Work record management device, work record management system, and work record management program

Abstract

Information required for analyzing a future failure cause has been sometimes insufficient in a failure cause analysis system when inputting content of executed work in free description sentences and recording the content.　Provided is a work record management device for managing a work record indicating content of work for a target system, the work record management device comprising a work recording unit that includes: an interface unit that receives the work record indicating work of a worker for the target system and described in a free description format; a storage unit that stores a work record data model indicating a record item to be described as the work record; a work record structuring unit that structures the received work record by using a generative AI; and a work record confirmation unit that identifies insufficient information in the work record by using the record item.

Description

Work record management device, work record management system, and work record management program

[0001] The present invention relates to the management of work records in work on target systems such as various plants such as manufacturing factories, power plants, and water supply facilities, equipment, and various equipment such as elevators, railways, and power conversion devices. Among these, it particularly relates to the input of work records and the cause analysis of failure events. Note that the work includes various types of work such as maintenance work, management work, and construction work.

[0002] Conventionally, when detecting an abnormality in a plant, a technique for assisting work by showing work records in past events in operation has been disclosed (for example, Patent Document 1). In the technique described in Patent Document 1, when receiving abnormality information, an artificial intelligence platform searches an event database for past events highly related to the abnormal event and provides the search results.

[0003] Japanese Unexamined Patent Application Publication No. 2020-201764

[0004] However, in the conventional technique, the ease of input of work records to be referred to and the quality of the input content have not been particularly considered. In particular, in order to reduce the burden on on-site workers, it is desirable, for example, that workers can input work content in a free description form such as voice. Furthermore, when inputting work records, the work itself is delayed. Therefore, by inputting by voice, work efficiency can also be improved.

[0005] On the other hand, when utilizing artificial intelligence technologies represented by the rapidly developing generative AI technology in recent years, it is desirable that the work records contain all the necessary information for artificial intelligence learning without omission. Input in a free description form has the advantage of a low burden on workers, but there is a problem that compared with the form of inputting into a format, there is a high possibility of leakage of necessary information, that is, insufficient information.

[0006] The present invention has been made in consideration of such circumstances, and when a worker inputs a work record in a free description form, the purpose is to identify insufficient information and realize more appropriate input of work records.

[0007] To solve the above problems, the present invention provides record items that indicate the items to be recorded in the work record, analyzes the work record generated in a free-text format using generation AI, and identifies missing information in the work record using the record items.

[0008] A typical embodiment of the present invention is a work record management device for managing work records that indicate the content of work performed on a target system, the work record management device having a work record unit comprising: an interface unit that indicates the work performed by an operator on the target system and accepts work records written in a free-form format; a storage unit that stores a work record data model indicating the record items to be recorded as the work record; a work record structuring unit that structures the received work record using generating AI; and a work record confirmation unit that identifies missing information in the work record using the record items.

[0009] Furthermore, the present invention also includes a work record management system including a work record management device, a work record management program for operating a work record management device which is a computer, and a storage medium for storing the program. Moreover, a work record management method executed by the work record management device or work record management system is also an embodiment of the present invention.

[0010] According to the present invention, it is possible to appropriately identify missing information indicating missing items in the content of work records generated in a free-form description format. Problems, configurations, and effects other than those described above will be clarified by the following description of embodiments.

[0011] A diagram showing the configuration of a fault analysis system in one embodiment of the present invention. A diagram showing the details of the work record input / output unit 5 in one embodiment of the present invention. A diagram showing the details of the fault analysis input / output unit 6 in one embodiment of the present invention. A diagram showing the details of the work record unit 2 in one embodiment of the present invention. A diagram showing the details of the cause / countermeasure estimation unit 3 in one embodiment of the present invention. A diagram showing the work record data model 21 in one embodiment of the present invention. A diagram showing work record input information 511, which is one specific example in one embodiment of the present invention. A diagram showing work record input information 512, which is another specific example in one embodiment of the present invention. A flowchart showing the structuring process for work record input information 51 in the work record structuring unit 22 in one embodiment of the present invention. A diagram showing the work record content mapping instruction statement 2241 in one embodiment of the present invention. A diagram showing one specific example of the work record content mapping result in one embodiment of the present invention. A flowchart showing the process of identifying missing information in the work record in the work record confirmation unit 13 in one embodiment of the present invention. A diagram showing one specific example of the item to be confirmed 23B1 in one embodiment of the present invention. A flowchart showing the process of generating a missing inquiry statement in the missing inquiry statement generation unit 24 in one embodiment of the present invention. A diagram showing the missing information inquiry statement generation instruction 2431 in one embodiment of the present invention. A diagram showing a missing information query 52 in one embodiment of the present invention. A diagram showing a missing information response result 2461 in one embodiment of the present invention. A diagram showing a work record 2471 with the missing information response result 2461 added in one embodiment of the present invention. A diagram showing a work record content mapping result 22521 in one embodiment of the present invention. A diagram showing a work record 251 stored in the work record database 25 in one embodiment of the present invention. A diagram showing the user interface 53 of the work record input / output unit 5 in one embodiment of the present invention. A diagram showing fault status information 611 in one embodiment of the present invention. A flowchart showing the cause estimation process and countermeasure generation process in the estimated cause / countermeasure generation unit 31 in one embodiment of the present invention. A diagram showing an estimated cause / countermeasure report generation instruction in one embodiment of the present invention. A diagram showing the estimated cause / countermeasure report generation result 3151 in one embodiment of the present invention. A diagram showing the user interface 63 of the fault analysis input / output unit 6 in one embodiment of the present invention.A diagram showing work record input information 513 in one embodiment of the present invention. A diagram showing the work record content correspondence result 2253 in one embodiment of the present invention. Hardware configuration diagram of input / output device 4 in one embodiment of the present invention. Hardware configuration diagram of fault analysis device 1 in one embodiment of the present invention.

[0012] Hereinafter, one embodiment of the present invention will be described with reference to the drawings. In this embodiment, as an example of a work record management system for managing work records, a fault analysis system for analyzing faults in the target system will be used for explanation.

