Abnormal event handling support device and abnormal event handling support method

Abstract

To support handling of an abnormal event on the basis of a document relating to the past abnormal event.SOLUTION: An abnormal event handling support device 100 includes: a reception unit 111 that receives identification information for an abnormal event occurrence apparatus and occurrence abnormality information as a content / status of the abnormal event; a similar event extraction unit 112 that refers to an event report database 140 storing a report including abnormal event information for the abnormal event that occurred in the past to extract a first abnormal event of the abnormal event information similar to the occurrence abnormal information, and refers to an apparatus information database 130 to extract a second abnormal event as an abnormal event of an apparatus similar to apparatus information for the abnormal event occurrence apparatus, and determines the abnormal event of both the first abnormal event and the second abnormal event as a similar abnormal event; a prompt generation unit 113 that generates a prompt to be input to a language model, including a content of the report of the occurrence abnormal information and the similar abnormal event; and an answer acquisition unit 114 that inputs the prompt to the language model to acquire an answer to output the answer.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an abnormal event response support device and an abnormal event response support method that support responses to abnormal events that occur in equipment. [Background technology]

[0002] When providing services using facilities and equipment, the most time-consuming and labor-intensive tasks are analyzing the causes of abnormal events, including malfunctions, breakdowns, and failures in the facilities and equipment, identifying the scope of the impact, and creating countermeasures, including recovery. Conventionally, this has involved referencing a database of past abnormal events to find similar abnormal events and using this information when analyzing the causes and creating countermeasures.

[0003] Meanwhile, systems / services (interactive artificial intelligence services, chat services) that use large-scale language models (LLMs) to conduct question-and-answer sessions using natural language are being offered. LLMs are natural language processing models trained on large amounts of text data, and answers can be obtained by inputting prompts, which are text containing requests, demands, or questions, into the LLMs. However, because the text data used for training is large, there is a problem in that unintended answers may be obtained.

[0004] Adding reference information to a prompt makes it easier to obtain an answer that takes the reference information into consideration. The sentence generation device described in Patent Document 1 extracts sentences similar to the input question from a pre-registered sentence database and includes them in the prompt of a large-scale language model. This makes it possible to obtain an answer in accordance with the reference information contained in the sentence database.

[0005] The method of registering documents and other information to be used as reference in a text database, extracting portions of text that are related to (highly similar to) the question from this database, and using them as additional information for the prompt is called RAG (Retrieval-augmented Generation).RAG is attracting attention as a method for incorporating internal data into large-scale language models without training the large-scale language model itself. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] Patent No. 7325152 Summary of the Invention [Problem to be solved by the invention]

[0007] It is believed that by asking questions about current abnormal events to a system that uses a large-scale language model and RAG and stores reports of past abnormal events in a text database, it will be possible to more efficiently analyze the causes and create countermeasures based on similar past abnormal events.

[0008] If the current event is a fairly general anomaly, there is a high possibility of obtaining an appropriate answer. However, if the anomaly occurs on a specific device or under specific conditions, including text obtained through a text-based search in the prompt may not necessarily present appropriate past anomalies. In particular, if there is a large amount of text (reports of anomalies), there is a high possibility that appropriate information for dealing with the anomaly, such as cause analysis or countermeasure creation, will not be obtained.

[0009] The present invention has been made in view of the above background, and has as its object to provide an abnormal event response support device and an abnormal event response support method that support the response to abnormal events based on documents related to past abnormal events. [Means for solving the problem]

[0010] In order to solve the above-mentioned problems, an abnormal event response support device according to the present invention includes: a receiving unit that receives identification information of an abnormal event occurrence device, which is a device in which an abnormal event has occurred, and occurred abnormality information, which is content / status of the abnormal event; a similar event extraction unit that refers to an event report database that stores reports including abnormal event information related to abnormal events that have occurred in devices in the past, to extract a first abnormal event, which is an abnormal event described in the report including abnormal event information similar to the occurred abnormality information, and refers to a device information database that stores device information related to the devices, to extract a second abnormal event, which is an abnormal event of a device similar to the device information of the abnormal event occurrence device, and classifies abnormal events that are both the first abnormal event and the second abnormal event as similar abnormal events; a prompt generation unit that generates a prompt to be input to a language model, which is a question related to the abnormal event that has occurred, and which includes the occurred abnormality information and the content of the report of the similar abnormal event; and an answer acquisition unit that inputs the prompt to the language model, and acquires and outputs an answer to the prompt from the language model. [Effects of the Invention]

[0011] According to the present invention, it is possible to provide an abnormal event response support device and an abnormal event response support method that support responses to abnormal events based on documents related to past abnormal events. Problems, configurations, and effects other than those described above will become clear from the description of the following embodiments. [Brief explanation of the drawings]

