Defect occurrence prediction system and defect occurrence prediction method

Abstract

To provide a defective occurrence prediction system that can reduce occurrence of defective products, a learning instruction method for an interactive AI, and a program.SOLUTION: A defective occurrence prediction system 10 includes: a plurality of sensors SEN1, SEN2, ..., and SENn that measure a state of a manufacturing apparatus for manufacturing an article; operation units SW1, SW2, ..., and SWn to be operated by a user having recognized a defect occurred in the article; and a transmission unit 50 that transmits learning data to an interactive AI configured based on a large-scale language model.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present invention is a defect occurrence prediction system. And defect occurrence prediction method Regarding. [Background technology]

[0002] Patent Document 1 describes a malfunction prediction system. This malfunction prediction system is characterized by including a status acquisition unit that acquires data related to the operating status at each timing of a monitoring target operating device in association with timing information, a status prediction unit that predicts future operating statuses for a plurality of periods of different lengths from the present according to the operating statuses acquired by the status acquisition unit, a third storage unit that stores a first correspondence relationship between the operating status and the occurrence of a predetermined malfunction in the operating device, and a malfunction prediction unit that calculates a predetermined index related to the possibility of malfunction occurrence based on the predicted operating status and the first correspondence relationship. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent Publication No. 2022-11723 Summary of the Invention [Problem to be solved by the invention]

[0004] The present invention relates to Using conversational AI A defect prediction system that can reduce the occurrence of defective products And defect occurrence prediction method The purpose is to provide. [Means for solving the problem]

[0005] The invention described in claim 1 is a method for measuring the state of a manufacturing device for manufacturing an article. and output the measurement data A plurality of sensors for detecting the presence of a signal; a data processing unit that performs data cleansing on the measurement data input at a predetermined first period and outputs the data as information A representing a state of the manufacturing apparatus; an operation unit operated by a user who has recognized a defect occurring in the article; A time identification unit for identifying a time ts at which the operation unit was operated and acquiring information B which is a time at which a defect occurred in the article and is based on the operation of the operation unit; an interactive AI configured based on a large-scale language model; For conversational AI, The information A, the information B, information C about the situation in which the defect occurred, which is recorded after the defect occurred, and information D about the method of dealing with the defect, which is recorded after the defect occurred. A transmission unit that transmits learning data; a defect occurrence confirmation unit that transmits a prompt inquiring whether or not there is a possibility that the product may be defective; Equipped with The transmission unit transmits the information A to the conversational AI that has been trained using the learning data at a predetermined second period, and The conversational AI, The defect occurrence confirmation unit notifies the prompt When is entered, the corresponding study data Based on Returning a Response This is a defect prediction system.

[0006]

[0007] The invention described in claim 2 is the defect occurrence prediction system described in claim 1, The conversational AI trained using the training data is An instruction to correct the time difference between the time tm when the defect actually occurred in the article and the time ts. Department Prepare further.

[0008] The invention described in claim 3 is a defect occurrence prediction system according to claim 1, wherein the time tm when a defect actually occurs in the article and 、 The time ts and 、 Time correction to correct the time difference that occurred between Department Prepare further.

[0009] The invention described in claim 4 is 2 or 3 In the defect occurrence prediction system described, a processing unit (excluding those performed by humans) for comparing the time ts with the defect occurrence time tr, the information B being a defect occurrence time tr described in a report created after the occurrence of a defect, and an alarm output unit (excluding those performed by humans) for outputting an alarm to a user notifying that the defect occurrence time tr and the time ts do not substantially coincide with each other when a time difference between the time ts compared by the processing unit and the defect occurrence time tr is greater than a predetermined value; It further comprises:

[0010] The invention described in claim 5 is According to any one of claims 1 to 3 In the defect occurrence prediction system described, The length of the second period is equal to or less than the length of the first period. .

[0011] The invention described in claim 6 is the defect occurrence prediction system described in any one of claims 1 to 3, wherein the length of the third period is equal to or less than the length of the first period.

