Maintenance information generating device, maintenance information generating method, and maintenance information generating program

Abstract

To generate maintenance information allowing for easily recognizing work to be executed for maintenance.SOLUTION: A maintenance information generation device comprises a sentence data input unit, an analysis unit, and a resin tree diagram generation unit. The sentence data input unit receives input of text data of a repair history where failures, causes, and treatments are registered. The analysis unit analyzes the test data of the repair history to extract important words relating to maintenance and uses the important words to classify the events, the causes, and the treatments. The tree diagram generation unit uses results from the analysis unit to form a tree diagram showing the events, the causes, and the treatments in a tree-shaped form.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a technique for generating maintenance information for a device. [Background technology]

[0002] Patent Document 1 describes a text analysis device. The text analysis device described in Patent Document 1 extracts text data by vectorizing text using distributed representation. The text analysis device described in Patent Document 1 also performs clustering on the extracted text data.

[0003] The writing analysis device described in Patent Document 1 extracts representative sentences from clusters that have been subjected to clustering. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2020-166426 Summary of the Invention [Problem to be solved by the invention]

[0005] When maintaining a device, referring to the repair history is effective and leads to improved maintainability.

[0006] Repair history is often stored as written records written by the worker. Therefore, when an event such as a device failure occurs, the worker refers to the repair history stored in written form to carry out the work. In this case, it takes the worker time to search for the repair history based on the event, which leads to a decrease in the efficiency of maintenance work.

[0007] Here, for example, even if the technology of Patent Document 1 is adopted for maintenance work, it is possible to extract individual elements of events, etc., but it is not easy to know from these elements what work should be done in maintenance, etc.

[0008] An object of the present invention is to provide a technique for generating maintenance information that enables a worker to easily understand the work that should be performed during maintenance. [Means for solving the problem]

[0009] In order to achieve the above object, the maintenance information generating device of the present invention is configured as follows.

[0010] The maintenance information generation device includes a text data input unit, an analysis unit, and a tree diagram generation unit. The text data input unit accepts input of text describing events, causes, and actions. The analysis unit analyzes the text data to extract important words related to maintenance and uses the important words to classify the events, causes, and actions. The tree diagram generation unit uses the results of the analysis unit to create a tree diagram that shows the relationships between events, causes, and actions in a tree-like structure.

[0011] This configuration allows events, causes, and actions to be properly extracted from text data such as maintenance repair history, and then displayed in a tree-like tree diagram. This allows workers to easily understand the causes of events and the actions to be taken for those causes, and to easily grasp the actions to be taken for those events.

[0012] The maintenance information generating device has the following configuration. The analysis unit performs clustering on important words and classifies events into multiple categories. With this configuration, events can be properly classified.

[0013] The analysis unit performs clustering on important words and classifies the causes of each event into multiple items. With this configuration, causes can be properly classified for each event.

[0014] The analysis unit performs clustering on key words and classifies actions into multiple categories for each cause category. With this configuration, actions can be properly classified for each cause.

[0015] The maintenance information generating device includes a status data input unit. The status data input unit accepts input of status data indicating the operating status of an event. The analysis unit associates the status data with the event, its cause, and the aforementioned action. The tree diagram generation unit adds the status data to the tree diagram using the association results from the analysis unit. With this configuration, the status at the time of the event occurrence can be confirmed using the tree diagram.

[0016] The tree diagram generator generates a tree diagram for each event and layers the tree diagram for each event. This configuration makes it easy to understand the relationship between multiple events and their causes and treatments.

[0017] The event is a failure of the equipment. With this configuration, the operator can easily understand the cause of the equipment failure and the measures to be taken to address the cause, and can easily grasp the measures to be taken to address the failure. [Effects of the Invention]

[0018] According to the present invention, maintenance information can be generated that allows easy understanding of the work that needs to be performed during maintenance. [Brief explanation of the drawings]

