the position of a specific component (such as the pressure valve) in each formation, which helps the warning model select the correct component position according to the actual formation situation during operation, so as to perform accurate simulation and calculation.

[0088] On the other hand, it is convenient for the subsequent warning model to make correct decisions and execute corresponding logics according to the existence or non-existence of the corresponding variables during operation. Specifically, there may be multiple calculation methods based on different data or conditions in the warning model. For example, one calculation method can rely on the data of the temperature sensor, while another calculation method may rely on the data of the pressure sensor. During the operation of the warning model, by checking whether specific variables (such as the data of the temperature sensor or the pressure sensor) exist in model_signals_collection, the warning model can select the corresponding calculation method or logic path to execute.

[0089] In step S2, the warning model can identify including: when the warning model identifies the first variable, determining that the first operation data corresponding to the first variable is associated with the first formation type; and when the warning model identifies the second variable, determining that the second operation data corresponding to the second variable is associated with the second formation type. In the present invention, the warning model may be a single model or a combination of multiple warning models, and no specific limitation is made in this regard.

[0090] It should be noted that when collecting the input data of the warning model, due to reasons such as measurement errors, data entry errors, special events or conditions, the data values of some collection points in the input data may deviate significantly from other values. These significantly deviated data points may affect the prediction results of the warning model, resulting in false alarms or the inability to send warning messages.

[0091] When processing the abnormal data values in the first operation data, the following two methods can be adopted: one is to synchronously process the abnormal data values while updating the first operation data using the first variable; the other is to process the signal abnormal values after the update of the first operation data using the first variable is completed. It can be selected according to the actual situation, and no specific restrictions are made on this.

[0092] As Figure 4 shown, before step S3, the method may further include the following steps.

[0093] Step P31, traverse the several data values corresponding to each variable in the first operation data, calculate the Z value corresponding to each data value, and generate a first array according to the Z value. The first array is used to mark the abnormal situation of each data value;

[0094] Step P32, count the number of elements continuously marked as abnormal values in the first array, and determine the corresponding second array according to the number of elements; among them, the array lengths of the first array and the second array are equal;

[0095] Step P33, update the first operation data according to the difference between the first array and the second array.

[0096] In this way, through the abnormal update operation of the abnormal data values in the first operation data, the updated first operation data conforms to the normal operation conditions of the rail vehicle, avoiding false alarms or missed reports of warning information.

[0097] In step P31, the first operation data can be the initial operation data, that is, the initial first operation data that has not been updated using the first variable; it can also be the updated first operation data.

[0098] In one embodiment, an abnormal value processing operation is performed on the updated first operation data. Specifically, the abnormal data values in the first operation data are identified and updated to normal values.

[0099] In a specific embodiment, the several data values corresponding to a single variable can form an array, and in this way, the data values corresponding to multiple variables can form a two-dimensional matrix.

[0100] The Z-value, also known as the Z-score or standard score, is a standardized value used to quantify the distance between a data point and the mean. The Z-value is the ratio of the difference between the sample mean and the population mean divided by the standard deviation. The formula for calculating the Z-value is Z = (x - μ) / σ, where x is a data point, μ is the population mean, and σ is the population standard deviation.

[0101] In the present invention, x can represent a data value under a certain variable, μ is the mean of all data values that meet the conditions under this variable, and σ is the standard deviation of all data values that meet the conditions under this variable.

[0102] In a specific embodiment, generating the first array according to the Z-value in step P31 includes the following steps.

[0103] Step P311, determine whether the Z-value corresponding to each data value is greater than a preset threshold;

[0104] If so, jump to step P312A and mark the data value at the corresponding position as an outlier;

[0105] If not, jump to step P312B and mark the data value at the corresponding position as a normal value;

[0106] Step P313, arrange the marking results of the data values according to the position distribution of the data values in the first operation data to obtain the first array.

[0107] In this way, by generating a first array that can accurately reflect the abnormal state of the data values, the abnormal and normal values in the train signal are effectively identified and distinguished, providing a reliable basis for subsequent data processing and analysis.

[0108] In a specific embodiment, the preset threshold is 2, the outlier is 1, and the normal value is 0. For example, for the variable TB_Ref_Speed, multiple data values are collected in the first operation data, that is, TB_Ref_Speed = [v1, v2, v3, v4, v5]. Calculate the Z-value corresponding to each data value and compare it with the preset threshold. When the Z-value is greater than the preset threshold, the data value at this position is marked as 1; when the Z-value is less than or equal to the preset threshold, the data value at this position is marked as 0, and the first numerical value m = [1, 1, 0, 1, 0] is obtained.