[0012] [Figure 1] 1 is a functional block diagram of an abnormal event response support device according to a first embodiment. [Figure 2] FIG. 2 is a data configuration diagram of a device information database according to the first embodiment. [Figure 3] FIG. 2 is a data configuration diagram of a knowledge base according to the first embodiment. [Figure 4] 10 is a prompt template according to the first embodiment. [Figure 5]4 is a flowchart of an abnormal event response support process according to the first embodiment. [Figure 6] FIG. 4 is a diagram illustrating the screen configuration of an abnormal event response support screen according to the first embodiment. [Figure 7] FIG. 4 is a diagram illustrating the screen configuration of an abnormal event response support screen according to the first embodiment. [Figure 8] FIG. 10 is a functional block diagram of a malfunction response support device (abnormal event response support device) according to a second embodiment. [Figure 9] FIG. 11 is a data configuration diagram of a device information database according to the second embodiment. [Figure 10] FIG. 10 is a data configuration diagram of a knowledge base according to the second embodiment. [Figure 11] 10 is a flowchart of a malfunction response support process according to the second embodiment. [Figure 12] FIG. 11 is a diagram illustrating the screen configuration of a malfunction handling support screen according to the second embodiment. [Figure 13] 10 is a prompt template according to the second embodiment. [Figure 14] FIG. 11 is a diagram illustrating the screen configuration of a malfunction handling support screen according to the second embodiment. [Figure 15] FIG. 2 is a hardware configuration diagram illustrating an example of a computer that realizes the functions of the abnormal event response support device and the malfunction response support device according to the above-described embodiments. DETAILED DESCRIPTION OF THE INVENTION

[0013] <<Outline of the abnormal event response support device>> An abnormal event response support device in a mode (embodiment) for carrying out the present invention will be described below. The abnormal event response support device has a knowledge base of reports and device information about abnormal events that have occurred in the past in facilities or devices (hereinafter simply referred to as devices). The report includes, for example, the model (model number) and identification number of the device in which the abnormal event occurred, the date and time the abnormal event occurred, the state / condition of the device, the cause, the impact of the abnormal event, cause analysis and recovery procedures, etc. Device information includes, for example, a log of events that occurred in the device and measurement values ​​(device state quantities) of sensors attached to the device or equipment that includes the device.

[0014] The abnormal event response support device accepts input of the identification information and status of the equipment in which the abnormal event occurred (the abnormal event-occurring equipment), the circumstances of the abnormal event, etc. (also referred to as related information or abnormal event information). Next, the abnormal event response support device acquires device information for the equipment. Next, the abnormal event response support device acquires abnormal events related to the same equipment, the same type (model), or the same type of equipment as the abnormal event-occurring equipment. The abnormal event response support device further narrows down the abnormal events to those with device information similar to the acquired device information. The abnormal event response support device also narrows down the abnormal events to those related to reports similar to the related information. Next, the abnormal event response support device generates a prompt including the content of the narrowed-down abnormal event report, the input related information, and a question inquiring about how to deal with the abnormal event. Next, the abnormal event response support device inputs the prompt into a language model, acquires a response to the prompt from the language model, and outputs it as a countermeasure.

[0015] According to such an abnormal event response support device, it is possible to obtain a countermeasure for a currently occurring abnormal event based on an abnormal event that occurred in the past, thereby improving the efficiency of creating the countermeasure. Hereinafter, the facilities and devices that are the targets of response support by the abnormal event response support device will be referred to as target devices or simply devices.

[0016] <Configuration of the abnormal event response support device> 1 is a functional block diagram of an abnormal event response support device 100 according to the first embodiment. The abnormal event response support device 100 is a computer, and includes a control unit 110, a storage unit 120, and an input / output unit 180. User interface devices such as a display, keyboard, and mouse are connected to the input / output unit 180. The input / output unit 180 also includes a communication device, and is capable of communicating with a language model server 310 and a site 320.

[0017] The language model server 310 is a language model service server that, upon receiving a request / demand / question written in a language (text), generates an answer using a large-scale language model (language model) and sends it back. The language model service is also generally called an interactive artificial intelligence service / chat service.

[0018] Site 320 is a facility such as a factory or power plant, and includes equipment 330 (target equipment). Equipment 330 is equipped with sensor 340, and information on events that have occurred in equipment 330 and measurements of the state quantities of equipment 330 taken by sensor 340 are transmitted to abnormal event response support device 100.

[0019] <Abnormal event response support device: memory unit> The storage unit 120 is configured to include storage devices such as a ROM (Read Only Memory), a RAM (Random Access Memory), and an SSD (Solid State Drive). The storage unit 120 stores a device information database 130, an event report database 140, a knowledge base 150, and a program 128. The program 128 includes a description of the processing of the functional units provided in the control unit 110, which will be described later, and in particular includes a description of the abnormal event response support processing (see FIG. 5, which will be described later).

[0020] The contents stored in the storage unit 120 may be stored in an external storage device such as a cloud server and read as needed. The contents stored in the storage unit 120 may also be accessible from the abnormal event response support device 100 via a network.

[0021] <Memory section: device information database> 2 is a data configuration diagram of the device information database 130 according to the first embodiment. The device information database 130 includes a sensor information database 131 and an event database 132.

[0022] The sensor information database 131 is, for example, data in a table format, and records measured values ​​of state quantities related to target devices (see device 330 in FIG. 1). A row (record) of the sensor information database 131 indicates a measurement record, and includes columns (attributes) of measurement date and time, sensor identification information (denoted as "sensor ID" in FIG. 2), device identification information (denoted as "device ID" in FIG. 2), type (type of state quantity), and measured value. The state quantities are measured, for example, at regular intervals, and stored in the sensor information database 131.

