Maintenance record input support device

Abstract

To provide a maintenance record inputting support device capable of reducing a step for inputting a maintenance record and improving a re-usability of the maintenance record.SOLUTION: A maintenance record inputting support device comprises: keyword extracting means for inputting a device status of a maintenance target device and extracting a keyword which represents the device status; option estimation means for estimating an option of a maintenance item using the keyword and presenting the estimated option to an outside; and maintenance record forming means for performing maintenance based on the options presented outside, and forming a maintenance record from a maintenance result and a maintenance content freely described by a maintenance staff, and storing the formed maintenance record in a maintenance record database.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a maintenance record input support device that supports the input of maintenance records. [Background technology]

[0002] By properly recording information about maintenance performed at various sites and devices (hereinafter referred to as target devices), it can be used for the next monitoring inspection and its planning, and can also be used for managing the lifespan of devices, etc.

[0003] Regarding this point, for example, Patent Document 1 proposes a configuration in which, with the aim of reducing the burden on workers when entering failure conditions, "some of the information displayed on the inspection input screen before the screen is switched is automatically transcribed into specified input item fields on the failure input screen. Also, for some of the input item fields, pre-registered options are provided to prompt the user to select and input, and the option selected and input is captured as input data. Furthermore, when the failure report comprehensive management unit of the equipment maintenance management device receives multiple failure reports that originate from the same source, a management number for the failure that becomes the comprehensive source is further assigned in addition to the management number that is uniquely assigned to the failure report unit, and when the response to the failure at the comprehensive source is completed, this is automatically reflected in all comprehensive failure reports that have been assigned the management number of the comprehensive source." [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2004-54348 A Summary of the Invention [Problem to be solved by the invention]

[0005] The method described in Patent Document 1 involves a selection system in which, when a maintenance worker inputs the failure status of the target equipment into a computer device, the computer device presents multiple phenomena, causes, countermeasures, etc., and the maintenance worker is asked to select from these the one that appears to be most similar. Maintenance records are then automatically generated based on this, thereby reducing the labor required to create the maintenance records.

[0006] However, when there are too many options, it takes time to select the appropriate one, and conversely, when there are too few options, the relevant option may not exist, and even inexperienced maintenance workers may not be able to select the appropriate option.

[0007] In view of the above, an object of the present invention is to provide a maintenance record input support device that can reduce the number of steps required for inputting maintenance records and improve the reusability of the maintenance records. [Means for solving the problem]

[0008] In view of the above, the present invention provides a keyword extraction means for inputting the device status of a device to be maintained and extracting keywords that represent the device status, an option estimation means for estimating options for maintenance items using the keywords and presenting them to the outside, and a maintenance record creation means for performing maintenance based on the options presented to the outside, creating a maintenance record from the maintenance results and the maintenance content freely written by the maintenance technician, and storing the record in a maintenance record database. An option estimation model construction and update means for generating an option estimation model by learning using the maintenance records; Equipped with The option estimation means refers to the option estimation model using the keyword, estimates the options of the maintenance items, and presents them to the outside. This is a "maintenance record input support device." Effect of the Invention

[0009] According to the present invention, it is possible to reduce the number of steps required for inputting maintenance records and to improve the reusability of maintenance records. [Brief description of the drawings]

[0010] [Figure 1] FIG. 1 is a diagram showing an example of a basic configuration of a maintenance record input support device according to a first embodiment of the present invention. [Diagram 2]FIG. 2 is a diagram showing the relationship between a maintenance worker M1 and a user interface UI1. [Diagram 3] A diagram showing an example of the configuration of maintenance record data D in the maintenance record database DB. [Figure 4] FIG. 2 is a diagram showing an example of the configuration of an option estimation model construction / update means 11. [Diagram 5] FIG. 11 is a diagram showing an example of a basic configuration of a maintenance record input support device according to a second embodiment of the present invention. [Figure 6] FIG. 11 is a diagram showing an example of the basic configuration of a maintenance record input support device according to a third embodiment of the present invention. [Figure 7] FIG. 11 is a diagram showing an example of the basic configuration of a maintenance record input support device according to a fourth embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] Hereinafter, embodiments of the present invention will be described with reference to the drawings. EXAMPLES

[0012] In the first embodiment, a basic configuration example of a maintenance record input support device according to the present invention will be described.

[0013] Fig. 1 shows an example of the basic configuration of a maintenance record input support device according to a first embodiment of the present invention. In this figure, the maintenance record input support device 1 is composed of a computer device, and during maintenance work on a target device, a maintenance worker M1 inputs input information I such as maintenance information of the target device while looking at display information O displayed on a monitor screen of a user interface UI1 of the maintenance record input support device 1. The input information is recorded as maintenance record information D in a maintenance record database DB.

[0014] Data processing in the maintenance record input support device 1 is executed in the calculation unit of a computer device according to a program, but the processing contents are represented as means in Fig. 1. The main processing means here are a keyword extraction means 13, an option estimation means 12, an option estimation model construction and update means 11, a maintenance record creation means 10, etc., and the details of these will be described later.

[0015] 2 is a diagram showing the relationship between the maintenance worker M1 and the user interface UI1. With reference to this diagram, input information I of the maintenance worker M1 and display information O presented by the user interface UI1 will be specifically explained.

[0016] First, the maintenance worker M1 inputs the device status of the target device that is the target of maintenance as input information I1. The input information I1 here may be any of the following: an error code I1a that occurs in the target device, image information I1b of the target device for maintenance, and fault state information I1c of the target device in free text. Alternatively, it may be fault state information 11c in free text including the error code I1a. In the case of FIG. 1, image information I1b of the target device for maintenance is input by the maintenance worker M1, and in the case of FIG. 2, fault state information 11c in free text is input by the maintenance worker M1.

