KR102527319B1 - System and method for machine learning prognostics and health management(phm) based on feature vector data of rolling stock parts - Google Patents

Abstract

A machine learning failure diagnosis system and method based on feature vector information of parts of a railway vehicle and an operating environment are disclosed. A method for diagnosing a vehicle failure includes generating component feature vector information using state sensor information received from state sensors installed on vehicle parts or tracks; generating operation feature vector information using environmental sensor information received from an environment sensor installed in an operating section of a vehicle including the track and operating performance information of the vehicle; determining priorities for fault diagnosis of the component feature vector information and the operating feature vector information using a verification algorithm; setting a failure diagnosis environment for the parts of the vehicle by using the part feature vector information, the operating feature vector information, and priorities for fault diagnosis of the part feature vector information and the operating feature vector information; generating a training information set and a test information set using the component feature vector information and the operating feature vector information; and outputting a failure diagnosis result by applying the training information set and the test information set to a machine learning model for failure diagnosis.

Description

Machine learning fault diagnosis system and method based on vector information on parts and operating environment characteristics of railway vehicles

The present invention relates to a system and method for diagnosing a vehicle failure through machine learning using feature vector information of parts of a railroad vehicle and an operating environment.

Maintenance work for conventional high-speed rail vehicles and freight vehicles is carried out as a preventive measure by the judgment of maintenance workers on the site at the time of storage at the repair shop, and maintenance history management is also used independently as information.

Conventional vehicle fault diagnosis device (No. 2011-0053906) is in the state where the interface of the vehicle electric field required to convert a vehicle using gasoline or diesel into a vehicle using LPG is in the equipped electronic control unit , The electronic control unit for the LPG monitors the vehicle operation information input to the electronic control unit that is being equipped to determine whether the modified vehicle is out of order. A conventional fault diagnosis device for a vehicle identifies a fault condition and informs a manager of the fault condition in real time or transmits status information such as temperature obtained from a sensor for fault diagnosis, and indicates the current state before a fault occurs. However, there is a limitation in not being able to provide failure diagnosis results for target parts requiring repair.

In addition, the standard for the diagnostic value used in the field of maintenance of railway vehicles and facilities is determined by the experience of the operator, and there is a problem in that the linkage between the manufacturer's useful life information and the operator's history management information is low.

Therefore, there is a demand for a method and apparatus capable of verifying the results of selecting feature vector information from sensors attached to high-speed trains and freight cars from a failure diagnosis point of view.

The present invention can provide an apparatus and method for minimizing maintenance work time due to periodic maintenance by improving the accuracy of fault diagnosis for parts of a railway vehicle.

In addition, the present invention can provide a device and method capable of safely operating a railway vehicle by determining a maintenance point for a part by predicting the remaining life of the part using a machine learning model.

A method for diagnosing a vehicle failure according to an embodiment of the present invention includes generating component feature vector information using state sensor information received from state sensors installed on parts of a vehicle or a track; generating operation feature vector information using environmental sensor information received from an environment sensor installed in an operating section of a vehicle including the track and operating performance information of the vehicle; determining priorities for fault diagnosis of the component feature vector information and the operating feature vector information using a verification algorithm; setting a failure diagnosis environment for the parts of the vehicle by using the part feature vector information, the operating feature vector information, and priorities for fault diagnosis of the part feature vector information and the operating feature vector information; generating a training information set and a test information set using the component feature vector information and the operating feature vector information; and outputting a failure diagnosis result by applying the training information set and the test information set to a machine learning model for failure diagnosis.

According to one embodiment of the present invention, maintenance work time due to periodic maintenance can be minimized by improving the accuracy of fault diagnosis for parts of a railway vehicle.

In addition, according to one embodiment of the present invention, by predicting the remaining life of parts using a machine learning model, it is possible to safely operate a railway vehicle by determining a maintenance point for parts.

1 is a diagram showing a failure diagnosis system according to an embodiment of the present invention.
2 is an example of a state sensor according to an embodiment of the present invention.
3 is an example of an environmental sensor according to an embodiment of the present invention.
4 is a diagram illustrating a failure diagnosis server according to an embodiment of the present invention.
5 is an example of a process of generating feature vector information according to an embodiment of the present invention.
6 is an example of a process of managing feature vector information according to an embodiment of the present invention.
7 is an example of a process of diagnosing a failure and predicting a remaining life according to an embodiment of the present invention.
8 is a flowchart illustrating a failure diagnosis method according to an embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. A failure diagnosis method according to an embodiment of the present invention may be performed by a failure diagnosis system.

