JP2022034392A - Fault diagnosis device, fault diagnosis method, and learning completed model - Google Patents

Abstract

To readily perform fault diagnosis which requires no trial and error at a time of fault diagnosis even without the knowledge of fault diagnosis.SOLUTION: A fault diagnosis device includes: a learning device 2 that produces a learning completed model 3 which is employed in fault diagnosis after subjected to machine learning using as input data an ontology model 6 which represents correlations among plural components constituting an object apparatus, and a fault mechanism database 7 containing plural fault report writings 8 each containing fault information on a fault occurring in the object apparatus, that is, a fault factor, fault region, and fault event; a fault diagnosis unit 4 that inputs a fault report 9 which contains at least one of pieces of fault information to the learning completed model 3, and extracts fault report writings 8 which exhibit higher degrees of consistence with the fault report 9 as fault diagnosis from the fault mechanism database 7; and a display unit that displays the fault report writings 8 extracted by the fault diagnosis unit.SELECTED DRAWING: Figure 1

Description

The present invention relates to a failure diagnosis device for performing failure diagnosis of various devices, a failure diagnosis method, and a learned model used for failure diagnosis.

Devices such as work vehicles may fail, and when a failure occurs, it is necessary to promptly confirm the exact symptom of the failure, investigate the cause of the failure, and deal with the failure. For example, a worker who receives a failure report (warranty application, complaint, etc.) imagines the range of influence and phenomenon expected from the contents of the report while considering the structure of the work platform, and guesses the location of the failure and the cause of the failure. And perform failure diagnosis. In addition, various auxiliary tools may be used when performing such a failure diagnosis.

For example, a work machine such as a combine is equipped with various actuators and sensors, and a controller for inputting a detection signal of the sensor and controlling the operation of the actuator is mounted. The checker device shown in Patent Document 1 is connected to a controller to operate an actuator and confirm a detection signal of a sensor to perform a failure diagnosis to identify a failure location.

Further, as shown in Patent Document 2, control data is acquired from a vehicle, an operation is performed to manifest data fluctuations due to the occurrence of a mechanical or control abnormality (failure), and the abnormality is diagnosed from the calculation result. It may be done.

Further, as shown in Patent Document 3, a plurality of event information including various elements is stored in the server, elements related to a failure are input, and based on event information having a predetermined number or more of elements matching the input elements. In some cases, analysis information is created and the cause of failure is predicted from the analysis information.

Further, as shown in Patent Document 4, failure information is obtained based on the acquired current operation data and historical operation data of the railway vehicle, a maintenance solution corresponding to the failure information is acquired, and the railway is based on the maintenance solution. Vehicle maintenance may be performed.

Further, as shown in Patent Document 5, failure events that are easy for general workers to understand while considering the causal relationship of failure are organized in a tree form from MFM information, MFM incidental information, and influence ripple rules. In some cases, FTA information and FMEA information, which are the information obtained, are generated, and failure diagnosis is performed using the FTA information and FMEA information.

Further, as shown in Patent Document 6, a target ontology is selected, a target ontology data and a diagnostic ontology data in which the target ontology is associated with each other are extracted, and an FTA case having an appropriate diagnostic range and diagnostic content is created. It may be displayed and a failure diagnosis may be performed.

Japanese Unexamined Patent Publication No. 2004-199500 Japanese Unexamined Patent Publication No. 2012-198144 Japanese Unexamined Patent Publication No. 2017-117193 Special Table 2019-531523 Gazette Japanese Patent No. 3808893 Japanese Unexamined Patent Publication No. 2000-322125

However, in order to understand the structure of a device such as a work vehicle and perform a failure diagnosis in consideration of this, it is necessary to understand the structure of the device to a considerable extent or more. In addition, the functions of devices such as work vehicles have been improved, and the structure has become more complicated. Even if the auxiliary tools shown in Patent Documents 1 to 4 above are used, it is easy if the structure of the work vehicle is not understood to some extent. It is in a situation where failure diagnosis cannot be performed. Further, in the auxiliary tools shown in Patent Documents 5 and 6 above, information on a tree structure organized in consideration of the structure of a device such as a work vehicle is provided, and failure diagnosis is performed based on the information, but the structure is When it becomes complicated, even if organized information is provided, it is often not easy to make a failure diagnosis while confirming it. Therefore, there is a demand for failure diagnosis that does not require trial and error in failure diagnosis and can be easily performed without knowledge for failure diagnosis.

The failure diagnosis device according to one embodiment of the present invention is an ontology model showing the mutual relationship of a plurality of parts constituting the target device, and failure information about a failure that has occurred in the target device, such as a failure factor, a failure site, and a failure site. A learning device that generates a trained model for performing failure diagnosis by machine learning using a failure mechanism database having a plurality of failure report sentences including failure events as input data, and at least one of the failure information. A failure diagnosis unit that inputs a failure report including the above into the trained model and extracts the failure report text having a high degree of agreement with the failure report as the failure diagnosis from the failure mechanism database, and the failure. It is provided with a display unit for displaying the failure report text extracted by the diagnosis unit.

Further, the failure diagnosis method according to one embodiment of the present invention includes an ontology model showing the mutual relationship of a plurality of parts constituting the target device, and failure information about a failure that has occurred in the target device, which is a failure factor and a failure site. , And a process of generating a trained model for performing failure diagnosis by machine learning as input data with a failure mechanism database having a plurality of failure report sentences including failure events, and at least one of the failure information. A step of inputting a failure report including one into the trained model and extracting the failure report text having a high degree of agreement with the failure report by a predetermined degree or more from the failure mechanism database as the failure diagnosis. It is provided with a step of displaying the failure report text.

Further, the trained model according to one implementation of the present invention is an ontology model showing the mutual relationship of a plurality of parts constituting the target device, and failure information about a failure that has occurred in the target device, which is a failure factor and a failure site. , And a failure mechanism database having a plurality of failure report sentences including failure events are generated by machine learning as input data, and a failure report including at least one of the failure information is input as a failure diagnosis. The computer is made to function to extract the failure report text having a high degree of agreement with the failure report from the failure mechanism database.

