Maintenance history presentation system and maintenance history presentation method

Abstract

To extract thoroughly, from a maintenance history database, maintenance records that are targets of failure analysis and present the extracted maintenance records.SOLUTION: A maintenance history presentation system 1 includes: a maintenance history database 4 which stores maintenance records; a text analysis AI 5 which acquires, based on an anomaly code name indicating an anomaly code which is a target of failure analysis, maintenance records extracted from the maintenance history database, and extracts, based on similarities of contents recorded in the entire text, from the maintenance history database, maintenance records which include descriptions that are similar to the anomaly code name and are descriptions different from the anomaly code name, as maintenance records that are the same as or similar to the anomaly code; and a presentation unit 7 which presents results extracted by the text analysis AI.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] The present invention relates to a maintenance history presentation system and a maintenance history presentation method. [Background technology]

[0002] For example, there is provided a technique for performing failure analysis for a failure that occurs in equipment or the like (see, for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP2020-201764A Summary of the Invention [Problem to be solved by the invention]

[0004] As a method of failure analysis, there is a method of extracting a maintenance history in which an abnormality code name indicating an abnormality code to be analyzed is recorded from a maintenance history database. On the other hand, there are cases where a maintenance history in which a natural sentence input by a person in charge is recorded instead of an abnormality code name and the correlation between the abnormality code name and the maintenance history is relatively high even if there is no co-occurrence of the character string between the abnormality code name and the maintenance history is stored. A method of extracting only maintenance history in which an abnormality code name exactly matches cannot extract a maintenance history in which an abnormality code name is not recorded. Therefore, even if a maintenance history identical or similar to an abnormality code is stored, it is not possible to extract a maintenance history identical or similar to the abnormality code, and it is inevitable that the maintenance history to be analyzed will be omitted. Under these circumstances, there is a demand for extracting and presenting the maintenance history to be analyzed without omission from the maintenance history database.

[0005] The present invention has been made in consideration of the above-mentioned circumstances, and its purpose is to provide a maintenance history presentation system and a maintenance history presentation method that can extract and present maintenance history that is the subject of failure analysis from a maintenance history database without any omissions. [Means for solving the problem]

[0006] According to the invention described in claim 1, the system includes a maintenance history database (4) that stores maintenance history, a text analysis AI (5) that acquires maintenance history extracted from the maintenance history database based on an abnormality code name indicating an abnormality code that is the subject of failure analysis, and extracts from the maintenance history database maintenance history that is similar to the abnormality code name but has a different notation recorded that is different from the abnormality code name based on the similarity of the recorded content of the entire text, as maintenance history that is identical or similar to the abnormality code, and a presentation unit (7) that presents the extraction results by the text analysis AI.

[0007] Based on the similarity of the recorded contents of the entire text, the maintenance history in which a notation similar to the abnormality code name but different from the abnormality code name is recorded is extracted from the maintenance history database as the maintenance history identical or similar to the abnormality code, and the extracted result is presented. By extracting and presenting the maintenance history in which a notation similar to the abnormality code name but different from the abnormality code name is recorded from the maintenance history database as the maintenance history identical or similar to the abnormality code, the maintenance history to be the subject of the failure analysis can be properly presented without omission from the maintenance history database. [Brief description of the drawings]

[0008] [Figure 1] A functional block diagram showing the overall configuration of one embodiment. [Diagram 2] Diagram explaining the process of failure prediction [Diagram 3] Diagram explaining the process of failure analysis [Figure 4] FIG. 1 is a diagram for explaining an aspect of extracting a maintenance history. [Diagram 5] FIG. 1 is a diagram for explaining an aspect of extracting treatment candidates; DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] An embodiment will be described below with reference to the drawings. As shown in FIG. 1, the failure prediction presentation system 1 is a system for predicting failures of equipment, for example. The failure prediction presentation system 1 includes a stratification unit 2, a failure prediction AI (Artificial Intelligence) 3, a maintenance history database (DB) 4, a first text analysis AI 5 (corresponding to the text analysis AI), a second text analysis AI 6, and a display unit 7 (corresponding to the presentation unit). The stratification unit 2, the failure prediction AI 3, the maintenance history DB 4, the first text analysis AI 5, and the second text analysis AI 6 function as a control unit of the failure prediction presentation system 1, and are mainly composed of a microcomputer. The control unit of the failure prediction presentation system 1 realizes a failure prediction presentation method. The AIs used in the failure prediction AI 3, the first text analysis AI 5, and the second text analysis AI 6 perform information processing using algorithms that mimic human intelligence, and perform processes such as learning, prediction, and extraction.