1 is a diagram showing a failure diagnosis system according to an embodiment of the present invention.

As shown in FIG. 1 , a failure diagnosis system according to an embodiment of the present invention may include a failure diagnosis server 110 , a state sensor 120 , and an environment sensor 130 .

The state sensor 120 may be attached to main parts of a passenger railway vehicle and a logistics railway vehicle, or may be installed on a track on which the vehicle moves. For example, the main parts of a railway vehicle may be a wheel, a motor block, a switchboard, a bogie, a bearing, and the like. In addition, the condition sensor 120 may acquire condition sensor information such as vibration, temperature, ultrasonic waves, and load cells for each part of the vehicle and transmit the information to the fault diagnosis server 110 .

The environmental sensor 130 may be installed on a track along which a passenger rail vehicle and a logistics rail vehicle move, and measure temperature, humidity, snow, rain, and cooling conditions for a track condition and an operating section of the railroad vehicle. In addition, the environmental sensor 130 may transmit environmental sensor information generated according to the measurement result to the failure diagnosis server 110 .

The fault diagnosis server 110 may generate component feature vector information and operation feature vector information from the state sensor information received from the state sensor 120 and the environment sensor information received from the environment sensor 130 , respectively. In addition, the failure diagnosis server 110 may improve the accuracy and efficiency of maintenance work by performing failure diagnosis by applying the part feature vector information and the operation feature vector information to the machine learning model. The specific configuration and operation of the failure diagnosis server 110 will be described in detail with reference to FIGS. 4 to 7 below.

The fault diagnosis system according to an embodiment of the present invention can improve fault diagnosis accuracy for parts of a railway vehicle. In addition, the failure diagnosis system according to an embodiment of the present invention can minimize maintenance work time due to periodic maintenance because the failure diagnosis accuracy of parts of a railway vehicle is improved compared to the conventional failure diagnosis system.

In addition, the fault diagnosis system according to an embodiment of the present invention predicts the remaining life of parts using a machine learning model, thereby determining a maintenance point for parts, thereby enabling safe operation of the railway vehicle. In addition, since the failure diagnosis system according to an embodiment of the present invention can predict the remaining life of parts, parts can be replaced and repaired before failure occurs. Accordingly, parts supply and demand costs and parts repair costs can be reduced.

2 is an example of a state sensor according to an embodiment of the present invention.

The state sensor 120 may be installed on each of the parts of the vehicle to sense the parts. In addition, the condition sensor 120 may transmit condition sensor information generated by sensing parts of the vehicle to the communicator 210 installed in each vehicle in a wired or wireless manner.

At this time, the communicator 210 may transmit the state sensor information received from the state sensors 120 to the wireless repeater 220 . At this time, the wireless repeater 220 may collect state sensor information received from the communicator 210 and the communicators 230 installed in other vehicles and transmit the collected state sensor information to the fault diagnosis server 110 .

3 is an example of an environmental sensor according to an embodiment of the present invention.

As shown in FIG. 3, the environmental sensor 310 may be installed within a certain distance from the track, or may measure temperature, humidity, snow, rain, and cooling conditions for the track condition and the operating section of the railroad vehicle at the lower part of the track. .

In addition, the repeater 320 may collect environmental sensor information generated by the environmental sensors 310 according to measurement results and transmit the collected information to the failure diagnosis server 110 .

4 is a diagram illustrating a failure diagnosis server according to an embodiment of the present invention.

As shown in FIG. 4, the failure diagnosis server 110 includes a component feature vector information generating device 410, an operation feature vector information generating device 420, an operation plan and performance management device 430, and a feature vector information management device ( 440), and a failure diagnosis device 450. At this time, the component feature vector information generating device 410, the operation feature vector information generating device 420, the operation plan and performance management device 430, the feature vector information management device 440, and the failure diagnosis device 450 are mutually connected. It may be each module for executing a program included in another process or one process.

The component feature vector information generating device 410 may generate component feature vector information using state sensor information received from state sensors 120 installed on vehicle parts or tracks.