As described above, the trained model is generated by inputting the ontology model of the target device and the failure report text regarding the past failure that occurred in the target device.

Therefore, the failure diagnosis using this trained model is a failure diagnosis considering the mutual relationship of a plurality of parts constituting the target device. As a result, the operator does not need to make trial and error in the failure diagnosis, and even if he / she is not familiar with the configuration of the target device, he / she can directly obtain the failure diagnosis result considering the configuration of the target device. can.

In addition, the failure diagnosis result can be output as a failure report text including failure information including a failure cause, a failure site, a failure event, and the like. Therefore, the failure diagnosis result is output in the form of a failure report text as a candidate for appropriate information corresponding to the failure, and the operator can easily understand the failure diagnosis result without any knowledge for failure diagnosis. , It is possible to perform a failure diagnosis with high accuracy only by judging whether or not the failure report text is appropriate as a text for a failure.

Further, the failure report text may further include a failure mode that causes the failure event as the failure information.

With such a configuration, the failure report text registered in the failure mechanism database records the failure mode that causes the failure event that occurred at the failure site. Therefore, the trained model is generated in consideration of the failure mode, and the failure mode is described in the failure report text output by the failure diagnosis. As a result, the failure diagnosis is performed in consideration of the failure mode, and more detailed failure report text is output, so that the operator can understand the failure diagnosis result more easily and perform the failure diagnosis with high accuracy. Can be done.

In addition, the description failure report text may further include a countermeasure policy for the failure.

With such a configuration, the failure report text output by the failure diagnosis describes the countermeasure policy for the failure, and the operator can easily respond to the failure such as repair or design change.

Further, the failure report may be a word associated with each of the failure information.

The failure report input at the time of failure diagnosis is a word related to the failure, and it is not necessary to have a deep understanding of the structure of the target device, and the failure diagnosis can be easily performed. Further, since the input word is associated with the failure information, the relationship with the failure actually occurring becomes clear, and accurate failure diagnosis can be expected. From the above, failure diagnosis can be easily and accurately performed.

Further, a failure report input unit for displaying the failure report input screen for individually inputting the failure information of the failure report may be provided on the display unit.

By inputting the failure report on the failure report input screen, the failure report can be easily input.

Further, in the failure diagnosis, a plurality of the failure report sentences may be extracted in descending order of the degree of coincidence with the failure report, and the extracted failure report sentences may be displayed in the order of the degree of coincidence with the failure report. ..

With such a configuration, as a result of the failure diagnosis, a plurality of failure report sentences having a high possibility of corresponding to the failure actually occurring are displayed in the order of high possibility of matching the failure, so that the number of workers is multiple. The diagnosis result can be judged by comparing the trouble report texts of the above, and the fault diagnosis can be performed more easily and accurately only by selecting the fault report text that is judged to be the most appropriate.

Further, in the failure diagnosis, the matching probability between the failure report and the failure report text may be calculated, and the extracted failure report text and the matching probability may be displayed.

With such a configuration, since the displayed plurality of failure report sentences are displayed together with the matching probabilities, the operator can judge the diagnosis result while referring to the matching probabilities, which makes it easier and more accurate. Failure diagnosis can be performed.

Further, the trained model may be generated by further performing machine learning using any word related to the failure occurring in the target device as input data and the failure report text corresponding to the failure as teacher data.

In the failure report input at the time of failure diagnosis, the terms and expressions used by the operator may vary, and even if the words have the same meaning, the failure report using words different from the words described in the failure report text is used. If you enter it, it may not be possible to accurately determine the degree of agreement between the failure report and the failure report text.

With the above configuration, the trained model can be machine-learned by using a wide range of words by a machine-learning operator, and it becomes possible to determine a synonym for one word. Therefore, in machine learning as well, it is possible to determine the degree of agreement between the failure report and the failure report sentence while determining the analogy of the words used in the failure report and the failure report sentence. As a result, even if a wide range of words are used in the failure report, the failure report text having a high degree of agreement with the failure report can be extracted, and the failure diagnosis can be performed more easily and accurately.

Further, in addition to the machine learning for the failure diagnosis, machine learning for natural language processing is performed, and the trained model may have information related to the natural language processing.

With the above configuration, the trained model supports natural language processing, so even if a wide range of words are used in failure reporting, failure diagnosis can be compared and examined including word synonyms. It is possible to extract a failure report sentence having a high degree of agreement with the failure report, and it is possible to perform a failure diagnosis more easily and accurately.

Further, a step of inputting the failure report into the trained model to newly generate a failure report sentence may be further provided.

Using the trained model as described above, not only the failure diagnosis determines the consistency between the failure report and the failure report text, but also the failure report text is registered in the failure mechanism database in consideration of the content of the failure report. A sentence according to the structure of can be generated as a new failure report sentence. Therefore, by using the trained model as described above, it is possible not only to perform a failure diagnosis with high accuracy but also to create a failure report text regarding a new failure.

Further, the failure diagnosis determination unit may further include a failure diagnosis determination screen for displaying an appropriate failure diagnosis determination screen on the display unit for determining an appropriate failure report sentence from the failure report text extracted by the failure diagnosis unit.

This makes it possible to easily determine an appropriate failure report text from a plurality of fault report texts.

Further, the target device is a work vehicle that performs a predetermined work at the work site, and may include at least one of the work content, the work site environment, and the usage mode of the work vehicle as the failure factor.

With such a configuration, it is possible to easily and accurately perform a failure diagnosis even for a work vehicle. Further, it is possible to input information such as the content of the work to be performed, the environment of the work site where the work is performed, and the usage mode of the work vehicle, which are peculiar to the work vehicle performing the work at the work site, as a failure factor. Therefore, it is possible to generate a trained model using a failure mechanism database containing information specific to the work vehicle as input data, and to input a failure report containing information specific to the work vehicle to perform failure diagnosis. As a result, it is possible to perform failure diagnosis with particularly high accuracy in the work vehicle.