[0012] The invention described in claim 7 includes a plurality of sensors that measure the state of a manufacturing device that manufactures an article and each output measurement data; a data processing unit that performs data cleansing on the measurement data input at a predetermined first period and outputs the data as information A representing the state of the manufacturing device; an operation unit that is operated by a user who has recognized a defect that has occurred in the article; a time identification unit that identifies a time ts at which the operation unit is operated and obtains information B, which is the time at which the defect occurred in the article and is based on the operation of the operation unit; an interactive AI configured based on a large-scale language model; and a processing unit that provides the interactive AI with the information A, the information B, information C regarding the situation in which the defect occurred that is recorded after the defect occurred, and information regarding the defect that is recorded after the defect occurred. A defect occurrence prediction method using a defect occurrence prediction system including a transmission unit that transmits learning data including information D regarding a countermeasure, and a defect occurrence confirmation unit that transmits a prompt inquiring whether or not there is a possibility of a defect occurring in the item, the defect occurrence prediction method including the steps of: the transmission unit transmitting the learning data to the interactive AI; the interactive AI learning the learning data; and the transmission unit transmitting the information A to the interactive AI that has learned using the learning data at a predetermined second period, and the interactive AI that has learned using the learning data returning a response based on the learning data when the prompt is input from the defect occurrence confirmation unit at a predetermined third period. Effect of the Invention

[0013] According to the present invention, Using conversational AI A defect prediction system that can reduce the occurrence of defective products And defect occurrence prediction method can be provided. [Brief description of the drawings]

[0014] [Figure 1A] FIG. 2 is an explanatory diagram showing a defect occurrence prediction function of a defect occurrence prediction system according to an embodiment of the present invention. [Figure 1B] FIG. 2 is an explanatory diagram showing a chatbot function of the defect occurrence prediction system. [Diagram 2] FIG. 2 is a configuration diagram of the defect occurrence prediction system. [Diagram 3] 4 is a functional block diagram of a first processing unit included in the defect occurrence prediction system. FIG. [Figure 4] FIG. 2 is an explanatory diagram showing a flow from the occurrence of a defect to the creation of a report in the defect occurrence prediction system. [Diagram 5] 4 is a functional block diagram of a second processing unit included in the defect occurrence prediction system. FIG. [Figure 6] FIG. 2 is an explanatory diagram showing the relationship between the interactive AI and the learning instruction unit used in the defect occurrence prediction system. [Figure 7] FIG. 1 is a flow diagram showing the operation of an interactive AI during a learning period. [Figure 8] 4 is a flowchart showing the operation of a second processing unit included in the defect occurrence prediction system. FIG. [Figure 9] FIG. 1 is a flow diagram of the learning process of an interactive AI. [Figure 10] FIG. 1 is a flow diagram showing operations during execution of an interactive AI. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0015] Next, embodiments of the present invention will be described with reference to the attached drawings to facilitate understanding of the present invention. Note that in the drawings, parts that are not relevant to the description may be omitted.

[0016] As shown in FIGS. 1A and 1B, a defect occurrence prediction system 10 according to an embodiment of the present invention can provide two main functions by using an interactive AI 20 provided as a cloud service.

[0017] First, the defect occurrence prediction system 10 can present the user with a necessary countermeasure based on the state of the manufacturing equipment 12 before a defect occurs in a machine part (an example of an article) manufactured by the manufacturing equipment 12. Secondly, the defect occurrence prediction system 10 can return an appropriate response to an inquiry from a user based on the learning data.

[0018] Here, the conversational AI 20 is, for example, ChatGPT ("CHATGPT" is an internationally registered trademark) provided by OpenAI, Inc., and is configured based on large language models. The conversational AI 20 includes not only the conversational AI itself, but also the means for making the conversational AI function.

[0019] The manufacturing apparatus 12 is, for example, a machine part manufacturing apparatus such as a die-casting machine, an injection molding machine, a blow molding machine, a hydraulic press, a machining center, etc. The manufacturing apparatus 12 may be a food manufacturing apparatus that manufactures food, or more specifically, may be an apparatus that manufactures any article.

[0020] As shown in FIG. 2, the defect occurrence prediction system 10 includes a plurality of sensors SEN1, SEN2, ... SENn, a plurality of defect occurrence identification switches SW1, SW2, ... SWn, a first processing unit 30, a second processing unit 40, a crawler unit 50, and a learning instruction unit 60. At least the functions of the first processing unit 30, the second processing unit 40, the crawler unit 50 and the learning instruction unit 60 are realized by a computer program.

[0021] A plurality of sensors SEN1, SEN2, ..., SENn are attached to the manufacturing equipment 12 and can measure the state of the manufacturing equipment 12. Each of the sensors SEN1, SEN2, ..., SENn is, for example, a pressure gauge, a thermometer, a vibration meter, etc., and outputs measurement data.

