CN117077050A

CN117077050A - Rail vehicle early warning method and device, electronic equipment and storage medium

Info

Publication number: CN117077050A
Application number: CN202310870319.9A
Authority: CN
Inventors: 陈美霞; 张金磊; 滑瑾; 吕红强; 黄涛; 齐玉玲; 黄盼
Original assignee: CRRC Nanjing Puzhen Co Ltd
Current assignee: CRRC Nanjing Puzhen Co Ltd
Priority date: 2023-07-14
Filing date: 2023-07-14
Publication date: 2023-11-17

Abstract

The disclosure provides a rail vehicle early warning method and device, electronic equipment and storage medium, wherein the method comprises the following steps: acquiring sensing data of a sensor of the railway vehicle within a preset time period; performing anomaly detection on the sensing data in the preset time period by using a preset anomaly detection model to obtain an anomaly detection result corresponding to the anomaly detection model; determining whether the sensor is abnormal or not according to the occurrence frequency of the abnormal characterization of the abnormal detection result; and outputting early warning information in response to the sensor abnormality. Through the method, the electronic equipment can determine the occurrence frequency of the abnormality by utilizing the multiple abnormality detection results, determine whether the abnormality exists or not and early warn based on the frequency, improve the early warning accuracy and improve the running stability of the railway vehicle.

Description

Rail vehicle early warning method and device, electronic equipment and storage medium

Technical Field

The disclosure relates to the technical field of rail transit, but not limited to, and in particular relates to a rail vehicle early warning method and device, electronic equipment and a storage medium.

Background

The rail vehicle braking system is used as an important component of the rail vehicle, and can control the speed of the vehicle during the running process of the rail vehicle so as to ensure the running safety of the rail vehicle. Common faults of rail vehicle brake systems are often manifested as brake spin, coasting, axle lock-up, sensor faults, etc. Common faults of the sensor are mainly caused by factors such as short circuit of internal lines, short circuit of connecting lines or looseness of the sensor. Therefore, the problem of how to perform fault detection on the rail vehicle is of great concern.

Disclosure of Invention

In view of this, the embodiments of the present disclosure desire to provide a rail vehicle early warning method and apparatus, a computer device, and a storage medium.

In a first aspect, an embodiment of the present disclosure provides a rail vehicle early warning method, the method including:

acquiring sensing data of a sensor of the railway vehicle within a preset time period;

performing anomaly detection on the sensing data in the preset time period by using a preset anomaly detection model to obtain an anomaly detection result corresponding to the anomaly detection model;

determining whether the sensor is abnormal or not according to the occurrence frequency of the abnormal characterization of the abnormal detection result;

and outputting early warning information in response to the sensor abnormality.

In some embodiments, the method further comprises:

dividing the sensing data in the preset time length according to time windows to obtain sensing data corresponding to a plurality of time windows;

performing anomaly detection on the sensing data in the preset time period by using a preset anomaly detection model to obtain an anomaly detection result corresponding to the anomaly detection model, wherein the anomaly detection result comprises:

and carrying out anomaly detection on the sensing data corresponding to the time windows by using the anomaly detection model to obtain an anomaly detection result corresponding to the anomaly detection model.

In some embodiments, the method further comprises:

acquiring historical sensing data of the sensor;

determining a threshold range corresponding to the abnormality detection model and representing abnormality according to the historical sensing data;

the step of using the anomaly detection model to perform anomaly detection on the sensing data corresponding to the multiple time windows to obtain an anomaly detection result corresponding to the anomaly detection model includes:

and determining an abnormality detection result corresponding to the abnormality detection model according to the sensing data corresponding to the plurality of time windows and the threshold range representing the abnormality corresponding to the abnormality detection model.

In some embodiments, the anomaly detection model includes a sensor disconnect detection sub-model;

the determining, according to the sensing data corresponding to the plurality of time windows and the threshold range representing the abnormality corresponding to the abnormality detection model, an abnormality detection result corresponding to the abnormality detection model includes:

determining the maximum duration of consistent adjacent sensing data in each time window according to the sensing data corresponding to each time window;

determining a disconnection value corresponding to the time window according to the ratio of the maximum duration to the duration corresponding to the time window;

Determining the disconnection value corresponding to the preset duration according to the disconnection value corresponding to each time window;

determining that the detection result corresponding to the sensor disconnection detection sub-model represents the abnormality in response to the disconnection value of the preset duration being within the threshold range of the representation abnormality corresponding to the sensor disconnection detection sub-model; the threshold range of the characterization abnormality corresponding to the sensor disconnection detection sub-model is a range determined based on the average value of the historical disconnection values.

In some embodiments, the anomaly detection model further comprises an extremum detection sub-model;

determining extreme value abnormal data quantity of sensing data which is not in a preset extreme value threshold range in the time window according to the sensing data corresponding to each time window;

determining an extreme value anomaly value corresponding to each time window according to the extreme value anomaly data quantity corresponding to each time window and the sensing data total quantity corresponding to the time window;

according to the extreme value abnormal value corresponding to each time window, determining the extreme value abnormal value corresponding to the preset duration;

Determining that the detection result corresponding to the extremum detection sub-model represents the abnormality in response to the extremum abnormality value corresponding to the preset duration being within the threshold range of the representation abnormality corresponding to the extremum detection sub-model; the threshold range corresponding to the extremum detection sub-model and used for representing the abnormality is a range determined based on the average value of the historical extremum abnormality values.

In some embodiments, the anomaly detection model further comprises an outlier detection sub-model;

for the sensing data corresponding to each time window, determining a difference value of the median value of each sensing data and preset sensing data in the time window, and determining an outlier data quantity of the sensing data of which the difference value is not in a preset difference threshold value range;

determining an outlier value corresponding to a time window according to the outlier data quantity corresponding to the time window and the sensing data total quantity corresponding to the time window;

determining an outlier corresponding to the preset duration according to the outlier corresponding to each time window;

Determining that the detection result corresponding to the outlier detection sub-model represents the abnormality in response to the outlier value corresponding to the preset duration being within the threshold range of the representation abnormality corresponding to the outlier detection sub-model; the threshold range of the characterization abnormality corresponding to the outlier detection sub-model is a range determined based on the mean of the historical outlier values.

In some embodiments, the anomaly detection model further comprises a transition detection sub-model;

determining differential values of adjacent sensing data in each time window according to the sensing data corresponding to each time window, and determining jump data quantity of which the differential values are not in a preset differential threshold range;

determining a jump rate value corresponding to a time window according to the jump data quantity corresponding to the time window and the total differential data quantity corresponding to the time window;

determining the jump rate value corresponding to the preset duration according to the jump rate value corresponding to each time window;

determining that the detection result corresponding to the jump detection sub-model represents the abnormality in response to the jump value corresponding to the preset duration being within the threshold range of the representation abnormality corresponding to the jump detection sub-model; the threshold range corresponding to the jump detection sub-model and used for representing the abnormality is a range determined based on the average value of the historical jump values.

In some embodiments, the method further comprises:

determining the abnormal times of the abnormal representation of the abnormal detection results in a plurality of abnormal detection results obtained according to the sensing data in a plurality of preset time periods in a preset period;

determining the occurrence frequency of the abnormality corresponding to the abnormality detection model according to the ratio of the abnormality times to the total abnormality detection times in the preset period;

the determining whether the sensor is abnormal according to the occurrence frequency of the abnormal characterization of the abnormal detection result comprises the following steps:

and determining that the sensor is abnormal in response to the occurrence frequency of the abnormality corresponding to the abnormality detection model being within a preset abnormality rate threshold range.

In some embodiments, the method further comprises:

acquiring historical disconnection value data of the sensor disconnection detection sub-model;

performing data transformation on the historical disconnection value data; wherein the transformed historical disconnect value data conforms to a normal distribution;

according to the historical disconnection value data conforming to normal distribution, determining a mean value and a variance corresponding to the historical disconnection value data;

and determining a threshold range of the characterization abnormality corresponding to the sensor disconnection detection sub-model based on the mean value and the variance.

In some embodiments, the sensors of the rail vehicle include a plurality of pressure sensors;

and processing the sensing data corresponding to the pressure sensors by using the abnormality detection model aiming at each pressure sensor to obtain an abnormality detection result of the abnormality detection model aiming at each pressure sensor.

In a second aspect, embodiments of the present disclosure provide a rail vehicle warning device, the device comprising:

the first acquisition module is used for acquiring sensing data of a sensor of the railway vehicle within a preset duration;

the detection module is used for carrying out anomaly detection on the sensing data in the preset time length by using a preset anomaly detection model to obtain an anomaly detection result corresponding to the anomaly detection model;

the first determining module is used for determining whether the sensor is abnormal or not according to the occurrence frequency of the abnormal representation of the abnormal detection result;

and the output module is used for responding to the abnormality of the sensor and outputting early warning information.

In some embodiments, the rail vehicle warning device further comprises:

The dividing module is used for dividing the sensing data in the preset duration according to time windows to obtain sensing data corresponding to a plurality of time windows;

the detection module is further configured to perform anomaly detection on the sensing data corresponding to the multiple time windows by using the anomaly detection model, so as to obtain an anomaly detection result corresponding to the anomaly detection model.