[0013] First, Figure 1 is a diagram showing the configuration of the fault analysis system in this embodiment. The fault analysis system includes a fault analysis device 1, an input / output device 4, and a large-scale language model device 7, which are connected to each other via a network or the like.

[0014] First, the fault analysis device 1 is an example of a work record management device for managing work records that indicate work performed on a target system, and it analyzes the cause of failures and proposed countermeasures for the target system, such as a plant or equipment. The input / output device 4 performs input and output to users such as workers. This input and output includes input of work records and output of missing information and the results of fault analysis. The input / output device 4 then transmits the input work records to the fault analysis device 1. The input / output device 4 also receives missing information and the results of fault analysis from the fault analysis device 1. Furthermore, the large-scale language model device 7 performs analysis of the work records received from the fault analysis device 1, generates text, estimates the cause of failures, and formulates countermeasures. The details of each device will be explained below. Missing information refers to record items that require user confirmation or supplementation.

[0015] The fault analysis device 1 consists of a work record unit 2, a cause / countermeasure estimation unit 3, and an interface unit 8. First, the work record unit 2 structures the work records received from the work record input / output unit 5 (described later) via the interface unit 8 according to a predetermined data model. The work record unit 2 also identifies any missing information in the received work records.

[0016] The work record unit 2 then transmits at least the missing information to the input / output device 4 via the interface unit 8. In response, the work record input / output unit 5 of the input / output device 4 outputs missing information query information, which includes the missing information. As a result, the user can be inquired about any missing information in the work record input. The missing information query information may be generated in sentence format or in word format indicating the missing information (record items). In this embodiment, a missing information query statement shown in sentence format is used as an example of missing information query information. Furthermore, the missing information query information may be the missing information itself.

[0017] Alternatively, the work record unit 2 may generate a missing information inquiry statement containing the missing information and transmit it to the input / output device 4 via the interface unit 8, or the work record unit 2 may transmit the missing information to the input / output device 4 via the interface unit 8, and the work record input / output unit 5 may use this to generate a missing information inquiry statement. Furthermore, if there is no missing information, the work record unit 2 saves the work record and notifies the cause / countermeasure estimation unit 3 of the work record in response to an inquiry from the cause / countermeasure estimation unit 3.

[0018] Furthermore, the cause / countermeasure estimation unit 3 estimates the cause and / or formulates countermeasures based on the fault conditions received from the fault analysis input / output unit 6 (described later) via the interface unit 8. In other words, the cause / countermeasure estimation unit 3 identifies the estimated cause and / or countermeasures that indicate the estimated cause and / or countermeasures. Then, the cause / countermeasure estimation unit 3 transmits the estimated cause and / or countermeasures to the fault analysis input / output unit 6 via the interface unit 8.

[0019] Furthermore, the interface unit 8 acts as a point of contact with other devices such as the input / output device 4 and the large-scale language model device 7, as well as with the user, and performs input and output with them. For this reason, the interface unit 8 can be implemented using a communication device, an input device, and an output device. The input device can be implemented using a device that accepts text information, such as a keyboard or mouse, as well as a device that accepts voice information, such as a microphone. The output device can be implemented using a display device, such as a monitor, as well as a device that outputs voice information, such as a speaker. Moreover, the input device and the output device may be configured as an integrated unit, such as a touch panel.

[0020] As described above, the fault analysis device 1 has the function of identifying missing information in the work records of the target system and performing fault analysis. For this reason, the fault analysis device 1 can be implemented using a computer such as a server. An example of this implementation will be explained later.

[0021] Furthermore, the input / output device 4 consists of a work record input / output unit 5, a fault analysis input / output unit 6, and a communication unit 9.

[0022] First, the work record input / output unit 5 receives work record input from the user. The received work record is then transmitted to the work record unit 2 of the fault analysis device 1 via the communication unit 9. The work record input / output unit 5 also receives missing information from the work record unit 2 via the communication unit 9 and outputs a missing information inquiry statement including this information. As a result, it is possible to inquire with the user about missing work records.

[0023] Furthermore, the fault analysis input / output unit 6 receives input of the fault situation from the user. The received fault situation is then transmitted via the communication unit 9 to the cause / countermeasure estimation unit 3 of the fault analysis device 1. The fault analysis input / output unit 6 also receives estimated cause and countermeasure proposals from the cause / countermeasure estimation unit 3 via the communication unit 9 and outputs them. As a result, the estimated cause and countermeasure proposals are presented to the user.

[0024] Furthermore, the communication unit 9 communicates with other devices such as the fault analysis device 1 and the large-scale language model device 7. The work record input / output unit 5 and the fault analysis input / output unit 6 serve as points of contact with the user. For this reason, these can be implemented as input devices and output devices, or they may be implemented as a single input device and output device.

[0025] Input devices can include devices that accept text information, such as keyboards and mice, as well as devices that accept audio information, such as microphones. Output devices can include display devices, such as monitors, as well as devices that output audio information, such as speakers. Furthermore, the input and output devices may be integrated into a single unit, such as a touch panel.

[0026] The work record input / output unit 5 and the fault analysis input / output unit 6 may perform processing on the information being input / output, as well as on the input / output itself. For example, the work record input / output unit 5 may merge additional user input regarding missing information with the work record input information 51.

[0027] Furthermore, the input / output device 4 can be implemented using a computer such as a terminal device represented by a smartphone, tablet, or PC. In addition, the input / output device 4 can also be implemented using a headset specifically designed for input and output.

[0028] Furthermore, the large-scale language model device 7 performs various language processing tasks, such as analyzing work records, using the large-scale language model. For this reason, the large-scale language model device 7 can be implemented using a computer such as a server. Moreover, the large-scale language model device 7 and the fault analysis device 1 may be implemented in a single device.

[0029] Next, the configuration details of the input / output device 4 and the fault analysis device 1 will be described. Figure 2 is a diagram showing the details of the work record input / output unit 5 in this embodiment. In Figure 2, the work record input / output unit 5 receives work record input information 51 indicating work records input by the user. The work record input / output unit 5 also outputs a missing information query statement 52. As a result, the user can check the missing information query statement 52, in particular the missing information.