[0023] The event database 132 is, for example, data in a table format, and records events that occur in target devices. A row (record) of the event database 132 indicates an event, and includes columns (attributes) of the occurrence date and time, device identification information (denoted as "device ID" in FIG. 2), type, event (event name), and status (device status when the event occurred). An event that occurs in a target device is notified to the abnormal event response support device 100 and stored in the event database 132.

[0024] <Memory section: incident report database> Returning to Fig. 1, the description of the storage unit 120 will continue. Reports of individual abnormal events are stored in the event report database 140. The reports include abnormal event information such as identification information of the abnormal event, the date and time of occurrence, identification information and model (model number) of the device in which the abnormal event occurred, the cause of the abnormal event, the impact of the abnormal event (scope of impact), the procedure for analyzing the cause, and the recovery procedure.

[0025] As described above, the device information (contained in the device information database 130) includes either the measured values ​​of the state quantities of the device or the events that have occurred in the device. The report (contained in the event report database 140) includes the cause of the abnormal event, the response (cause analysis procedures and recovery procedures), and / or the impact. The report includes abnormal event information including identification information of the equipment in which the abnormal event occurred.

[0026] <Memory section: knowledge base> 3 is a data configuration diagram of the knowledge base 150 according to the first embodiment. The knowledge base 150 stores sensor data knowledge 151, event knowledge 152, incident report knowledge 153, and association information 154.

[0027] The sensor data knowledge 151 stores data obtained by vectorizing the records of the sensor information database 131, which indicate the state quantities of the equipment when an abnormal event occurs, using a predetermined method. By vectorizing the records, the similarity between the records can be calculated. Event knowledge 152 stores data obtained by vectorizing records of event database 132, which indicate events that occurred in devices when abnormal events occurred, using a predetermined method. By vectorizing the records, the similarity between the records can be calculated.

[0028] Event report knowledge 153 stores data that is an abnormal event report or abnormal event information included in the report, which has been vectorized using a predetermined method. Vectorization is the conversion of text (sentences) into numerical vectors. By vectorizing, it is possible to calculate the similarity of the text. If the size of the report is large and exceeds the input size to the language model server 310 (described later), it is divided into blocks called chunks and then vectorized.

[0029] The association information 154 stores, in association with each other, the vectorized records of the sensor information database 131 stored in the sensor data knowledge 151 and the identification information of the abnormal event related to the record. The association information 154 also stores, in association with each other, the vectorized records of the event database 132 stored in the event knowledge 152 and the identification information of the abnormal event related to the record. The association information 154 also stores, in association with each other, the vectorized abnormal event reports stored in the event report knowledge 153 or the abnormal event information included in the reports and the identification information of the abnormal event reported in the reports.

[0030] As described above, the event report knowledge 153 stores vectorized data of the reports stored in the event report database 140 or the abnormal event information included in the reports. The sensor data knowledge 151 stores vectorized data of measured values ​​of state quantities of devices stored in the device information database 130. The event knowledge 152 stores vectorized data of events that have occurred in devices stored in the device information database 130.

[0031] <Abnormal event response support device: control unit> Returning to Fig. 1, the control unit 110 will be described. The control unit 110 is configured to include a CPU (Central Processing Unit) and is equipped with a reception unit 111, a similar event extraction unit 112, a prompt generation unit 113, and an answer acquisition unit 114. The control unit 110 may also be configured to include a GPU (Graphics Processing Unit), an FPGA (Field Programmable Gate Array), an ASIC (Application Specific Integrated Circuit), etc.

[0032] <Control Unit: Reception Unit> The reception unit 111 receives input of related information (occurred abnormality information) related to the abnormal event that has occurred. The related information includes, for example, the identification information of the device in which the abnormal event has occurred, its status, the details / status of the abnormal event (e.g., the contents displayed on the panel, abnormal noise, etc.), and the identification information of the device in which the abnormal event has occurred and the details of the abnormal event issued by the system monitoring the target device.

[0033] As described above, the abnormal event response support device 100 includes a reception unit 111 that receives identification information of an abnormal event occurrence device, which is a device in which an abnormal event has occurred, and occurred abnormality information, which is the content / status of the abnormal event.

[0034] <Control unit: Similar event extraction unit> The similar event extraction unit 112 extracts past abnormal events (identification information of the abnormal events) that are similar to the abnormal event received by the reception unit 111. To explain in more detail, the similar event extraction unit 112 first refers to the association information 154 in the event report database 140 or the knowledge base 150 (see FIG. 3 ) to obtain identification information of all abnormal events that have occurred in the past, and creates a list of the abnormal events.

[0035] Next, the similar event extraction unit 112 narrows down similar abnormal events (identification information of abnormal events) by searching for reports that include the device where the abnormal event occurred, with reference to the event report database 140. The similar event extraction unit 112 may also narrow down similar abnormal events by searching for reports that include devices of the same model or type (function) as the device where the abnormal event occurred.

[0036] The similar event extraction unit 112 may further narrow down the identification information of the narrowed-down abnormal events in the following manner. First, the similar event extraction unit 112 acquires data from the event knowledge 152 that is similar to the vectorized data of the records in the event database 132 related to the abnormal event-occurring device at the time of the abnormal event occurrence. Next, the similar event extraction unit 112 acquires the identification information of the abnormal event (second abnormal event) associated with the relevant data by referring to the association information 154. Subsequently, the similar event extraction unit 112 uses the acquired identification information to further narrow down the identification information of the narrowed-down abnormal events. The similar event extraction unit 112 may further narrow down the identification information of the narrowed-down abnormal events in the same manner for the vectorized data of the records in the sensor information database 131 related to the abnormal event-occurring device at the time of the abnormal event occurrence.