It is a figure which shows the schematic structure of the failure diagnosis apparatus. It is a figure which shows the schematic structure of the failure report text. It is a figure which illustrates the failure report text. It is a figure which shows the model composition example of the trained model. It is a block diagram which illustrates the structure of the failure diagnosis apparatus. It is a figure which illustrates the failure report input screen. It is a figure which illustrates the processing flow of the failure diagnosis method. It is a figure which illustrates the failure diagnosis determination screen.

In addition to work vehicles such as agricultural work machines and construction machines, various devices are composed of mechanical parts, electronic parts, and the like. Even if these devices are manufactured with sufficient durability and reliability, failures may inevitably occur. When a failure occurs, the failure location is identified, the failure diagnosis is performed by analyzing the failure mechanism, and then the failure location is repaired and design changes are considered to suppress the occurrence of future failures.

Hereinafter, a failure diagnosis device for diagnosing a failure occurring in such a device such as a work vehicle (hereinafter referred to as “target device”) will be described with reference to the drawings.

[Outline of failure diagnosis]
First, a schematic configuration in which the failure diagnosis device performs failure diagnosis will be described with reference to FIG.

As shown in FIG. 1, the failure diagnosis device 1 is configured such that the learning device 2 performs machine learning to generate a learned model 3, and the failure diagnosis unit 4 performs a failure diagnosis using the learned model 3.

The learning device 2 is equipped with AI (Artificial intelligence) 5. Machine learning is performed by inputting the ontology model 6 of the target device and the failure mechanism database 7 into the AI 5 of the learning device 2 as input data, and the learned model 3 is generated.

The ontology model 6 is information indicating the mutual relationship of a plurality of parts constituting the target device. The target device is composed of mechanical parts, electronic parts, and the like, and these are organically linked. The ontology model 6 is information showing the parts included in the target device, their connection relations, the connection mode, the mode of change of other parts when a certain part operates, the influence range of each part and the operation, and the like.

The failure mechanism database 7 is a database in which a plurality of failure report sentences 8 are stored. The failure mechanism database 7 is a database in which the failure status, which is the analysis result of the failure that has occurred in the past in the target device, is summarized as the failure report text 8 including the failure information and the failure information. Further, the failure report text 8 may appropriately include failure information not only about the target device but also about a device similar to the target device or a device having a common configuration with the target device. The failure report text 8 will be described in detail later.

When a failure occurs in the target device, a failure report 9 indicating the failure status is input to the failure diagnosis unit 4. In the failure report 9, various information related to the failure report 9 may be input to the failure diagnosis unit 4 one by one by screen input or the like, but the failure report 9 created as one data is input to the failure diagnosis unit 4. You may enter it directly. The failure report 9 created as one data is failure information indicating the symptom of the failure when the target device has a failure, and is the information produced by the user of the target device after confirming the failure status. be. For example, the failure report 9 created as one piece of data can be a warranty application form, a complaint report, or the like for requesting a warranty from a manufacturer or a retailer when a failure occurs in the target device. .. Specifically, as shown in FIG. 2, the information included in the failure report 9 includes a failure factor 13, a failure site 14, a failure event 15, and the like, and includes at least one of these. The failure factor 13, the failure site 14, the failure event 15, and the like are failure information described in the failure report text 8, and details will be described later. The failure report 9 and the failure report text 8 may further include a failure mode 18. The failure diagnosis unit 4 performs failure diagnosis by inputting the failure report 9 into the generated learned model 3.

[Failure report text]
The failure report text 8 shows failure information regarding the mechanism that led to the failure, including the cause and symptom of the failure, the location of the failure, and the like. A failure is that there is a cause of the failure, and the cause affects a certain part and causes a defect in the part.

As shown in FIGS. 2 and 3, the failure report text 8 includes failure information 13 which is a failure factor 13, a failure site 14, and a failure event 15, and further includes a failure outline 16. The failure report text 8 may further include a functional block 17, a failure mode 18, a countermeasure policy 19, and the like. In the failure report sentence 8, the cause of the failure corresponds to the failure factor 13, and the component in which the failure occurs corresponds to the failure portion 14. The phenomenon that occurs in the component corresponds to the failure event 15, and the effect on the component corresponds to the failure mode 18.

As described above, the failure factor 13 is the cause of the failure and can be divided into an internal factor and an external factor. The internal factor is the stress that occurs or increases under certain conditions due to the structure of the target product. For example, an internal factor is engine vibration or the like. The external factor is stress that occurs or increases under certain conditions due to the usage method, usage environment, manufacturing / transportation, etc. of the target product. For example, an external factor is rocking of the vehicle during high-load operation. Failure occurs due to the increase and occurrence of this stress in internal and external factors.

The failure portion 14 is a portion of a part or the like in which a failure has occurred due to the failure factor 13. For example, the failure portion 14 is a “welded portion of the fuel filter bracket”, a “fuel filter”, a “fuel hose”, or the like.

The failure event 15 is information indicating what the failure portion 14 has become due to the failure. For example, in the failure event 15, information such as "damage", "dropping", and "fuel leakage" is described. This information corresponds to the case where the welded portion of the fuel filter bracket is damaged and the fuel filter falls off, or the fuel leaks from the fuel hose. Further, the failure event 15 can also include events that occur as a result of these events. For example, when the fuel hose and the radiator fan come into contact with each other and interfere with each other to generate an abnormal noise, it is possible to describe "radiator fan", "engine room", "abnormal noise generation" and the like.

The functional block 17 is a functional block that includes a failure portion 14, and includes relationships with related parts and the like. For example, it is shown that the functional block 17 related to the above-exemplified failure site 14 is an engine auxiliary device (fuel system) and includes a fuel filter and a bracket.