[0030] Furthermore, the work record input information 51 is entered by the user, who is a worker, and indicates the work record for the target system. The work record input information 51 includes information such as the details of routine maintenance on the target system, such as plants and equipment, and information indicating the failure status, cause, and countermeasures taken in the event of a failure.

[0031] Furthermore, the missing information inquiry message 52 includes the missing information in the work record input information 51 and is information that requests additional input from the user. The work details that the user additionally inputs are added to the work record input information 51 by the work record input / output unit 5.

[0032] Figure 3 shows the details of the fault analysis input / output unit 6 in this embodiment. In Figure 3, the fault analysis input / output unit 6 receives fault status information 61 input from the user. The fault analysis input / output unit 6 also outputs estimated cause and proposed countermeasure information 62. As a result, the user can check the estimated cause and proposed countermeasure information 62.

[0033] Furthermore, the failure status information 61 shows the failure status entered by the user, who is a worker. For example, it is the failure status entered by the user when dealing with a new failure that has occurred in the target system.

[0034] Furthermore, the estimated cause and proposed countermeasure information 62 is generated by the fault analysis device 1 in response to the fault status information 61 and includes estimated causes and / or countermeasures estimated based on past work records.

[0035] Next, the details of the fault analysis device 1 will be described. Figure 4 is a diagram showing the details of the work record unit 2 in this embodiment. In Figure 4, the work record unit 2 includes a work record data model 21, a work record structuring unit 22, a work record verification unit 23, a missing query statement generation unit 24, and a work record database 25 (work record DB). Of these, it is desirable that the work record data model 21 and the work record database 25 be stored in the storage unit of the fault analysis device 1.

[0036] First, the work record data model 21 is a data model that defines the record items that should be included in the work record. The specific details of this will be described later using Figure 6.

[0037] Furthermore, the work record structuring unit 22 structures the work record, which is the work record input information 51 transmitted from the input / output device 4, according to the work record data model 21. The work record verification unit 23 then checks the structured work record for any missing information indicating missing items that are missing from the record items included in the work record data model 21.

[0038] In other words, the work record verification unit 23 structures the received work record and identifies missing information in the work record using the record items. If there is missing information, the missing information query generation unit 24 generates missing information query information that includes the missing information. If there is no missing information, the work record verification unit 23 saves the structured work record to the work record database 25 in the storage unit.

[0039] Furthermore, the deficiency inquiry statement generation unit 24 generates a deficiency inquiry statement 52 that includes deficiency information in the structured work record and transmits it to the input / output device 4 via the interface unit 8. As mentioned above, the deficiency inquiry statement 52 is an example of deficiency inquiry information, so the deficiency inquiry statement generation unit 24 can be called a deficiency inquiry information generation unit.

[0040] The work record database 25 stores work records that have been confirmed by the work record verification unit 23 to be free of missing information. Additionally, work records are retrieved in response to requests from the cause / countermeasure estimation unit 3. Note that the storage unit described above, particularly the work record database 25, may be implemented as a separate device from the fault analysis device 1, for example, as a file server connected via a network.

[0041] Next, Figure 5 shows the details of the cause and countermeasure estimation unit 3 in this embodiment. In Figure 5, the cause and countermeasure estimation unit 3 has an estimated cause and countermeasure proposal generation unit 31. Therefore, the function of the estimated cause and countermeasure proposal generation unit 31 may be implemented by the estimated cause and countermeasure proposal generation unit 31 itself.

[0042] The estimated cause and proposed countermeasure generation unit 31 estimates the cause of the failure in the target system and / or formulates a proposed countermeasure based on the failure status information 61 and the work records in the work record database 25, and transmits the estimated cause and proposed countermeasure information 62 via the interface unit 8.

[0043] FIG. 6 is a diagram showing the work record data model 21 in the present embodiment. In FIG. 6, the work record data model 21 has record items necessary as work records. In the present embodiment, as an example of the work record data model 21, record items necessary for performing failure cause analysis and countermeasure plan formulation by few-shot learning using a generation AI and record item IDs for identifying these are associated. And the record items include #, date and time, device ID, failure situation, failure location, cause, cause location, countermeasure, countermeasure location, and person in charge. Note that these are examples, and the record items may include at least one of these, and other items may be added.

[0044] # indicates a number for identifying the corresponding work record or the work corresponding thereto. Date and time indicates the date and time when the corresponding work record was input or the work indicated by the work record was performed. Device ID identifies the target system that is the target of the work indicated by the corresponding work record.

[0045] Further, the failure situation and the failure location respectively indicate the situation of the failure in the target system that is the target of the work and the location where the failure occurred in the target system. Also, the cause and the cause location respectively indicate the cause of the failure and the location that is the cause of the failure in the target system.

[0046] Further, the countermeasure and the countermeasure location respectively indicate the countermeasure as work and the location of the target system where the countermeasure was taken. Furthermore, the person in charge indicates the person in charge who took the countermeasure that is the target work. This person in charge is a worker, but the input of the work record may be performed by someone other than the worker (person in charge). In this case, it is desirable to use the worker and the reporter as the person in charge. Of course, the worker may be the reporter.

[0047] With the above, the description of the record items is completed, but these are merely examples, and the record items may be any items necessary as work records and can be set as appropriate. In particular, in the case of maintenance work such as inspection, the failure situation, failure location, cause, cause location, countermeasure, and countermeasure location can be omitted, and the maintenance content and maintenance location can be used.

[0048] Next, a specific example of the work record input information 51 will be described. FIG. 7 is a diagram showing work record input information 511 which is a specific example of the work record input information 51 in the present embodiment. As the work record input information example 511, the work date and time, the target device ID, the work report created and input in a free description format, and the reporter are input. In particular, as the work report, "The gas pressure is lower than the set value and an error has occurred (pressure gauge P55). The gas supply tank T1 was empty, so the tank was replaced." is shown. In this example, among the recording items shown in FIG. 6, the failure situation, the failure site, the cause, the cause site, the countermeasure, and the countermeasure site are recorded. Note that the creation and input in the free description format are also referred to as freely described below.