[0037] The similar event extraction unit 112 may further narrow down the identification information of the narrowed-down abnormal events as follows: First, the similar event extraction unit 112 acquires data from the event report knowledge 153 that is similar to the vectorized data of the related information received by the reception unit 111. Next, the similar event extraction unit 112 references the association information 154 to acquire identification information (first abnormal event) of the abnormal event associated with the data. Next, the similar event extraction unit 112 uses the acquired identification information to further narrow down the identification information of the narrowed-down abnormal events.

[0038] In the above explanation, the similar event extraction unit 112 narrows down all abnormal events to those similar to the abnormal event that has occurred, using devices, events, sensor data (measurements), and related information. Some of this narrowing down may be eliminated, or the order of narrowing down may be changed. For example, similar abnormal events for devices and similar abnormal events for related information may be acquired, and the intersection / intersection / common part of the two abnormal events may be narrowed down to an abnormal event (similar abnormal event).

[0039] Furthermore, the similar event extraction unit 112 extracts past abnormal events similar to the abnormal event that has occurred by referring to the knowledge base 150 (see FIG. 3 ), but this is not limitative. The similar event extraction unit 112 may also extract similar abnormal events (similar abnormal events) by referring to the device information database 130 or the event report database 140.

[0040] As described above, the abnormal event response support device 100 includes a similar event extraction unit 112 that refers to the event report database 140 that stores reports including abnormal event information related to abnormal events that have occurred in equipment in the past, and extracts a first abnormal event, which is an abnormal event described in a report that includes abnormal event information similar to the occurred abnormality information (related information).

[0041] The similar event extraction unit 112 refers to the device information database 130 that stores device information related to devices, and extracts a second abnormal event that is an abnormal event of a device similar to the device information of the abnormal event-occurring device. The similar event extraction unit 112 determines an abnormal event that is both a first abnormal event and a second abnormal event as a similar abnormal event.

[0042] The similar event extraction unit 112 refers to the event report database 140 and extracts a third abnormal event, which is an abnormal event that occurred in the abnormal event-occurring device, a device of the same type as the abnormal event-occurring device, or a device of the same type as the abnormal event-occurring device. The similar event extraction unit 112 determines an abnormal event that is the first abnormal event, the second abnormal event, and the third abnormal event as a similar abnormal event.

[0043] The similar event extraction unit 112 refers to the event report knowledge 153 and extracts a first abnormal event that is an abnormal event in a report that includes abnormal event information similar to the occurred abnormality information. Furthermore, the similar event extraction unit 112 refers to the event knowledge 152 or the sensor data knowledge 151 to extract a second abnormal event that is an abnormal event of a device similar to the device information of the device in which the abnormal event occurred.

[0044] <<Controller: Prompt Generator>> The prompt generation unit 113 generates a prompt to be input to the language model server 310, based on the related information received by the reception unit 111 and the report of the abnormal event extracted by the similar event extraction unit 112. If the similar event extraction unit 112 extracts multiple abnormal events, a prompt is generated for each of the abnormal events.

[0045] FIG. 4 shows a prompt template 200 according to the first embodiment. The prompt generation unit 113 generates a prompt by embedding related information in "{context1}" and the text of the extracted report on the abnormal event in "{context2}." The prompt template 200 includes a request / demand related to dealing with the abnormal event, such as "Please briefly summarize the similarities and differences." The prompt may be considered a question asking about the cause, similarities, and differences.

[0046] <<Control unit: Response acquisition unit>> Returning to Figure 1, the explanation of the control unit 110 will continue. The answer acquisition unit 114 sends the prompt created by the prompt generation unit 113 to the language model server 310 (see Figure 1), and receives, acquires, and outputs an answer to the request / demand / question included in the prompt (see Figure 7 described below). Note that acquiring an answer to a prompt using the language model server 310 is also referred to as inputting the prompt into a language model and acquiring an answer, acquiring an answer to a prompt using a language model, or acquiring an answer using a language model. Furthermore, if there are multiple prompts, the answer acquisition unit 114 acquires an answer for each prompt.

[0047] In other words, the answer acquisition unit 114 uses the language model server 310 to acquire an answer to the request / demand / question included in the prompt, which includes the cause of the abnormal event, similarities, and differences (comparisons). This answer becomes a countermeasure or a proposed measure for dealing with the abnormal event. The prompt may include a question about the impact (scope and extent) of the abnormal event, and the answer may include the impact.

[0048] As described above, the abnormal event response support device 100 includes a prompt generation unit 113 that generates a prompt (see FIG. 4) to be input to a language model (see language model server 310), the prompt being a question related to the abnormal event that has occurred and including the content of the report on the occurred abnormal event and similar abnormal event. The abnormal event response support device 100 also includes a response acquisition unit 114 that inputs a prompt to the language model, and acquires and outputs a response to the prompt from the language model.

[0049] The prompt is a question that inquires about the cause, the response, or the impact of the abnormal event that has occurred in the abnormal event-occurring device. The response includes any of the cause, response, and impact of the abnormal event that occurred in the abnormal event-occurring device.