The failure mode 18 is information indicating how the failure portion 14 or the like is broken. The failure mode 18 has a causal relationship with the failure factor 13, and occurs in association with the failure factor 13. That is, a failure event 15 which is a failure event occurs in the failure portion 14 due to the failure mode 18 caused by the failure factor 13. Here, the failure mode 18 is a failure of a component that requires repair / replacement, and is defined as destruction of the component / destruction of the interface between the components / irreversible change in physical properties. For example, the failure mode 18 is described as "fatigue failure", "contact / interference", and "disconnection".

The failure outline 16 is information in which the failure status is described as a sentence corresponding to the failure information such as the failure factor 13, the failure site 14, and the failure event 15. As the failure report sentence 8 output as a result of the failure diagnosis, only this failure summary 16 may be output. When the failure information is the failure factor 13, the failure site 14, and the failure event 15 exemplified above, the failure outline 16 is, for example, "fuel filter bracket due to engine vibration and vibration of the vehicle body during high load operation. The welded part was damaged due to fatigue failure, and the fuel filter fell off. The fuel hose contacted and interfered with the radiator fan, and an abnormal noise was generated in the engine room. The fuel hose was cut and fuel leaked. " ..

The countermeasure policy 19 describes the measures taken for the failure when the failure occurs, the content of repairing the failure, and the matters of design changes in response to the failure. For example, the countermeasure policy 19 is "design the stress to be lower than the reference value by shifting the resonance point of the fuel filter bracket from the engine rotation range or lowering the amplitude level at the time of resonance." Further, the countermeasure policy 19 can include design ToDo, experimental ToDo, and the like, which are matters to be performed at the time of product development. Further, the countermeasure policy 19 may include the progress status, deliverables, etc. of the design ToDo, the experimental ToDo, and the like.

[Trained model]
As described above, the ontology model 6 describes the parts constituting the target device, the connection relations of the parts, and the relational relations between the parts such as the components intervening in the propagation of the operation between the parts, and the object centering on the parts. The configuration of the device is described. Further, in the failure report sentence 8, the failure situation is described as a sentence including the failure information and the failure information.

Therefore, the trained model 3 machine-learned using the failure mechanism database 7 having a plurality of failure report sentences 8 and the ontology model 6 as input data is the ontology model in addition to the failure situation acquired from the failure report sentence 8. The relationship of the parts related to the failure acquired from 6 is modeled. Therefore, in the trained model 3, the past failure status associated with all the parts included in the functional block of the target device is modeled based on the failures that have occurred in the past.

An example of such a model of the trained model 3 will be described with reference to FIG. 2 with reference to FIG. Here, a model in which the portion related to the failure of the fuel system illustrated in FIG. 3 is modeled will be described as an example.

As shown in FIG. 4, the trained model 3 is a model in which the failure situation includes words corresponding to failure information, etc., and sentences including failure information regarding the event that has occurred, related parts, the configuration related to the propagation of the event, and the like. It has been transformed. Of these, the connection relationship of parts, the configuration related to the propagation of events, and the like are extracted from the ontology model 6, and the events that have occurred, the related parts, the details of failures, and the like are extracted from the failure mechanism database 7.

Specifically, since the failure factor 13 is the vibration of the engine and the shaking of the vehicle during high load operation, and the functional block 17 in which the failure occurs is the fuel system of the engine auxiliary equipment, the learned model 3 is , The effects of engine vibration and vehicle vibration during high load operation on the fuel system are modeled.

The trained model 3 includes a failure report sentence 8 relating to a failure situation shown in FIG. 3 as one failure of the fuel system. In this case, the text of the failure report sentence 8 states, "The welded part of the fuel filter bracket is fatigued and broken due to the vibration of the engine and the shaking of the vehicle during high load operation, the fuel filter falls off, and the fuel hose and radiator fan are damaged. Contact / interference will occur and the fuel hose will break, causing fuel leakage. "

The trained model 3 further models the interrelationships of fuel system components from the ontology model 6. As shown in FIG. 4, the structure of the fuel system modeled from the ontology model 6 is a configuration in which the fuel filter is held by the fuel filter bracket and fuel is supplied from the fuel filter via the fuel hose.

By incorporating such an ontology model 6, as shown by the arrow in FIG. 4, “fatigue fracture” and “damage” in the failure report text 8 are associated with the welded portion of the fuel filter bracket and “fall off”. Is associated with the fuel filter, "contact / interference" is associated with the fuel hose, and "cutting" is associated with the fuel hose. Further, "contact / interference" is associated with "radiator fan" and "abnormal noise generation", and "abnormal noise generation" and "engine room" are associated with each other. Therefore, a trained model 3 is generated in which the device configuration of the fuel system is added to the failure report sentence 8 indicating the failure status of the fuel system.

As a result, the learned model 3 relating to one of the fuel system failures shown in FIG. 3 includes the fuel system structure in addition to the fuel system failure status. Further, the trained model 3 is machine-learned by inputting the failure mechanism database 7 to the failure status of other fuel systems. Further, the trained model 3 is machine-learned by inputting an ontology model 6 of another functional block and a failure mechanism database 7 related to a failure situation. As a result, the trained model 3 is a model that can deal with a wide range of failures by modeling various failure situations of the target device and the device configuration.

[Failure diagnosis device]
Next, a configuration example of the failure diagnosis device 1 will be described with reference to FIGS. 1 and 2 with reference to FIGS. 5 and 6.

The failure diagnosis device 1 includes a control unit 20, a learning device 2, a failure diagnosis unit 4, an input unit 22, a storage unit 23, and a display unit 24.

The control unit 20 controls the operation of the failure diagnosis device 1. The input unit 22 inputs the ontology model 6 and the failure mechanism database 7, which are input data for machine learning. The input ontology model 6 and the failure mechanism database 7 are stored in the storage unit 23 under the control of the control unit 20.