[0049] Also, FIG. 8 is a diagram showing work record input information 512 which is another specific example of the work record input information 51 in the present embodiment. Similarly to FIG. 7, in the work record input information example 512, the work date and time, the target device ID, the freely described work report, and the reporter are input. In particular, as the work report, "Check because the gas pressure has decreased with the pressure gauge P55. The packing of the valve V2 has deteriorated and is leaking." is shown. In this example, among the recording items shown in FIG. 6, the failure situation, the failure site, the cause, and the cause site are recorded, but the countermeasure and the countermeasure site are not recorded. Thus, in the work record input information example 512, the countermeasure and the countermeasure site are insufficient information. Below, the identification of this insufficient information and the processing necessary for this identification will be described.

[0050] FIG. 9 is a flowchart showing the structuring process for the work record input information 51 in the work record structuring unit 22 in the present embodiment. First, the process starts from step S221. Also, in step S222, the work record structuring unit 22 reads the work record input information 51 transmitted from the input / output device 4 and received by the interface unit 8. Note that in step S222, the work record input information 51 stored in the storage unit in advance may be read, or the work record input information 51 received by the input unit provided in the failure analysis device 1 may be read.

[0051] Furthermore, in step S223, the work record structuring unit 22 reads the work record data model 21 from the storage unit. If multiple work record data models 21 are stored, the work record structuring unit 22 will extract the work record data model 21 corresponding to the read work record input information 51 from the multiple work record data models 21. In this case, the work record structuring unit 22 can extract from the work record data model 21 according to the type of work transmitted from the input / output device 4.

[0052] Furthermore, in step S224, the work record structuring unit 22 generates a work record content mapping instruction statement based on the read work record input information 51 and the work record data model 21. The work record structuring unit 22 then transmits the work record content mapping instruction statement to the large-scale language model device 7 via the interface unit 8. In response, the large-scale language model device 7 generates a work record content mapping result according to the work record content mapping instruction statement.

[0053] Here, we will explain the work record content mapping instruction statement and the work record content mapping result. First, Figure 10 is a diagram showing the work record content mapping instruction statement 2241 in this embodiment. The work record content mapping instruction statement 2241 consists of an instruction prompt 2241-1 for the large-scale language model device 7, a free-form description of the work report 2241-2 referenced from the instruction prompt, and a data model 2241-3 for structuring the work report.

[0054] Furthermore, in data model 2241-3, the report entity is specified to be split into multiple items as needed (curly braces {} within square brackets []). Therefore, if multiple work items are entered in a free-text work report, the splitting process is instructed. The work record structuring unit 22 identifies the free-text text 2241-2 from the read work record input information 51 and identifies data model 2241-3 from the read work record data model 21. In addition, a predetermined instruction prompt 2241-1 can be used. As a result, a work record content mapping instruction statement 2241 is generated.

[0055] Next, we will explain the work record content mapping results 2251 and 2252, which are generated according to the work record input information 511 and work record input information 512, respectively, which are specific examples of work record input information 51.

[0056] First, Figure 11 shows a work record content mapping result 2251, which is a specific example of the work record content mapping result in this embodiment. This work record content mapping result 2251 is generated when work record input information 511 is received. For this purpose, the large-scale language model device 7 receives work record content mapping instruction statement 2241 based on the work record input information 511 and the work record data model 21, and generates the work record content mapping result 2251 based on this. At this time, the large-scale language model device 7 converts the work record entered by the user in free-form text into structured data.

[0057] Next, Figure 12 shows a work record content mapping result 2252, which is another specific example of the work record content mapping result in this embodiment. This work record content mapping result 2252 is generated when work record input information 512 is received. For this purpose, the large-scale language model device 7 receives the work record input information 512 and a work record content mapping instruction statement 2242 based on the work record data model 21, and generates the work record content mapping result 2252 based on these. At this time, the large-scale language model device 7 converts the work record entered by the user in free-form text into structured data.

[0058] Here, the structured result of the work record content mapping 2252 shows that the countermeasures and the parts affected are "N / A (Not Available)" (missing information). This is because the work record input information 512 entered by the user states, "The gas pressure was low on pressure gauge P55, so it was checked. The packing of valve V2 had deteriorated and was leaking," but the implemented countermeasures and the parts affected were not entered. This concludes the explanation of the work record content mapping results, and we will now return to the explanation of the flowchart.

[0059] In step S225, the work record structuring unit 22 receives the work record content mapping result from the large-scale language model device 7 via the interface unit 8. Then, in step S226, the process ends.

[0060] By adopting the configuration of this embodiment as described above, the free-form text entered by the user is structured based on a predefined work record data model, thereby reducing the burden on the user when entering work reports. Furthermore, it becomes possible to input using free-form text. For this reason, by using a device capable of handling voice information, such as a headset, as the input / output device 4, it becomes possible to input work records by voice even when both hands are occupied and the keyboard or touch panel cannot be used, such as when performing maintenance work. As a result, the degree of freedom for the user is improved, and the efficiency of work can be increased.

[0061] Furthermore, by adopting the configuration of this embodiment described above, the free-form text entered by the user can be structured based on the predefined work record data model 21. This has the effect of enabling the detection of missing information that should be entered when utilizing work record data.

[0062] Next, we will explain the process for identifying missing information in the work record. Figure 13 is a flowchart showing the process for identifying missing information in the work record in the work record verification unit 23 in this embodiment.

[0063] First, processing begins from step S231. Then, in step S232, the work record confirmation unit 23 reads the work record content mapping result received in step S225. As this work content mapping result, the work record content mapping results 2251 and 2252 shown in Figures 11 and 12 can be used.

[0064] Furthermore, in step S233, the work record confirmation unit 23 reads the work record data model 21 corresponding to the read work record content mapping result from the storage unit or the like. Then, in step S234, it sets the record item ID to 1. Here, if multiple work record content mapping results were read in step S232, the record item IDs are set from 1 in any order (ascending order).