[0050] When the similar event extraction unit 112 extracts multiple similar abnormal events, the prompt generation unit 113 generates a prompt for each of the multiple similar abnormal events, and the answer acquisition unit 114 acquires and outputs an answer for each of the multiple prompts.

[0051] <<Abnormal event response support processing>> Fig. 5 is a flowchart of the abnormal event response support process according to the first embodiment. With reference to Fig. 5, the abnormal event response support process will be described, which searches for past cases (reports) similar to the abnormal event that has occurred, and outputs countermeasures (proposed countermeasures) including comparisons based on the cases.

[0052] In step S11, the receiving unit 111 receives input of related information relating to the abnormal event that has occurred. Fig. 6 is a diagram showing the screen configuration of an abnormal event handling support screen 210 according to the first embodiment. An area 211 displays the device in which the abnormal event occurred. When a "related information" button 212 is clicked, a screen for inputting related information is displayed. A user, who is an operator responding to the abnormal event, inputs related information into the screen.

[0053] Area 213 displays options for searching for past cases (reports) of abnormal events. In FIG. 6, it is specified to narrow down abnormal events by equipment, events, and sensor data (measurements). When the "Launch Cause Analysis Assistant" button 214 is clicked, past cases of abnormal events searched based on related information and the options specified in area 213 are collected (extracted).

[0054] Returning to FIG. 5, the abnormal event response support process will be further described. In step S12, the similar event extractor 112 creates a list of all abnormal events that have occurred in the past. In step S13, the similar event extraction unit 112 narrows down the list created in step S12 by devices where abnormal events have occurred. For example, the similar event extraction unit 112 refers to the event report database 140, searches for reports that include the device where the abnormal event occurred, devices of the same model, or devices of the same type (function), and extracts abnormal events, thereby narrowing down the list created in step S12.

[0055] In step S14, the similar event extraction unit 112 further narrows down the list from step S13 using events and measurement values ​​of state quantities (sensor data). For example, the similar event extraction unit 112 narrows down the list to abnormal events similar to events (see event database 132 in FIG. 2) related to the abnormal event-occurring device at the time of the abnormal event occurrence or measurement values ​​of state quantities (see sensor information database 131).

[0056] In step S15, the similar event extraction unit 112 narrows down the list from step S14 using the related information. More specifically, the similar event extraction unit 112 narrows down the list to abnormal events similar to the related information received in step S11. In step S16, the prompt generation unit 113 starts the process of repeating steps S17 to S18 for each abnormal event in the list of step S15. Hereinafter, the abnormal event that is the target of this repeated process will be referred to as a target abnormal event.

[0057] In step S17, the prompt generating unit 113 generates a prompt (see FIG. 4) including a request / demand / question regarding how to deal with the abnormal event to be processed. In step S18, the answer acquisition unit 114 uses the language model server 310 (language model) to acquire an answer (a response to the abnormal event) to the request / demand / question included in the prompt created in step S17.

[0058] In step S19, the response acquisition unit 114 outputs the response acquired in step S18 to an abnormal event response support screen (see FIG. 7, which will be described later). 7 is a diagram showing the screen layout of the abnormal event response support screen 220 according to the first embodiment. Area 221 displays answers for each collected past abnormal event, including an outline of the cause and similarities and differences with the currently occurring abnormal event. Clicking the "Details" button 222 displays a report of the past abnormal event.

[0059] <Features of the abnormal event response support device> The abnormal event response support device 100 searches for reports of past abnormal events that are similar to the abnormal event that has occurred in terms of equipment, events, related information, etc. (see steps S13 to S15 in FIG. 5), and displays an outline of the cause and a comparison with the abnormal event that has occurred (see FIG. 7). By referring to this information, workers responding to the abnormal event can efficiently create and implement countermeasures.

[0060] The abnormal event response support device 100 narrows down the abnormal events using a knowledge base 150 that vectorizes the data stored in the device information database 130 and the event report database 140 (see steps S14 to S15 in FIG. 5). This allows the abnormal event response support device 100 to narrow down the abnormal events at high speed. Furthermore, when there are multiple similar abnormal events, the abnormal event response support device 100 creates prompts for each abnormal event and obtains an answer. Because the causes, similarities, and differences are obtained based on each individual abnormal event, high accuracy of answers can be expected.

[0061] <<Variation: Language Model>> In the above-described embodiment, the answer acquisition unit 114 acquires answers to prompts using the language model server 310. Alternatively, the answer acquisition unit 114 may generate and acquire answers to prompts using a language model. The language model is, for example, a pre-trained large-scale language model. The language model may also be an existing large-scale language model that has been fine-tuned using text data related to the target device, such as specifications or manuals. Note that the language model itself is data and is not a functional unit that processes data, but it may be described as receiving a prompt as input and outputting an answer.

[0062] Second Embodiment The abnormal event response support device 100 according to the first embodiment acquires, as device information (see device information database 130 in FIG. 2) relating to the device where the abnormal event occurred, past abnormal events relating to events and state quantities (sensor data) similar to the event and state quantity of the device when the abnormal event occurred, and displays an outline of the cause of the abnormal event and a comparison with the abnormal event that has just occurred. Basic product information such as the manufacturer and user (customer) of the device may be used as device information instead of events and state quantities.