[0019] [Figure 1] FIG. 1 is a functional block diagram of a maintenance information generating device according to this example. [Figure 2] FIG. 2 is a diagram showing an example of text data indicating a repair history. [Figure 3] FIG. 3 is a table showing an example of important words extracted from the content of the problem. [Figure 4] FIG. 4 is a diagram showing an example of classification of fault contents. [Figure 5] FIG. 5 is a table showing an example of important words extracted from the cause of a failure. [Figure 6] FIG. 6 is a diagram showing an example of classification of fault causes. [Figure 7] FIG. 7 is a table showing an example of important words extracted from a treatment. [Figure 8] FIG. 8 is a diagram showing an example of treatment classification. [Figure 9] FIG. 9 is a diagram showing an example of a tree diagram. [Figure 10] FIG. 10 is a diagram showing an example of a layered tree diagram. [Figure 11] FIG. 11 is a flowchart showing the process of generating a tree diagram. [Figure 12] FIG. 12 is a functional block diagram of a maintenance information generating device according to a modified example. [Figure 13] FIG. 13 is a diagram illustrating an example of the state data. [Figure 14] FIG. 14 is a diagram showing an example of a tree diagram of a modified example. [Figure 15] FIG. 15 is a flowchart showing a modified example of a process for generating a tree diagram. DETAILED DESCRIPTION OF THE INVENTION

[0020] Hereinafter, an embodiment of the present invention will be described.

[0021] <1. Application Examples> FIG. 1 is a functional block diagram of a maintenance information generating device according to this example.

[0022] Text data in which repair history is written in text is input to the text data input unit 21. The text data includes, for example, the details of past failures, the causes of the failures, and the measures taken.

[0023] The key word extraction unit 31 of the analysis unit 30 extracts key words related to the problem content (problem content key words) from the text data. The event classification unit 32 of the analysis unit 30 clusters the problem content key words and classifies the problem content into multiple problem items.

[0024] The important word extraction unit 31 of the analysis unit 30 extracts important words related to the cause of the failure (important failure cause words) from the text data for each failure item. The cause classification unit 33 of the analysis unit 30 clusters the important failure cause words and classifies the failure causes into multiple cause items.

[0025] The key word extraction unit 31 of the analysis unit 30 extracts key words related to actions (key action words) from the text data for each cause item. The cause classification unit 33 of the analysis unit 30 clusters the key action words and classifies the actions into multiple action contents.

[0026] Using the above-mentioned classification results, the analysis unit 30 associates the fault items, cause items, and action contents with each other, and outputs them to the tree diagram generation unit 40.

[0027] The tree diagram generating unit 40 uses the fault items, cause items, and action contents input from the analyzing unit 30 and generates a tree diagram by referring to the set relationships.

[0028] With this configuration and processing, the maintenance information generating device 10 can generate a tree diagram that allows easy understanding of the relationship between failures, causes, and treatments from repair histories written in text. This allows workers to easily understand the work that needs to be done during maintenance by using this tree diagram (maintenance information).

[0029] <2.Configuration example> 1, the maintenance information generation device 10 includes a text data input unit 21, an analysis unit 30, and a tree diagram generation unit 40. The analysis unit 30 includes an important word extraction unit 31, an event classification unit 32, a cause classification unit 33, and a treatment classification unit 34.

[0030] (Text data input section 21) Text data indicating a repair history is input to the text data input unit 21. FIG. 2 is a diagram showing an example of text data indicating a repair history. As shown in FIG. 2, the repair history includes, for example, a job number, date and time, location, fault content, cause of the fault, and treatment. The job number indicates a serial number assigned to each job. The date and time indicates the date and time when the job was performed. The location indicates the location where the job was performed. The fault content indicates the fault phenomenon confirmed by the worker. The fault cause indicates the cause (mechanical cause, software cause, etc.) of the fault phenomenon confirmed by the worker. The treatment indicates the content of the work performed by the worker in response to the cause.

[0031] For example, in the case of task No. 1 in Figure 2, the date and time is "January 2, 2020 (2020 / 1 / 2)" and the location is "Station A." The problem is "IC card not returned." The cause of the problem is "motor damage." The corrective action is "motor replacement."

[0032] In this way, each item in the repair history is stored as a character or a string of characters (for example, text data).

[0033] The text data input unit 21 outputs the input text data to the analysis unit 30.

[0034] (Analysis Department 30) When the key word extraction unit 31 of the analysis unit 30 acquires the text data, it extracts key words for each of "content of the fault," "cause of the fault," and "measure." For example, the key word extraction unit 31 extracts key words by performing morphological analysis.

[0035] (Classification of the problem) First, the key word extraction unit 31 extracts key words from the problem details of each task number. FIG. 3 is a table showing an example of key words extracted from the problem details. For example, as shown in FIGS. 2 and 3, "IC card" and "return" are extracted as key words from task number 1. Furthermore, "change" and "return" are extracted as key words from task number 2. Furthermore, "ticket" and "not" are extracted as key words from task number 3.