The operating feature vector information generating device 420 may generate operating feature vector information by using the environmental sensor information received from the environmental sensor 130 installed in the operating section of the vehicle including the track and the operating performance information of the vehicle.

The operation plan and performance management device 430 may store and manage operation plan information and operation performance information for each railroad vehicle. For example, the operation plan and performance management device 430 may be a railway vehicle operation plan and performance management system. Also, the operation plan and performance management unit 430 may transmit operation plan information and operation performance information for each railroad vehicle to the operation feature vector information generation unit 420 according to the request of the operation feature vector information generation unit 420 .

For example, the operation performance information for each railroad vehicle may include at least one of a moving distance, a required time, and a moving track section of an operating section. In addition, the operation plan information for each railroad vehicle may include at least one of the rotation angle and distance according to the number of curve sections and the movement direction of the curve section, the number of brakes used, the movement speed at the time of brake use, and the time required to reach a stop or deceleration section. .

The feature vector information management device 440 may distribute, store and manage the part feature vector information generated by the component feature vector information generating device 410 and the operating feature vector information generated by the operating feature vector information generating device 420. .

At this time, the feature vector information management device 440 may determine priorities for fault diagnosis of component feature vector information and operating feature vector information by using a verification algorithm. Next, the feature vector information management device 440 may set a failure diagnosis environment for parts of a vehicle by using parts feature vector information, operation feature vector information, and priorities for fault diagnosis of parts feature vector information and operation feature vector information. . Next, the feature vector information management device 440 may generate a training information set and a test information set by using the part feature vector information and the operational feature vector information in a vehicle component failure diagnosis environment.

The failure diagnosis apparatus 450 may output a failure diagnosis result by applying the training information set and the test information set to a machine learning model for failure diagnosis.

5 is an example of a process of generating feature vector information according to an embodiment of the present invention.

The state sensor information receiving unit 510 of the component feature vector information generating device 410 may receive state sensor information from the state sensor 120 .

In this case, the part feature vector information generation unit 511 of the part feature vector information generating device 410 may generate part feature vector information by extracting feature values necessary for fault diagnosis from state sensor information. For example, the part feature vector information may include at least one of the severity, failure probability, and remaining life of parts based on a value generated based on a threshold value that may be a failure factor, the number of measurements, an accumulated value, and a generated result value. there is.

At this time, the part feature vector information generation unit 511 may receive the refinement details of the part feature vector information from the part feature vector information refinement details extractor 630 of the feature vector information management device 440 .

Also, the information transmitter 513 of the part feature vector information generation device 410 may transmit the generated part feature vector information to the feature information storage/manager 610 of the feature vector information management device 440 . At this time, the time at which the information transmitter 513 transmits the part feature vector information to the feature information storage/manager 610 may be during a spare time of the operation start time or when the operation is completed. In addition, the information transmitter 513 may store railway vehicle operation performance information from the operation plan and performance management device 430 in the operation plan and performance management device 430 during a spare time of operation start time or when operation is completed.

The environment sensor information receiving unit 521 of the operation feature vector information generating device 420 may receive environment sensor information from the environment sensor 130 .

The vehicle operation performance information reception unit 522 of the operation feature vector information generation device 420 is configured to operate plan for each rolling stock in the plan performance management unit 531 of the operation plan and performance management unit 430 and the rolling stock operation environment management unit 532. Information and operational performance information can be inquired. In addition, the vehicle operation performance information receiving unit 522 may receive operation plan information and operation performance information for each railroad vehicle inquired from the operation plan and performance management device 430 .

The operation feature vector information generation unit 523 of the operation feature vector information generation device 420 includes the environment sensor information received by the environment sensor information receiver 521 and the operation plan for each railroad vehicle received by the vehicle operation performance information receiver 522. Operating feature vector information may be generated using at least one of information and operating performance information. In this case, the operating feature vector information generating unit 523 may generate operating feature vector information by classifying environmental sensor information according to time zones in which the environmental sensor information is generated. For example, the operation feature vector information includes line state values such as temperature and humidity received from the environmental sensor 130 and the circumference of a curved section during a moving section according to operation, deceleration, brake pressure for stopping, stopping distance, and moving distance. It can be generated by assigning weight values that affect failures to reference values such as , time, and speed.