The learning device 2 is equipped with AI5, and has been learned by inputting the ontology model 6 and the failure mechanism database 7 stored in the storage unit 23 into AI5 as input data and performing machine learning according to the control of the control unit 20. Generate model 3. The generated trained model 3 is stored in the storage unit 23.

The failure mechanism database 7 contains a large number of failure report sentences 8. If the words used in these failure report sentences 8 are not unified, it is not possible to appropriately compare with the input failure report 9 at the time of failure diagnosis using the generated learned model 3. Therefore, the failure report sentence 8 included in the failure mechanism database 7, which is the input data for machine learning, is created in advance by standardizing the words, expressions, and syntax to be used. Further, even if the failure report sentence 8 is created using standardized words, expressions, and syntax, there may be some variations. Even in such a case, it can be expected that the trained model 3 can be a model that can judge a certain degree of similarity in terms of words, expressions, and syntax used by machine learning a large number of failure report sentences 8.

When performing a failure diagnosis, the failure diagnosis device 1 may include a failure report input unit 27 for displaying the failure report input screen 30 on the display unit 24. The failure report 9 is input to the failure diagnosis device 1 in a mode of selecting an item displayed on the failure report input screen 30. The display unit 24 is, for example, a touch panel, and has a configuration in which information can be input from the displayed failure report input screen 30.

On the failure report input screen 30, the failure factor 13, the failure site 14, the failure event 15, and the like included in the failure report 9 can be individually input as the failure report 9, and the failure mode 18 can also be input.

The failure report input screen 30 displays an item switch 31 that can be selectively input as an item for inputting an internal factor and an external factor among the failure factors 13, a failure portion 14, a failure event 15, and a failure mode 18. When the item switch 31 corresponding to the item to be input is selected by touching the screen or the like, the pull-down menu 32 corresponding to the item is displayed. In the pull-down menu 32, the candidates to be input to the item are listed, and the information of the item is input by selecting one or more of the displayed candidates. In FIG. 6, the item switch 31 corresponding to the internal factor is selected, and “engine vibration” is selected from the candidates displayed in the pull-down menu 32. Then, by selecting the item switch 31 corresponding to each of the internal factor and the external factor, the failure part 14, the failure event 15, and the failure mode 18 in order, all the information related to the failure report 9 can be input. .. If there is no information to be input for that item, select "None" from the candidates in the pull-down menu 32, and if there is no corresponding information in the candidates, select "Other" and select any option. You can enter information. When "Other" is selected and arbitrary information is input, a keyboard may be displayed on the display unit 24, and arbitrary information may be input from the keyboard. Arbitrary information may be input from the input means.

The information input in this way is associated with the corresponding item and stored in the storage unit 23 as a failure report 9.

Then, according to the control of the control unit 20, the failure report 9 stored in the storage unit 23 is input to the learned model 3 stored in the storage unit 23, and the failure diagnosis is performed. As a failure diagnosis, the failure diagnosis unit 4 compares the failure report sentence 8 modeled in the trained model 3 with the failure report 9, and outputs one or a plurality of failure report sentences 8 having a high degree of coincidence. Specifically, the failure report sentence 8 having a degree of agreement with the failure report 9 and a predetermined degree or more is selected. For example, the failure diagnosis unit 4 extracts and outputs a word included in the failure report 9 or a word similar to the word in the failure report sentence 8 containing a predetermined number or more or a predetermined ratio or more. When comparing the failure report sentence 8 and the failure report 9, the description of the failure report 9 is compared with the failure report sentence 8 in consideration of synonyms and similar expressions by grasping the word / context by natural language processing. Processing may be performed. Further, since the trained model 3 is modeled on the ontology model 6, the failure diagnosis unit 4 can perform a failure diagnosis in consideration of the parts related to the parts corresponding to the words included in the failure report 9. There is also.

The display unit 24 displays the failure report text 8 output by the failure diagnosis unit 4 according to the control of the control unit 20. The display unit 24 may display only the portion of the failure report sentence 8 corresponding to the failure summary 16. When a plurality of failure report sentences 8 are output, the display unit 24 displays the plurality of failure report sentences 8 in descending order of the degree of coincidence.

The failure diagnosis unit 4 may calculate the matching probability between the failure report sentence 8 and the failure report 9. In this case, the display unit 24 displays the match probability in association with the failure report sentence 8 together with the failure report sentence 8.

As described above, since machine learning is performed using the failure mechanism database 7 consisting of the failure report sentence 8 as input data, the failure diagnosis result is directly used as the failure report sentence 8 including at least the failure summary 16 which is a sentence indicating the failure status. Can be output as a target. Therefore, even if the worker is not familiar with the structure of the target device such as a work vehicle, there is no need for trial and error in the failure diagnosis, so that the failure diagnosis can be easily performed.

In particular, the output failure report sentence 8 is output in the form of a sentence consisting of words included in the failure report 9 input at the time of failure diagnosis. Therefore, the worker can easily understand the failure diagnosis result even if he / she is not familiar with the structure of the target device and does not have sufficient knowledge of the failure diagnosis, and can write the failure report text that is judged to be the most appropriate. Just by selecting, you can perform failure diagnosis with high accuracy.

Further, as described above, the failure report text 8 can include a countermeasure policy. In this case, the failure report text 8 output as the failure diagnosis result describes the failure status and the countermeasures against the failure. Therefore, the operator can quickly confirm the method of dealing with the failure, the method of repairing the faulty part, or the matters for which the design change should be considered in response to the fault, just by checking the failure diagnosis result. It is possible to deal with various failures. Further, it is also possible to detect the problem of the current target device by referring to the failure report sentence 8 at the time of developing a new product and refer to it for the development of a new product.

[Failure diagnosis method]
Next, a failure diagnosis method will be described with reference to FIGS. 1 and 2 with reference to FIG. 7. In the following description, the method of performing the failure diagnosis by the above-mentioned failure diagnosis device 1 may be used, but the failure diagnosis method is not limited to the case where the failure diagnosis device 1 is used, and other device configurations are used. , Or software, or a mixture of these may be implemented.