In some embodiments, the rail vehicle warning device further comprises:

the second acquisition module is used for acquiring historical sensing data of the sensor;

the second determining module is used for determining a threshold range of the characterization abnormality corresponding to the abnormality detection model according to the historical sensing data;

the detection module is further configured to determine an abnormality detection result corresponding to the abnormality detection model according to the sensing data corresponding to the plurality of time windows and the threshold range representing the abnormality corresponding to the abnormality detection model.

In some embodiments, the detection module includes a sensor disconnect detection sub-module;

the sensor disconnection detection submodule is used for determining the maximum duration of consistent adjacent sensing data in each time window according to the sensing data corresponding to each time window; determining a disconnection value corresponding to the time window according to the ratio of the maximum duration to the duration corresponding to the time window; determining the disconnection value corresponding to the preset duration according to the disconnection value corresponding to each time window; determining that the detection result corresponding to the sensor disconnection detection sub-model represents the abnormality in response to the disconnection value of the preset duration within the threshold range of the representation abnormality corresponding to the sensor disconnection detection sub-model; the threshold range of the characterization abnormality corresponding to the sensor disconnection detection sub-model is a range determined based on the average value of the historical disconnection values.

In some embodiments, the detection module further comprises an extremum detection sub-module;

the extremum detection submodule is used for determining extremum abnormal data quantity of sensing data which is not in a preset extremum threshold range in the time window according to the sensing data corresponding to each time window; determining an extreme value anomaly value corresponding to each time window according to the extreme value anomaly data quantity corresponding to each time window and the sensing data total quantity corresponding to the time window; according to the extreme value abnormal value corresponding to each time window, determining the extreme value abnormal value corresponding to the preset duration; determining that the detection result corresponding to the extremum detection sub-model represents the abnormality in response to the extremum abnormality value corresponding to the preset duration being within the threshold range of the representation abnormality corresponding to the extremum detection sub-model; the threshold range corresponding to the extremum detection sub-model and used for representing the abnormality is a range determined based on the average value of the historical extremum abnormality values.

In some embodiments, the detection module further comprises an outlier detection sub-module;

the outlier detection sub-module is used for determining a difference value between each sensing data and a median value of preset sensing data in each time window according to the sensing data corresponding to each time window, and determining an outlier data volume of the sensing data of which the difference value is not in a preset difference threshold value range; determining an outlier value corresponding to a time window according to the outlier data quantity corresponding to the time window and the sensing data total quantity corresponding to the time window; determining an outlier corresponding to the preset duration according to the outlier corresponding to each time window; determining that the detection result corresponding to the outlier detection sub-model represents the abnormality in response to the outlier value corresponding to the preset duration being within the threshold range of the representation abnormality corresponding to the outlier detection sub-model; the threshold range of the characterization abnormality corresponding to the outlier detection sub-model is a range determined based on the mean of the historical outlier values.

In some embodiments, the detection module further comprises a transition detection sub-module;

the jump detection submodule is used for determining the differential value of adjacent sensing data in each time window according to the sensing data corresponding to each time window, and determining the jump data quantity of which the differential value is not in a preset differential threshold range; determining a jump rate value corresponding to a time window according to the jump data quantity corresponding to the time window and the total differential data quantity corresponding to the time window; determining the jump rate value corresponding to the preset duration according to the jump rate value corresponding to each time window; determining that the detection result corresponding to the jump detection sub-model represents the abnormality in response to the jump value corresponding to the preset duration being within the threshold range of the representation abnormality corresponding to the jump detection sub-model; the threshold range corresponding to the jump detection sub-model and used for representing the abnormality is a range determined based on the average value of the historical jump values.

In some embodiments, the rail vehicle warning device further comprises:

the third determining module is used for determining the abnormal times of the abnormal representation of the abnormal detection results among a plurality of abnormal detection results obtained according to the sensing data in a plurality of preset time periods in a preset period;

A fourth determining module, configured to determine, according to a ratio of the number of anomalies to a total number of anomalies detected in the preset period, a frequency of occurrence of anomalies corresponding to the anomaly detection model;

the first determining module is further configured to determine that the sensor is abnormal in response to the occurrence frequency of the abnormality corresponding to the abnormality detection model being within a preset abnormality rate threshold range.

In some embodiments, the sensor disconnect detection sub-module is further for obtaining historical disconnect value data for the sensor disconnect detection sub-model; performing data transformation on the historical disconnection value data; wherein the transformed historical disconnect value data conforms to a normal distribution; according to the historical disconnection value data conforming to normal distribution, determining a mean value and a variance corresponding to the historical disconnection value data; and determining a threshold range of the characterization abnormality corresponding to the sensor disconnection detection sub-model based on the mean value and the variance.

the detection module is further used for processing the sensing data corresponding to the pressure sensors by using the abnormality detection model aiming at each pressure sensor to obtain an abnormality detection result of the abnormality detection model aiming at each pressure sensor.

In a third aspect, an embodiment of the present disclosure provides an electronic device, including a memory and a processor, the memory storing a computer program executable on the processor, the processor implementing the method described in the first aspect when executing the program.

In a fourth aspect, embodiments of the present disclosure provide a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the method described in the first aspect.

The technical scheme provided by the embodiment of the disclosure can comprise the following beneficial effects:

in the embodiment of the disclosure, the electronic device firstly uses the abnormality detection model to perform abnormality detection on the sensing data within a preset time period to obtain an abnormality detection result, then determines the occurrence frequency of the abnormality of the sensor according to the abnormality detection result, judges whether the sensor is abnormal according to the abnormality frequency, and can output early warning information when the sensor is abnormal. On the one hand, compared with the case that whether the pressure sensor is abnormal or not is determined and early-warned only through one abnormal detection result, in the embodiment of the disclosure, the electronic device determines the occurrence frequency of the abnormality based on the multiple abnormal detection results, and determines whether the abnormality is actually present or not and early-warned based on the occurrence frequency of the abnormality, so that erroneous judgment caused by unstable sensing data but not abnormal data or erroneous judgment caused by instability of an abnormality detection model can be reduced, and the accuracy of the sensor abnormality judgment can be improved to improve the accuracy of early-warning; on the other hand, the electronic equipment outputs early warning information when the sensor is abnormal, so that maintenance personnel of the railway vehicle can carry out corresponding overhaul based on the output early warning information, and the running stability of the railway vehicle is improved. It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the disclosure and together with the description, serve to explain the technical aspects of the disclosure.

Fig. 1 is a schematic implementation flow diagram of a rail vehicle early warning method according to an embodiment of the disclosure;

fig. 2 is a histogram of disconnection values corresponding to a sensor disconnection detection sub-model according to an embodiment of the present disclosure;

FIG. 3 is a probability density distribution plot of sensor off-rate values after normal correction provided by an embodiment of the present disclosure;

FIG. 4 is a sensor disconnect anomaly profile provided by an embodiment of the present disclosure;

fig. 5 is a histogram corresponding to an outlier detection result provided in an embodiment of the present disclosure;

FIG. 6 is a histogram of outliers corresponding to a sensor outlier detection sub-model provided in an embodiment of the disclosure;

FIG. 7 is a probability density distribution plot of sensor outliers after normal correction provided by an embodiment of the present disclosure;

FIG. 8 is an outlier anomaly profile provided by an embodiment of the present disclosure;

FIG. 9 is a graph of a sensor transition anomaly probability density provided by an embodiment of the present disclosure;

FIG. 10 is a graph of a sensor jump anomaly profile provided by an embodiment of the present disclosure;

FIG. 11 is an overall flow chart of sensor anomaly detection provided by an embodiment of the present disclosure;

fig. 12 is a schematic diagram of a rail vehicle early warning device according to an embodiment of the disclosure;

fig. 13 is a schematic diagram of a hardware entity of a computer device in an embodiment of the disclosure.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the present disclosure more apparent, the technical solutions of the present disclosure are further elaborated below in conjunction with the drawings and the embodiments, and the described embodiments should not be construed as limiting the present disclosure, and all other embodiments obtained by those skilled in the art without making inventive efforts are within the scope of protection of the present disclosure.

In the following description, reference is made to "some embodiments" which describe a subset of all possible embodiments, but it is to be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments and can be combined with one another without conflict.

The terms "first/second/third" and "first/second/third" in reference to the present disclosure are merely distinguishing between similar objects and not representing a particular ordering of objects, it being understood that the "first/second/third" may be interchanged with a particular order or precedence, as allowed, to enable embodiments of the disclosure described herein to be implemented in other than those illustrated or described herein.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. The terminology used herein is for the purpose of describing the present disclosure only and is not intended to be limiting of the present disclosure.

In the related art, the fault detection mode of the rail vehicle braking system is limited to means such as centralized checking and regular maintenance, and the hidden trouble of rail vehicle braking cannot be found early and early warned due to lack of an online fault detection and early warning method, so that the rail vehicle braking efficiency is low, the normal operation of the rail vehicle is influenced, and even potential safety hazards are possibly caused to the operation of the rail vehicle.