At this time, the operation feature vector information that can act as a failure factor of vehicle parts is an input for failure diagnosis for parts attached to the outside of the railway vehicle, such as wheels, bogies, and brake pads, which have a lot of correlation with respect to parts that have a lot of influence on each part. Feature vector information can be provided so that failure diagnosis procedures can be performed by using it as information.

Also, the information transmitter 524 of the operating feature vector information generation device 420 may transmit the generated operating feature vector information to the feature information storage/manager 610 of the feature vector information management device 440 . In this case, the time at which the information transmitter 524 transmits the operation feature vector information to the feature information storage/manager 610 may be during a spare time of the operation start time or when the operation is completed.

6 is an example of a process of managing feature vector information according to an embodiment of the present invention.

The feature information storage/management unit 610 of the feature vector information management device 440 may receive part feature vector information and operation feature vector information from the information transmitter 513 and the information transmitter 524, respectively. The feature information storage/management unit 610 may create an index on the received component feature vector information and operation feature vector information, distribute and store them in a plurality of servers or terminals, and manage them.

The feature government analysis unit 620 of the feature vector information management device 440 determines the priority for fault diagnosis of the part feature vector information and operation feature vector information distributed and stored by the feature information storage/management unit 610 using a verification algorithm. can For example, the verification algorithm may be a Chi-squared Test algorithm.

The part feature vector information refinement details extraction unit 630 of the feature vector information management device 440 receives details of part feature vector information that will not be used for machine learning among part feature vector information from the feature government analysis unit 620. can And, the part feature vector information refinement details extractor 630 feeds back the received refinement details of the part feature vector information to the part feature vector information generator 511, so that the part feature vector information generator 511 can perform machine learning It is possible to avoid generating component feature vector information that will not be utilized.

That is, the part feature vector information refinement details extractor 630 feeds back part feature vector information that will not be used for machine learning to the part feature vector information generator 511, so that the part feature vector information generator 511 can perform machine learning It is possible to generate part feature vector information optimized for .

The operating feature vector information refinement details extraction unit 640 of the feature vector information management device 440 receives details of operating feature vector information that will not be used for machine learning among the operating feature vector information from the feature government analysis unit 620. can In addition, the operating feature vector information refinement details extractor 640 feeds back the refinement details of the received operating feature vector information to the operating feature vector information generator 523, so that the operating feature vector information generator 523 performs machine learning. It is possible to avoid generating operational feature vector information that will not be utilized.

That is, the operating feature vector information refinement details extractor 640 feeds back operational feature vector information that will not be used for machine learning to the operating feature vector information generator 523 so that the operating feature vector information generator 523 can perform machine learning It is possible to generate operating feature vector information optimized for .

The operation feature vector information refinement details extraction unit 640 of the feature vector information management device 440 extracts the refinement details of the operation feature vector information from the feature government analysis unit 620 and transmits the operation feature vector information generation unit 523. can

The failure diagnosis environment setting unit 650 of the feature vector information management device 440 distributes and stores the part feature vector information stored by the feature information storage/manager 610, the operating feature vector information, and the part determined by the feature government analysis unit 620. A failure diagnosis environment for parts of a vehicle may be set using the feature vector information and the priorities for failure diagnosis of the operation feature vector information.

Specifically, the failure diagnosis environment setting unit 650 priorities for failure diagnosis of component feature vector information and operation feature vector information, operation feature vector information, or component feature vector input from a manager who wants to perform failure diagnosis in a manual mode. Part feature vector information and operating feature vector information to be used in the training information set and the test information set may be determined using the information and operating feature vector information.

In addition, the failure diagnosis environment setting unit 650 includes a generation ratio of part feature vector information and operation feature vector information to be used in the training information set and the test information set, the period during which the feature vector information is generated, and machine learning for the failure diagnosis target part. You can choose an algorithm.

The training information set generation unit 660 of the feature vector information management device 440 receives the part feature vector information and operation feature vector information selected by the failure diagnosis environment setting unit 650 from the feature information storage/management unit 610 for training. A set of information can be created.

The test information set generation unit 670 of the feature vector information management device 440 receives the part feature vector information and operation feature vector information selected by the failure diagnosis environment setting unit 650 from the feature information storage/management unit 610 and tests them. A set of information can be created.

7 is an example of a process of diagnosing a failure and predicting a remaining life according to an embodiment of the present invention.