[0063] Second Embodiment: Configuration of the malfunction response support device 8 is a functional block diagram of a malfunction response support device 100A (abnormal event response support device) according to the second embodiment. Compared to the abnormal event response support device 100 according to the first embodiment (see FIG. 1), a device information database 130A, a malfunction report database 140A, a knowledge base 150A, and a similar malfunction extraction unit 112A are different.

[0064] Second Embodiment: Device Information Database FIG. 9 is a data configuration diagram of a device information database 130A according to the second embodiment. The device information database 130A includes a product basic information database 136. Information related to the manufacture and user of each target device is recorded in the product basic information database 136. Rows (records) of the product basic information database 136 include columns (attributes) of operation date, product name, model number, device identification information (denoted as "device ID" in FIG. 9), lot, part, user (e.g., customer), and usage location (site). The lot is the identification information of the production lot of the device. The part is the model number, identification information, or lot of the part that makes up the device.

[0065] Second Embodiment: Defect Report Database Returning to Fig. 8, the description of storage unit 120 will continue. Reports of identified malfunctions (abnormal events) related to devices are stored in malfunction report database 140A (event report database). The reports include malfunction information (abnormal event information) such as malfunction identification information, the type (model number) of the device, lot, malfunction details (status), cause of the malfunction, impact, and countermeasures (first aid and recovery procedures / repair procedures).

[0066] The defect information may be, for example, information about a defect in a device discovered by the device manufacturer (supplier), or information about a defect that occurred in a device and that has been found to have the potential to occur in the same type of device. The defect information may also be information that the manufacturer has confirmed about defects pointed out by customers and summarized the causes, effects, and countermeasures. Note that the defects already identified in the defect report database 140A may be considered to be defects (abnormal events) that occurred in the past.

[0067] <Second embodiment: knowledge base> 10 is a data configuration diagram of a knowledge base 150A according to the second embodiment. The knowledge base 150A stores basic product knowledge 155, defect report knowledge 153A, and association information 154A.

[0068] The product basic knowledge 155 stores data obtained by vectorizing, using a predetermined method, records in the product basic information database 136 corresponding to devices for which defects have already been identified. The defect report knowledge 153A stores data in which a defect report or defect information included in the report is vectorized using a predetermined method. The vectorization method is the same as that of the event report knowledge 153 according to the first embodiment.

[0069] Association information 154A stores, in association with each other, records of vectorized product basic information database 136 stored in product basic knowledge 155 and identification information of defects related to the records. Association information 154A also stores, in association with each other, vectorized defect reports stored in defect report knowledge 153A or defect information included in the reports and identification information of defects reported in the reports.

[0070] <Control unit: Similar defect extraction unit> The similar defect extraction unit 112A extracts discovered defects (defect identification information) that are similar to the related information of the defect received by the reception unit 111. To explain in more detail, the similar defect extraction unit 112A first refers to the association information 154A in the defect report database 140A or the knowledge base 150A (see FIG. 10 ) to obtain the identification information of all discovered defects, and creates a list of defects.

[0071] The similar defect extraction unit 112A narrows down the acquired defect identification information as follows. First, the similar defect extraction unit 112A acquires data from the product basic knowledge 155 that is similar to data obtained by vectorizing records in the product basic information database 136 related to the defect-occurring device (the abnormal event-occurring device). Next, the similar defect extraction unit 112A refers to the association information 154A to acquire identification information (second abnormal event) of the defect associated with the data. Then, the similar defect extraction unit 112A uses the identification information to narrow down the acquired defect identification information.

[0072] The similar defect extraction unit 112A may further narrow down the identification information of the narrowed-down defects as follows. First, the similar defect extraction unit 112A acquires data from the defect report knowledge 153A that is similar to the vectorized data of the related information received by the receiving unit 111. Next, the similar defect extraction unit 112A refers to the association information 154A to acquire identification information (first abnormal event) of the defect associated with the data. Next, the similar defect extraction unit 112A uses the identification information to further narrow down the identification information of the narrowed-down defects.

[0073] As in the first embodiment, some of the defects may be omitted or the order of narrowing down may be changed. Also, as in the first embodiment, similar defects may be extracted by referring to the device information database 130A or the defect report database 140A without using the knowledge base 150A.

[0074] As described above, the device information (contained in the device information database 130A) includes any of the device model (model number), identification information, lot, parts, user, and use location. The event report knowledge (fault report knowledge 153A) stores vectorized data of reports stored in the event report database (fault report database 140A) or abnormal event information included in the reports. The product basic knowledge 155 stores vectorized data of data stored in the equipment information database 130A, including any of the equipment identification information, model, lot, parts, user, and use location.

[0075] The similar event extraction unit (similar fault extraction unit 112A) refers to the event report knowledge (fault report knowledge 153A) and extracts a first abnormal event that is an abnormal event in a report that includes abnormal event information similar to the occurred abnormality information. Furthermore, the similar event extraction unit (similar failure extraction unit 112A) refers to the product basic knowledge 155 and extracts a second abnormal event that is an abnormal event of a device similar to the device information of the device in which the abnormal event occurred.

[0076] <<Fault handling support processing>> Fig. 11 is a flowchart of the malfunction response support process according to the second embodiment. With reference to Fig. 11, the malfunction response support process for finding and comparing identified cases (reports) similar to the malfunction that has occurred will be described.