[0010] The classification unit 2 reads equipment-related documents such as operation diagrams, timing charts, equipment 3D (three-dimensional) drawings, equipment wiring diagrams, spare parts lists, change point management, process FMEA (Failure Mode and Effects Analysis), and problem countermeasure lists by image recognition processing, OCR processing, etc. The equipment-related documents include several drawings and tables containing information related to the equipment's software and information related to the hardware.

[0011] The operation diagram shows the operation of the equipment, and includes information such as sensor data names and flow charts. The timing chart includes information showing the operation of the equipment in chronological order. The equipment 3D drawing includes information showing part types and part positions. The equipment wiring diagram includes information showing the electrical wiring. The spare parts list includes information showing the frequency of part replacement. The change point management includes information on the frequency of part replacement and maintenance status. The process FMEA includes information showing expected malfunctions in the process. The problem countermeasures table includes information showing countermeasures for problems in the process.

[0012] When the stratification unit 2 reads the equipment-related documents, it extracts information from the written information of the equipment-related documents read by the equipment-related documents using information extraction software, and stratifies the sensor data name, the faulty part name, and the abnormality code name from the extracted information. The faulty part name includes information indicating the part to which the sensor data indicated by the sensor data name is related and the part that causes the failure. The abnormality code name may be called a failure mode, and includes information indicating the abnormality code induced by the faulty part indicated by the faulty part name. The stratification unit 2 rearranges the information linked by keyword reference in the order of the sensor data name, the faulty part name, and the abnormality code name, and stratifies it according to the maintenance thinking. The maintenance thinking is the thinking that a maintenance worker may have when performing maintenance. In other words, the stratification unit 2 stratifies the sensor data name, the faulty part name, and the abnormality code name from the information related to the software and the information related to the hardware of the equipment extracted from the equipment-related documents, following the thinking of a maintenance worker who performs maintenance, who identifies the sensor data from the sensor data name, predicts the part to which the sensor data is related and the part that causes the failure, and predicts the abnormality code induced by the faulty part.

[0013] The failure prediction AI 3 has a learning function and a prediction function. When the failure prediction AI 3 receives and acquires the stratification result by the stratification unit 2, the failure prediction AI 3 learns the acquired stratification result and predicts a failure.

[0014] The maintenance history DB4 stores maintenance history in units of phenomenon, cause, and treatment. The maintenance history stored in the maintenance history DB4 includes maintenance history corresponding to an abnormality code name, maintenance history not corresponding to an abnormality code name, and maintenance history corresponding to a faulty part name. Maintenance history corresponding to an abnormality code name is maintenance history in which an abnormality code is recorded in the phenomenon column. Maintenance history not corresponding to an abnormality code name is maintenance history in which, for example, natural language is recorded in the phenomenon column instead of an abnormality code. Maintenance history corresponding to a faulty part name is maintenance history in which a faulty part is recorded in the cause column.

[0015] The first text analysis AI5 extracts maintenance histories from the maintenance history DB4 based on the abnormality code name indicating the abnormality code to be analyzed as a process for extracting maintenance histories. The first text analysis AI5 extracts maintenance histories from the maintenance history DB4 that are similar to the abnormality code name and have a different notation recorded based on the similarity of the recorded contents of the entire text as maintenance histories that are the same as or similar to the abnormality code.