The machine learning execution unit 720 of the failure diagnosis device 450 uses the training information set 701 received from the training information set generator 660 and the test information set 702 received from the test information set generator 670. It can be stored in memory by applying ETL (Extraction, Transformation, Loading) logic according to the machine learning model. When machine learning is completed, the machine learning performer 720 may delete information stored in the memory before applying the machine learning model to the next part.

At this time, the machine learning model used by the machine learning performer 720 may be a machine learning model for fault diagnosis for each railway vehicle component set in the fault diagnosis environment setting unit 650 or a machine learning model selected by a manager.

In addition, the machine learning performer 720 may apply a machine learning model to information stored in the memory and output failure diagnosis result information 703 . At this time, the machine learning performer 720 may classify the failure diagnosis result information 703 into classes (ML1, ML2, ..., MLn) for each of 2 or 3 or more diagnosis levels, and output the results.

In addition, the machine learning performer 720 may output feature vector information that is inconsistent with the failure diagnosis prediction.

The learning result analyzer 730 may operate in a mode in which a manager manually selects a machine learning model. At this time, the learning result analysis unit 730 may sort the performance results of the machine learning performed by the machine learning unit 720 using a plurality of machine learning models in order of comparison result value and priority for each item, and provide the result to the manager. there is. For example, the performance result may include at least one of accuracy, sensitivity, specificity, and f-value for each machine learning model.

At this time, the learning result analysis unit 730 selects an optimal machine learning model among machine learning models for fault diagnosis for each part, function-based, probability-based, tree structure (5-5), etc. After sequentially performing the machine learning models included in the machine learning model list, a reference value for selecting a machine learning model with high accuracy may be provided.

The failure diagnosis machine learning selection unit 710 may select and register the machine learning model selected by the manager as the machine learning model to be used by the machine learning execution unit 720 according to the comparison result of the learning result analysis unit 730 .

At this time, the machine learning execution unit 720 applies ETL (extraction, transformation, loading) logic to the training information set 701 according to the machine learning model registered by the failure diagnosis machine learning selection unit 710 and performs machine learning. can do.

The failure diagnosis result information 703 may be information transmitted and provided to a failure diagnosis expert and a maintenance worker. For example, the failure diagnosis result information 703 may be a result value classified into normal, good, inspection target, failure, and the like.

The fault diagnosis result analyzer 750 may check a misclassified value among part feature vector information and operation feature vector information according to the classification result value of the fault diagnosis result information 703 . In addition, the failure diagnosis result analysis unit 750 may provide erroneously classified part feature vector information or operation feature vector information to a manager.

The result verification unit 740 inputs whether the failure diagnosis result information 703 matches the actual inspection result or maintenance work result value and the diagnosis class value from the diagnosis specialist and maintenance worker who have received the fault diagnosis result information 703 receive and register.

The feature vector information refinement details management unit 760 may create and manage feature vector information refinement details by changing a class value of feature vector information that is inconsistent with the failure diagnosis result information 703 . The feature vector information refinement details management unit 760 may use the feature vector information that is inconsistent with the failure diagnosis result information 703 as input information by changing the class value of the feature vector information that does not match the failure diagnosis result information 703. .

At this time, the feature vector information refinement details management unit 760 may distribute and store the feature vector information having the class value changed in the feature information storage/management unit 770 .

The fault diagnosis apparatus 450 can provide fault diagnosis results consistent with the judgments of experts and maintenance workers by repeatedly performing fault diagnosis through the above process.

In addition, the remaining life analysis unit 780 of the failure diagnosis device 450 may order operation feature vector information based on the current failure diagnosis result value and operating distance, feature vector information mapped based on part feature vector information, and operation performance. there is. Next, the remaining life analysis unit 780 may inquire a location range of feature vector information based on a failure diagnosis result value. Next, the remaining life analysis unit 780 may generate a value of a possible remaining movement distance.

Also, the remaining life analysis unit 780 may receive vehicle dispatch plan and operation period information 704 from the operation plan and performance management device 430 .

At this time, the remaining life analysis unit 780 is operated based on a point in time when the remaining travel distance value is smaller than the railroad vehicle allocation plan information standard travel distance value included in the vehicle allocation plan and operation period information 704 and the remaining travel distance value. You can create a possible duration value.

Also, the remaining life analysis unit 780 may output remaining life analysis result information 705 including the remaining movable distance and the operable period value according to the operable period value.