[0077] In step S31, the receiving unit 111 receives input of related information relating to the malfunction that has occurred. FIG. 12 is a diagram showing the screen layout of a fault handling support screen 260 according to the second embodiment. Area 261 displays information related to the device in which the fault occurred. Area 262 is used to input the details / status (related information) of the fault that the customer inquired about. When the "Search for fault information" button 263 is clicked, the searched cases of identified faults are collected, and the details, impact, and countermeasures for the fault that occurred are compiled with reference to those cases.

[0078] Returning to FIG. 11, the description of the malfunction handling support process will be continued. In step S32, the similar defect extracting section 112A creates a list of all defects that have been identified. In step S33, the similar defect extracting section 112A narrows down the list of step S32 by the product basic information.

[0079] In step S34, the similar event extraction unit 112 narrows down the list from step S33 using the related information received in step S31 (see area 262 in FIG. 12) (to defects similar to the related information). Steps S35 to S38 are similar to steps S16 to S19.

[0080] 13 shows a prompt template 250 according to the second embodiment. In step S36, the prompt generation unit 113 generates a prompt by embedding related information in "{context1}" and the extracted text of the defect report in "{context2}." Note that the prompt template 250 includes a request / demand / question regarding how to deal with the defect, such as "Please briefly summarize the details of the defect, its impact, and countermeasures."

[0081] 14 is a diagram showing the screen layout of a fault handling support screen 270 according to the second embodiment. Area 271 displays the details, impact, and countermeasures of the fault that has occurred, with reference to collected and identified faults. Clicking the "Details" button 272 displays a report on the fault.

[0082] <Features of the troubleshooting support device> The malfunction response support device 100A searches for reports of identified malfunctions that are similar to the malfunction that has occurred in terms of basic product information, related information, etc. (see steps S33 and S34 in FIG. 11), and displays the details, effects, and countermeasures of the malfunction (see FIG. 14). By referring to this information, the person in charge of the inquiry desk handling the malfunction can efficiently respond to the customer with countermeasures.

[0083] Other variations Although several embodiments of the present invention have been described above, these embodiments are merely illustrative and do not limit the technical scope of the present invention. For example, in the repeated processing of steps S16 to S18 in the abnormal event response support processing shown in Fig. 5, the processing order of the abnormal events to be processed in the repeated processing of steps S14 to S15 may be in descending order of similarity. Furthermore, the display order of the answers in area 221 shown in Fig. 7 may also be in descending order of similarity. Similarly, in the fault response support processing shown in Fig. 11, the processing order of the abnormal events to be processed (faults to be processed) in the repeated processing of steps S35 to S37 and the display order of the answers in area 271 shown in Fig. 14 may also be in descending order of similarity in steps S33 to S34.

[0084] The present invention can take on various other embodiments, and various modifications such as omissions and substitutions can be made without departing from the spirit of the present invention. These embodiments and modifications are included in the scope and spirit of the invention described in this specification, etc., and are also included in the invention described in the claims and their equivalents.

[0085] <Hardware configuration> The abnormal event response support device 100 and the malfunction response support device 100A according to the above-described embodiments are realized by, for example, a computer 900 configured as shown in FIG. 15 . FIG. 15 is a hardware configuration diagram showing an example of the computer 900 that realizes the functions of the abnormal event response support device 100 and the malfunction response support device 100A according to the above-described embodiments. The computer 900 includes a CPU 901, a ROM 902, a RAM 903, an SSD 904, an input / output interface 905 (referred to as an input / output I / F (Interface) in FIG. 15 ), a communication interface 906 (referred to as a communication I / F in FIG. 15 ), and a media interface 907 (referred to as a media I / F in FIG. 15 ). The computer 900 may include an HDD (Hard Disc Drive) instead of the SSD 904, or may include an HDD in addition to the SSD 904.

[0086] The CPU 901 operates based on a program stored in the ROM 902 or the SSD 904, and performs control by the control unit 110 in Fig. 1. The ROM 902 stores a boot program executed by the CPU 901 when the computer 900 starts up, programs related to the hardware of the computer 900, and the like.

[0087] The CPU 901 controls an input device 910 such as a mouse or keyboard, and an output device 911 such as a display or printer, via an input / output interface 905. The CPU 901 acquires data from the input device 910 via the input / output interface 905, and outputs generated data to the output device 911.

[0088] The SSD 904 stores programs executed by the CPU 901 and data used by the programs. The communication interface 906 receives data from other devices (not shown) (e.g., the language model server 310) via a communication network and outputs the data to the CPU 901, and also transmits data generated by the CPU 901 to other devices via the communication network.

[0089] The media interface 907 reads a program or data stored in the recording medium 912 and outputs it to the CPU 901 via the RAM 903. The CPU 901 loads the program from the recording medium 912 onto the RAM 903 via the media interface 907 and executes the loaded program. The recording medium 912 is an optical recording medium such as a DVD (Digital Versatile Disk), a magneto-optical recording medium such as an MO (Magneto Optical disk), a magnetic recording medium, a conductive memory tape medium, a semiconductor memory, or the like.