[0022] Each of the defect occurrence identification switches (an example of an operation unit) SW1, SW2, ..., SWn is, for example, a push button switch. Each of the defect occurrence identification switches SW1, SW2, ..., SWn is operated by a user who recognizes that a defect has occurred in a manufactured mechanical part, and can generate, for example, a pulse-like time-specific signal. The defect occurrence identification switches SW1, SW2, ... SWn are provided according to the type of defect, specifically, there are provided defect occurrence identification switches which are pressed when a mechanical part is scratched, defect occurrence identification switches which are pressed when a chip occurs, etc.

[0023] The first processing unit 30 is, for example, a terminal used by a user, and receives as input the measurement data output by each of the sensors SEN1, SEN2, . . . SENn and the time identification signals output by each of the defect occurrence identification switches SW1, SW2, . . . SWn. The measurement data output from each of the sensors SEN1, SEN2, . . . SENn is input to the first processing unit 30 at a predetermined period Tc1. This period Tc1 is, for example, 60 seconds.

[0024] As shown in FIG. 3, the first processing unit 30 has a data processing unit 302 and a time identifying unit 304. The data processing unit 302 pre-processes the measurement data output by each of the sensors SEN1, SEN2, . . . SENn. This pre-processing is, for example, data cleansing including noise removal. The pre-processed measurement data is called processed measurement data (obtained by multiple sensors). Made by Manufacturing equipment status Represents The information is output as the information A) and stored in the first storage unit 32 (see FIG. 2).

[0025] The time identifying unit 304 identifies the time ts at which the defect occurrence identifying switches SW1, SW2, ..., SWn were activated (an example of information B that is the time at which a defect occurred in an article) based on the time identifying signals output from the defect occurrence identifying switches SW1, SW2, ..., SWn. This time ts is also the time at which an operator recognized that a defect occurred in a manufactured machine part. The identified time ts is stored in the first storage unit 32 and is also transmitted to the second processing unit 40.

[0026] Here, a defect in a machine part occurs due to a malfunction of the manufacturing equipment 12, and time ts is the time when an operator recognizes that a defect has occurred in a manufactured machine part. Therefore, as shown in Fig. 4, time ts may have a time TL1 shift (delay of time TL1) from time tm (an example of information B) when a malfunction of the manufacturing equipment 12 occurs and the defect actually occurs in the machine part. For example, if the manufacturing equipment 12 is a die-casting machine, there may be a time TL1 shift between time tm when a defect actually occurs in a machine part manufactured by the die-casting machine and time ts when the defect is recognized.

[0027] The second processing unit 40 is, for example, a terminal used by a user, and can convert defect countermeasure information recorded as audio data or video data into text data. Generally, after a defect occurs in a mechanical part, a user creates a report (one example of a record) to record how to deal with the defect, as shown in Fig. 4. The defect handling information mentioned above is the information recorded in this report. The report may be recorded on an electronic medium not only as text data, but also as audio data or video data.

[0028] The defect handling information recorded in the report includes at least the following information: (1) Time of defect occurrence tr (an example of information B) The defect occurrence time tr is the time when an operator recognizes that a defect has occurred in a manufactured machine part, and is the time recorded in the report. Therefore, the defect occurrence time tr may differ from the time ts when the defect occurrence identification switches SW1, SW2, ... SWn are activated by a time TL2 due to factors such as misunderstanding on the part of the report writer. (2) Circumstances under which the defect occurred (an example of information C on the circumstances under which the defect occurred, recorded after the defect occurred) The defect occurrence status is information about the status in which the defect occurred, and includes at least information about the type of defect. (3) How to deal with defects (an example of information D on how to deal with defects, recorded after the occurrence of a defect) The defect handling method is information on the method of handling the defect that has occurred.

[0029] As shown in FIG. 5, the second processing unit 40 has a conversion processing unit 402 and an alarm output unit 404. The conversion processing unit 402 can convert the defect action information recorded in the report into text data. The defect action information converted into text data is stored in the second storage unit 42 (see FIG. 2).

[0030] The alarm output unit 404 (see FIG. 5) The conversion processing unit 402 The time tr at which the defect occurred, which is included in the defect handling information described above, is compared with the time ts at which the defect occurrence specific switches SW1, SW2, ..., SWn were operated. Results If the difference between the time TL2 and the time TL3 is greater than a predetermined value, an alarm is output. After the alarm is output, if the difference in time TL2 between the defect occurrence time tr and the time ts is greater than a predetermined value, the defect occurrence time tr is corrected to the time ts, as will be described later in detail.