8 is a flowchart illustrating a failure diagnosis method according to an embodiment of the present invention.

In step 810, the component feature vector information generating device 410 may generate component feature vector information using state sensor information received from the state sensors 120.

In operation 820 , the operating feature vector information generating device 420 may generate operating feature vector information by using the environment sensor information received from the environment sensor 130 and the operating performance information of the vehicle.

In operation 830, the feature vector information management device 440 may determine priorities for fault diagnosis of component feature vector information and operating feature vector information by using a verification algorithm.

In step 840, the feature vector information management device 440 may set a failure diagnosis environment for parts of a vehicle by using part feature vector information, operation feature vector information, and priorities for fault diagnosis.

In step 850, the feature vector information management device 440 may generate a training information set and a test information set using part feature vector information and operational feature vector information according to the failure diagnosis environment.

In step 860, the failure diagnosis apparatus 450 may output a failure diagnosis result by applying the training information set and the test information set to a machine learning model for failure diagnosis.

The present invention can minimize maintenance work time due to periodic maintenance by improving the accuracy of fault diagnosis for parts of a railway vehicle.

In addition, the present invention predicts the remaining life of parts using a machine learning model, thereby determining a maintenance point for parts, thereby enabling safe operation of the railway vehicle.

Meanwhile, the method according to the present invention is written as a program that can be executed on a computer and can be implemented in various recording media such as magnetic storage media, optical reading media, and digital storage media.

Implementations of the various techniques described herein may be implemented in digital electronic circuitry, or in computer hardware, firmware, software, or combinations thereof. Implementations may be a computer program product, i.e. an information carrier, e.g. a machine-readable storage, for processing by, or for controlling the operation of, an information processing device, e.g. a programmable processor, computer, or plurality of computers. It can be implemented as a computer program tangibly embodied in a device (computer readable medium) or a radio signal. A computer program, such as the computer program(s) described above, may be written in any form of programming language, including compiled or interpreted languages, and may be written as a stand-alone program or in a module, component, subroutine, or computing environment. It can be deployed in any form, including as other units suitable for the use of. A computer program can be deployed to be processed on one computer or multiple computers at one site or distributed across multiple sites and interconnected by a communication network.

Processors suitable for processing a computer program include, by way of example, both general and special purpose microprocessors, and any one or more processors of any kind of digital computer. Generally, a processor will receive instructions and information from read only memory or random access memory or both. Elements of a computer may include at least one processor that executes instructions and one or more memory devices that store instructions and information. Generally, a computer may include, receive information from, send information to, or both, one or more mass storage devices that store information, such as magnetic, magneto-optical disks, or optical disks. It can also be combined to become. Information carriers suitable for embodying computer program instructions and information include, for example, semiconductor memory devices, for example, magnetic media such as hard disks, floppy disks and magnetic tapes, compact disk read only memory (CD-ROM) ), optical media such as DVD (Digital Video Disk), magneto-optical media such as Floptical Disk, ROM (Read Only Memory), RAM (RAM) , Random Access Memory), flash memory, EPROM (Erasable Programmable ROM), EEPROM (Electrically Erasable Programmable ROM), and the like. The processor and memory may be supplemented by, or included in, special purpose logic circuitry.

In addition, computer readable media may be any available media that can be accessed by a computer, and may include both computer storage media and transmission media.

Although this specification contains many specific implementation details, they should not be construed as limiting on the scope of any invention or what is claimed, but rather as a description of features that may be unique to a particular embodiment of a particular invention. It should be understood. Certain features that are described in this specification in the context of separate embodiments may also be implemented in combination in a single embodiment. Conversely, various features that are described in the context of a single embodiment can also be implemented in multiple embodiments individually or in any suitable subcombination. Further, while features may operate in particular combinations and are initially depicted as such claimed, one or more features from a claimed combination may in some cases be excluded from that combination, and the claimed combination is a subcombination. or sub-combination variations.

Similarly, while actions are depicted in the drawings in a particular order, it should not be construed as requiring that those actions be performed in the specific order shown or in the sequential order, or that all depicted actions must be performed to obtain desired results. In certain cases, multitasking and parallel processing can be advantageous. Further, the separation of various device components in the embodiments described above should not be understood as requiring such separation in all embodiments, and the program components and devices described may generally be integrated together into a single software product or packaged into multiple software products. You have to understand that you can.