[0036] The extraction of important words is realized, for example, by the following method. As a first method, a plurality of types of important words (extraction reference words) to be used during maintenance are stored in advance in the important word extraction unit 31. If there is a word identical to an extraction reference word in the text data, the important word extraction unit 31 extracts this word as an important word.

[0037] As a second method, the key word extraction unit 31 records the number of times an extracted word has been extracted, and when the number of times the extracted word has been extracted reaches or exceeds a detection threshold, the extracted word is set as a key word.

[0038] The key word extraction unit 31 outputs the extracted multiple (multiple types of) key words to the event classification unit 32.

[0039] The event classification unit 32 performs clustering on the extracted multiple important words. For example, the k-means method is used as the clustering method. By performing clustering, the event classification unit 32 classifies the fault content into multiple fault items, for example, as shown in FIG. 4. FIG. 4 is a diagram showing an example of classification of fault content. For example, in this example, based on the clustering result, the fault content is classified into multiple fault items such as "The 'card' is not returned," "The 'change' is not returned," and "The 'ticket' is not dispensed."

[0040] Furthermore, the event classification unit 32 associates a task number with each problem item, and outputs a list of task numbers (task number information) for each problem item to the key word extraction unit 31.

[0041] (Classification of causes of failure) The key word extraction unit 31 uses task number information based on the fault item to extract key words from the fault cause of each task number included in this task number information. Figure 5 is a table showing an example of key words extracted from the fault cause. For example, as shown in Figures 2 and 5, "motor" and "damage" are extracted as key words from task number 1. Also, "reading unit" and "dirt" are extracted as key words from task number 9.

[0042] The method for extracting key words for the cause of a problem uses the same concept as the method for extracting key words for the problem item described above.

[0043] The key word extraction unit 31 outputs the extracted multiple (multiple types of) key words to the cause classification unit 33.

[0044] The cause classification unit 33 performs clustering on the extracted multiple important words, similar to the event classification unit 32. By performing clustering, the cause classification unit 33 classifies the fault causes into multiple cause items, for example, as shown in FIG. 6. FIG. 6 is a diagram showing an example of classification of fault causes. For example, in this example, the fault causes are classified based on the clustering results into multiple fault causes such as "Sensor - Defective," "Motor - Breakdown," and "Insertion Port - Defective." These are classified in association with the fault item "Card not returned."

[0045] The cause classification unit 33 also associates a task number with each cause item, and outputs a list of task numbers (task number information) for each cause item to the key word extraction unit 31.

[0046] The cause classification unit 33 performs the above processing for each of the multiple fault items classified by the event classification unit 32.

[0047] (Classification of treatment) The key word extraction unit 31 uses task number information based on the cause of the fault to extract key words from the actions of each task number included in this task number information. FIG. 7 is a table showing an example of key words extracted from actions. For example, as shown in FIGS. 2 and 7, "reader" and "cleaning" are extracted as key words from task number 9. Also, "sensor" and "replacement" are extracted as key words from task number 11.

[0048] The method for extracting important words for measures uses the same concept as the method for extracting important words for the above-mentioned problem items and problem causes.

[0049] The key word extraction unit 31 outputs the extracted multiple (multiple types of) key words to the action classification unit .

[0050] The action classification unit 34 performs clustering on the extracted multiple important words, similar to the event classification unit 32 and the cause classification unit 33. By performing clustering, the action classification unit 34 classifies actions into multiple action details, for example, as shown in FIG. 8. FIG. 8 is a diagram showing an example of action classification. For example, in this example, based on the clustering results, the actions are classified into multiple action details such as "sensor" and "replacement," "reader" and "cleaning." Note that these are classified in association with the fault item "card not returned" and the fault cause "sensor failure."

[0051] The treatment classification unit 34 counts and stores the number of times the treatment is performed for each of the classified treatment contents.

[0052] The treatment classification unit 34 performs the above-mentioned treatment for each of the multiple failure causes classified by the cause classification unit 33.

[0053] The analysis unit 30 outputs the classification results of the fault items, cause items, and action contents, and the number of times each action content has been executed, in a state in which they are associated with each other, to the tree diagram generation unit 40.

[0054] (Tree diagram generation unit 40) The tree diagram generating unit 40 generates a tree diagram using the classification results of the fault items, cause items, and corrective actions associated with each other through the above analysis by the analyzing unit 30, and the number of times each corrective action has been performed.