[0016] The second text analysis AI6, as a process for extracting treatment candidates, inputs the maintenance history extracted from the maintenance history DB4 based on the abnormality code name indicating the abnormality code to be the subject of the failure analysis or the faulty part name indicating the faulty part, and extracts treatment candidates for the abnormality code or the faulty part from the maintenance history DB4 based on the recorded contents of the treatment in the input maintenance history.

[0017] The display unit 7 displays the abnormal code prediction result and the failed part prediction result as the prediction results by the failure prediction AI 3. The display unit 7 also displays the maintenance history as the extraction result by the first text analysis AI 5. Furthermore, the display unit 7 displays the candidate treatments for the abnormal code and the candidate treatments for the failed part as the extraction results by the second text analysis AI 6.

[0018] Next, the operation of the above-mentioned configuration will be described with reference to Figures 2 to 5. Here, the process flow of failure prediction and the process flow of failure analysis will be described.

[0019] (1) Flow of failure prediction process (see Figure 2) As shown in Figure 2, the stratification unit 2 reads the operation diagram and timing chart, extracts the master behavior of the sensor data from the read information using information extraction software, and stratifies the sensor data name (S1). The stratification unit 2 reads the equipment 3D drawing, extracts the positional relationships between parts and the part names from the read information using information extraction software, and stratifies the faulty part name (S2). The stratification unit 2 reads the equipment wiring diagram, extracts the electrical connection relationships between parts and the part names from the read information using information extraction software, and stratifies the faulty part name (S3).

[0020] The classification unit 2 reads the spare parts list and change point management, and extracts information on part replacement frequency and part maintenance status from the read information using information extraction software, and classifies the names of the failing parts (S4).The classification unit 2 reads the process FMEA and problem countermeasure table, and extracts parts that are likely to fail within the process from the read information, and classifies the names of the failing parts (S5).

[0021] The stratification unit 2 refers to the faulty part name by keyword (S6), and stratifies the information linked by the keyword reference according to maintenance thinking by rearranging the order of the sensor data name, the faulty part name, and the abnormality code name (S7), and outputs the stratification result to the failure prediction AI 3 (S8).

[0022] When the failure prediction AI 3 receives and acquires the stratification results from the stratification unit 2, it learns the acquired stratification results, predicts failures, and outputs the prediction results to the display unit 7. When the display unit 7 receives the prediction results from the failure prediction AI 3, it displays the abnormality code prediction result and the failed part prediction result as the input prediction results.

[0023] (2) Failure analysis process flow (see Figures 3 to 5) 3, the failure prediction AI 3 predicts an abnormality code name (S11), and extracts maintenance history from the maintenance history DB 4 using the predicted abnormality code name as a keyword (S12). The failure prediction AI 3 outputs the extracted maintenance history to the first text analysis AI 5 (S13).

[0024] The first text analysis AI5 inputs and acquires the maintenance history from the failure prediction AI3 (corresponding to a maintenance history acquisition step), and extracts the maintenance history that is similar to the abnormality code name and has a different notation recorded based on the similarity of the entire recorded content of the text as the maintenance history that is the same or similar to the abnormality code (S14, corresponding to a maintenance history extraction step), and outputs the extraction result to the display unit 7. When the display unit 7 inputs the extraction result from the first text analysis AI5, it displays the maintenance history that is the same or similar to the abnormality code as the input extraction result (corresponding to an extraction result presentation step).

[0025] As shown in FIG. 4, the first text analysis AI5 extracts a maintenance history in which "abnormality code A" is recorded in the phenomenon column, identifies "a1" as the cause of the extracted maintenance history, and identifies "a111" as the treatment. The first text analysis AI5 extracts a maintenance history in which another "abnormality code A" is recorded, identifies "a2" as the cause of the extracted maintenance history, and identifies "a222" as the treatment. "a1" and "a2" are information indicating the cause of "abnormality code A", such as code information. "a111" and "a222" are information indicating the treatment for "abnormality code A", such as code information.

