Internal restraint device

Abstract

Provided is an internal restraint device capable of performing internal restraint at an appropriate timing. An internal restraint device according to the present disclosure comprises a restraint execution unit that detects the procedure of inspection work carried out by a maintenance person, and performs restraint on the maintenance person when the detected procedure of the inspection work deviates from a regular procedure. The internal restraint device may comprise a work state estimation unit that estimates the state of the inspection work being carried out by the maintenance person. The restraint execution unit may select a restraint implement for use in restraint and the content of the restraint in accordance with the estimation result of the work state estimation unit.

Description

internal check device

[0001] The present disclosure relates to an internal check system.

[0002] Japanese Patent Application Laid-Open No. 2009-102499 discloses a maintenance work management device that can prevent unperformed work items from being erroneously recorded as completed. This maintenance work management device includes a work structure data acquisition unit that acquires work structure data that shows inspection targets in building facility maintenance work hierarchically represented by higher-level inspection targets and lower-level inspection targets included in the higher-level inspection targets, a determination method data acquisition unit that acquires determination method data that shows a determination method for determining whether a work item has been performed, and a validity determination unit that determines the validity of information indicating that a work item has been performed. If the determination method data does not include a determination method for a work item, the validity determination unit determines the validity of the information indicating that the work item has been performed in accordance with a determination method for another work item that corresponds to an inspection target included in a higher-level inspection target of the inspection target corresponding to the work item in the work structure data and at the lowest level.

[0003] International Publication No. 2017-187564

[0004] Conventional technology has the problem that it is not possible to check at any time other than when the implementation results are entered, which makes it easy for work to be reworked.

[0005] The present disclosure has been made to solve the above-mentioned problems, and an object of the present disclosure is to provide an internal check device that can perform internal checks at appropriate times.

[0006] The internal check device disclosed herein is equipped with a check execution unit that detects the inspection procedure performed by a maintenance worker and checks the maintenance worker if the detected inspection procedure deviates from the correct procedure.

[0007] According to the present disclosure, it is possible to provide an internal check device that can perform internal checks at appropriate timing.

[0008] FIG. 1 is a diagram showing an internal check device according to embodiment 1. FIG. 2 is a diagram for explaining an example of a method by which a work state estimation unit estimates a work state. FIG. 3 is a diagram showing an example of the execution of internal check. FIG. 4 is a diagram showing an example of how checks are performed by the internal check device. FIG. 5 is a diagram showing an internal check device according to embodiment 2. FIG. 6 is a diagram showing an example of internal checks performed in embodiment 2. FIG. 7 is a diagram showing example information displayed on a supervisor terminal. FIG. 8 is a diagram showing an internal check device according to embodiment 3. FIG. 9 is a diagram showing an internal check device according to embodiment 4. FIG. 10 is a diagram showing an internal check device according to embodiment 5. FIG. 11 is a diagram showing an internal check device according to embodiment 6. FIG. 12 is a diagram showing an example of a configuration for realizing the functions of the internal check device of the present disclosure.

[0009] Hereinafter, embodiments will be described with reference to the drawings. Common or corresponding elements in each drawing are designated by the same reference numerals, and descriptions thereof will be simplified or omitted. The configurations shown in the following embodiments are examples of the technical ideas of the present disclosure, and may be combined with other known technologies, or multiple technical ideas described in the present disclosure may be combined. Furthermore, it is also possible to omit or modify part of the configuration without departing from the gist of the present disclosure.

[0010] 1 is a diagram showing an internal check device according to embodiment 1. The internal check device 1 of this embodiment performs internal checks to prevent, for example, a maintenance worker inspecting an elevator or facility such as a power plant from fraudulently declaring that work has been performed when it has not.

[0011] The facility is equipped with a facility control device 90. A maintenance worker who inspects the facility uses an inspection terminal 91. The inspection terminal 91 may be, for example, a notebook computer or a tablet terminal.