In this regard, the disclosure provides a rail vehicle early warning method, fig. 1 is a schematic implementation flow diagram of the rail vehicle early warning method provided in an embodiment of the disclosure, and as shown in fig. 1, the method includes the following steps:

s101, acquiring sensing data of a sensor of the railway vehicle in a preset time period;

s102, carrying out anomaly detection on the sensing data in the preset time period by using a preset anomaly detection model to obtain an anomaly detection result corresponding to the anomaly detection model;

S103, determining whether the sensor is abnormal or not according to the occurrence frequency of the abnormal representation of the abnormal detection result;

s104, responding to the sensor abnormality and outputting early warning information.

The rail vehicle early warning method disclosed by the embodiment of the invention can be applied to electronic equipment. The electronic device may be a computer, a server, or the like. For example, the electronic device may be a control device directly mounted on a railway vehicle such as a subway, a high-speed rail, or the like, such as a computer, a server, or the like on the railway vehicle; it is also possible to provide control devices which establish a communication connection with the rail vehicle and control the rail vehicle, for example a computer or a server in a rail vehicle operation center room, etc. The fault detection and the early warning of the rail vehicle are realized through the rail vehicle early warning method.

In step S101, the manner in which the electronic device obtains the sensing data of the rail vehicle may be to directly obtain the sensing data of the corresponding duration according to the preset duration; or dividing the preset time length into a plurality of time windows, and then acquiring the sensing data of the time length corresponding to each time window. For example, the duration of one time window may be set to 1000 seconds in general.

In embodiments of the present disclosure, the sensors of the rail vehicle may include various types of sensors, such as pressure sensors, temperature sensors, etc., and the corresponding sensed data may be pressure data, temperature data, etc. In addition, the various types of sensors may also be different types of pressure sensors, such as a brake reservoir pressure sensor, a total wind pressure sensor, a brake reservoir pressure sensor, an air suspension pressure sensor, a parking brake pressure sensor, etc., and the corresponding sensing data may be a brake reservoir pressure, a total wind pressure, a brake reservoir pressure, an air suspension pressure, a parking brake pressure, etc. Embodiments of the present disclosure are described with respect to different types of pressure sensing data.

When the electronic equipment acquires the pressure sensing data, the electronic equipment can acquire the pressure sensing data aiming at various types of pressure sensors, and the acquired various pressure sensing data are classified and stored according to types, so that the electronic equipment can detect and early warn aiming at the pressure sensors of different types; the pressure sensors at different positions of the railway vehicle can be acquired, and the acquired pressure sensing data are classified and stored according to the positions, so that the electronic equipment can detect and early warn the pressure sensors at different positions of the railway vehicle conveniently; the pressure sensing data of the pressure sensor can be acquired according to the type and the position, the acquired pressure sensing data is stored in a classified mode, and the electronic equipment can detect and early warn according to the pressure sensors in different positions and different types of the railway vehicle.

In step S102, the electronic apparatus may set the abnormality detection model in advance, and the abnormality detection model may be set to be single or plural. It should be noted that, from the perspective of the performance of the historical fault conditions of the rail vehicle, the rail vehicle may have multiple types of faults, and the severity of the different fault types is not the same. For example, serious failure faults are mainly manifested by sudden malfunctions of the sensor measurement, which remain constant throughout; the slight faults are mainly represented by random open faults, impact faults, deviation faults, drift faults, self-interference faults and the like.

In the embodiment of the disclosure, the anomaly detection model may be a model that synthesizes each fault detection function, so that the sensing data can be comprehensively detected by using each fault detection function to obtain a comprehensive anomaly detection result; in addition, the anomaly detection model can be further subdivided, and is configured to be specific to anomaly detection submodels of different fault types so as to more clearly locate the fault condition of the sensor, wherein each anomaly detection submodel is, for example, a sensor disconnection detection submodel, an extremum detection submodel, an outlier detection submodel, a jump detection submodel, and the like.

In the embodiment of the disclosure, the method for performing anomaly detection on the sensing data in the preset duration by using the preset anomaly detection model by the electronic device may be to directly count the anomaly sensing data in the preset duration, and obtain an anomaly detection result based on the anomaly sensing data; or firstly counting the abnormal sensing data in each time window in the preset time period, obtaining the abnormal detection result corresponding to each time window based on the abnormal sensing data, and then determining the abnormal detection result in the preset time period according to the abnormal detection result corresponding to each time window.

In the embodiment of the disclosure, the anomaly detection result corresponding to the anomaly detection model may be represented by using the probability of occurrence of the anomaly, or may be represented by using the number of occurrences of the anomaly, where when the probability of occurrence of the anomaly or the number of occurrences of the anomaly is within a preset threshold range for representing the anomaly, it is determined that the anomaly detection result corresponding to the anomaly detection model is an anomaly, otherwise, the anomaly detection result is an anomaly. For example, if the threshold range of the abnormality detection model for characterizing the abnormality is that the number of abnormalities is greater than 10, then the detection result is normal when the number of abnormalities of the sensor is 5 times and the detection result is abnormal when the number of abnormalities of the sensor is 15 times in the detection duration of one day; for another example, if the threshold range of the abnormality detection model corresponding to the characteristic abnormality is that the abnormality probability is greater than 5%, the characteristic detection result is normal when the sensor abnormality probability is 2% and the characteristic detection result is abnormal when the sensor abnormality probability is 10% in the detection period of one day.

In step S103, the electronic device may determine, according to the abnormality detection result, the occurrence frequency of the sensor abnormality, and if the occurrence frequency of the abnormality is within the preset abnormality rate threshold range, determine that the sensor is abnormal, otherwise determine that the sensor is normal. The preset anomaly rate threshold range may be determined according to a historical anomaly rate of the sensor.

In the embodiment of the disclosure, a plurality of ways of determining the occurrence frequency of the abnormality may be to determine the ratio of the number of times of detecting the abnormality in the preset duration to the total number of times of detecting the abnormality by the abnormality detection model as the occurrence frequency of the abnormality of the sensor; or determining the average value or the median value of the abnormal rates corresponding to the time windows in the preset time period as the occurrence frequency of the sensor abnormality based on the abnormal rates corresponding to the time windows in the preset time period; the method can also comprise the steps of determining a plurality of preset time periods as a detection period, determining the abnormal times representing the abnormality according to the abnormal detection results of the plurality of preset time periods, and determining the ratio of the abnormal times to the total abnormal detection times in the detection period as the occurrence frequency of the sensor abnormality.

In addition, the electronic equipment can also directly acquire the self-checking result of the sensor, and the sensor abnormality is determined according to the self-checking result. For example, the self-checking result of the pressure sensor includes failure information such as a brake cylinder pressure sensor failure, a total wind pressure sensor failure, a brake cylinder pressure sensor failure, an air suspension pressure sensor failure, a parking brake pressure sensor failure, etc., and the electronic device can directly determine that the pressure sensor is abnormal according to the pressure sensor failure information.

In step S104, the electronic device may output the early warning information when the abnormality detection model detects an abnormality; and the early warning information can be output when the sensor self-detection result is obtained to represent the sensor abnormality. The early warning information can be arranged corresponding to the abnormality detection model and/or the sensor type, can be composed of a sensor type name, can also be composed of a sensor type name and abnormality information, and can also be composed of a sensor position, a sensor name and abnormality information. For example, the early warning information may be an abnormality of the brake cylinder pressure sensor, or an abnormality of the brake cylinder pressure sensor of the 1 frame of the subway 8 car.

In the embodiment of the disclosure, the electronic device firstly uses the abnormality detection model to perform abnormality detection on the sensing data within a preset time period to obtain an abnormality detection result, then determines the occurrence frequency of the abnormality of the sensor according to the abnormality detection result, judges whether the sensor is abnormal according to the abnormality frequency, and can output early warning information when the sensor is abnormal. On the one hand, compared with the case that whether the pressure sensor is abnormal or not is determined and early-warned only through one abnormal detection result, in the embodiment of the disclosure, the electronic device determines the occurrence frequency of the abnormality based on the multiple abnormal detection results, and determines whether the abnormality is actually present or not and early-warned based on the occurrence frequency of the abnormality, so that erroneous judgment caused by unstable sensing data but not abnormal data or erroneous judgment caused by instability of an abnormality detection model can be reduced, and the accuracy of the sensor abnormality judgment can be improved to improve the accuracy of early-warning; on the other hand, the electronic equipment outputs early warning information when the sensor is abnormal, so that maintenance personnel of the railway vehicle can carry out corresponding overhaul based on the output early warning information, and the running stability of the railway vehicle is improved.

In some embodiments, the method further comprises:

In the embodiment of the disclosure, the electronic device may divide the sensing data in the preset duration into the durations of the time windows and obtain the sensing data of the time windows. The corresponding time durations of the time windows may be set to be the same or different. The sensing data of the time windows can be sensing data with preset duration and are obtained by directly dividing the sensing data with the preset duration according to the duration of the time windows; or after dividing the sensing data with preset duration according to the duration of the time window, extracting part of the time windows from the sensing data with divided time windows as sensing data of a plurality of time windows. For example, when the preset time period is set to 1 day and the time window length is set to be fixed for 20 minutes, the sensing data of 1 day may be directly divided into sensing data of 72 time windows; the sensing data of one time window can be obtained every other hour in 1 day, namely, the sensing data of 24 time windows are selected among the sensing data of 72 time windows, so that the method is beneficial to reducing the calculated amount under the condition of ensuring the accuracy of the early warning method of the railway vehicle.