[0065] In step S235, the work record verification unit 23 checks whether a value has been entered for the record item ID in the read work record content mapping result. If a value has been entered (YES), the process proceeds to step S237; if the value is N / A (not entered: NO), the process proceeds to step S236. In step S236, the work record verification unit 23 adds the record item IDs for which no value was entered to the items requiring verification.

[0066] In step S237, the work record confirmation unit 23 determines whether the current record item ID is the last item. If it is the last item, the unit proceeds to step S239; otherwise, it proceeds to step S238. In other words, the work record confirmation unit 23 determines whether each record item is not entered (N / A) or is entered.

[0067] In step S238, 1 is added to the record item ID (+1) and the process proceeds to step S235. Steps S235 to S238 may be omitted. Furthermore, steps S235 to S238 may be executed, and step S234 may be omitted.

[0068] Furthermore, in step S239, the work record verification unit 23 determines whether or not there are items that need to be checked. In other words, the work record verification unit 23 determines whether there is missing information, which is a record item that has not been entered for each record item. If there are items that need to be checked (missing information), the process proceeds to step S23A; otherwise, the process proceeds to step S23B. For example, in the case of the work record content correspondence result 2252 shown in Figure 12, it is determined that the countermeasures and countermeasure parts are items that need to be checked among the record items.

[0069] In step S23A, the work record verification unit 23 records the work record content matching result in the work record database 25 and proceeds to step S23C. Also, in step S23B, the work record verification unit 23 outputs the items requiring verification to the missing inquiry statement generation unit 24 and proceeds to step S23C. Finally, in step S23C, the process ends.

[0070] Here, we will describe a specific example of an item requiring confirmation that is output in step S23B. Figure 14 is a diagram showing a specific example of an item requiring confirmation 23B1 in this embodiment. In this example, item requiring confirmation 23B1 is a confirmation item for the work record input information 512, and it indicates that additional input of information on the implemented countermeasures and the parts of the countermeasures is required, that is, these are missing information that require confirmation. In this example, the work report of the work record input information 512 is also included as reference information. Reference information is optional, and other items may be used. Furthermore, other record items that are not missing information, such as the cause of the corresponding work record content mapping result, may be used as reference information. Furthermore, the work record content mapping result itself may be used as item requiring confirmation 23B1.

[0071] By adopting the configuration of this embodiment described above, only work reports that can be used as training data for failure analysis from among the work reports entered by users in free-form text format are stored in the work report database, which has the effect of enabling failure cause estimation using high-quality training data.

[0072] Next, the process for generating the missing query statement 52 using the item to be confirmed 23B1 will be explained. Figure 15 is a flowchart showing the process for generating the missing query statement in the missing query statement generation unit 24 in this embodiment.

[0073] First, processing begins from step S241. Then, in step S242, the missing inquiry statement generation unit 24 reads the transmitted items requiring confirmation 23B1.

[0074] Furthermore, in step S243, the missing information query generation unit 24 generates a missing information query generation instruction using the item to be confirmed 23B1 and transmits it to the large-scale language model device 7 via the interface unit 8. For example, the missing information query generation unit 24 generates a missing information query generation instruction that includes the contents of the item to be confirmed 23B1 and a generation instruction (instruction prompt), and transmits it.

[0075] Here, we will explain the instruction for generating a missing information query statement. Figure 16 shows the missing information query statement generation instruction 2431 in this embodiment. In Figure 16, the missing information query statement generation instruction 2431 includes an instruction prompt 2431-1 for the large-scale language model device 7, a reference work report 2431-2, and missing information 2431-3. As a result, the large-scale language model device 7 generates a missing information query statement in response to the missing information query statement generation instruction 2431.

[0076] Then, in step S244, the deficiency query statement generation unit 24 receives the deficiency information query statement, which is the result of generating the deficiency information query statement, from the large-scale language model device 7 via the interface unit 8. Now, the deficiency information query statement will be explained. Figure 17 is a diagram showing the deficiency information query statement 52 in this embodiment. As shown in Figure 17, the deficiency information query statement 52 is a question (in sentence form) to the user regarding deficiency information.

[0077] Next, in step S245, the deficiency inquiry statement generation unit 24 transmits the deficiency information inquiry statement 52 to the input / output device 4 via the interface unit 8. In response, the work record input / output unit 5 of the input / output device 4 outputs the deficiency inquiry statement 52 to the user, thereby enabling the user to inquire about deficiency information. The work record input / output unit 5 then receives the deficiency information response result to the deficiency information inquiry from the user, that is, additional input regarding the deficiency information. In response, the work record input / output unit 5 transmits this to the fault analysis device 1 via the communication unit 9.

[0078] Then, in step S246, the deficiency inquiry statement generation unit 24 reads the deficiency information response result for the deficiency information inquiry statement 52 received from the user via the interface unit 8. Here, this response result will be explained. Figure 18 is a diagram showing the deficiency information response result 2461 for the deficiency information in this embodiment. The deficiency information response result 2461 includes the implemented countermeasures and the parts of the countermeasures taken as the user's response to the deficiency information inquiry statement 2441.

[0079] Furthermore, in step S247, the missing information query generation unit 24 sends the work record with the missing information response result 2461 added to it to the work record structuring unit 22. Here, the work record with the missing information response result 2461 added will be explained. Figure 19 shows the work record 2471 with the missing information response result 2461 added in this embodiment. In addition to the work report initially entered by the user, the work record 2471 also contains the missing information response result 2461 (bottom of the figure) that was added in step S246.

[0080] The work record structuring unit 22 then checks if there is any missing information in the work record (additional information) added by the user. If there is no missing information, the work record structuring unit 22 structures the additional information. In response, the work record confirmation unit 23 adds the work record, including the structured additional information, to the storage unit. If there is missing information, the work record structuring unit 22 queries for the missing information again. This verification can be performed as explained in Figure 13. When adding to the storage unit, the work record structuring unit 22 performs data structuring based on the work record 2471 with the missing information response result 2461 added, and generates the work record content mapping result 22521. In this way, the work record confirmation unit 23 stores the work record corresponding to the missing information response result 246 to the missing information query generation unit 24 in the storage unit.