[0026] When the first text analysis AI5 identifies the cause and treatment for "abnormal code A", it searches for a maintenance history that matches the identified cause and treatment. In this case, the first text analysis AI5 extracts the maintenance history of "XX failure" in which "a1" is recorded in the cause column and "a111" is recorded in the treatment column, and extracts the maintenance history of "YY corrosion" in which "a2" is recorded in the cause column and "a222" is recorded in the treatment column. That is, the first text analysis AI5 extracts the maintenance history of "XX failure" and "YY corrosion" as maintenance history identical or similar to the maintenance history of "abnormal code A". The display unit 7 displays the maintenance history of "XX failure" and "YY corrosion" extracted by the first text analysis AI5 in this manner as maintenance history identical or similar to the maintenance history of "abnormal code A".

[0027] Furthermore, the failure prediction AI 3 predicts the name of the failing part (S15), and extracts the maintenance history from the maintenance history DB 4 using the predicted failing part name as a keyword (S16).

[0028] The second text analysis AI6 extracts candidate treatments for the abnormality code and the faulty part from the maintenance history DB4 based on the recorded contents of the treatment in the maintenance history (S17), and outputs the extraction result to the display unit 7. In this case, the second text analysis AI6 extracts candidate treatments for the abnormality code and the faulty part from the maintenance history DB4 based on criteria such as frequency, length of equipment downtime, or cost of treatment. When the display unit 7 receives the extraction result from the second text analysis AI6, it displays the candidate treatments for the abnormality code and the faulty part as the input extraction result.

[0029] As shown in FIG. 5, when the second text analysis AI6 uses, for example, frequency as a criterion for candidate treatments for abnormality codes and faulty parts, if the treatment "aaa1" was the most common in the past and the treatment "aaa2" was the next most common, it extracts "aaa1" as the first candidate treatment and "aaa2" as the second candidate treatment. The display unit 7 displays the candidate treatments extracted by the second text analysis AI6 in this manner, for example, using frequency as a criterion. When the second text analysis AI6 uses the length of equipment downtime as a criterion, it extracts the treatment with the longest equipment downtime in the past as the first candidate treatment, and extracts the treatment with the second longest equipment downtime as the second candidate treatment. When the second text analysis AI6 uses the cost of treatment as a criterion, it extracts the treatment with the lowest cost in the past as the first candidate treatment, and extracts the treatment with the second lowest cost as the second candidate treatment. "n1" and "n2" indicate the number of times corresponding to frequency, and "n1" is a value greater than "n2".

[0030] In the above, the maintenance history is extracted from the maintenance history DB4 using the abnormality code name as a keyword, and the maintenance history is extracted from the maintenance history DB4 using the faulty part name as a keyword, thereby extracting candidate treatments for the abnormality code and the faulty part from the maintenance history DB4. However, candidate treatments for either the abnormality code or the faulty part may be extracted from the maintenance history DB4. That is, for example, steps S15 to S16 may be omitted, and the candidate treatments for the abnormality code may be extracted from the maintenance history DB4 simply by extracting the maintenance history from the maintenance history DB4 using the abnormality code name as a keyword. For example, steps S11 to S14 may be omitted, and the candidate treatments for the faulty part may be extracted from the maintenance history DB4 simply by extracting the maintenance history from the maintenance history DB4 using the faulty part name as a keyword.

[0031] As described above, according to this embodiment, the following advantageous effects can be obtained. In the failure prediction and presentation system 1, the software-related information and hardware-related information of the equipment extracted from the equipment-related documents are stratified according to a maintenance philosophy, the stratification results are learned to predict failures, and the prediction results are displayed. By learning the stratification results obtained by stratifying the information extracted from the equipment-related documents according to a maintenance philosophy and predicting failures, it is possible to promptly and appropriately display knowledge of both the hardware and software related to the failure when it occurs.

[0032] In addition, based on the similarity of the recorded contents of the entire text, the maintenance history in which the abnormality code name is similar to the abnormality code name and a different notation is recorded is extracted from the maintenance history database as the maintenance history identical or similar to the abnormality code, and the extraction result is displayed. By extracting and displaying the maintenance history in which the abnormality code name is similar to the abnormality code name and a different notation is recorded from the maintenance history database as the maintenance history identical or similar to the abnormality code, the maintenance history to be the subject of the failure analysis can be properly extracted and displayed from the maintenance history database without any omissions.