First, an ontology model 6 showing the structure of a target device such as a work vehicle is generated, and a failure mechanism database 7 that collects failure report sentences 8 showing a failure status that has occurred in the past for the target device is generated.

Next, the generated model 3 is generated by machine learning the AI5 using the generated ontology model 6 and the failure mechanism database 7 as input data (step # 1).

When a failure occurs in the target device (step # 2 Yes), the operator creates a failure report 9 using the information within the range that can be confirmed for the failure information such as the failure factor 13, the failure site 14, and the failure event 15. (Step # 3). The failure report 9 may be a sentence including a failure factor 13, a failure part 14, a failure event 15, and the like, but may be a word associated with the failure factor 13, the failure part 14, the failure event 15, and the like.

Next, the fault diagnosis is performed by inputting the created fault report 9 into the trained model 3 (step # 4). As a result of the failure diagnosis, one or a plurality of failure report sentences 8 having a high degree of agreement with the failure report 9 are output from the failure report sentences 8 stored in the failure mechanism database 7.

Next, the output failure report text 8 is displayed (step # 5). The displayed failure report text 8 includes at least a failure summary 16. When a plurality of failure report sentences 8 are output, the failure report sentences 8 are displayed in descending order of the degree of matching. Further, the matching probability of each failure report sentence 8 with the failure report 9 may be displayed in association with each other.

Next, the worker confirms the displayed failure report sentence 8, and whether the content of the displayed failure report sentence 8 matches the failure actually occurring, or which failure report sentence 8 is the most one. By examining whether or not the failure is done, the failure report text 8 that is judged to correspond to the failure that actually occurs is selected (step # 6).

Then, the worker confirms the detailed status of the failure from the selected failure report sentence 8 and considers the response to the failure. Alternatively, when the countermeasure policy 19 is indicated in the failure report sentence 8, the countermeasure is taken according to the countermeasure policy 19 (step # 7).

Even with the above-mentioned failure diagnosis method, as in the case of the failure diagnosis device 1, even if the worker is not familiar with the structure of the target device such as a work vehicle and the failure diagnosis, the diagnosis result includes at least the failure outline 16. Since it can be directly confirmed as the failure report text 8 and there is no need for trial and error in the failure diagnosis, the failure diagnosis can be easily performed.

[Judgment of failure diagnosis]
As a result of the failure diagnosis performed by the failure diagnosis device 1 and the failure diagnosis method, one or a plurality of failure report sentences 8 are displayed on the display unit 24. From the displayed failure report text 8, the fault diagnosis worker may determine which fault report text 8 is appropriate, but in the fault diagnosis device 1, the fault diagnosis determination unit 28 assists in the determination. May be provided. Hereinafter, a specific configuration example for determining an appropriate failure report sentence 8 will be described with reference to FIGS. 2, 4, and 8.

The failure diagnosis device 1 displays the failure diagnosis determination screen 35 on the display unit 24 under the control of the failure diagnosis determination unit 28. The failure diagnosis determination screen 35 displays the failure event candidate 38 and the component connection relationship 36. In the failure event candidate 38, the failure unit 39, which is a failure event related to a specific component in the ontology model 6, is extracted from the failure report text 8 listed as a candidate by the failure diagnosis and listed. At this time, the unit 39 of the failure may be arranged in descending order of the degree of matching with reference to the degree of matching of the corresponding failure report sentence 8, or may be displayed together with the matching probability. In the component connection relationship 36, a plurality of components corresponding to the candidate of the failure report sentence 8 are listed as component candidates 37 for this specific component. Further, in the component connection relationship 36, for one specific component selected from the component candidates 37, the component connected to the upstream side in the ontology model 6 is separately displayed as the component candidate 37. The component candidates 37 displayed at this time may be displayed in order not only for one upstream component but also for a plurality of upstream components.

When a plurality of failure report sentences 8 are listed as candidates by the failure diagnosis, first, the most downstream part of the ontology model 6 in the plurality of failure report sentences 8 listed as candidates by the failure diagnosis is set as the component candidate 37, and the failure diagnosis is performed. They are listed and displayed in the connection relationship 36 of the parts on the determination screen 35. At the same time, the failure unit 39 corresponding to the most downstream component candidate 37 is displayed on the failure diagnosis determination screen 35 as the failure event candidate 38.

Next, the operator selects the failure unit 39 that matches the actual failure situation occurring in the target device from the failure units 39 displayed in the failure event candidate 38. Such selection is made by touching the screen of the corresponding failure unit 39 or the like. When one failure unit 39 is selected, among the component candidates 37 corresponding to the most downstream in the connection relationship 36 of the parts, the component corresponding to the selected failure unit 39 changes the character color, blinks, or the like. More emphasized. Then, in the component candidate 37 corresponding to the upstream side of the ontology model 6, the components connected to this component are listed and displayed. At this stage, the fault report text 8 including the selected fault unit is narrowed down.

When the failure unit 39 corresponding to the most downstream component is selected, it is included in the narrowed-down failure report text 8 in the ontology model 6 and is upstream of the component corresponding to the selected failure unit 39. The failure unit 39 related to the component is listed and displayed in the failure event candidate 38. The operator selects, among the displayed failure units 39, the failure unit 39 that matches the actual failure status occurring in the target device.

Such processing is repeated up to the most upstream component of the ontology model 6, and the failure report sentence 8 is narrowed down. When the failure report sentence 8 is narrowed down to one up to the most upstream part of the ontology model 6, it is determined that the failure report sentence 8 is the failure report sentence 8 corresponding to the failure actually occurring. can do. If a plurality of failure report sentences 8 remain even after reaching the most upstream part of the ontology model 6, the operator considers these failure report sentences 8 and further investigates to obtain an appropriate failure. Judge the report sentence 8.