[0055] FIG. 9 is a diagram showing an example of a tree diagram. As shown in FIG. 9, a tree diagram 101 is generated for one fault item. As shown in FIG. 9, the tree diagram 101 is shown in a tree-like form, with the fault item at the apex, multiple cause items are connected, and multiple corrective actions are connected to the multiple cause items. For example, in the case of FIG. 9, the fault item "card not returned" is connected to the cause items "sensor failure," "motor failure," and "insertion slot failure." Then, the corrective actions "sensor replacement" and "reader cleaning" are connected to the cause item "sensor failure," and the corrective action "insertion slot replacement" is connected to the cause item "insertion slot failure."

[0056] In this way, by using the configuration of the maintenance information generating device 10 of this example, it is possible to generate a tree diagram 101 in which the relationship between the details of the failure, the cause of the failure, and the measures taken is written in a tree-like form from the repair history entered in text. This allows the worker to easily understand the work that needs to be done during maintenance by referring to the tree diagram 101.

[0057] Furthermore, the tree diagram 101 lists the number of times each action has occurred. This allows the worker to easily understand the number of times each action has occurred and can refer to it during maintenance. This improves the efficiency of maintenance work.

[0058] Furthermore, the tree diagram 101 also lists the location of the problem, the type of problem, and the total number of times it has occurred, separately from the main tree diagram, allowing the worker to easily grasp the tendency of the problem to occur.

[0059] In this example, a tree diagram is generated for one type of problem item. However, it is preferable that the tree diagram generation unit 40 generates a tree diagram for each of the extracted multiple problem items. This allows the operator to easily understand the maintenance work that should be performed for all problem items.

[0060] Fig. 10 is a diagram showing an example of a layered tree diagram. As shown in Fig. 10, the tree diagram generation unit 40 generates a tree diagram 101 for the problem item "card not returned." The tree diagram generation unit 40 generates a tree diagram 102 for the problem item "change not given." The tree diagram generation unit 40 generates a tree diagram 103 for the problem item "ticket not issued." The tree diagram generation unit 40 generates a tree diagram 104 for the problem item "characters on receipt are garbled."

[0061] The tree diagram generator 40 generates a tree diagram 100 in which tree diagrams 101, 102, 103, and 104 of multiple fault items are layered, as shown in Fig. 10. This allows the operator to grasp multiple fault items, multiple fault causes, and multiple countermeasures in an organized manner, further improving the efficiency of maintenance work.

[0062] <3. Example of operation> Fig. 11 is a flowchart showing the process of generating a tree diagram. Note that the specific content of each process shown in Fig. 11 is described in the description of the configuration above. Therefore, in the following, explanations that overlap with the description of the configuration will be omitted.

[0063] The text data input unit 21 inputs text data such as repair history (S11). The key word extraction unit 31 extracts key words related to the fault content (event) from the text data (S12). The event classification unit 32 clusters the key words related to the fault content (event) and classifies them into multiple fault items (event items) (S13).

[0064] The key word extraction unit 31 extracts key words related to the cause of the failure from the text data (S14). This process is performed for each failure content. The cause classification unit 33 clusters the key words related to the cause of the failure and classifies them into multiple cause items (S15).

[0065] The key word extraction unit 31 extracts key words related to actions from the text data (S16). This process is performed for each cause of failure. The action classification unit 34 clusters the key words related to actions and classifies them into multiple action contents (S17).

[0066] The tree diagram generating unit 40 generates a tree diagram from the classified fault contents (events), fault causes, and measures (S18).

[0067] By performing such processing, the maintenance information generating device 10 can generate a tree diagram that is effective for maintenance from the text data of the repair history.

[0068] <4. Modifications> Next, a modified example of the maintenance information generating device will be described. This modified example differs from the above example in that status data (such as the operating status of a component related to a fault) is added to the tree diagram.

[0069] Fig. 12 is a functional block diagram of a maintenance information generating device according to a modified example. Fig. 13 is a diagram showing an example of status data. Fig. 14 is a diagram showing an example of a tree diagram according to a modified example. Fig. 15 is a flowchart showing the process of generating a tree diagram according to a modified example.

[0070] 12, the maintenance information generation device 10A includes a text data input unit 21, a condition data input unit 22, an analysis unit 30A, and a tree diagram generation unit 40A. The text data input unit 21 is similar to that of the maintenance information generation device 10 described above, and a description thereof will be omitted.