On the other hand, the embodiments of the present invention disclosed in this specification and drawings are only presented as specific examples to aid understanding, and are not intended to limit the scope of the present invention. In addition to the embodiments disclosed herein, it is obvious to those skilled in the art that other modified examples based on the technical idea of the present invention can be implemented.

110: fault diagnosis server
120: state sensor
130: environment sensor

Claims (10)

generating component feature vector information by using state sensor information received from state sensors installed on parts of a vehicle or a track;
generating operation feature vector information using environmental sensor information received from an environment sensor installed in an operating section of a vehicle including the track and operating performance information of the vehicle;
outputting a failure diagnosis result by applying the component feature vector information and the operating feature vector information to a machine learning model for failure diagnosis;
ordering feature vector information mapped on the basis of the failure diagnosis result and operating distance, operating feature vector information, component feature vector information, and operating performance;
generating a remaining movable distance by inquiring a location range of the feature vector information based on a result of the failure diagnosis; and
Generating an operable period based on a point in time when the remaining possible movement distance is less than the movement distance based on the railroad vehicle allocation plan information included in the vehicle allocation plan and operation period information and the remaining movement distance
Vehicle fault diagnosis method comprising a. According to claim 1,
The part feature vector information,
A method for diagnosing a vehicle failure including at least one of severity, failure probability, and remaining lifespan of parts based on a value generated based on a threshold value that may be a failure factor, the number of measurements, an accumulated value, and a generated result value. According to claim 1,
The operating feature vector information is
The line condition values such as temperature and humidity received from the environmental sensor and the reference values such as the circumference of the curved section during the movement section according to the operation, deceleration, brake pressure for stopping, stopping distance, moving distance, time, speed, etc. A method for diagnosing vehicle failures created by assigning weight values that affect According to claim 1,
determining priorities for fault diagnosis of the component feature vector information and the operating feature vector information using a verification algorithm; and
Setting a failure diagnosis environment for parts of the vehicle by using the parts feature vector information, the operating feature vector information, and priorities for fault diagnosis of the part feature vector information and the operating feature vector information.
Fault diagnosis method of a vehicle further comprising a. According to claim 4,
The step of setting the failure diagnosis environment,
A method for diagnosing vehicle failures that selects the generation ratio of part feature vector information and operational feature vector information to be used in the training information set and test information set, the period during which feature vector information is generated, and the machine learning algorithm for the part subject to failure diagnosis. According to claim 1,
In the step of outputting the failure diagnosis result,
generating a training information set and a test information set using the component feature vector information and the operating feature vector information; and
Outputting a failure diagnosis result by applying a training information set and a test information set to a machine learning model for failure diagnosis
Vehicle fault diagnosis method comprising a. According to claim 1,
Checking incorrectly classified part feature vector information or operating feature vector information among the part feature vector information and the operating feature vector information according to the failure diagnosis result and providing the information to a manager
Fault diagnosis method of a vehicle further comprising a. According to claim 1,
Further comprising the step of performing machine learning using machine learning models for fault diagnosis, sorting the execution results and providing them to a manager,
As a result of the above,
At least one of accuracy, sensitivity, specificity, and f-value for each machine learning model,
The machine learning model for fault diagnosis is
A method for diagnosing a breakdown of a vehicle, which is a machine learning model selected by a manager according to the execution result from among machine learning models for failure diagnosis. delete a part feature vector information generating device for generating part feature vector information by using state sensor information received from state sensors installed on a part of a vehicle or a track;
an operating feature vector information generating device for generating operating feature vector information using environmental sensor information received from an environmental sensor installed in an operating section of a vehicle including the track and operating performance information of the vehicle; and
A fault diagnosis device that outputs a fault diagnosis result by applying the part feature vector information and the operating feature vector information to a machine learning model for fault diagnosis.
including,
The fault diagnosis device,
Ordering feature vector information mapped based on operation feature vector information, part feature vector information, and operation performance based on the failure diagnosis result and operating distance;
Based on the failure diagnosis result, a position range of the feature vector information is inquired to generate a remaining movable distance;
A vehicle fault diagnosis server that generates an operable period based on a point in time when the remaining possible travel distance is less than the standard travel distance and remaining travel distance included in the vehicle allocation plan and operation period information.

Images