[0031] The crawler unit (an example of a transmitting unit) 50 (see Figure 2) is, for example, a terminal used by a user, and can communicate with the first processing unit 30 and the second processing unit 40, respectively, and can acquire information stored in the first memory unit 32 and the second memory unit 42, respectively. The crawler unit 50 transmits at least the following information to the interactive AI 20 as fine-tuning learning data after performing necessary processing. (1) Processed measurement data stored in the first storage unit 32 (an example of information A) (2) Defect handling information stored in the second storage unit 42 (examples of information B, information C, and information D)

[0032] The learning instruction unit 60 is, for example, a terminal used by a user, and can instruct the interactive AI 20 on a learning method via an API (Application Programming Interface) provided by the interactive AI 20. The API is executed via a cloud computing service 22, as shown in FIG. In detail, the learning instruction unit 60 can divide the linguistic feature space into a first space in which tokens extracted from defect handling information are vectorized and mapped, and a second space in which tokens are not mapped, and instruct the interactive AI 20 to generate an answer by regarding tokens in the first space as more relevant tokens than tokens in the second space.

[0033] In addition, the learning instruction unit 60 can instruct the interactive AI 20 to correct the time lag TL1 (see Figure 4) that occurs between the time tm and the time ts when a defect actually occurs in the mechanical part, for the processed measurement data. However, instead of the learning instruction unit 60 instructing to correct the deviation in time TL1, the defect occurrence prediction system 10 may be provided with a time correction unit (not shown) for correcting the deviation in time TL1 that occurs between the time tm and the time ts when a defect actually occurs in the mechanical part for the processed measurement data.

[0034] In this way, according to the defect prediction system 10, the processed measurement data indicating the state of the manufacturing equipment 12 and defect handling information are transmitted as learning data to the interactive AI 20. As a result, the interactive AI 20 can learn 1) the relationship between the state of the manufacturing equipment 12 and the occurrence timing of each type of defect, such as scratches or chips, and 2) the defect occurrence circumstances and defect handling methods.

[0035] Next, the operation of the defect occurrence prediction system 10 (defect occurrence prediction method) will be described separately for the learning period and the execution period of the interactive AI 20.

[0036] During the learning period of the interactive AI 20, the sensors SEN1, SEN2, . . . SENn constantly measure the state of the manufacturing equipment 12, and the first processing unit 30 stores the processed measurement data in the first memory unit 32.

[0037] Furthermore, when a defect is found in a manufactured mechanical part, the first processing unit 30 stores the time ts for each type of defect in the first storage unit 32, and the user creates a report as text data, audio data, or video data, as shown in Fig. 7. When a predetermined amount of report is created by repeating this process, the defect handling information included in the report created as audio data or video data based on the user's operation is converted into text data by the second processing unit 40 and stored in the second storage unit 42.

[0038] Here, the conversion process by the conversion processing unit 402 included in the second processing unit 40 is executed in accordance with the following steps, as shown in FIG.

[0039] (Step SA1) The report is loaded by the user.

[0040] (Step SA2) The conversion processing unit 402 determines whether the input data is voice data or not. If the input data is voice data, step SA3 is executed, whereas if the input data is not voice data, step SA4 is executed.

[0041] (Step SA3) The conversion processing unit 402 converts the voice data into text data. Then, step SA6 is executed.

[0042] (Step SA4) The conversion processing unit 402 determines whether the input data is video data or not. If the input data is video data, step SA5 is executed, whereas if the input data is not video data, step SA6 is executed.

[0043] (Step SA5) The conversion processing unit 402 separates the audio data contained in the video data.

[0044] (Step SA6) The conversion processing unit 402 compares the defect occurrence time tr contained in the text data converted in step SA3 with the time ts at which the defect occurrence specifying switches SW1, SW2, . . . SWn were activated.

[0045] (Step SA7) The conversion processing unit 402 judges whether the compared defect occurrence time tr and the time ts at which the defect occurrence specifying switches SW1, SW2, . . . SWn were activated substantially coincident with each other. If the difference in time TL2 between the defect occurrence time tr and the time ts is equal to or smaller than a predetermined value, the defect occurrence time tr and the time ts are deemed to substantially coincide with each other, and step SA8 is executed. If the difference is greater than the predetermined value, step SA9 is executed.