[0012] The internal check device 1 includes a check execution unit 2 that detects the inspection procedure performed by a maintenance worker and checks the maintenance worker if the detected inspection procedure deviates from the correct procedure. The internal check device 1 may detect the inspection procedure based on data input by the maintenance worker to the inspection terminal 91, for example, or may detect the inspection procedure based on equipment control signal data from the equipment control device 90. The check execution unit 2 checks the worker by, for example, asking a question such as "Aren't the work procedures inaccurate?" via a speaker provided in the equipment or a speaker or screen of the inspection terminal 91. This allows checks to be performed at the appropriate time, preventing the occurrence of rework.

[0013] The internal check device 1 of this embodiment includes a work state estimation unit 3 that estimates the status of inspection work being performed by a maintenance worker. The check execution unit 2 may select the check device to be used for check and the check content according to the estimation result of the work state estimation unit 3. The work state estimation may be performed using pre-constructed logic or a machine learning method. The work state estimation unit 3 may estimate the work state using a work estimation model 10. For example, as on-site data, equipment control signal data or inspection terminal data is collected in the on-site data collection unit 6. The work state estimation unit 3 estimates the work state at that time (e.g., work A in progress, work A completed, all work completed) based on characteristic changes in the on-site data. The work state estimation unit 3 outputs the result as a work state estimation result.

[0014] If the facility is an elevator, the restraining device used for restraint may be selected from, for example, an intercom inside the car, a speaker on the car, an inspection terminal 91, etc.

[0015] FIG. 2 is a diagram illustrating an example of a method by which the work state estimation unit 3 estimates the work state. As shown in FIG. 2 , for example, the start of work may be estimated from the ON of an inspection mode signal, which can be acquired as equipment control signal data. The start of work may also be estimated from data, which can be acquired as inspection terminal data, indicating that the inspection terminal 91 has been connected to the equipment. It may also be estimated that a certain work is being performed from characteristics of changes in the equipment control signal data or the inspection terminal data (such as ON / OFF signals for safety devices, equipment operation on the inspection terminal, etc.). The completion of all work may also be estimated from the OFF of an inspection mode signal, which can be acquired as equipment control signal data. The completion of work may also be estimated from the completion of input of inspection results to the inspection terminal 91.

[0016] The work state estimating unit 3 in this embodiment may estimate the state of the inspection work being performed by the maintenance worker along with the estimation accuracy.

[0017] For example, an inspection of the internal structure of a motor in an elevator pit is performed with the power turned off, so no control signal change data is recorded. On the other hand, a motor inspection is a task that takes at least 10 minutes. For this reason, if it is detected that there is only a 5-minute time between when a maintenance worker enters the pit and turns the power off and then turns it back on, it is weakly assumed that the inspection has not been carried out. In other words, it is estimated that "motor internal structure inspection" has been "not carried out" and has a "low probability."

[0018] As another example, if a door inspection should be performed before the inspection of the equipment's safety devices, and there is no record of a change in the door's ON / OFF signal in the data, it is strongly assumed that the door inspection has not been performed. In other words, it is estimated that "door inspection" has not been performed and there is a "high probability."

[0019] As another example, when an inspection of a facility's safety devices is carried out, the change in the safety device ON / OFF signal is recorded in the data. Based on the data, if there is an ON / OFF change, it is assumed that the safety device inspection has been carried out, and the result is "safety device inspection" "implemented" and "high probability."

[0020] As another example, even if an inspection of an oil pan in an elevator pit is carried out, it will not be recorded as control signal change data. On the other hand, the fact that a maintenance worker has entered the area where the oil pan is located will be recorded in the control signal data. Since the maintenance worker has been near the oil pan, it is weakly inferred that the inspection has been carried out. In other words, it is inferred that "oil pan inspection" has been "carried out" and "with low accuracy."