For example, when the failure report text 8 illustrated in FIG. 4 is appropriate, the failure unit displayed first is the failure unit A corresponding to the failure situation for the most downstream component of the ontology model 6. It is displayed including the unit 39 of the failure that "the fuel hose is cut and the fuel leaks". When "fuel hose is cut and fuel leaks" is selected and the next failure unit 39 is displayed, "abnormal noise has occurred in the engine room" corresponding to the failure unit B is selected. Finally, the failure report sentence 8 including "the fuel filter has fallen off" corresponding to the unit C of the failure remains, and the appropriate failure report sentence 8 is extracted by this.

By determining an appropriate failure report sentence 8, even if a plurality of failure report sentences 8 are displayed as candidates, the failure diagnosis can be easily performed.

[Another Embodiment]
(1) In the failure diagnosis device 1, a failure report sentence generation unit 26 may be further provided. The failure report sentence generation unit 26 newly generates a failure report sentence 8 corresponding to the failure report 9 by inputting the failure report 9 into the trained model 3.

The trained model 3 is generated using the failure report sentence 8 (fault mechanism database 7) and the ontology model 6 as input data. Therefore, in the trained model 3, the structure of the target device is modeled in addition to the failure report sentence 8. By inputting the failure report 9 into such a trained model 3, the failure report 9 can be analyzed in consideration of the structure of the target device while referring to the past failure report sentence 8. As a result, a new failure report sentence 8 corresponding to the input failure report 9 can be generated with reference to the description content and the document structure of the past failure report sentence 8.

The failure report sentence 8 is generated not only by the failure diagnosis device 1, but also by inputting the failure report 9 into the trained model 3 into the failure diagnosis method described in the above embodiment to generate the failure report sentence 8. May be added.

The failure report text 8 is created by a worker or the like based on the diagnosis result after performing a failure diagnosis when a failure occurs. At this time, the words and expressions used differ depending on the worker, and the knowledge about the structure of the target device and the knowledge about the failure diagnosis differ depending on the worker, so that the completeness and expressions of the created failure report sentence 8 are not constant. was there. As described above, the failure report 9 can be input to the trained model 3 to generate a new failure report sentence 8. Therefore, the consistency with the words and expressions used in the already existing failure report sentence 8 is achieved, and the degree of completion of the failure report sentence 8 tends to be constant. As a result, even an operator who does not have high knowledge about the structure of the target device and knowledge about the failure diagnosis can easily generate the failure report sentence 8 at the same level as the other failure report sentences 8.

Further, the failure mechanism database 7 including the failure report sentence 8 generated by inputting the failure report 9 may be input to the trained model 3, machine learning may be performed again, and the trained model 3 may be updated. The failure report sentence 8 generated by inputting the failure report 9 into the trained model 3 has the same criteria as the already machine-learned failure report sentence 8, and the words, expressions, and syntax used are standardized. Therefore, in the updated trained model 3, the standardization of words, expressions, and syntax used is improved, and the accuracy of failure diagnosis can be further improved.

(2) In each of the above embodiments, not only machine learning is performed using the failure mechanism database 7 and the ontology model 6 as input data, but also words corresponding to failure information related to any failure are used as input data to deal with the failure. Further machine learning may be performed using the failure report sentence 8 as teacher data.

The existing failure report sentence 8 stored in the failure mechanism database 7 is organized in advance, and the words, expressions, and syntax used are standardized. The separately prepared failure information regarding an arbitrary failure may have the same meaning as the word or expression used in the existing failure report sentence 8, but may have a different word or expression. By adding the above machine learning, the trained model 3 becomes more versatile in terms of words, expressions, and syntax, and even if different words, expressions, and syntax are used in failure diagnosis, failure reporting 9 is appropriate. It is possible to compare and verify the degree of agreement between the above and the failure report sentence 8. As a result, failure diagnosis can be performed more easily and accurately.

(3) In each of the above embodiments, machine learning related to natural language processing may be further added as machine learning. As a result, the trained model 3 is modeled so that synonyms for words, expressions, and the like can be determined. As a result, it becomes possible to appropriately compare and verify the degree of coincidence between the failure report 9 and the failure report sentence 8, and it becomes possible to perform the failure diagnosis more easily and accurately.

(4) The configuration is not limited to the configuration in which the failure report 9 is input via the failure report input screen 30, and the dataized failure report 9 may be input from the input unit 22 and stored in the storage unit 23. At the time of failure diagnosis, the failure report 9 stored in the storage unit 23 is read out, and the failure diagnosis unit 4 performs the failure diagnosis. The failure report 9 may be a sentence including a word corresponding to at least one of the failure factor 13, the failure site 14, and the failure event 15. Further, the failure report 9 may be a word corresponding to at least one of the failure factor 13, the failure site 14, and the failure event 15, which are not written and are failure information. Further, the failure report 9 may further include a failure mode 18 as failure information. The fault report 9 to be input does not necessarily have to be converted into data, and words corresponding to the fault report 9 may be input one by one. Even in such a case, a word similar to the word included in the input failure report 9 is considered in the failure diagnosis. Therefore, the failure diagnosis can be performed without strictly limiting the words used in the failure report 9, and the failure diagnosis can be easily performed from this as well.

(5) In each of the above embodiments, the failure diagnosis device 1 includes a processor such as a CPU in addition to the AI5. The failure diagnosis device 1 may be configured by hardware, but at least a part of the functions may be configured by software. In this case, the software is stored in the storage unit 23 and is operated by the processor. Further, the configuration of the functional blocks in the failure diagnosis device 1 is not limited to the above configuration, and each functional block may be appropriately aggregated or subdivided.

Further, the AI 5 is not limited to the case where the failure diagnosis device 1 is provided, and the failure diagnosis device 1 may be provided in a state in which communication is possible via a network or the like. In this case, the input data of machine learning and the failure report 9 of the failure diagnosis are transmitted to the AI5 via the network, and the failure report sentence 8 which is the failure diagnosis result is transmitted from the AI5 to the failure diagnosis device 1 via the network.