In the embodiment of the disclosure, the electronic equipment divides the sensing data with preset duration, performs abnormality detection on the sensing data corresponding to a plurality of time windows by using the abnormality detection model, and obtains the corresponding abnormality detection result, so that the data volume during single abnormality detection can be reduced, the storage pressure and the calculation pressure of the electronic equipment can be reduced, the occurrence of the conditions of clamping, breakdown and the like caused by excessive data of the electronic equipment can be reduced, the stability of the electronic equipment in running the rail vehicle early warning method can be improved, and the computer resources can be saved.

In some embodiments, the method further comprises:

acquiring historical sensing data of the sensor;

In the embodiment of the disclosure, the electronic device may acquire historical sensing data, where the historical sensing data may be data in a historical time window corresponding to the anomaly detection model, for example, the detection times and the anomaly times of the historical time window, the anomaly rate corresponding to the historical time window, and the like); the data may also be data in a history preset time period corresponding to the abnormality detection model, for example, the number of times of detection and the number of abnormalities in the history preset time period, the abnormality rate corresponding to the history preset time period, and the like).

In the embodiment of the disclosure, the electronic device determines, according to the historical sensing data, a plurality of methods for characterizing the abnormal threshold range corresponding to the abnormal detection model, which may be to determine, according to the data of a plurality of historical time windows, the threshold range corresponding to the time window, and use the determined threshold range as the threshold range corresponding to the detection model for characterizing the abnormal; or determining a threshold range corresponding to the preset time length and representing the abnormality according to the data of the plurality of historical preset time lengths, and taking the threshold range as the threshold range corresponding to the abnormality detection model and representing the abnormality.

In the embodiment of the disclosure, the threshold range for representing the abnormality may be measured by using the occurrence probability of the abnormality, or may be measured by using the occurrence frequency of the abnormality, and the threshold range for representing the abnormality may be fitted based on the historical sensing data to obtain the average value and the variance of the sensing data in the historical sensing data, and the threshold range for representing the abnormality corresponding to the sensor abnormality detection model is determined based on the average value and the variance of the sensing data. For example, a range outside n (e.g., n=3) times the standard deviation of the sensing data mean corresponding to the abnormality detection model may be determined as the threshold range characterizing the abnormality; the number of abnormalities outside the n (e.g., n=3) times the standard deviation range of the average value of the number of abnormalities corresponding to the abnormality detection model may be determined as the threshold range representing the abnormality.

In the embodiment of the disclosure, the electronic device determines, according to the sensing data corresponding to the plurality of time windows and representing the threshold range of the abnormality corresponding to the abnormality detection model, that the abnormality detection result corresponding to the abnormality detection model is abnormal if the number of abnormalities or the abnormality rate corresponding to the plurality of time windows is within the threshold range representing the abnormality, or determines that the abnormality detection result corresponding to the abnormality detection model is normal. The abnormal times corresponding to the time windows can be obtained by counting the abnormal times in each time window; the abnormal rate corresponding to the time windows can be obtained by calculating the abnormal rate corresponding to each time window and obtaining the abnormal rate corresponding to the time windows by means of calculating the mean value or the median value; the number of abnormality occurrences in each time window may be counted and the total number of abnormality occurrences may be accumulated, and the ratio of the total number of abnormality occurrences to the total number of abnormality detections may be used as the abnormality rate corresponding to the plurality of time windows.

In the embodiment of the disclosure, the electronic device determines the threshold range of the characterization abnormality corresponding to the abnormality detection model according to the historical sensing data, and compared with the mode of presetting the fixed threshold range of the abnormality, the threshold range of the characterization abnormality of the embodiment of the disclosure is determined according to the actual historical sensing data, so that the abnormality detection of the sensing data can be more accurate, and the accuracy of the abnormality judgment of the sensor can be further improved.

In the embodiment of the disclosure, the sensor disconnection detection submodel is used for evaluating whether disconnection abnormality occurs to the sensor. It should be noted that, the electronic device generally collects the sensing data of the sensor at a predetermined sampling interval (for example, at 500 ms), for example, when the data duration of one time window is 20 minutes, then 2400 sensing data of the total time window, that is, the data of the sensor, is updated every 500 ms. Thus, if the sensor data indication value of the sensor remains unchanged for a long time, it can be understood that the sensor is in an off state.

In the embodiment of the disclosure, the electronic device may first count, for each time window, a duration in which adjacent sensing data in the time window is maintained consistent, determine a maximum duration in which adjacent sensing data in each time window is maintained consistent, then use a ratio of the maximum duration to a duration corresponding to the time window as a disconnection value of each time window, and use an average value or a median value of disconnection values corresponding to the time windows as a disconnection value corresponding to a preset duration.

Exemplary, the electronic device may acquire sensing data of the sensor over a plurality of time windows, and determine that the constant duration of the corresponding adjacent sensing data is t according to the first time window ₁ Determining that the constant duration of the corresponding adjacent sensing data is t according to the second time window ₂ …, determining the constant time length of the corresponding adjacent sensing data to be t according to the nth time window _n Assuming that the duration of the time window is t, the disconnection values of the sensors corresponding to the time windows are t respectively ₁ t，t ₂ t，…，t _n t, determining the disconnection value of the preset time period as (t) according to the disconnection value of each time window ₁ +t ₂ +…+t _n )nt。

Fig. 2 is a histogram of disconnection values corresponding to a sensor disconnection detection sub-model according to an embodiment of the present disclosure, where the histogram of disconnection values is obtained by counting the disconnection values of a plurality of sensors such as a brake cylinder pressure sensor, a total wind pressure sensor, a parking brake pressure sensor, etc. in a plurality of preset durations, and as shown in fig. 2, the abscissa of the histogram represents the disconnection values, and the ordinate represents the number of times of occurrence of different disconnection values, if the histogram of disconnection values of the sensors in fig. 2 is fitted, a probability density distribution map corresponding to the disconnection values may be in quasi-normal distribution, but the brake cylinder pressure sensor and the parking brake pressure sensor are likely to have obvious left-offset distribution.

In the embodiment of the disclosure, for the non-standard normal distribution condition that the probability distribution is left or right, the electronic device can perform normal correction transformation on the disconnection value, so that the disconnection value of the sensor tends to the standard normal distribution, wherein the normal correction transformation method can adopt logarithmic transformation, power transformation (Box-Cox transformation, box-Cox) and the like. Fig. 3 is a probability density distribution diagram of a normal corrected sensor disconnection value according to an embodiment of the present disclosure, and fig. 3 is a probability density distribution obtained by Box-Cox conversion for a plurality of sensors such as a brake cylinder pressure sensor, a total wind pressure sensor, a parking brake pressure sensor, etc., respectively, where the probability density distribution of each sensor disconnection value is close to a standard normal distribution as shown in fig. 3.

Fig. 4 is a distribution diagram of abnormal disconnection of a sensor provided by an embodiment of the present disclosure, fig. 4 is a result of testing a plurality of sensors such as a brake cylinder pressure sensor, a total wind pressure sensor, a parking brake pressure sensor, and the like on a historical data set, as shown in fig. 4, a disconnection abnormal test result of 700 days in the past is presented, an abscissa represents the historical nth day, an ordinate represents a test result of a disconnection detection sub-model on the nth day, wherein a 1 bit of a numerical value represents a fault detected on the same day, a 0 bit represents a fault not detected on the same day, and the accuracy of abnormal detection of the disconnection detection sub-model of the sensor is 97.86% according to statistics that the number of false positives of less than 30 days exists in the test result in the historical data of 1400 days.

In the embodiment of the disclosure, when the disconnection value of the preset duration is within the threshold range of the characterization abnormality corresponding to the sensor disconnection detection sub-model, the electronic device determines that the detection result corresponding to the sensor disconnection detection sub-model is characterized as abnormal. The threshold range of the characteristic abnormality corresponding to the sensor disconnection detection sub-model is determined based on the historical disconnection rate value, a corresponding probability density distribution diagram can be obtained by carrying out distribution statistics according to the historical disconnection rate value, the average value of the disconnection rate value is calculated based on the distribution diagram, the threshold range of the characteristic abnormality is determined, for example, the probability range out of n (for example, n=3) standard deviation ranges of the average value of the disconnection rate value can be determined as the threshold range of the characteristic abnormality corresponding to the sensor disconnection detection sub-model.

In the embodiment of the disclosure, the electronic equipment subdivides the anomaly detection model into sub-models of different anomaly types, so that corresponding anomaly detection rules and anomaly detection thresholds are set in a targeted manner according to the detection characteristics of different anomaly types, thereby being beneficial to more clearly positioning the fault condition of the sensor and further improving the early warning accuracy. In addition, in the embodiment of the disclosure, when the electronic device performs abnormality detection through the sensor disconnection detection submodel, the disconnection value is determined based on the maximum duration of the consistency of the adjacent sensing data in the time window, and compared with the method for determining the disconnection value by selecting the duration of the consistency of all the adjacent sensing data, the calculation amount of the abnormality detection of the electronic device is reduced, and the early warning efficiency is improved.