[0081] In this embodiment, it is desirable that the work record confirmation unit 23 transmits to the input / output device 4 via the interface unit 8 that the work record content mapping result 22521 has been generated. As a result, the work record input / output unit 5 outputs a message indicating that it has been generated and a request for confirmation. The work record input / output unit 5 then confirms the work record content mapping result 22521 according to the user's response to the confirmation request.

[0082] Here, we will explain in detail the work record content mapping result 22521 and the work record 251 stored in the work record database 25, which are the processing results of the above process. First, Figure 20 is a diagram showing the work record content mapping result 22521 in this embodiment. The work record content mapping result 22521 has the same configuration as the work record content mapping results in Figures 11 and 12, and the missing information (N / A) is resolved by the additional information provided by the user. Figure 21 is a diagram showing the work record 251 stored in the work record database 25 in this embodiment. Figure 21 shows that the work record database 25 is composed of multiple work records (#1, #2) from which the missing information has been resolved. This concludes the explanation of each piece of information. The information is registered in this way, and the process ends in step S248.

[0083] Next, we will describe the user interface from the input of work records in each of the above processes (flowcharts) to the generation of the work record content mapping result 22521 and the request for confirmation thereof. Figure 22 is a diagram showing the user interface 53 of the work record input / output unit 5 in this embodiment.

[0084] In Figure 22, the user interface 53 displays the following items. The work report mode 5301 indicates that the current input screen is the work report input screen. In the user interface 53 of this embodiment, it is displayed in a chat format, with a speech bubble from the left of the drawing indicating user input, and a speech bubble from the right of the drawing indicating AI output from the system side (fault analysis device 1). The date and time input field 5302 is for entering the current date and time. In addition to manual input by the user, the date and time information may also be automatically entered as the current time. The equipment ID input field 5303 accepts input of an equipment ID that identifies the system to be worked on. The equipment ID may be obtained by activating the camera of the input / output device 4 via barcode reading 5304 and reading the barcode attached to the equipment.

[0085] Furthermore, the worker ID is entered in the worker ID input field 5305. The worker ID may also be obtained by activating the camera of the input / output device 4 using the barcode reader 5306 and reading the barcode attached to the worker's badge, work clothes, helmet, etc. The user input field 5307 and the support AI output field 5308 indicate user input in the speech bubble on the left side of the screen and AI output in the speech bubble on the right side of the screen, respectively.

[0086] Furthermore, user input 5309 indicates free-form input of a work report by the user. AI output 5310 indicates an output of a query for missing information regarding user input 5309. User input 5311 indicates the user's response to AI output 5310 regarding missing information.

[0087] Furthermore, the AI ​​output 5312 structures the data based on user inputs 5309 and 5311, and if there is no missing information, it displays the information to be registered in the database, allowing the user to confirm it. The registration permission button 5313 is a button to allow registration in the database based on the contents of the AI ​​output 5312. The registration denial button 5314 is a button to instruct the system not to register the data in the database if the contents of the AI ​​output 5312 differ from the user's intention.

[0088] The user input and AI output described above are presented as screen images and text for illustrative purposes, but they may also be treated as audio information.

[0089] By adopting the configuration of this embodiment described above, it is possible to detect when there is insufficient information in the work report entered by the user in free-form text format that is not suitable for use as training data for failure analysis, and to input the missing information. This has the effect of enabling failure cause estimation using high-quality training data, even with work reports entered in free-form text or voice input.

[0090] In the above, we added missing information and supplemented the work records. Next, we will explain how to estimate the cause of the failure in the target system and generate countermeasures. In this embodiment, we use failure status information indicating the status of the failure, which is entered by the user, and the supplemented work records.

[0091] Figure 23 shows the fault status information 611 in this embodiment. The fault status information 611 is received by the fault analysis input / output unit 6 in response to input from the user. In Figure 23, the fault status information 611 contains the fault status and the location where the fault occurred. Data structuring and verification may be performed on the fault status information, and if there is missing information, additional input may be requested.

[0092] Next, Figure 24 is a flowchart showing the cause estimation process and the countermeasure proposal generation process in the estimated cause / countermeasure proposal generation unit 31.

[0093] First, processing begins at step S311. Then, in step S312, the estimated cause / countermeasure generation unit 31 reads the fault status information 611 transmitted from the input / output device 4 and received by the interface unit 8. For this purpose, the fault status information 611 is received by the fault analysis input / output unit 6 as described above and transmitted to the fault analysis device 1 via the communication unit 9.

[0094] Furthermore, in step S313, the estimated cause / countermeasure generation unit 31 reads work records corresponding to the failure status information 611 from the work record database 25. At this time, the estimated cause / countermeasure generation unit 31 searches the work record database 25 using work items such as faulty equipment included in the failure status information 611 as keys, and can identify the relevant work record.

[0095] Furthermore, in step S314, the estimated cause / countermeasure generation unit 31 transmits an instruction to generate an estimated cause / countermeasure report to the large-scale language model device 7 via the interface unit 8. To this end, the estimated cause / countermeasure generation unit 31 generates an instruction to generate an estimated cause / countermeasure report using the read failure status information 611 and work records.

[0096] Here, we will explain the instruction for generating an estimated cause and proposed countermeasure report. Figure 25 is a diagram showing the instruction 3141 for generating an estimated cause and proposed countermeasure report in this embodiment. The instruction 3141 for generating an estimated cause and proposed countermeasure report consists of an instruction prompt 3141-1 for the large-scale language model device 7, a newly occurring fault situation 3141-2, past work records (history) 3141-3 learned in Few-Shot Learning, and an output format 3141-4. The fault situation 3141-2 is generated based on the loaded fault situation information 611. The past work records (history) 3141-3 are generated from the loaded work records. The instruction prompt 3141-1 and the output format 3141-4 may be predetermined or determined according to the fault situation information 611.

[0097] Furthermore, in step S315, the estimated cause / countermeasure proposal generation unit 31 receives the estimated cause / countermeasure proposal report generation result from the large-scale language model device 7 via the interface unit 8. At this time, the large-scale language model device 7 estimates the cause of the failure, generates a countermeasure proposal, and creates the estimated cause / countermeasure proposal report generation result in response to the estimated cause / countermeasure proposal report generation instruction 3141. The estimated cause / countermeasure proposal report generation result will now be explained.