[0033] In addition, the system acquires a maintenance history extracted from the maintenance history database based on the abnormality code name indicating the abnormality code and the faulty part name indicating the faulty part to be analyzed, and extracts candidate treatments for the abnormality code and candidate treatments for the faulty part from the maintenance history database based on the recorded contents of the treatments in the acquired maintenance history, and displays the extraction results. By extracting and displaying candidate treatments for the abnormality code and candidate treatments for the faulty part from the maintenance history database, it is possible to appropriately extract and display candidate treatments suitable for the target of the failure analysis.

[0034] In the above, in "(2) Processing flow of failure analysis", a configuration is exemplified in which the maintenance history is extracted from the maintenance history DB4 using the abnormality code name or the faulty part name predicted by the failure prediction AI 3 as a keyword, but a configuration without using the failure prediction AI 3 is also possible. In other words, the maintenance history may be extracted from the maintenance history DB4 using the abnormality code name or the faulty part predicted by a method different from the failure prediction AI 3 as a keyword.

[0035] Although the present disclosure has been described based on the embodiment, it is understood that the present disclosure is not limited to the embodiment or structure. The present disclosure also includes various modifications and modifications within the equivalent range. In addition, various combinations and forms, as well as other combinations and forms including only one element, more than one element, or less than one element, are also within the scope and concept of the present disclosure.

[0036] The control unit and the method described in the present disclosure may be realized by a special-purpose computer provided by configuring a processor and a memory programmed to execute one or more functions embodied in a computer program. Alternatively, the control unit and the method described in the present disclosure may be realized by a special-purpose computer provided by configuring a processor with one or more dedicated hardware logic circuits. Alternatively, the control unit and the method described in the present disclosure may be realized by one or more special-purpose computers configured by a combination of a processor and a memory programmed to execute one or more functions and a processor configured with one or more hardware logic circuits. Also, the computer program may be stored in a computer-readable non-transitory tangible recording medium as instructions executed by a computer. [Explanation of symbols]

[0037] In the drawing, 1 is a failure prediction presentation system, 2 is a stratification section, 3 is a failure prediction AI, 4 is a maintenance history DB, 5 is a first text analysis AI (text analysis AI), 6 is a second text analysis AI, and 7 is a display section (presentation section).

Claims

1. A maintenance history database (4) for storing maintenance history; A text analysis AI (5) that obtains a maintenance history extracted from the maintenance history database based on an abnormality code name indicating an abnormality code to be analyzed, and extracts a maintenance history that is similar to the abnormality code name and has a different notation recorded based on the similarity of the entire text recorded content from the maintenance history database as a maintenance history that is the same as or similar to the abnormality code; A maintenance history presentation system comprising: a presentation unit (6) that presents the extraction results by the text analysis AI.

2. The maintenance history database stores maintenance history in units of phenomenon, cause, and treatment. A maintenance history presentation system as described in claim 1, wherein the text analysis AI extracts maintenance history from the maintenance history database that has a similar cause or treatment to a maintenance history in which a phenomenon is recorded with an abnormality code name, as maintenance history that is identical or similar to the abnormality code.

3. A fault prediction AI (3) is provided to predict an abnormality code; 3. The maintenance history presentation system according to claim 1, wherein the text analysis AI uses the fault code predicted by the fault prediction AI as the fault code to be subjected to fault analysis.

4. A maintenance history acquisition step for acquiring a maintenance history extracted from the maintenance history database based on an abnormality code name indicating an abnormality code to be subjected to failure analysis; a maintenance history extraction procedure for extracting, from the maintenance history database, a maintenance history that is similar to the abnormality code name and has a different notation recorded based on the similarity of the entire text recorded content, as a maintenance history that is identical or similar to the abnormality code; and an extraction result presentation step of presenting the extraction result obtained by the maintenance history extraction step.

Images