In the disclosed embodiment, the extremum detecting sub-model evaluates whether the sensed data is outside the normal range of values of the sensor. It should be noted that, under the condition that the sensor is working normally, the sensing data will be in the normal data range, and when the sensing data is too small, the sensing data can be regarded as the minimum value beyond the normal data range, and when the sensing data is too large, the sensing data can be regarded as the maximum value beyond the normal data range.

In the embodiment of the disclosure, the electronic device may determine, as the sensor extremum abnormal data, sensing data that is not within a preset extremum threshold range in each time window, calculate to obtain an extremum abnormal data amount corresponding to each time window, determine, as the extremum abnormal value corresponding to each time window, a ratio of the extremum abnormal data amount corresponding to each time window to the total sensing data amount in the time window, and then use an average value or a median value of the extremum abnormal values corresponding to each time window as the extremum abnormal value corresponding to the preset duration. The preset extremum threshold range can be determined according to historical sensing data, the determining mode can be that the historical sensing data of the sensor is subjected to square statistics, sensing data corresponding to a peak value of the square statistics is determined to be a reference value when the sensor works normally, and sensing data in a certain range of the reference value can be determined to be the preset extremum threshold range; and the normal sensing data representing the normal operation of the sensor in the history sensing data can be obtained, the average value is obtained to obtain the average value of the normal sensing data, and the preset extremum threshold range is determined according to the average value of the normal sensing data.

The electronic device may obtain sensing data of the sensor in a plurality of time windows, and determine that the corresponding extreme value abnormal data amount is x according to the first time window ₁ Determining the corresponding extreme value abnormal data quantity as x according to the second time window ₂ …, determining the corresponding extreme value abnormal data amount as x according to the nth time window _n Assuming that the total data amount of the sensor data of each time window is x, the extremum of the sensor corresponding to each time window is differentThe constant values are x respectively ₁ x，x ₂ x，…，x _n x, determining the extreme value anomaly value of the preset time length as (x) according to the extreme value anomaly value of each time window ₁ +x ₂ +…+x _n )nx。

In the embodiment of the disclosure, when the extremum anomaly value of the preset duration is within the threshold range of the characterization anomaly corresponding to the extremum detection sub-model, the electronic device determines that the detection result corresponding to the extremum detection sub-model is characterized as anomaly. The threshold range of the characteristic abnormality corresponding to the extremum detection sub-model is a range determined based on the mean value of the historical extremum abnormality values, the extremum abnormality values corresponding to a plurality of preset time periods can be subjected to histogram statistics, a probability density distribution diagram corresponding to the extremum abnormality values is obtained through fitting, the mean value of the extremum abnormality values of the distribution diagram is calculated, and the probability range outside n (for example, n=3) standard deviation ranges of the mean value of the extremum abnormality values can be determined as the threshold range of the characteristic abnormality corresponding to the extremum detection sub-model.

In the embodiment of the disclosure, the electronic device performs abnormality detection through the extremum detection submodel, and can detect the condition that the sensing data of the sensor exceeds the preset extremum threshold range, generally speaking, the maximum value or the minimum value in the sensing data is likely to be the data generated when the sensor works abnormally, and the method for performing abnormality detection through the maximum value or the minimum value is simple and efficient, so that the calculation amount of abnormality detection of the electronic device is reduced, and the early warning efficiency is improved.

In the disclosed embodiment, the outlier detection submodel is an evaluation of whether the sensed data is far from the median of the sensed data. It should be noted that, under the normal condition of the sensor, the sensing data generally fluctuates within a certain range of the median data, and when the difference between the sensing data and the median is too large, the sensing data can be considered as outlier data. The method for determining the median of the sensing data is various, namely, the historical sensing data is ordered, and the middle number is taken as the median of the sensing data; the data in the middle area of the sorted sensing data can be averaged, and the average value is used as the median of the sensing data.

In the embodiment of the disclosure, the electronic device may calculate, according to the sensing data in each time window, a difference value between each sensing data and a median value in preset sensing data, and if the difference value is within a preset difference threshold range, the sensing data corresponding to the difference value belongs to normal data; if the difference value is not in the preset difference threshold value range, the sensing data corresponding to the difference value belongs to outlier data, and the outlier data volume corresponding to each time window is counted according to the judging result of the difference value in the time window.

In the embodiment of the disclosure, the preset difference threshold range may be determined according to a difference value corresponding to the historical sensing data. Firstly, determining a historical difference value based on the obtained historical sensing data and the median value of the historical sensing data, carrying out square statistics on the historical difference value, determining the difference value corresponding to the peak value of the square statistics as the mean value of the difference value, and determining the data in a certain range of the mean value of the difference value to a preset extremum threshold range.

In the embodiment of the disclosure, the electronic device determines a ratio of an outlier data volume corresponding to each time window to a sensing data total volume in the time window as an outlier value corresponding to each time window, and then takes a mean value or a median value of the outliers corresponding to each time window as an outlier value corresponding to a preset duration.

In the embodiment of the disclosure, when an outlier value of a preset duration is within a threshold range of the characterization abnormality corresponding to the outlier detection sub-model, the electronic device determines that the detection result corresponding to the outlier detection sub-model is characterized as abnormal. The threshold range of the characteristic abnormality corresponding to the outlier detection sub-model is a range determined based on the mean value of the historical outlier values, histogram statistics can be performed on outlier values corresponding to a plurality of preset time periods, a probability density distribution diagram corresponding to the outlier values is obtained through fitting, the mean value of the outlier values of the distribution diagram is calculated, and the probability range outside n (for example, n=3) standard deviation ranges of the mean value of the outlier values can be determined as the threshold range of the characteristic abnormality corresponding to the outlier detection sub-model.

In general, most of the sensing data in a time window is distributed near the median of the sensing data, so the probability of the difference value of 0 in the time window is high, but the data with the difference value of 0 is not focused for outlier detection, and even interferes with the result of outlier detection. Fig. 5 is a histogram corresponding to an outlier detection result provided in the embodiment of the present disclosure, where outlier ratio is obtained by performing outlier detection (Median Absolute Deviation, MAD) on a plurality of sensors such as a brake cylinder pressure sensor, a total wind pressure sensor, a parking brake pressure sensor, etc., and the outlier ratio is 0 when the difference value is 0. As shown in fig. 5, the abscissa of the histogram represents the outlier ratio, and the ordinate represents the number of occurrences of different outlier values, it can be seen that each sensor has a large amount of interference data distributed around the outlier ratio of 0.

In the embodiment of the disclosure, the difference value of the sensing data can be cleaned by excluding the range corresponding to the difference value of the disturbance anomaly detection from the preset difference threshold range, so that the disturbance of the invalid sensor to the outlier detection caused by the invalid difference value is reduced, and the accuracy of the outlier detection submodel is improved. Fig. 6 is an outlier histogram corresponding to a sensor outlier detection sub-model according to an embodiment of the present disclosure, where the outlier histogram is obtained by performing outlier detection after data cleaning on a plurality of sensors, such as a brake cylinder pressure sensor, a total wind pressure sensor, a parking brake pressure sensor, etc., respectively, and as shown in fig. 6, an interference value near 0 after cleaning has been deleted. In addition, as shown in fig. 6, the outlier ratio distribution diagram of the brake cylinder pressure sensor and the total wind pressure sensor has a significant left-hand distribution.

In the embodiment of the disclosure, for the non-standard normal distribution situation such as left deviation or right deviation shown in fig. 6, the electronic device may perform normal correction transformation on the outlier, so that the outlier of the sensor tends to be standard normal distribution, where a method of normal correction transformation may use logarithmic transformation, box-Cox transformation, and the like. Fig. 7 is a probability density distribution diagram of sensor outliers after normal correction according to an embodiment of the present disclosure, and fig. 7 is a probability density distribution diagram obtained by Box-Cox conversion for a plurality of sensors such as a brake cylinder pressure sensor, a total wind pressure sensor, a parking brake pressure sensor, etc., where the probability density distribution of each sensor outlier is close to a standard normal distribution, as shown in fig. 7.

Fig. 8 is an outlier anomaly distribution diagram provided by an embodiment of the disclosure, fig. 8 is a result of testing a plurality of sensors such as a brake cylinder pressure sensor, a total wind pressure sensor, a parking brake pressure sensor and the like on a historical data set, as shown in fig. 8, an outlier anomaly test result of 700 days in the past is presented, an abscissa represents a historical nth day, an ordinate represents a test result of an outlier detection sub-model on the nth day, wherein a value of 1 bit represents a fault detected on the same day, a value of 0 bit represents a fault not detected on the same day, and the accuracy of outlier detection of the sensor outlier detection sub-model is 96.43% when statistics shows that the number of false positives is less than 50 days in the test result in the historical data of 1400 days.

In the embodiment of the disclosure, the electronic device performs abnormality detection through the outlier detection sub-model, and can detect the outlier of the sensing data far away from the median of the sensing data aiming at the sensor, generally speaking, the outlier with overlarge deviation from the median of the sensing data in the sensing data is likely to be the data generated when the sensor works abnormally, and the method for performing abnormality detection through the outlier detection sub-model is simple and efficient, so that the calculation amount of the abnormality detection of the electronic device is reduced, and the early warning efficiency is improved.