[0021] The internal check device 1 includes an ideal work incompleteness estimation unit 4 that estimates an ideal work incompleteness state that represents work that should have been performed up to that point in time and work that should be performed after that point in time. The ideal work incompleteness estimation unit 4 estimates the ideal work incompleteness state based on the work state estimated by the work state estimation unit 3, the regular inspection procedure stored in the work flow definition database 7, and on-site data. In this embodiment, the ideal work incompleteness estimation unit 4 may estimate the ideal work incompleteness state with estimation accuracy. The ideal work incompleteness estimation unit 4 outputs the estimated result as an ideal work incompleteness state estimation result.

[0022] The check selection unit 8 selects a check method based on the work state estimation result, the desired work incomplete state estimation result, and the estimation accuracy of these. The check execution unit 2 executes the selected check method.

[0023] Figure 3 is a diagram showing an example of internal checks being executed. In the example at the top of Figure 3, it is estimated with low certainty that a door inspection has not been carried out, and with high certainty that a safety device inspection is currently being carried out. In this case, a question is asked over a facility speaker, for example, "Has the door inspection not been carried out yet?" Checks using non-assertive language like this are called "indirect checks."

[0024] The example in the bottom part of Figure 3 shows a case where it is estimated with high accuracy that safety device inspection has been completed, and the next inspection, inspection X, has not been performed, but the maintenance worker inputs that inspection X has been completed into the system. In this case, a warning is issued from the inspection terminal saying, "Inspection X has not been performed." This type of deterrence using definitive language is called "direct deterrence." Furthermore, in this case, the input indicating that inspection X has been performed is deleted.

[0025] The internal check device 1 of this embodiment mechanically detects when there is a clear procedural omission during work, when an operation has clearly been omitted, or when an operation has clearly been omitted but has been performed, or when there is a possibility that an operation has been omitted but has been performed. When this detection occurs, a voice message is issued over the elevator speaker, for example, saying, "Based on the data, we have detected the possibility that an operation has been omitted." This makes the maintenance worker aware that related data is being collected or monitored.

[0026] If there is a task that may have been marked as completed but not yet performed, a screen that requires the maintenance worker to promise to perform the task is displayed, for example, on the inspection terminal 91. Even when the supervisor who oversees the maintenance worker checks, the task may have been marked as completed, but the screen on the supervisor's terminal 5 shows that the task has been performed and that the worker has finally confirmed it, and when the supervisor approves the task, the supervisor asks, "Did you really do it?" The supervisor may then display on the screen of the supervisor's terminal 5 the person in charge or the site with a history of marking the task as completed but not yet performed, and the supervisor may visit the site for a quality patrol. In this way, this embodiment can apply psychological pressure to prevent the occurrence of misconduct by marking the task as completed but not yet performed.

[0027] In the example shown in FIG. 1, the inspection result database 9 stores information on whether each inspection work has been performed or not, which is received from the inspection terminal 91 .

[0028] In the example shown in FIG. 1 , the work flow definition database 7 stores the regular procedures for inspection work as an inspection flow definition. The work state estimation unit 3 estimates the work state at that time, along with the estimation accuracy, from the on-site data collected by the on-site data collection unit 6 and the inspection flow definition. The work state may be, for example, a certain work in progress, a certain work completed, or all work completed. The ideal work completion / non-completion estimation unit 4 estimates the ideal work completion / non-completion state, along with the estimation accuracy, from the work state estimated by the work state estimation unit 3, the on-site data, and the inspection flow definition.

[0029] 4 is a diagram showing an example of a check method executed by the internal check device 1. The check method definition database 11 defines the check device to be used for check and the check content. The check selection unit 8 selects the check device and check content defined in the check method definition database 11, and the check execution unit 2 executes check in accordance with the selected check device and check content.

[0030] The check method definition database 11 stores information that defines the check method corresponding to, for example, the work status, the work that differs from the expected work status, and the estimated accuracy of the work. The check selection unit 8 references the definition and selects the check method (check device, check content) corresponding to the work phase, the work status, and the estimated accuracy (step S1). The check execution unit 2 executes check via a device corresponding to the check method (step S2). For example, if work is in progress, the check is issued by voice from the facility speaker, and if work has been completed, the check is issued by display on a terminal such as a personal computer.