[0071] Status data of parts provided in the device to be treated is input to the status data input unit 22 (S21 in FIG. 15). As shown in FIG. 13, the status data includes, for example, the job number, date and time, and the status of the parts (e.g., the motor rotation speed of part A, the motor rotation speed of part B). For example, in the case of job No. 1 in FIG. 13, the date and time is "January 2, 2020 (2020 / 1 / 2) 10:00:00," the motor rotation speed of part A is "12300," and the motor rotation speed of part B is "52300."

[0072] In this way, each item of the status data is input as a character or a character string (for example, text data). The status data input unit 22 outputs the input status data to the analysis unit 30A.

[0073] The analysis unit 30A further includes a relevance detection unit 35 in addition to the components of the analysis unit 30 described above.

[0074] The association detection unit 35 refers to the task number and date and time of the text data and the task number, date and time of the status data, and associates the fault item, cause item, and remedial action content with the status of the part (FIG. 15: S22).

[0075] The analysis unit 30A outputs the associated fault item, cause item, action content, and part status to the tree diagram generation unit 40A.

[0076] The tree diagram generating unit 40A generates a tree diagram 101A in which the state of the part is added for each remedial action (FIG. 15: S18A). For example, as shown in FIG. 14, the tree diagram generating unit 40A writes "operating state at the time of failure" and "Motor A: XXX rotations" based on the state data near the remedial action "sensor replacement" in a tree formed by the fault item "card not returned," the cause item "sensor failure," and the remedial action "sensor replacement."

[0077] With this configuration, the operator can understand the state of the parts when a failure occurs and when a repair is performed from the tree diagram 101A. This increases the amount of information available for maintenance, improving maintainability.

[0078] In the above explanation, a device failure is used as an example of an event. explanation However, the event is not limited to a failure as long as the cause and treatment are associated with it.

[0079] Furthermore, the correspondence between the configuration according to the present invention and the configuration according to the above-described embodiment can be described as follows: <Additional Notes> The maintenance information generation device (10) includes a text data input unit (21), an analysis unit (30), and a tree diagram generation unit (40). The text data input unit (21) accepts input of text describing events, causes, and actions. The analysis unit (30) analyzes the text data to extract important words related to maintenance and classifies the events, causes, and actions using the important words. The tree diagram generation unit (40) uses the results of the analysis unit (30) to form a tree diagram (101) that shows the relationships between events, causes, and actions in a tree-like structure. [Explanation of symbols]

[0080] 10, 10A: Maintenance information generating device 21: Text data input section 22: Status data input section 30, 30A: Analysis department 31: Important word extraction unit 32: Event classification section 33: Cause classification department 34: Treatment classification section 35: Relevance detection unit 40, 40A: Tree diagram generation section 100, 101, 101A, 102, 103, 104: Tree diagram

Claims

1. a text data input unit that accepts input of text data describing an event, a cause, and a treatment; an analysis unit that analyzes the text data to extract important words related to maintenance and classifies events, causes, and measures using the important words; a tree diagram generating unit that uses the results of the analysis unit to generate a tree diagram that shows the relationship between the event, the cause, and the treatment in a tree form; the analysis unit performs clustering on the important words to classify the events into a plurality of categories, and for each of the classified event categories, classifies the causes into a plurality of categories, and further classifies the measures into a plurality of categories for each of the classified cause categories. Maintenance information generating device.

2. a status data input unit that receives input of status data indicating an operation status related to the event; the analysis unit associates the status data with the event, the cause, and the treatment; the tree diagram generation unit adds the state data to the tree diagram using the association result by the analysis unit. The maintenance information generating device according to claim 1 .

3. The maintenance information generating device a text data input step of accepting input of text data describing an event, a cause, and a treatment; an analysis step of analyzing the text data to extract important words related to maintenance and classifying events, causes, and measures using the important words; a graphical representation step of forming a tree diagram using the results of the analysis to show the relationship between the events, the causes, and the treatments in a tree-like form; Run The analysis step is a step of performing clustering on the important words to classify the events into multiple categories, classifying the causes into multiple categories for each of the classified event categories, and further classifying the actions into multiple categories for each of the classified cause categories, in this maintenance information generation method.

4. a text data input step of accepting input of text data describing an event, a cause, and a treatment; an analysis step of analyzing the text data to extract important words related to maintenance and classifying events, causes, and measures using the important words; a graphical representation step of forming a tree diagram using the results of the analysis to show the relationship between the events, the causes, and the treatments in a tree-like form; on the computer, The analysis step is a step of performing clustering on the important words to classify the events into multiple categories, classifying the causes into multiple categories for each of the classified event categories, and further classifying the actions into multiple categories for each of the classified cause categories.

Images