[0046] (Step SA8) The defect handling information is stored in the second storage unit 42 as text data.

[0047] (Step SA9) The alarm output unit 404 (see FIG. 5) outputs an alarm indicating that the defect occurrence time tr and the time ts do not substantially coincide with each other.

[0048] (Step SA10) The user corrects the defect occurrence time tr (an example of information B) included in the text data to the time ts. After that, step SA6 is executed again.

[0049] By repeating the above steps, the defect handling information converted into text data is stored in the second storage unit 42. After that, as described above, the crawler unit 50 transmits the following information to the interactive AI 20 as learning data. (1) Processed measurement data stored in the first storage unit 32 (an example of information A) (2) Defect handling information stored in the second storage unit 42 (examples of information B, information C, and information D)

[0050] When the interactive AI 20 receives learning data from the crawler unit 50, a loop process for performing deep learning is executed as shown in FIG. In addition, steps SB1, SB2 and step SC1 、 SC 2 are executed in parallel.

[0051] (Step SB1) The interactive AI 20 receives the processed measurement data stored in the first storage unit 32 .

[0052] (Step SB2) Based on instructions from the learning instruction unit 60, the received processed measurement data is corrected for the time TL1 deviation that occurs between the time tm when the defect actually occurred in the mechanical part and the time ts, and then the processed measurement data is recorded in a memory unit not shown.

[0053] However, as mentioned above, if the defect occurrence prediction system 10 is equipped with a time correction unit (not shown) for correcting the time difference TL1 that occurs between the time tm when a defect actually occurs in the mechanical part and the time ts for the processed measurement data, the time difference TL1 has already been corrected by this time correction unit in the processed measurement data, so the time-corrected processed measurement data is recorded as is in the memory unit (not shown).

[0054] (Step SC1) The interactive AI 20 receives the defect handling information. In addition, since the above-mentioned step SA10 is executed, the defect occurrence time tr included in the defect handling information substantially coincides with the time ts at which the defect occurrence specifying switches SW1, SW2, . . . SWn are operated.

[0055] (Step SC2) The above-mentioned first space is created in the linguistic feature space based on an instruction from the learning instruction unit 60. That is, the linguistic feature space is divided into a first space and a second space.

[0056] When all learning data has been received from the crawler unit 50 and the loop processing has been completed, the interactive AI 20 completes learning. Thereafter, as shown in FIG. 7, the accuracy with which the interactive AI 20 predicts the occurrence of defects is verified.

[0057] In this way, by repeating the series of steps of defect occurrence, handling, report creation, conversion of defect handling information into text data, learning of the interactive AI 20 (sending of learning data by the crawler unit 50), and accuracy verification, the interactive AI 20 learns 1) the relationship between the state of the manufacturing equipment 12 and the timing of occurrence of each type of defect such as scratches or chips, and 2) the circumstances under which defects occur and how to handle the defects, thereby completing fine-tuning of the interactive AI 20 during the learning period.

[0058] During the execution period of the interactive AI 20, prediction of defect occurrence is performed at any time, as shown in FIG. The crawler unit 50 transmits the processed measurement data at a predetermined period Tc2 to the interactive AI 20. This period Tc2 is, for example, 60 seconds, and is preferably equal to or shorter than the period Tc1 during which the first processing unit 30 acquires the measurement data output by each of the sensors SEN1, SEN2, ..., SENn.

[0059] Meanwhile, a prompt is input to inquire whether there is a possibility of a defect occurring at a predetermined period Tc3, and the interactive AI 20 returns a response based on the learned information to the user terminal 80 used by the user or the user's smartphone 82. This period Tc3 is, for example, 60 seconds, and is preferably equal to or shorter than the length of the period Tc1.

[0060] In addition, when a prompt is input from the user terminal 80 or the smartphone 82, for example, inquiring about a method for preventing the occurrence of a defect, the interactive AI 20 returns a response based on the defect handling information that it has learned.

[0061] As described above, the conversational AI 20 generates answers by regarding tokens in the first space as more relevant tokens than tokens in the second space, thereby reducing the likelihood of returning answers that are inaccurate. It is preferable that the user terminal 80 and the smartphone 82 are connected to the interactive AI 20 via a VPN (Virtual Private Network) connection.

[0062] In this manner, according to the defect occurrence prediction system 10 of this embodiment, the possibility of defects occurring in the machine parts manufactured by the manufacturing apparatus 12 is reduced.