[0031] The check method shown in Fig. 4 is stored in the check method definition database 11. In the example of Fig. 4, when the work status is during safety device inspection, the work that differs from the expected work-not-performed state is that the previous work has not been performed, and the estimated accuracy is 90% or higher, a check voice saying "XX work has not been performed" is output using a speaker of the building equipment or inspection terminal 91 as a check device.

[0032] As another example in Figure 4, if the work status is that all work has been completed and inspection results have been input, and an operation that differs from the expected state in which work has not been performed is that of an oil pan inspection that has not been performed (input as performed but estimated as not having been performed), and the estimation accuracy is 60% or higher, a message will be output on the screen or speaker of the inspection terminal 91 saying, "Have you forgotten to perform work XX? Is the input correct?" or a voice output will be displayed.

[0033] In this embodiment, direct checks are performed when the accuracy of the estimation made by the work status estimation unit 3 or the accuracy of the estimation made by the ideal work execution / non-execution estimation unit 4 is high, and indirect checks are performed when the accuracy of the estimation is low. This makes it possible to effectively check work that is estimated to have not been performed with low accuracy. As a result, the fraud prevention effect is improved.

[0034] This embodiment enables checks that are appropriate for each phase, such as when work is in progress, when a certain task is completed, or when all work has been completed, thereby improving work efficiency and preventing fraud. For example, if a task that should have been performed before the completion of a certain task is missed, instructions can be issued from the facility speaker to perform the task, minimizing the effort of moving the work location to redo the task. In addition, by receiving instructions during work, maintenance personnel feel that they are being constantly watched, improving the prevention of fraud.

[0035] 5 to 7, a second embodiment will be described, focusing on differences from the first embodiment described above, and explanations of commonalities will be simplified or omitted. Elements common to or corresponding to the elements described above will be denoted by the same reference numerals.

[0036] 5 is a diagram showing an internal check device 1 according to embodiment 2. As shown in FIG. 5, the internal check device 1 according to this embodiment includes a check history database 12 that records the check history of each maintenance personnel. The check execution unit 2 can select a check method according to the check history.

[0037] The check history database 12 records the check history of each maintenance worker (such as the number of times an attempt was made to enter an item as performed when the estimated result was that it had not been performed, or the number of times the same task was forgotten).

[0038] Fig. 6 is a diagram showing an example of internal checks executed in the second embodiment. In the example shown in Fig. 6, for a maintenance worker with little check history, a check voice message saying "X work has not been performed" is simply output, whereas for a maintenance worker with a lot of check history, a notification is also sent to the supervisor terminal 5. This allows for more effective checks. In the example of Fig. 6, when the check selection unit 8 selects a check method and the check execution unit 2 executes the check, the history is recorded in the check history database.

[0039] In this embodiment, the percentage of inspection work that has not been performed that has been entered as performed for each maintenance worker may be displayed on the supervisor terminal 5 based on past check history. FIG. 7 is a diagram showing an example of information displayed on the supervisor terminal 5. In the example of FIG. 7, the names of the inspectors (maintenance workers) are displayed in descending order of the percentage of inspection work that has not been performed, such as "Possible fraud: 70%" and "Possible fraud: 50%," and a site patrol priority ranking is generated. In this way, it is possible to prioritize sites where there are concerns about inspection quality as site patrol targets based on past check history. Since high-risk sites can be prioritized, this leads to improved patrol efficiency and inspection quality.

[0040] Embodiment 3 Next, a third embodiment will be described with reference to Fig. 8. The description will focus on the differences from the first embodiment described above, and the description of the commonalities will be simplified or omitted. Elements that are common to or correspond to the elements described above will be denoted by the same reference numerals.