[0109] As Figure 5 shown, in a specific embodiment, step P32 may specifically include the following steps.

[0110] Step P321, initialize the outlier count marker p, and the outlier count marker p is used to represent the number of consecutive outlier-marked elements in the first array;

[0111] Step P322, traverse each element in the first array, and determine whether the value of the current element is an outlier;

[0112] If so, jump to step P323A, increment the outlier count flag p by 1, and use the result of the outlier count flag p as the element value at the corresponding position in the second array;

[0113] If not, jump to step P323B, set the outlier count flag p to 0, and use the result of the outlier count flag p as the element value at the corresponding position in the second array.

[0114] In this way, the number of consecutive outliers at each position is intuitively shown through the second array, providing important reference information for subsequent data analysis and processing.

[0115] As Figure 6 shown, in a specific embodiment, step P33 may specifically include the following steps.

[0116] Step P331, shift each element in the second array one position to the left, and set the last element to empty, generating a third array, where the number of elements in the third array is equal to that in the second array;

[0117] Step P332, calculate the difference between the second array and the third array to obtain a fourth array, where the number of elements in the fourth array is equal to that in the third array;

[0118] Step P333, according to the positions of the elements with outlier values in the fourth array, determine that the data values at the corresponding positions in the first operation data are outliers, and update and determine the abnormal data values.

[0119] In this way, the continuous change trend of abnormal data values is effectively identified, and the abnormal data values in the operation data are accurately updated accordingly, improving the accuracy and efficiency of data processing.

[0120] In a specific embodiment, in step P333, determining that the data values at the corresponding positions in the first operation data are outliers and updating and determining the abnormal data values may specifically include the following steps.

[0121] Step P3331, determine whether the abnormal data value is at the first position or the last position in the first operation data;

[0122] If not, jump to step P3332, obtain the third data value and the fourth data value adjacent to the abnormal data value; according to the third data value and the fourth data value, use the linear filling method to determine the estimated value of the abnormal data value, and update the estimated value to the corresponding abnormal data value;

[0123] If not, jump to step P3333. When the abnormal data value is at the first position in the first operation data, fill the abnormal data value with the data value at the next position.

[0124] Step P3334, when the abnormal data value is at the last position in the first operation data, fill the abnormal data value with the data value at the position before the last position.

[0125] In step P3332, linear filling is an interpolation method based on a linear relationship. It assumes that there is a certain linear relationship between data points and uses this relationship to estimate the value of the position data point. Specifically, the linear filling method will consider the two known data values adjacent to the abnormal data value (i.e., the third data value and the fourth data value), as well as their respective position information. According to the calculation formula of linear interpolation, the value at the position where the abnormal data value is located can be estimated.

[0126] To facilitate the understanding of the above process of abnormal data value processing, for example, as shown in Figures 7(a) and 7(b), assume that according to the data values of the vehicle running speed collected in the first operation data, TB_Ref_Speed = [v1, v2, v3, v4, v5, v6, v7, v8, v9] is determined. Calculate the Z value corresponding to each data value and compare it with the preset threshold to obtain the first numerical array m = [1, 1, 0, 1, 0, 0, 1, 1, 1]. Statistically analyze the continuously abnormal data values in the first data to obtain the second array t = [1, 2, 0, 1, 0, 0, 1, 2, 3].

[0127] Shift each element in the second array t one position to the left and set the last element to empty (i.e., set it to NaN) to obtain the third array t(shift(-1)) = [2, 0, 1, 0, 0, 1, 2, 3, NaN]. Calculate the difference between the second array t and the third array t(shift(-1)) to obtain the fourth array t - t(shift(-1)) = [-1, 2, -1, 1, 0, -1, -1, -1, NaN].

[0128] In the fourth array, the data value at the position where the element value is 1 is the abnormal value to be processed. For example, the fourth element value in the fourth data is 1, indicating that the fourth element (v4) in the corresponding TB_Ref_Speed is an abnormal value and needs to be processed for abnormalities. The data value v4 is not at the first position and the last position in the array TB_Ref_Speed. Then, according to the data values v3 and v5 adjacent to the data value v4, the linear filling method is used to determine the corresponding estimated value. For example, the estimated value v4’ is calculated using the linear interpolation calculation formula, and the updated TB_Ref_Speed = [v1, v2, v3, v4’, v5, v6, v7, v8, v9] is obtained.

[0129] Similarly, similar processing is also performed on the abnormal data values in the second operation data. The specific operation steps can refer to the embodiments and specific examples described above, as well as the processes shown in FIGS. 7(a) and 7(b), and will not be elaborated here.