In the embodiment of the disclosure, the jump detection submodel is used for evaluating whether jump abnormality occurs to the sensor. Generally, if the difference between adjacent sensing data in the time window is not within the preset difference threshold range, it can be understood that the sensor has a jump abnormal condition.

In the embodiment of the present disclosure, the electronic device may first perform differential computation on adjacent sensing data in each time window, count, according to differential values of adjacent sensing data in each time window, data amounts of which differential values are not in a preset differential threshold range as jump data amounts, then use a ratio of the jump data amounts to differential data amounts corresponding to the time windows as jump values of each time window, and use a mean value or a median value of jump values corresponding to each time window as a jump value corresponding to a preset duration.

In the embodiment of the disclosure, the preset differential threshold range may be determined according to the historical sensing data, for example, the differential value corresponding to the historical sensing data of the sensor is counted in a distributed manner, and the probability density function distribution of the differential value is usually in a normal form with extremely high kurtosis. And determining the upper and lower boundaries (namely a preset differential threshold range) of the jump variation constant differential value as the upper and lower standard deviations or corresponding rounding values of the deviation mean values by 3 times according to the distribution statistics.

It should be noted that, the sensing data acquired by the electronic device may be discontinuous, if the rail vehicle is stopped briefly within a time window, the sensing data of the time window may be discontinuous, but when measuring whether the jump value of the sensor is abnormal, the sensing data is generally required to be based on continuous sensing data, so that adjacent sensing data with too long time interval can be cleaned, so as to improve the detection precision of the jump detection submodel of the sensor. For example, the differential data cleansing method may adopt a method of performing differential calculation on adjacent sensing data, calculating a time interval and a differential value of the adjacent sensing data, and deleting the differential value data with the time interval exceeding 10 seconds.

In the embodiment of the disclosure, when the jump rate value of the preset duration is within the threshold range of the characterization abnormality corresponding to the jump detection sub-model, the electronic device determines that the detection result corresponding to the jump detection sub-model is characterized as abnormal. The threshold range of the characteristic anomaly corresponding to the jump detection sub-model is a range determined based on the historical jump value, probability distribution statistics is carried out on the historical jump value to obtain a probability density distribution diagram, the average value of the jump value is calculated based on the distribution diagram, the threshold range of the characteristic anomaly is determined, for example, the probability range out of n (for example, n=3) standard deviation ranges of the average value of the jump value is determined as the threshold range of the characteristic anomaly corresponding to the sensor jump detection sub-model.

Fig. 9 is a probability density distribution diagram of a sensor jump abnormality provided by an embodiment of the present disclosure, and fig. 9 is a probability density distribution diagram obtained by respectively carrying out probability distribution statistics on a plurality of sensors such as a brake cylinder pressure sensor, a total wind pressure sensor, a parking brake pressure sensor, etc., as shown in fig. 9, an abscissa represents a jump abnormality value, an ordinate is a value without a specific meaning, and an area obtained by integrating based on abscissa data is used for representing a probability of occurrence of the jump value.

Fig. 10 is a distribution diagram of sensor jump anomalies provided by an embodiment of the present disclosure, fig. 10 is a result of testing a plurality of sensors such as a brake cylinder pressure sensor, a total wind pressure sensor, a parking brake pressure sensor and the like on a historical data set, as shown in fig. 10, a jump anomaly test result of 700 days in the past is presented, an abscissa represents a historical nth day, an ordinate represents a test result of a jump detection sub-model on the nth day, wherein a value 1 bit represents a fault detected on the same day, a value 0 bit represents a fault not detected on the same day, and the accuracy of sensor jump detection sub-model anomaly detection is 99.29% when statistics shows that the number of false reports is less than 10 days in the test result in the historical data of 1400 days.

In the embodiment of the disclosure, the electronic device detects the abnormality through the jump detection sub-model, and can detect whether the jump degree of the sensing data aiming at the sensor is abnormal, generally speaking, when the jump degree in the sensing data is too large, the data is likely to be generated when the sensor works abnormally, and the method for detecting the abnormality through the jump detection sub-model is simple and efficient, so that the calculated amount of the abnormality detection of the electronic device is reduced, and the early warning efficiency is improved.

In some embodiments, the method further comprises:

In the embodiment of the disclosure, the railway vehicle detection device may perform abnormality detection multiple times in a preset period, obtain multiple abnormality detection results according to sensing data in multiple preset durations, count the number of abnormality times of the abnormality characterization of the abnormality detection results, and determine a ratio of the number of abnormality times to the total number of abnormality detection times in the preset period as the frequency of occurrence of the abnormality corresponding to the abnormality detection model. For example, the number of abnormalities in which the sensor is abnormal may be counted within 10 days based on the abnormality detection result obtained from the daily sensing data, and if the abnormality detection result is 1 occurrence within 10 days, the frequency of occurrence of the abnormality is 0.1.

In the embodiment of the disclosure, if the abnormality occurrence frequency corresponding to the abnormality detection model is within the preset abnormality threshold range, determining that the sensor is abnormal and outputting early warning information. The preset abnormal threshold range can be obtained by testing historical data in an abnormal detection model, or can be obtained by carrying out probability distribution statistics based on the occurrence frequency of the abnormality, and determining the abnormal threshold range based on the average value obtained by the probability distribution.

In the embodiment of the disclosure, the electronic device counts the probability of occurrence of the corresponding abnormality based on the abnormal results corresponding to the sensing data in a plurality of preset durations, so that the abnormality judgment can be comprehensively performed, and the accuracy of sensor abnormality detection and early warning is improved.

In some embodiments, the method further comprises:

If the probability density distribution diagram shows left or right state, the threshold range of the characterization abnormality corresponding to the sensor disconnection detection sub-model is determined directly through the probability density distribution diagram with the degree of deviation, which is easy to cause the deviation of the detection result of the sensor disconnection detection sub-model and is the decrease of the detection accuracy.

In this regard, in the embodiment of the present disclosure, the probability density distribution diagram may be made to trend toward the standard normal distribution by means of data transformation, and the threshold range of the characterization anomaly corresponding to the sensor off-detection sub-model may be determined based on the probability density distribution diagram of the standard normal distribution. In the embodiment of the disclosure, the electronic device may acquire the historical disconnection rate value data of the sensor disconnection detection sub-model within a plurality of preset time periods, for example, the disconnection rate value data in the past year may be acquired. The electronic device may perform data transformation on the historical disconnection rate value data, so that the transformed historical disconnection rate value data conforms to normal distribution, where the data transformation may be in a Box-Cox transformation, a logarithmic transformation, or the like. For example, the general formula for the Box-Cox transform is as follows:

Wherein y (lambda) is a new variable obtained after Box-Cox transformation, y is an original continuous dependent variable, and lambda is a transformation parameter. The above transformation requires that the original variable y takes a positive value, and if the original variable y takes a negative value, a constant a is added to all the original data to make (y+a) positive value, and then the above transformation is performed. The transformations made for different λ are different. The transform is logarithmic when λ=0, reciprocal when λ= -1, and square root when λ=0.5. The estimation of the parameter lambda in the Box-Cox transformation has two methods of maximum likelihood estimation and Bayes. By solving the lambda value, the corresponding transformation form can be determined.

In the embodiment of the disclosure, the electronic device determines a mean value and a variance corresponding to the historical disconnection value data according to the historical disconnection value data conforming to normal distribution, and determines a threshold range of the characterization abnormality corresponding to the sensor disconnection detection sub-model based on the mean value and the variance. For example, a data range outside n times the standard deviation above and below the mean (e.g., n=3) may be used as the threshold range for characterizing the abnormality corresponding to the corresponding disconnection detection sub-model.

In the embodiment of the disclosure, under the condition that the probability density of the historical disconnection value presents quasi-normal distribution with different degrees of skewness, the electronic equipment firstly enables the probability density distribution diagram to conform to the normal distribution in a data transformation mode, and then determines the threshold range of the characterization abnormality corresponding to the disconnection detection sub-model of the sensor based on the data conforming to the normal distribution, so that the deviation of the threshold range of the characterization abnormality caused when the disconnection value presents non-standard normal distribution can be reduced to a certain extent. The detection precision of the sensor disconnection detection sub-model is improved.

It should be noted that there are many types of pressure sensors on a railway vehicle, such as a brake cylinder pressure sensor, a total wind pressure sensor, and a parking brake pressure sensor, where the above different pressure sensors can perform pressure detection and fault early warning on the railway vehicle from multiple dimensions. The working principle of various pressure sensors and corresponding sensing data thresholds under normal working conditions may be different.

In this regard, in the embodiment of the present disclosure, the electronic device may set, for each pressure sensor, a threshold range corresponding to the corresponding abnormality detection model and representing an abnormality, and process, using the abnormality detection model, sensor data corresponding to the pressure sensor, to obtain an abnormality detection result of the abnormality detection model for each pressure sensor. The electronic equipment can acquire historical sensing data of various pressure sensors, and calculate the threshold range corresponding to various sensors and representing the abnormality after classifying the acquired sensing data.