[0098] Figure 26 shows the details of the estimated cause and proposed countermeasure report generation result 3151 in this embodiment. In Figure 26, the estimated cause and proposed countermeasure report generation result 3151 includes fault status information 3151-1 and estimated cause and proposed countermeasure 3151-2. Fault status information 3151-1 corresponds to fault status 3141-2 of the estimated cause and proposed countermeasure report generation instruction 3141. Estimated cause and proposed countermeasure 3151-2 shows the estimated cause of the fault and its proposed countermeasure, and in this example, (1) and (2) show multiple causes and their proposed countermeasures. Then, the process ends in step S316.

[0099] Next, we will explain the user interface for estimating the cause of the above-mentioned failures and generating countermeasures. Figure 27 is a diagram showing the user interface 63 of the failure analysis input / output unit 6 in this embodiment. In Figure 27, the user interface 63 displays the following items. First, the failure analysis mode 6301 indicates that the current input screen is the business analysis input / output screen. The date and time input field 6302 accepts input of the current date and time. The date and time information may be manually entered by the user, or it may be automatically entered as the current time.

[0100] Furthermore, the equipment ID input field 6303 accepts input of the equipment ID of the equipment to be worked on. The equipment ID may be entered by activating the camera of the input / output device 4 using the barcode reader 6304 and reading the barcode attached to the equipment. The worker input field 6305 accepts input of the worker's ID. The worker's ID may be entered by activating the camera of the input / output device 4 using the barcode reader 6306 and reading the barcode attached to the worker's badge, work clothes, helmet, etc.

[0101] The user input field 6307 and the support AI output field 6308 indicate that the speech bubble on the left side of the screen is for user input and the speech bubble on the right side of the screen is for AI output, respectively. User input 6309 accepts free-form input of the failure situation by the user. AI output 6310 shows the estimated cause and proposed countermeasures generated by the estimated cause and countermeasures generation unit 31, that is, the content of the estimated cause and countermeasures report generation result 3151.

[0102] Through the above process, the user (operator) can perform the countermeasures (work) indicated in the estimated cause / countermeasure report generation result 3151 for the target system where the failure occurred. The user then inputs work record input information 513, which is a record of the work performed based on the estimated cause / countermeasure report generation result 3151. In other words, the work record input / output unit 5 receives the input of work record input information 513 from the user. Now, let's explain the work record input information 513. Figure 28 is a diagram showing the work record input information 513 input by the user in this embodiment. In Figure 28, the work record input information 513 includes a freely written work report. This work report contains two work items. Note that the work report is not limited to two work items; it may contain one or three or more work items.

[0103] In this embodiment, the work record input / output unit 5 transmits the work record input information 513 to the fault analysis device 1 via the communication unit 9. The work record structuring unit 22 of the fault analysis device 1 then structures the work record input information 513.

[0104] Figure 29 shows the work record content mapping result 2253, which is a structured representation of the work record input information 513 in this embodiment. The work record structuring unit 22 divides and structures multiple (two in the example shown) work contents for the freely written work report in the work record input information 513.

[0105] By adopting the configuration of this embodiment described above, it becomes possible to record work reports entered by the user in free-form text, even when multiple work items are performed in a single task, by dividing the report into separate entries. This has the effect of enabling fault cause estimation using high-quality training data that can be used as training data for fault analysis.

[0106] Next, an example of the implementation of the input / output device 4 and the fault analysis device 1 in this embodiment will be described. First, Figure 30 is a hardware configuration diagram of the input / output device 4 in this embodiment. In Figure 2, the input / output device 4 can be implemented as a so-called smartphone and has a touch panel 401, a processing device 402, a communication device 403, and a storage device 404, which are connected to each other via a communication path.

[0107] First, the touch panel 401 has a configuration corresponding to the work record input / output unit 5 and the fault analysis input / output unit 6 in Figure 1, and receives user instructions and operations, and displays various information. Here, the touch panel 401 may be configured separately as an input device and an output device. Of the functions of the work record input / output unit 5 and the fault analysis input / output unit 6 in Figure 1, the information processing of the received information can be executed by the processing unit 402 according to the work record management program 405 described later.

[0108] Furthermore, the processing unit 402 can be implemented using a processor such as a CPU (Central Processing Unit), and it performs calculations according to the work record management program 405 stored in the storage device 404, which will be described later. The work record management program 405 is a type of application.

[0109] The work record management program 405 is composed of a work record input / output module 406 and a fault analysis input / output module 407, each with its own function. These modules may be implemented as individual programs or in combinations thereof.

[0110] Here, the processing unit 402 performs information processing on the information received by the work record input / output unit 5 and the fault analysis input / output unit 6, respectively, according to the work record input / output module 406 and the fault analysis input / output module 407.

[0111] Furthermore, the communication device 403 corresponds to the communication unit 9 in Figure 1 and communicates with other devices such as the fault analysis device 1 via the network 100 described later. The storage device 404 stores the work record management program 405, work record input information 51, fault status information 61, and estimated cause / countermeasure information 62. The work record management program 405 is stored in various storage media such as the storage device 404.

[0112] Furthermore, although omitted in Figure 30, it is desirable that the input / output device 4 be equipped with a camera for reading barcodes, a microphone for inputting audio information, and a speaker for outputting audio information. The microphone and speaker correspond to the work record input / output unit 5 and the fault analysis input / output unit 6 in Figure 1.

[0113] Next, Figure 31 is a hardware configuration diagram of the fault analysis device 1 in the embodiment of this present invention. In Figure 31, the fault analysis device 1 includes a processing unit 101, a communication device 102, a memory 103, and a sub-storage device 104, which are connected to each other via a communication path.

[0114] First, the processing unit 101 can be implemented using a processor such as a CPU, and it performs calculations according to the fault analysis program 105 stored in the secondary memory device 104, which will be described later. The fault analysis program 105 will be described later.