[0041] 8 is a diagram showing an internal check device 1 according to a third embodiment. As shown in FIG. 8, the internal check device 1 of this embodiment is provided with a post-check inspection result database 13 that records the post-check inspection results. In this embodiment, the post-check inspection result database 13 is used to compare the pre-check inspection results self-reported by the maintenance personnel with the post-check inspection results, and the results can be displayed as before and after on the supervisor terminal 5 or the inspection terminal 91.

[0042] In this embodiment, the supervisor can check the before and after on the supervisor terminal 5, which can be used for follow-up after the work is completed by the supervisor. This improves the check effect. Also, the inspector can check the before and after on the inspection terminal 91, which can be used for self-learning. This improves the check effect.

[0043] Embodiment 4 Next, a fourth embodiment will be described with reference to Fig. 9, focusing on differences from the first embodiment described above, and explanations of commonalities will be simplified or omitted. Elements common to or corresponding to the elements described above will be denoted by the same reference numerals.

[0044] FIG. 9 is a diagram illustrating an internal check device 1 according to a fourth embodiment. As shown in FIG. 9, the internal check device 1 according to this embodiment has a check effect database 14 that records the check method and effect of checks that have been executed. The effect of checks refers to the number of times checks were ignored, the number of times that unperformed work was prompted, and the number of times that unperformed work was prevented from being entered as completed. The check effect estimation unit 15 estimates whether the checks that have been executed have been effective. For example, if checks are ignored, the check effect estimation unit 15 estimates that the check effect was ineffective, and if unperformed work was prompted or the entry of unperformed work as completed was prevented. The estimation results of the check effect estimation unit 15 are recorded in the check effect database 14.

[0045] The internal check device 1 may have a model generation unit that generates a trained model for inferring effective check methods according to the work phase and the estimated accuracy of the work by performing machine learning using data recorded in the check effect database 14 as learning data. The internal check device 1 may also have an update unit that uses the trained model to automatically update the check method definition database 11 that defines the check methods to be executed. This makes it possible to implement more effective checks. In the example of FIG. 9 , the check method definition generation unit 16 generates new check methods using the trained model, and the check method definition database 11 stores the generated check methods.

[0046] The learning algorithm used by the model generation unit can be a known algorithm such as supervised learning, unsupervised learning, or reinforcement learning. As an example, a case where a neural network is applied will be described. The model generation unit learns effective deterrence methods, for example, by so-called supervised learning according to a neural network model. Here, supervised learning refers to a method in which a learning device is provided with pairs of input and result (label) data, and the device learns the features of the learning data and infers the result from the input.

[0047] A neural network is composed of an input layer consisting of multiple neurons, an intermediate layer (hidden layer) consisting of multiple neurons, and an output layer consisting of multiple neurons. The intermediate layer may be one layer, or two or more layers.

[0048] For example, in a three-layer neural network, when multiple inputs are input to the input layer, the values ​​are multiplied by a weight W1 before being input to the middle layer, and the result is then multiplied by a weight W2 before being output from the output layer. This output result changes depending on the values ​​of the weights W1 and W2.

[0049] In the present application, the neural network learns effective methods of deterrence through so-called supervised learning in accordance with learning data created based on a combination of the methods of deterrence that have been executed and the effects of the deterrence.

[0050] The trained model storage unit stores the trained model output from the model generation unit.

[0051] The check effect database 14 acquires the check method and the check effect that were executed. Note that, although the check method and the check effect that were executed are acquired simultaneously, it is sufficient if the check method and the check effect that were executed are input in association with each other, and the data on the check method and the check effect that were executed may be acquired at different times.

[0052] The deterrence method definition generation unit 16 learns effective deterrence methods through so-called supervised learning in accordance with learning data created based on the combination of deterrence methods executed and the effects of deterrence acquired by the deterrence effect database 14, and generates a learned model.

[0053] Furthermore, the check method definition generation unit 16 infers effective check methods obtained by utilizing the trained model. That is, by inputting an executed check method into this trained model, it is possible to output an effective check method inferred from the executed check method.