In the embodiment of the disclosure, the electronic device performs anomaly detection on each pressure sensor, can set a corresponding threshold range representing anomaly and anomaly detection rules on each pressure sensor, performs fitting analysis according to anomaly detection results of each pressure sensor, further tightens the discrimination scale of anomaly detection, and is beneficial to improving the accuracy of a sensor anomaly detection model and further improving the accuracy of fault early warning.

Fig. 11 is an overall flowchart of sensor anomaly detection provided in an embodiment of the disclosure, as shown in fig. 11, a rail vehicle fault prediction and early warning system obtains sensing data (including pressure data of various pressure sensors and fault data of various pressure sensors) from a rail vehicle control management system, performs anomaly detection on the sensing data by using an anomaly detection model (including anomaly detection sub-models such as sensor connection disconnection anomaly detection, sensor extremum anomaly detection, sensor outlier anomaly detection, sensor numerical jump anomaly detection, etc.), obtains an anomaly detection result corresponding to the anomaly detection model, determines whether the sensor is abnormal according to the occurrence frequency of the anomaly characterization anomaly of the anomaly detection result, and when determining that the sensor is abnormal, sends early warning information to the rail vehicle control management system. The rail vehicle control system can be a control system of the rail vehicle, and the rail vehicle fault prediction and early warning system can be electronic equipment which is in communication connection with the rail vehicle.

It should be noted that, with the improvement of the automation level of the subway, a significant part of the subway lines in the whole country realize unattended full-automatic operation, while the railway vehicle control management system (Train Control and Management System, TCMS) is a brain for running the full-automatic operation vehicle, and is also the only interface for data communication between the vehicle and the signal railway vehicle automatic control (Automatic Train Control, ATC) for collecting and controlling information such as traction assistance, braking, vehicle doors, air conditioning and the like for running the vehicle, and the operation data generated by the braking system can be collected in the TCMS system.

The fault prediction and health management (Prognostics and Health Management, PHM) technology is generally applied to a rail vehicle fault prediction and early warning system, various data information of the system can be acquired by using as few sensors as possible, the health state of the system can be estimated by means of an intelligent reasoning algorithm, and the fault of the system can be predicted before the system fault occurs. The method can realize the transition from the traditional maintenance to the intelligent maintenance, and advances the process of replacing the post-maintenance and preventive maintenance by the state maintenance.

As shown in fig. 11, the rail vehicle early warning method includes the following steps:

S1101, system initialization.

In this embodiment, when the system is initialized, the system includes configuration parameters, json file is configured, and the file includes data such as a threshold range of the characteristic abnormality that needs to be used by each anomaly detection sub-model.

S1102, starting an anti-interference module.

In the embodiment, the rail vehicle fault prediction and early warning system starts the anti-interference module, so that the influence of environmental factors is reduced, the system can work stably, and the functions of pressure sensing data acquisition, anomaly detection, fault judgment, early warning information transmission and the like are executed smoothly.

S1103, acquiring sensing data and judging the effectiveness.

In the embodiment, the rail vehicle fault prediction and early warning system acquires pressure sensing data through the data processing layer and judges the effectiveness of the data. The data processing layer can acquire pressure sensing data of various pressure sensors within a preset duration from a platform of the railway vehicle control management system, for example, can acquire sensing data of pressure sensors such as X-frame brake cylinder pressure of an X-frame brake cylinder; and fault data such as faults of the X-axis brake cylinder X pressure sensor of the X-axis frame X of the X-vehicle can also be obtained.

In addition, in this embodiment, the brake operation data conditions are adaptively set, particularly for the performance of the TCMS system of the railway vehicle, including the signal frequency of the adaptive TCMS system being 500ms, and the signal accuracy of the adaptive TCMS brake cylinder pressure, etc. being 0.1kPa.

S1104, inquiring the index of the sensing variable and the fault self-checking variable.

In the embodiment, the rail vehicle fault prediction and early warning system queries the sensing variable and the fault self-checking variable index through the data processing layer so as to classify the acquired pressure sensing data, and if the query result represents the pressure sensing data, the step S1105 is executed so as to facilitate the anomaly detection of each anomaly detection sub-model sent to the state monitoring layer after the further processing of the sensing data; if the query result represents the fault data of the pressure sensor, step S1112 is executed, so as to facilitate early warning according to the fault data.

S1105, sensing variable slice.

In this embodiment, the data volume acquired by the fault prediction and early warning system is relatively large, so that in order to reduce the calculation volume of the anomaly model and process the sensing data in time, the sensing variable is usually sliced, for example, 2000 sensing data can be used as one data packet. The sensor variable after slicing can be directly used for abnormality detection, for example:

Step S1106 is executed to perform sensor disconnection abnormality detection.

In this embodiment, the state monitoring layer may perform sensor disconnection anomaly detection using the sliced sensor variable and the sensor disconnection detection submodel in the embodiments of the present disclosure. For example, a sensor variable of 1 day is divided into 72 data packets, and the sensor disconnection detection sub-model performs abnormality detection for each data packet and outputs an abnormality result. The anomaly detection result can be represented by normal or anomaly, or can be represented by an anomaly rate corresponding to the data packet. In addition, the probability density distribution map of the off-abnormality rate obtained by the sensor off-abnormality detection is likely to exhibit a certain degree of left bias, so that the off-abnormality rate can be corrected by using Box-Cox transformation in this embodiment to obtain the probability density distribution map of the off-abnormality rate that tends to a standard normal distribution.

Step S1107 is executed to perform sensor extremum abnormality detection.

In this embodiment, the state monitoring layer may perform sensor extremum anomaly detection using the sliced sensor variable and the extremum detection submodel in the embodiments of the present disclosure. For example, the sensor variable for 1 day is divided into 72 data packets, and the extremum detecting sub-model performs anomaly detection for each data packet and outputs an anomaly result. The anomaly detection result can be represented by normal or anomaly, or can be represented by an anomaly rate corresponding to the data packet.

Step S1108 is performed to perform sensor outlier detection.

In this embodiment, the state monitoring layer may utilize the sliced sensing variables and the outlier detection submodel in the disclosed embodiments to perform sensor outlier detection. For example, a sensor variable of 1 day is divided into 72 data packets, and an outlier detection sub-model performs outlier detection on each data packet and outputs outlier results. The anomaly detection result can be represented by normal or anomaly, or can be represented by an anomaly rate corresponding to the data packet. In addition, the probability density distribution diagram of the outlier anomaly rate obtained by outlier detection is easy to present a left bias to a certain extent, so that the outlier anomaly rate can be corrected by using Box-Cox transformation in the embodiment, so that the probability density distribution diagram of the outlier anomaly rate which tends to be in standard normal distribution can be obtained.

Step S1109 may be performed after step S1105, and differential feature calculation.

In this embodiment, the data processing layer may perform differential value calculation on the adjacent sensing variables, so as to facilitate detection of sensor value jump anomalies by using the differential values.

Step S1110 is performed to disconnect the data filtering for a long time.

In this embodiment, the data is processed to calculate the time interval between adjacent sensing variables, and if the time interval is greater than 10 seconds, it is determined that the sensor is disconnected for a long time, and such data is generally determined as invalid data and is easy to interfere with the accuracy of detecting the numerical jump abnormality of the sensor, so that it is necessary to filter such data to obtain a filtered differential value.

Step S1111 is executed to perform sensor value jump abnormality detection.

In this embodiment, the state monitoring layer may perform sensor value transition anomaly detection using the filtered differential value and the transition detection submodel in the embodiments of the present disclosure. For example, the sensor variable for 1 day is divided into 72 data packets, and the transition detection sub-model performs anomaly detection for each data packet and outputs an anomaly result. The anomaly detection result can be represented by normal or anomaly, or can be represented by an anomaly rate corresponding to the data packet.

Step S1112 is performed to index the fault self-checking variable.

In this embodiment, the state monitoring layer may execute step S1113 to perform anomaly detection on the sensor self-checking signal according to the acquired fault information and the self-checking variable index table, so as to determine anomaly information corresponding to the fault information of the sensor.

Step S1114 is executed to calculate the anomaly rates of the plurality of sub-models.

In the embodiment, the state monitoring layer sends the 1-day abnormal detection results accumulated by the abnormal detection submodels to the health evaluation layer (including a sensor connection disconnection abnormal detection result, a sensor extremum abnormal detection result, a sensor outlier detection result and a sensor vertical jump abnormal detection result), so that the health evaluation layer can conveniently determine the corresponding abnormal rate according to the abnormal detection results within 1 day of each abnormal submodel. For example, the monitoring and evaluating layer may calculate the anomaly rate within 1 day by using the anomaly rates corresponding to the plurality of data packets of each anomaly sub-model, and store the anomaly rates of each anomaly sub-model into the historical anomaly rate. In addition, the sensor self-checking signal abnormal detection result does not need to calculate an abnormal rate, and the abnormal detection result can be directly sent to the fault early-warning layer, so that the fault early-warning layer sends early-warning information according to the detection result.

Step S1115 is executed to determine the mean-standard deviation anomaly rate for the plurality of sub-models.