[0115] Furthermore, the communication device 102 corresponds to the interface unit 8 in Figure 1, connects to the network 100, and communicates with other devices such as the input / output device 4 and the large-scale language model device 7. The functions of the input / output device 4 can be provided in the fault analysis device 1. In this case, the fault analysis device 1 would further include input devices such as a keyboard and microphone, and output devices such as a monitor and speakers.

[0116] Furthermore, the memory 103 and the sub-storage device 104 correspond to the storage units that store the work record data model 21 and the work record DB 25 in Figure 4. The memory 103 then displays the information used for processing by the fault analysis program 105 and the processing unit 101, which are stored in the sub-storage device 104. The sub-storage device 104 can be implemented as a so-called storage device and stores the fault analysis program 105, the work record data model 21, and the work record database 25. The sub-storage device 104 may also be implemented as various storage media such as an external HDD (Hard Disk Drive), SSD (Solid State Drive), or memory card, or it may be implemented as a separate device from the fault analysis device 1, such as a file server.

[0117] Here, the failure analysis program 105 consists of a work record module 106 and a cause / countermeasure estimation module 107. Note that each of these modules may be implemented as an individual program or a combination of some of them.

[0118] Here, the configuration shown in Figure 1, in which each module performs the same function, is as follows: Work recording module 106: Work recording unit 2 Cause and countermeasure estimation module 107: Cause and countermeasure estimation unit 3 Therefore, the processing unit 101 will execute the processing of the work recording unit 2 and the cause and countermeasure estimation unit 3 according to the failure analysis program 105. Note that each module constituting the failure analysis program 105 may be implemented as a program composed of independent modules or combinations of some modules.

[0119] This concludes the explanation of the implementation example of the input / output device 4 and fault analysis device 1 in this embodiment. In addition, although only one input / output device 4 is described in this embodiment, multiple input / output devices 4 may be used. In this case, the fault analysis device 1 may be constructed as a cloud system that can be used by multiple organizations such as companies. In this example, security is ensured by dividing and managing resources for the fault analysis device 1 for each organization.

[0120] It should be noted that the present invention is not limited to the embodiments described above, and various modifications are included. For example, the embodiments described above are described in detail for the purpose of explaining the present invention in an easy-to-understand manner, and are not necessarily limited to those having all the configurations described. Furthermore, it is possible to replace some of the configurations of one embodiment with other configurations. It is also possible to add the configurations of other specific examples to one specific example of this embodiment. Furthermore, it is possible to add, delete, or replace some of the configurations of this example with other configurations. In addition, some or all of the above configurations, functions, processing devices, parts, etc., may be realized in hardware, for example, by designing them as integrated circuits.

[0121] 1...Fault analysis device, 2...Work record unit, 3...Cause and countermeasure estimation unit, 4...Input / output device, 5...Work record input / output unit, 6...Fault analysis input / output unit, 7...Large-scale language model device, 8...Interface unit, 9...Communication unit, 21...Work record data model, 22...Work record structuring unit, 23...Work record verification unit, 24...Incomplete inquiry statement generation unit, 25...Work record database, 31...Estimated cause and countermeasure proposal generation unit.

Claims

1. A work record management device for managing work records that show the content of work performed on a target system, the work record management device having a work record unit comprising: an interface unit that shows the work performed by an operator on the target system and accepts work records written in a free-form format; a storage unit that stores a work record data model that shows the record items to be recorded as the work record; a work record structuring unit that structures the received work record using generating AI; and a work record confirmation unit that identifies missing information in the work record using the record items.

2. A work record management device according to claim 1, wherein the work record unit outputs the missing information via the interface unit.

3. A work record management device according to claim 2, wherein the work record unit further comprises a deficiency inquiry information generation unit that generates deficiency inquiry information including the deficiency information, and the work record management device outputs the deficiency inquiry information via the interface unit.

4. A work record management device according to claim 3, wherein the interface unit is connected to an input / output device used by the user, receives the work record input to the input / output device, outputs the missing information to the input / output device, and the input / output device outputs the missing information to the user.

5. A work record management device according to claim 1, wherein the work record data model uses one of the following as a record item: failure status, failure location, cause, cause location, countermeasure, and countermeasure location, relating to equipment failure in the target system.

6. A work record management device according to claim 1, wherein the work record structuring unit divides the work content if the work record contains two or more work contents.

7. A work record management device according to claim 3, wherein the work record confirmation unit stores the work record in the storage unit if the work record does not contain the missing information.

8. A work record management device according to claim 7, wherein the deficiency inquiry information generation unit receives the deficiency information response result for the deficiency information transmitted via the interface unit, and stores the work record corresponding to the deficiency information response result in the storage unit.

9. A work record management device according to claim 4, wherein the input / output device performs input from the user and output to the user in the form of voice information.

10. A work record management device according to claim 1, further comprising an estimated cause / countermeasure generation unit that estimates the cause of a failure in the target system and plans countermeasures for the failure based on the work records stored in the storage unit and the failure status of the target system received via the interface unit.

11. A work record management system comprising a work record management device for managing work records indicating the content of work performed on a target system and an input / output device for use by a user, wherein the work record management device comprises an interface unit connected to the input / output device that indicates the work performed by an operator on the target system and accepts work records described in a free-form format; a storage unit that stores a work record data model indicating the record items to be recorded as the work record; a work record structuring unit that structures the received work record using generating AI; and a work record confirmation unit that identifies missing information in the work record using the record items.

12. A work record management system according to claim 11, wherein the work record unit outputs the missing information to the input / output device via the interface unit.

13. A work record management program for causing a work record management device, which is a computer for managing work records that show the content of work performed on a target system, to function as a work record unit comprising: an interface unit that shows the work performed by an operator on the target system and accepts work records written in a free-form format; a storage unit that stores a work record data model that shows the record items to be recorded as the work record; a work record structuring unit that structures the received work record using generating AI; and a work record confirmation unit that identifies missing information in the work record using the record items.

14. A work record management program according to claim 13, wherein the work record unit outputs the missing information via the interface unit.

Images