[0090] For example, when the computer 900 functions as the abnormal event response support device 100 or the malfunction response support device 100A according to the above-described embodiments, the CPU 901 of the computer 900 executes a program 128 (see FIG. 1) loaded onto the RAM 903, thereby realizing the functions of the abnormal event response support device 100 or the malfunction response support device 100A. The CPU 901 reads the program from a recording medium 912 and executes it. Alternatively, the CPU 901 may read the program from another device via a communication network, or may install the program 128 from the recording medium 912 onto the SSD 904 and execute it. [Explanation of symbols]

[0091] 100 Abnormal event response support device 100A Malfunction Response Support Device (Abnormal Event Response Support Device) 111 Reception 112 Similar Event Extraction Unit 112A Similar defect extraction unit (similar event extraction unit) 113 Prompt Generation Unit 114 Answer acquisition part 130,130A Equipment Information Database 131 Sensor Information Database 132 Event Database 136 Product Basic Information Database 140 Incident Report Database 140A Malfunction Report Database (Event Report Database) 150,150A Knowledge Base 151 Sensor Data Knowledge 152 Event Knowledge 153 Incident Reporting Knowledge 153A Malfunction Reporting Knowledge 154,154A Association information 155 Basic product knowledge 210,220 Abnormal event response support screen 260,270 Troubleshoot support screen 310 Language Model Server (Language Model)

Claims

1. a receiving unit that receives identification information of an abnormal event occurrence device, which is a device in which an abnormal event has occurred, and occurred abnormality information, which is the content / status of the abnormal event; extracting a first abnormal event that is an abnormal event described in a report that includes abnormal event information similar to the occurred abnormal event information by referring to an event report database that stores reports that include abnormal event information related to abnormal events that have occurred in equipment in the past; extracting a second abnormal event that is an abnormal event of a device similar to the device information of the abnormal event-occurring device by referring to a device information database that stores device information related to the device; a similar event extraction unit that identifies an abnormal event that is both the first abnormal event and the second abnormal event as a similar abnormal event; a prompt generation unit that generates a prompt to be input to a language model, the prompt being a question related to the abnormal event that has occurred, the prompt including the content of the occurred abnormality information and the report of the similar abnormal event; an answer acquisition unit that inputs the prompt to the language model, and acquires and outputs an answer to the prompt from the language model; Abnormal event response support device.

2. The said report: Including any of the causes, responses, and effects of the abnormal event, The prompt may include: A question inquiring about any of the cause, countermeasure, and impact of the abnormal event that has occurred in the abnormal event occurrence device, The answer is, Including any of the causes, countermeasures, and effects of the abnormal event that occurred in the abnormal event-occurring device The abnormal event response support device according to claim 1 .

3. The device information includes: The measurement value of the state quantity of the device and the event that occurred in the device are included. The abnormal event response support device according to claim 1 .

4. The said report: abnormal event information including identification information of the abnormal event-occurring device; The similar event extraction unit extracting a third abnormal event that is an abnormal event that occurred in the abnormal event-occurring device, a device of the same type as the abnormal event-occurring device, or a device of the same type as the abnormal event-occurring device, by referring to the event report database; An abnormal event that is the first abnormal event, the second abnormal event, and the third abnormal event is defined as the similar abnormal event. The abnormal event handling support device according to claim 3.

5. The device information includes: Including the type, identification, lot, part, user, and location of the device The abnormal event response support device according to claim 1 .

6. When the similar abnormal events extracted by the similar event extraction unit are plural, The prompt generation unit generating the prompt for each of the plurality of similar abnormal events; The answer acquisition unit Obtaining and outputting the answer for each of the plurality of prompts. The abnormal event response support device according to claim 1 .

7. The similar event extraction unit extracting a first abnormal event that is an abnormal event in a report that includes abnormal event information similar to the occurred abnormality information, by referring to a report stored in the event report database or event report knowledge that stores vectorized data of abnormal event information included in the report; Extracting a second abnormal event, which is an abnormal event of a device similar to the device information of the abnormal event-occurring device, by referring to event knowledge that stores vectorized data of events that have occurred in the device and that are stored in the device information database, or sensor data knowledge that stores vectorized data of measured values ​​of state quantities of the device and that are stored in the device information database. The abnormal event handling support device according to claim 3.

8. The similar event extraction unit extracting a first abnormal event that is an abnormal event in a report that includes abnormal event information similar to the occurred abnormality information, by referring to a report stored in the event report database or event report knowledge that stores vectorized data of abnormal event information included in the report; A second abnormal event, which is an abnormal event of a device similar to the device information of the abnormal event-occurring device, is extracted by referring to basic product knowledge that stores vectorized data including any of the device identification information, model, lot, parts, user, and usage location stored in the device information database. The abnormal event handling support device according to claim 5.

9. The abnormal event response support device receiving identification information of an abnormal event occurrence device, which is a device in which an abnormal event has occurred, and information on an abnormal event that is the content / situation of the abnormal event; extracting a first abnormal event, which is an abnormal event described in a report including abnormal event information similar to the occurred abnormal event information, by referring to an event report database that stores reports including abnormal event information relating to abnormal events that have occurred in equipment in the past; extracting a second abnormal event that is an abnormal event of a device similar to the device information of the abnormal event-occurring device by referring to a device information database that stores device information related to the device; a step of determining an abnormal event that is both the first abnormal event and the second abnormal event as a similar abnormal event; generating a prompt to be input to a language model, the prompt being a question related to the abnormal event that has occurred, the prompt including the content of the abnormal event information and the report of the similar abnormal event; inputting the prompt into the language model, and obtaining and outputting an answer to the prompt from the language model. Methods for supporting abnormal event response.

Images