[0054] In this embodiment, the case where supervised learning is applied to the learning algorithm used by the model generation unit has been described, but the present invention is not limited to this. As for the learning algorithm, reinforcement learning, unsupervised learning, semi-supervised learning, etc. can also be applied in addition to supervised learning.

[0055] Fifth Embodiment Next, a fifth embodiment will be described with reference to Fig. 10. The description will focus on the differences from the first embodiment described above, and the description of the commonalities will be simplified or omitted. Elements that are common to or correspond to the elements described above will be given the same reference numerals.

[0056] 10 is a diagram showing an internal check device 1 according to a fifth embodiment. As shown in FIG. 10, the internal check device 1 of this embodiment includes a user interface 17 that allows definition information to be input to a check method definition database 11 that defines check methods, and a definition support unit 18 that, depending on the constraints of the check method to be defined and the system configuration, prevents definitions that do not satisfy the constraints or configuration from being input to the check method definition database 11. For example, the definition support unit 18 determines that a field device cannot be defined as a check device unless the work phase is currently in progress. This more reliably prevents errors from occurring in the definition of the check method.

[0057] Sixth Embodiment Next, a sixth embodiment will be described with reference to Fig. 11. The description will focus on the differences from the first embodiment described above, and the description of the commonalities will be simplified or omitted. Elements that are common to or correspond to the elements described above will be given the same reference numerals.

[0058] FIG. 11 is a diagram illustrating an internal check device 1 according to a sixth embodiment. As illustrated in FIG. 11 , the internal check device 1 of this embodiment includes an eye camera (wearable camera 19) equipped on a maintenance worker, a learning data generation unit 20, and a model update unit 21. The wearable camera 19 acquires video data during work. The learning data generation unit 20 estimates the time period during which each task will be performed from the acquired video data. Next, the learning data generation unit 20 extracts other on-site data other than the video (such as equipment control signal data and inspection terminal operation data) within the time period during which each task will be performed. The extracted data represents data changes during the performance of each task. The model update unit 21 uses the data generated by the learning data generation unit 20 as learning data for constructing an activity estimation model 10 and updates the activity estimation model 10.

[0059] This embodiment reduces the time and effort required to construct the activity estimation model 10. Furthermore, the estimation accuracy of the activity estimation model 10 improves as the amount of training data increases.

[0060] Work state estimation unit 3 may estimate what work is being done from the image captured by wearable camera 19.

[0061] FIG. 12 is a diagram showing an example of a configuration for realizing the functions of the internal check device 1 of the present disclosure. Each function of the internal check device 1 is realized, for example, by a processing circuit. The processing circuit may be dedicated hardware 600. The processing circuit may include a processor 601 and a memory 602. A portion of the processing circuit may be formed as dedicated hardware 600, and the processing circuit may further include a processor 601 and a memory 602. In the example shown in FIG. 12, a portion of the processing circuit is formed as dedicated hardware 600. Furthermore, in the example shown in FIG. 12, the processing circuit further includes a processor 601 and a memory 602 in addition to the dedicated hardware 600.

[0062] The processing circuitry of which at least one portion is dedicated hardware 600 may be, for example, a single circuit, multiple circuits, a programmed processor, parallel programmed processors, an ASIC, an FPGA, or a combination thereof.

[0063] When the processing circuit has at least one processor 601 and at least one memory 602, the functions of each part of the internal check device 1 are realized by software, firmware, or a combination of software and firmware.

[0064] The software and firmware are written as programs and stored in memory 602. The processor 601 realizes the functions of each unit by reading and executing the programs stored in memory 602. The processor 601 is also called a CPU (Central Processing Unit), central processing unit, processing unit, arithmetic unit, microprocessor, microcomputer, or DSP. The memory 602 may be, for example, a non-volatile or volatile semiconductor memory such as RAM, ROM, flash memory, EPROM, or EEPROM, or a magnetic disk, flexible disk, optical disk, compact disk, minidisk, DVD, etc.