In this embodiment, analysis may be generally performed based on data of the past year, and anomaly probability distribution statistics may be performed for the anomaly detection result of each sub-model, so as to determine an anomaly rate average value, and data outside a 3-time standard deviation range of the anomaly rate average value may be used as an anomaly rate discrimination threshold, and if the anomaly rate of the current day is within the anomaly rate discrimination threshold range, it is determined that a corresponding anomaly occurs in the sensor of the current day; if the abnormality rate of the current day is not within the abnormality rate judging threshold range, determining that the sensor has no corresponding abnormality in the current day.

In addition, the abnormality rate determination threshold may be calculated and updated periodically based on recent history data, and the updating may be performed by configuring the obtained abnormality rate determination threshold by a configuration parameter json.

Step S1116 is performed to determine whether the abnormality rate for 10 days exceeds the set threshold for 2 days or more? If yes, the fault early-warning layer sends abnormal early-warning information of the sensor; if not, the process continues to step S1116 until the data determination is completed for 10 days, and the operation is ended. In addition, the fault early warning layer can also send abnormal early warning information of the sensor directly according to the abnormal detection result of the sensor self-detection signal.

In addition, threshold ranges corresponding to the anomaly detection sub-models and used for characterizing anomalies

In the embodiment of the disclosure, the electronic device firstly performs abnormality detection on the sensing data within a preset time period by using each abnormality detection sub-model to obtain an abnormality detection result, then determines the occurrence frequency of the sensor abnormality within a preset period according to the abnormality detection result, judges whether the sensor is abnormal according to the abnormality frequency, and can output specific early warning information aiming at different abnormality detection sub-models when the sensor is abnormal. On the one hand, the electronic equipment determines the occurrence frequency of various anomalies based on the multiple anomaly detection results of the anomaly detection sub-models, determines whether the anomalies are actually present or not and gives an early warning based on the occurrence frequency of the various anomalies, and can reduce misjudgment caused by unstable sensing data but not anomalies or misjudgment caused by instability of the anomaly detection models, so that the accuracy of sensor anomaly judgment can be improved to improve the accuracy of early warning; on the other hand, the electronic equipment outputs early warning information when the sensor is abnormal, so that maintenance personnel of the railway vehicle can carry out corresponding overhaul based on the output early warning information, and the running stability of the railway vehicle is improved.

Fig. 12 is a schematic diagram of a rail vehicle early warning device according to an embodiment of the disclosure, and as shown in fig. 12, a rail vehicle early warning device 600 includes:

a first obtaining module 601, configured to obtain sensing data of a sensor of the rail vehicle within a preset duration;

the detection module 602 is configured to perform anomaly detection on the sensing data within the preset duration by using a preset anomaly detection model, so as to obtain an anomaly detection result corresponding to the anomaly detection model;

a first determining module 603, configured to determine whether the sensor is abnormal according to the occurrence frequency of the abnormality represented by the abnormality detection result;

and the output module 604 is used for responding to the abnormality of the sensor and outputting early warning information.

In some embodiments, the rail vehicle warning device 600 further comprises:

the detection module 602 is further configured to perform anomaly detection on the sensing data corresponding to the multiple time windows by using the anomaly detection model, so as to obtain an anomaly detection result corresponding to the anomaly detection model.

In some embodiments, the rail vehicle warning device 600 further comprises:

the detection module 602 is further configured to determine an anomaly detection result corresponding to the anomaly detection model according to the sensing data corresponding to the plurality of time windows and the threshold range representing the anomaly corresponding to the anomaly detection model.

In some embodiments, the detection module 602 includes a sensor disconnect detection sub-module;

In some embodiments, the detection module 602 further includes an extremum detection sub-module;

In some embodiments, the detection module 602 further comprises an outlier detection sub-module;

In some embodiments, the detection module 602 further includes a transition detection sub-module;

In some embodiments, the rail vehicle warning device 600 further comprises:

the first determining module 603 is further configured to determine that the sensor is abnormal in response to the occurrence frequency of the abnormality corresponding to the abnormality detection model being within a preset abnormality rate threshold range.

the detection module 602 is further configured to process, for each pressure sensor, sensing data corresponding to the pressure sensor by using the anomaly detection model, to obtain an anomaly detection result of the anomaly detection model for each pressure sensor.

Fig. 13 is a schematic diagram of a hardware entity of an electronic device according to an embodiment of the disclosure, as shown in fig. 13, the hardware entity of the electronic device 800 includes: a processor 801, a communication interface 802, and a memory 803, wherein: the processor 801 generally controls the overall operation of the electronic device 800. The communication interface 802 may enable the electronic device to communicate with other terminals or servers over a network.

The memory 803 is configured to store instructions and applications executable by the processor 801, and may also cache data (e.g., image data, audio data, voice communication data, and video communication data) to be processed or processed by various modules in the processor 801 as well as the electronic device 800, which may be implemented by a FLASH memory (FLASH) or a random access memory (Random Access Memory, RAM). Data may be transferred between processor 801, communication interface 802, and memory 803 via bus 804. Wherein the processor 801 is configured to perform some or all of the steps of the above method.

Accordingly, embodiments of the present disclosure provide a computer readable storage medium having stored thereon a computer program which, when executed by a processor, performs some or all of the steps of the above-described method.

It should be noted here that: the description of the storage medium and apparatus embodiments above is similar to that of the method embodiments described above, with similar benefits as the method embodiments. For technical details not disclosed in the embodiments of the storage medium and apparatus of the present disclosure, please refer to the description of the embodiments of the method of the present disclosure for understanding.

It should be appreciated that reference throughout this specification to "one embodiment" or "an embodiment" means that a particular feature, structure or characteristic described in connection with the embodiment is included in at least one embodiment of the present disclosure. Thus, the appearances of the phrases "in one embodiment" or "in an embodiment" in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. It should be understood that, in various embodiments of the present disclosure, the sequence numbers of the foregoing processes do not mean the order of execution, and the order of execution of the processes should be determined by their functions and internal logic, and should not constitute any limitation on the implementation of the embodiments of the present disclosure. The foregoing embodiment numbers of the present disclosure are merely for description and do not represent advantages or disadvantages of the embodiments.

It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

In the several embodiments provided in the present disclosure, it should be understood that the disclosed apparatus and method may be implemented in other ways. The above described device embodiments are only illustrative, e.g. the division of the units is only one logical function division, and there may be other divisions in practice, such as: multiple units or components may be combined or may be integrated into another system, or some features may be omitted, or not performed. In addition, the various components shown or discussed may be coupled or directly coupled or communicatively coupled to each other via some interface, whether indirectly coupled or communicatively coupled to devices or units, whether electrically, mechanically, or otherwise.

The units described as separate members may or may not be physically separate, and members displayed as units may or may not be physical units; can be located in one place or distributed to a plurality of network units; some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in each embodiment of the present disclosure may be integrated in one processing unit, or each unit may be separately used as one unit, or two or more units may be integrated in one unit; the integrated units may be implemented in hardware or in hardware plus software functional units.

Those of ordinary skill in the art will appreciate that: all or part of the steps for implementing the above method embodiments may be implemented by hardware related to program instructions, and the foregoing program may be stored in a computer readable storage medium, where the program, when executed, performs steps including the above method embodiments; and the aforementioned storage medium includes: a mobile storage device, a Read Only Memory (ROM), a magnetic disk or an optical disk, or the like, which can store program codes.

Alternatively, the above-described integrated units of the present disclosure may be stored in a computer-readable storage medium if implemented in the form of software functional modules and sold or used as separate products. Based on such understanding, the technical solutions of the present disclosure may be embodied essentially or in part in a form of a software product stored in a storage medium, including several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the methods described in the embodiments of the present disclosure. And the aforementioned storage medium includes: various media capable of storing program codes, such as a removable storage device, a ROM, a magnetic disk, or an optical disk.

The foregoing is merely an embodiment of the present disclosure, but the protection scope of the present disclosure is not limited thereto, and any person skilled in the art can easily think about the changes or substitutions within the technical scope of the present disclosure, and should be covered by the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure shall be subject to the protection scope of the claims.

Claims

1. A rail vehicle warning method, the method comprising:

and outputting early warning information in response to the sensor abnormality.

2. The method according to claim 1, wherein the method further comprises:

3. The method according to claim 2, wherein the method further comprises:

Acquiring historical sensing data of the sensor;

4. A method according to claim 3, wherein the anomaly detection model comprises a sensor off-detection sub-model;

5. A method according to claim 3, wherein the anomaly detection model further comprises an extremum detecting sub-model;

6. The method of claim 3, wherein the anomaly detection model further comprises an outlier detection sub-model;

7. The method of claim 3, wherein the anomaly detection model further comprises a transition detection sub-model;

8. The method according to claim 1, wherein the method further comprises:

9. The method according to claim 4, wherein the method further comprises:

10. The method of claim 1, wherein the sensors of the rail vehicle comprise a plurality of pressure sensors;

11. A rail vehicle warning device, the device comprising:

12. An electronic device comprising a memory and a processor, the memory storing a computer program executable on the processor, characterized in that the processor implements the steps of the method of any one of claims 1 to 10 when the program is executed.

13. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 10.