[0130] In step S3, the warning model is a mathematical or computational model used to predict and warn of potential risks or potential problems. In the present invention, several warning models can be extracted from the warning model library according to actual needs and added to the model container. Subsequently, based on the updated first operation data and second operation data, the warning models are run one by one, and thus various types of prediction results can be obtained. Based on the comparison between the prediction results and the warning criteria, warning information is determined and output. The warning information includes at least one of a fault code, a warning time, a train number, and a service life. The warning information obtained by running each warning model is stored in the warning queue. After all the warning models are run, the warning information in the warning queue is uniformly stored in the database, and subsequently, the warning information and the corresponding train formation in the database are queried to facilitate the front-end configuration display according to the train formation.

[0131] The above-mentioned various embodiments, examples, or specific examples provided by the present invention can be combined with each other, thereby finally forming multiple more optimal embodiments.

[0132] For example, FIGS. 8(a) and 8(b) show a schematic flow diagram of a method for warning of rail vehicle faults. The following summarizes its working process in combination with the content shown in FIGS. 8(a) and 8(b).

[0133] Obtain the operation data of the rail vehicle, and determine whether the vehicle is a mixed-formation vehicle according to the mixed-formation flag bit in the operation data; if so, based on the train number, determine the first operation data corresponding to the first formation type (such as 4 formations) and / or the second operation data corresponding to the second formation type (such as 6 formations).

[0134] Update the first operation data using the first variable corresponding to the 4-formation train, and / or update the second operation data using the second variable corresponding to the 8-formation train. At the same time, data processing operations such as calculating common intermediate variables (such as calculating the average value, maximum value, or minimum value) and variable data unit conversion can also be performed. At the same time, all variables are traversed to perform outlier inspection and filling operations to obtain the first operation data and / or second operation data that can be recognized by the warning model; input them into several warning models respectively to obtain corresponding warning results; determine whether the warning results meet the warning criteria; if so, store the corresponding warning information in the warning queue and perform subsequent processing based on the warning information in the warning queue.

[0135] The early warning method based on the configuration mode adopted by the present invention parses the operation data and configures variables for the corresponding operation data according to the formation type, which means that the early warning models for different projects or different data sets can use the same configuration logic, thus ensuring the unity of the models. This unity makes the early warning model easier to reuse, without the need for a large number of modifications to the code of the early warning model, saving labor and time costs.

[0136] An embodiment of the present invention provides a computer-readable storage medium.

[0137] In one embodiment, the computer-readable storage medium stores the computer program executed by the aforementioned processor, or the rail vehicle fault early warning method in any of the foregoing technical solutions.

[0138] When the processor executes the computer program, it can execute the description of the rail vehicle fault early warning method in any of the foregoing technical solutions. Therefore, it will not be elaborated here. In addition, the description of the beneficial effects of adopting the same method will not be elaborated either.

[0139] The computer-readable storage medium may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disc, etc.

[0140] In summary, the rail vehicle fault early warning method and storage medium provided by the present invention parse the operation data of the rail vehicle, determine the first operation data corresponding to the first formation type and the second operation data corresponding to the second formation type, and update these operation data using the corresponding variables to ensure that the data input into the early warning model meets the model requirements. In this way, the early warning model can perform more accurate early warnings on the faults of the rail vehicle, avoiding early warning errors caused by chaotic and mismatched data, not only improving the accuracy and applicability of the data, but also enabling the early warning model to handle rail vehicles of different formation types without the need to separately construct an early warning model for each formation type, with strong reusability and versatility.

[0141] It should be understood that although this specification is described according to embodiments, not every embodiment only contains an independent technical solution. This narrative way of the specification is only for clarity. Those skilled in the art should regard the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

[0142] The series of detailed descriptions listed above are only specific descriptions of the feasible implementation manners of the present invention, and they are not intended to limit the protection scope of the present invention. Any equivalent implementation manners or changes made without departing from the technical spirit of the present invention should be included within the protection scope of the present invention.

Claims

1. A rail vehicle fault warning method, characterized in that: include: Obtaining operation data of the rail vehicle, determining first operation data corresponding to the first constellation type and second operation data corresponding to the second constellation type; Using the first variable to update the first operation data, and using the second variable to update the second operation data, and inputting the updated first operation data and the updated second operation data into the early warning model; The early warning model is pre-trained to identify the first variable and the second variable; According to the output of the warning model, rail vehicle fault warning information is determined.