[0065] In this way, the processing circuit can realize the functions of the internal check device 1 by hardware, software, firmware, or a combination of these. Note that each function of the internal check device 1 may be realized by multiple devices working together, or may be realized by a single device. Furthermore, at least some of the functions of the internal check device 1 may be implemented on a server or the like on an external network.

[0066] Of the features of the above-described multiple embodiments, two or more features that can be combined may be combined and implemented.

[0067] DESCRIPTION OF SYMBOLS 1 Internal check device, 2 Check execution unit, 3 Work status estimation unit, 4 Work execution / non-execution estimation unit, 5 Supervisor terminal, 6 Field data collection unit, 7 Definition database, 8 Check selection unit, 9 Inspection result database, 10 Work estimation model, 11 Check method definition database, 12 Check history database, 13 Inspection result database, 14 Check effect database, 15 Check effect estimation unit, 16 Check method definition generation unit, 17 User interface, 18 Definition support unit, 19 Wearable camera, 20 Learning data generation unit, 21 Model update unit, 90 Equipment control device, 91 Inspection terminal, 600 Dedicated hardware, 601 Processor, 602 Memory

Claims

1. An internal check device having a check execution unit that detects the inspection procedures performed by a maintenance technician and checks the maintenance technician if the detected inspection procedures deviate from the correct procedures.

2. An internal check device as described in claim 1, further comprising a work status estimation unit which estimates the status of the inspection work being performed by the maintenance worker, and the check execution unit selects the check equipment to be used for check and the content of the check depending on the estimation result of the work status estimation unit.

3. An internal check device as described in claim 2, wherein the work status estimation unit estimates the status of the inspection work using equipment control signal data or inspection terminal data.

4. An internal check device as described in any one of claims 1 to 3, further comprising an ideal work not being performed estimation unit that estimates an ideal work not being performed state representing work that should have been performed up to that point and work that should be performed after that point, and the check execution unit selects the check equipment to be used for check and the content of the check depending on the estimation result of the ideal work not being performed estimation unit.

5. An internal check device as described in claim 2 or claim 3, wherein the work status estimation unit estimates the status of the inspection work being performed by the maintenance worker with an estimated accuracy, and the check execution unit performs direct check when the estimated accuracy is high, and performs indirect check when the estimated accuracy is low.

6. An internal check device as described in claim 4, wherein the ideal work incompletion estimation unit estimates the ideal work incompletion state with an estimated accuracy, and the check execution unit performs direct check when the estimated accuracy is high and performs indirect check when the estimated accuracy is low.

7. An internal check device as described in any one of claims 1 to 6, comprising a database that records the past check history of each maintenance personnel, and the check execution unit selects the method of check depending on the check history.

8. An internal check device as described in any one of claims 1 to 7, capable of displaying on a supervisor terminal the percentage of individual maintenance personnel who have entered unperformed inspection work as having been performed based on past check history.

9. An internal check device as described in any one of claims 1 to 8, capable of comparing the inspection results before check and the inspection results after check, as self-reported by the maintenance personnel, and displaying the results on a supervisor terminal or an inspection terminal.

10. An internal check device as described in any one of claims 1 to 9, comprising: a model generation unit that records the check method and effect executed as a check effect database, and generates a trained model for inferring an effective check method according to the work phase and the estimated accuracy of the work by performing machine learning using the check effect database as learning data; and an update unit that uses the trained model to automatically update a check method definition database that defines the check method to be executed.

11. An internal check device as described in any one of claims 1 to 10, comprising: a user interface that enables definition information to be input into a check method definition database that defines check methods; and a definition support unit that, depending on the constraints of the check method to be defined and the system configuration, makes it impossible to input definitions that do not satisfy the constraints or configuration into the check method definition database.

12. An internal check device as described in any one of claims 1 to 11, which acquires video data during work captured by an eye camera worn by the maintenance worker, estimates the time period during which each task is performed from the acquired video data, extracts other on-site data other than the video within the range of the time period during which each task is performed, and utilizes the extracted data as learning data for constructing a task estimation model.

Images