[0063] Although the embodiment of the present invention has been described above, the present invention is not limited to the above-mentioned embodiment, and all changes in conditions that do not depart from the gist of the present invention are within the scope of application of the present invention. In the above-described embodiment, the time tm when a defect actually occurs in a mechanical part, the time ts when the defect occurrence specifying switches SW1, SW2, . . . SWn are operated, and the defect occurrence time tr are all examples of the information B. [Explanation of symbols]

[0064] 10 Defect prediction system 12 Manufacturing equipment 20 Conversational AI 30 First processing section 32 First memory unit 40 Second processing section 42 Second memory section 50 Crawler section 60 Learning Instructions 80 User Terminals 82 Smartphone 302 Data Processing Section 304 Time identification part 402 Conversion processing section 404 Alarm output section SEN1, SEN2, ... SENn sensors SW1, SW2, ... SWn Defect occurrence identification switch

Claims

1. A plurality of sensors that measure the state of a manufacturing device that manufactures an article and output measurement data, respectively; a data processing unit that performs data cleansing on the measurement data input at a predetermined first period and outputs the data as information A representing a state of the manufacturing apparatus; an operation unit operated by a user who has recognized a defect occurring in the article; a time specifying unit for specifying a time ts at which the operation unit is operated and acquiring information B which is a time at which a defect occurs in the article and is based on the operation of the operation unit; A conversational AI based on a large-scale language model; A transmission unit that transmits learning data including the information A, the information B, information C on the situation in which the defect occurred that is recorded after the defect occurred, and information D on a method of dealing with the defect that is recorded after the defect occurred to the interactive AI; a defect occurrence confirmation unit that transmits a prompt inquiring whether or not there is a possibility that a defect will occur in the item; A defect occurrence prediction system in which the transmission unit transmits the information A to the interactive AI that has learned using the learning data at a predetermined second period, and when the prompt is input from the defect occurrence confirmation unit at a predetermined third period, the interactive AI that has learned using the learning data returns a response based on the learning data.

2. 2. The defect occurrence prediction system according to claim 1, A defect occurrence prediction system further comprising an instruction unit that instructs the interactive AI trained using the learning data to correct a time lag that occurs between the time tm when a defect actually occurs in the item and the time ts.

3. 2. The defect occurrence prediction system according to claim 1, The defect occurrence prediction system further comprises a time correction unit for correcting a time lag occurring between the time tm when a defect actually occurs in the article and the time ts.

4. 4. The defect occurrence prediction system according to claim 2, The information B is a defect occurrence time tr described in a report created after the defect occurs, A processing unit (excluding a processing unit performed by a human) that compares the time ts with the defect occurrence time tr; and an alarm output unit (excluding an alarm performed by a human) which outputs an alarm to a user, notifying the user that the defect occurrence time tr and the time ts do not substantially coincide, when the time difference between the time ts and the defect occurrence time tr compared by the processing unit is larger than a predetermined value.

5. In the defect occurrence prediction system according to any one of claims 1 to 3, A defect occurrence prediction system, wherein the length of the second period is equal to or less than the length of the first period.

6. A defect occurrence prediction system according to any one of claims 1 to 3, A defect occurrence prediction system, wherein the length of the third period is equal to or less than the length of the first period.

7. A plurality of sensors for measuring the state of a manufacturing device for manufacturing an article and outputting respective measurement data; a data processing unit that performs data cleansing on the measurement data input at a predetermined first period and outputs the data as information A representing a state of the manufacturing apparatus; an operation unit operated by a user who has recognized a defect occurring in the article; a time specifying unit for specifying a time ts at which the operation unit is operated and acquiring information B which is a time at which a defect occurs in the article and is based on the operation of the operation unit; A conversational AI based on a large-scale language model; A transmission unit that transmits learning data including the information A, the information B, information C on the situation in which the defect occurred that is recorded after the defect occurred, and information D on a method of dealing with the defect that is recorded after the defect occurred to the interactive AI; a defect occurrence confirmation unit that transmits a prompt inquiring whether or not there is a possibility that a defect will occur in the article, The transmission unit transmits the learning data to the conversational AI; The conversational AI learns the learning data; The transmission unit transmits the information A to the interactive AI that has learned using the learning data at a predetermined second period, and when the prompt is input from the defect occurrence confirmation unit at a predetermined third period, the interactive AI that has learned using the learning data returns a response based on the learning data.

Images