2. The fault warning method according to claim 1, characterized in that: The updating of the first operation data by using the first variable includes: Identify first data values ​​with the same attributes and other second data values ​​with different attributes in the first operation data, and use the first variable to update the first data value and the second data value respectively; a mapping relationship exists between the variable and the data value in the updated first operation data.

3. The fault warning method according to claim 2, characterized in that: The updating of the first data value and the second data value respectively by using the first variable comprises: Using the first variable, each data value in the first operation data is updated to obtain the first intermediate operation data; there is a mapping relationship between the variable and the data value in the first intermediate operation data; Identify and abstract variables with the same attributes in the first intermediate operation data to determine the second intermediate operation data; The first intermediate operation data and the second intermediate operation data are combined to obtain updated first operation data.

4. The fault warning method according to claim 3, characterized in that: The identifying and abstracting variables having the same attributes in the first intermediate operation data to determine the second intermediate operation data includes: Variables with the same prefix in the first intermediate operation data are identified, summarized into a unified abstract variable, and the corresponding second intermediate operation data is determined based on the abstract variable.

5. The fault warning method according to claim 2, characterized in that: The updating of the first data value and the second data value respectively by using the first variable comprises: Obtaining a plurality of first data values ​​having the same attribute in the first operation data; Using the first variable to perform abstract update processing on a plurality of first data values, second intermediate operation data is obtained; there is a mapping relationship between the variable and the data value in the second intermediate operation data; Using the first variable, respectively update other second data values ​​that do not have the same attribute to obtain first intermediate operation data; The first intermediate operation data and the second intermediate operation data are combined to obtain updated first operation data.

6. The fault warning method according to claim 1, characterized in that: Before inputting the updated first operation data and the updated second operation data into the early warning model, the method further includes: Traversing a number of data values ​​corresponding to each variable in the first operation data, calculating a Z value corresponding to each data value, and generating a first array according to the Z value, where the first array is used to mark an abnormal situation of each data value; Counting the number of elements in the first array that are continuously marked as abnormal values, and determining the corresponding second array according to the number of elements; wherein the first array and the second array have the same array length; The first operation data is updated according to the difference between the first array and the second array.

7. The fault warning method according to claim 6, characterized in that: The step of generating a first array according to the Z value comprises: Determine whether the Z value corresponding to each data value is greater than a preset threshold; If so, the data value at the corresponding position is marked as an outlier; If not, the data value at the corresponding position is marked as a normal value; According to the position distribution of the data values ​​corresponding to the first operating data, the marking results of the data values ​​are arranged to obtain a first array.

8. The fault warning method according to claim 6, characterized in that: The counting of the number of elements in the first array that are continuously marked as abnormal values, and determining the corresponding second array according to the number of elements, includes: Initialize an exception count mark p, where the exception count mark p is used to indicate the number of elements of consecutive exception marks in the first array; Traverse each element in the first array and determine whether the value of the current element is an abnormal value; If yes, then the exception count mark p is incremented by 1, and the result of the exception count mark p is used as the element value of the corresponding position of the second array; If not, the exception count flag p is set to 0, and the result of the exception count flag p is used as the element value of the corresponding position of the second array.

9. The fault warning method according to claim 6, characterized in that: The updating of the first operation data according to the difference between the first array and the second array includes: Shift each element in the second array left by one position and set the last element to null to generate a third array. The number of elements in the third array is equal to that in the second array. Calculate the difference between the second array and the third array to obtain a fourth array, where the number of elements in the fourth array is equal to that in the third array; According to the position where the element value in the fourth array is the abnormal value, it is determined that the data value at the corresponding position in the first operation data is the abnormal value, and the abnormal data value is updated and determined.

10. The fault warning method according to claim 9, characterized in that: Determining that the data value at the corresponding position in the first operation data is an abnormal value, and updating and determining the abnormal data value, comprises: Determine whether the abnormal data value is at the first position or the last position in the first operation data; If not, obtain a third data value and a fourth data value adjacent to the abnormal data value, determine an estimated value of the abnormal data value using a linear filling method based on the third data value and the fourth data value, and update the estimated value to the corresponding abnormal data value.

11. The fault warning method according to claim 10, characterized in that: After determining whether the abnormal data value is at the first position or the last position of the first operation data, the method further includes: If yes, when the abnormal data value is at the first position in the first operation data, the data value at the next position is used to fill the abnormal data value; When the abnormal data value is at the end position in the first operation data, the abnormal data value is filled with the data value of the previous position of the end position.

12. A computer storage medium storing a computer program, wherein when the computer program is executed, the device where the computer storage medium is located executes the steps of the rail vehicle fault warning method according to any one of claims 1 to 11.

Images