INDUSTRIAL APPLICABILITY The present invention can be applied to failure diagnosis for a failure that occurs in a target device of various work vehicles such as agricultural work vehicles and construction work vehicles.

1 Failure diagnosis device 2 Learning device 3 Learned model 4 Failure diagnosis unit 6 Ontology model 7 Failure mechanism database 8 Failure report text 9 Failure report 13 Failure cause 14 Failure part 15 Failure event 18 Failure mode 19 Countermeasure policy 24 Display unit 26 Failure report Sentence generator

Claims (26)

A failure that has an ontology model that shows the interrelationship of multiple components that make up the target device, and multiple failure report texts that include failure factors, failure sites, and failure events that are failure information about the failure that occurred in the target device. A learning device that generates a trained model for failure diagnosis by machine learning with a mechanism database as input data.
A failure report containing at least one of the failure information is input to the trained model, and the failure report text having a high degree of agreement with the failure report as the failure diagnosis is obtained from the failure mechanism database. The failure diagnosis unit to be extracted and
A failure diagnosis device including a display unit that displays the failure report text extracted by the failure diagnosis unit. The failure diagnosis device according to claim 1, wherein the failure report text further includes a failure mode that causes the failure event as the failure information. The failure diagnosis device according to claim 1 or 2, wherein the failure report text further includes a countermeasure policy for failure. The failure diagnosis device according to any one of claims 1 to 3, wherein the failure report is a word associated with each of the failure information. The failure diagnosis device according to any one of claims 1 to 4, further comprising a failure report input unit for displaying a failure report input screen for individually inputting the failure information of the failure report on the display unit. The failure diagnosis unit extracts a plurality of the failure report sentences in descending order of the degree of agreement with the failure report.
The failure diagnosis device according to any one of claims 1 to 5, wherein the display unit displays the extracted failure report texts in descending order of the degree of coincidence with the failure report. The failure diagnosis unit calculates the matching probability between the failure report and the failure report text, and obtains a probability of matching.
The failure diagnosis device according to claim 6, wherein the display unit displays the extracted failure report text and the match probability. The learning device generates the trained model by further performing machine learning using an arbitrary word related to a failure occurring in the target device as input data and a failure report sentence corresponding to the failure as teacher data. The failure diagnosis device according to any one of 7 to 7. The learning device performs machine learning for natural language processing in addition to machine learning for failure diagnosis.
The failure diagnosis device according to any one of claims 1 to 8, wherein the trained model has information related to the natural language processing. The failure diagnosis device according to any one of claims 1 to 9, further comprising a failure report sentence generation unit for inputting the failure report into the trained model and newly generating a failure report sentence. One of claims 1 to 10, further comprising a failure diagnosis determination unit for displaying a failure diagnosis determination screen for determining an appropriate failure report sentence from the failure report text extracted by the failure diagnosis unit on the display unit. The failure diagnostic device described in the section. The target device is a work vehicle that performs a predetermined work at the work site.
The failure diagnosis device according to any one of claims 1 to 11, wherein the failure cause includes at least one of a work content, a work site environment, and a usage mode of the work vehicle. A failure that has an ontology model that shows the interrelationship of multiple components that make up the target device, and multiple failure report texts that include failure factors, failure sites, and failure events that are failure information about the failure that occurred in the target device. The process of generating a trained model for failure diagnosis by machine learning with the mechanism database as input data, and
A failure report containing at least one of the failure information is input to the trained model, and the failure report text having a high degree of agreement with the failure report as the failure diagnosis is obtained from the failure mechanism database. The process of extraction and
A failure diagnosis method including a step of displaying the extracted failure report text. The failure diagnosis method according to claim 13, wherein the failure report text further includes a failure mode that causes the failure event as the failure information. The failure diagnosis method according to claim 13 or 14, wherein the failure report text further includes a countermeasure policy for failure. The failure diagnosis method according to any one of claims 13 to 15, wherein the failure report is a word associated with each of the failure information. In the failure diagnosis, a plurality of the failure report sentences are extracted in descending order of the degree of agreement with the failure report.
The failure diagnosis method according to any one of claims 13 to 16, wherein the extracted failure report texts are displayed in descending order of matching degree with the failure report. In the failure diagnosis, the matching probability between the failure report and the failure report text is calculated.
The failure diagnosis method according to claim 17, wherein the extracted failure report text and the match probability are displayed. Any of claims 13 to 18 in which the trained model is generated by further performing machine learning using any word related to the failure occurring in the target device as input data and the failure report text corresponding to the failure as teacher data. The failure diagnosis method described in item 1. In addition to the machine learning for failure diagnosis, machine learning for natural language processing is performed.
The failure diagnosis method according to any one of claims 13 to 19, wherein the trained model has information related to the natural language processing. The failure diagnosis method according to any one of claims 13 to 20, further comprising a step of inputting the failure report into the trained model and newly generating a failure report sentence. The target device is a work vehicle that performs a predetermined work at the work site.
The failure diagnosis method according to any one of claims 13 to 21, which includes at least one of the work content, the work site environment, and the usage mode of the work vehicle as the cause of the failure. A failure that has an ontology model that shows the interrelationship of multiple components that make up the target device, and multiple failure report texts that include failure factors, failure sites, and failure events that are failure information about the failure that occurred in the target device. Generated by machine learning with the mechanism database as input data
A failure report containing at least one of the failure information is input to the trained model, and the failure report text having a high degree of agreement with the failure report by a predetermined degree or more is extracted from the failure mechanism database as a failure diagnosis. A trained model to make a computer work to do. The learned model according to claim 23, wherein the failure report text further includes a failure mode that causes the failure event as the failure information. The trained model according to claim 23 or 24, wherein machine learning is further performed using an arbitrary word related to a failure occurring in the target device as input data and a failure report sentence corresponding to the failure as teacher data. In addition to the machine learning for failure diagnosis, machine learning for natural language processing is performed.
The trained model according to any one of claims 23 to 25, which has information relating to the natural language processing.

Images