[0017] Returning to Fig. 1, the keyword extraction means 13 converts the input image information I1b of the maintenance target device into keywords that represent the device state of the maintenance target device and extracts them by image recognition using the image information I1b. Similarly, the keyword extraction means 13 in Fig. 2 converts the input free description failure state information 11c into keywords that represent the device state of the maintenance target device and extracts them.

[0018] Next, the maintenance worker M1 presses an "option estimation" button (not shown) on the user interface UI1. In response to this, the option estimation means 12 in Fig. 1 uses an option estimation model obtained by the option estimation model construction and update means 11 (described later) to estimate options based on the failure status (device status) indicated by the keyword extracted by the keyword extraction means 13, and displays the estimated options on the monitor screen of the user interface UI1.

[0019] The display example in Figure 2 shows an example of the appropriate options being inferred for each item based on the input failure situation, with the items being "Phenomenon," "Cause," and "Countermeasure." When there are multiple options, such as for the cause, it is best to sort the options in order of likelihood and display multiple options.

[0020] Maintenance worker M1 performs maintenance work on the target device based on the options (display information O1: option estimation result) presented on the user interface UI1. At this time, if a "phenomenon" or "cause" different from the presented options is confirmed, or a different "measure" is taken, the maintenance worker M1 inputs input information I2 (option selection result) to correct the option and performs this correction work.

[0021] Returning to Figure 1, the maintenance record creation means 10 creates a maintenance record using the option selection result I2, which is the final maintenance result corrected by the maintenance worker M1 for the options presented by the computer device, and the image information I1b of the device to be maintained that was input at this time. The maintenance record created at this time includes information on the device status and maintenance results freely written by the maintenance worker M1. Finally, when the maintenance worker M1 presses the "save maintenance record" button (not shown) on the user interface UI1, the maintenance record input support device 1 saves the estimated results of the options and the correction results as maintenance record information D in the maintenance record database DB.

[0022] FIG. 3 shows an example of the configuration of maintenance record data D in the maintenance record database DB. The stored maintenance record data D is composed of at least the equipment status D2 freely entered by the maintenance worker M1 for each case D1, and the options D3 selected for each item in the maintenance record. The options D3 should be recorded for each item of the maintenance, such as "phenomenon," "cause," and "measure." An example of a maintenance record stored in the maintenance record database DB is a case D1-1 in which the equipment status D2 is freely entered as "XX packing and AA replacement, BB tank replacement," and the options D3 are "phenomenon 1," "cause 3," and "measure 1." These maintenance records are recorded and stored each time maintenance work is performed on the target equipment.

[0023] The option estimation model construction and update means 11 in Fig. 1 forms an option estimation model by learning using the accumulated maintenance record data D. Therefore, the more abundant the maintenance record data D is, the more accurate the option estimation model becomes.

[0024] Fig. 4 is a diagram showing an example of the configuration of the option estimation model construction / update means 11. The option estimation model construction / update means 11 constructs and updates an option estimation model that estimates appropriate options to be entered for each item of the maintenance record from the freely written device status based on the maintenance record data D accumulated in the maintenance record database DB. The option estimation model is, for example, a probability model that represents the relationship between options as shown in Fig. 4 and between options and freely written words.

[0025] According to the example of the probability model in FIG. 4, the relationship between the cause and the phenomenon and the relationship between the cause and the countermeasure can be obtained as a probability. i ca is the occurrence probability of cause i, P ij ef is the probability that effect j occurs when cause i occurs, P ij ac is the probability of taking action j when cause i occurs, P i w is the probability that words w1, w2, and w3 appear in the device state entered by the maintenance worker in free text during phenomenon i.

[0026] According to this model, as illustrated on the left side of Figure 4, if phenomenon 1 occurs, the probability of occurrence of words w2 is highest, followed by w3; if phenomenon 2 occurs, the probability of occurrence of word w1 is highest; and if phenomenon 3 occurs, the probability of occurrence of words w3 is highest, followed by w2.

[0027] 1, when the option estimation means 12 refers to the option estimation model using the keyword extracted by the keyword extraction means 13, if the word meaning the keyword corresponds to, for example, the word w3, it is found that there is a high probability that phenomenon 3, followed by phenomenon 1, has occurred, and further, it is possible to identify a cause of phenomenon 3 that is likely to cause this phenomenon 3, and then it is possible to identify measures that are likely to be taken against this cause in order of probability. In this way, the options for each item found in order of occurrence probability for each of "phenomenon," "cause," and "measure" are "phenomenon 1," "cause 3," and "measure 1," and it is possible to propose these to maintenance worker M1 as option estimation results O1.

[0028] According to the first embodiment, in order to reduce the man-hours required for creating a maintenance record, keywords can be extracted from, for example, an image of a device to be maintained input by a maintenance worker M1, and the contents to be entered in each item of the maintenance record can be estimated. This allows the creation of a maintenance record to be completed simply by inputting an image and confirming and correcting the estimated items. EXAMPLES

[0029] In the second embodiment, a method for constructing an option estimation model with higher accuracy will be described.

[0030] Fig. 5 shows an example of a configuration in which a model performance improvement unit 1A is further added to the maintenance record input support device 1 for this purpose. Note that in the example of Fig. 5, instead of the image information I1b of the maintenance target device in Fig. 1, fault condition information 11c is input in the form of free description including an error code I1a, but as described above, either may be used.

[0031] The model performance improvement unit 1A may be configured within the maintenance record input support device 1, but here, an example is shown in which it is configured in a separate computer device. A domain expert M2 accesses a user interface UI2 of the model performance improvement unit 1A and inputs information I13 of known inter-item dependency relationships from here via dependency relationship input means 14. Information I1d of inter-item dependency relationships is reflected in option estimation model construction and update means 11.

[0032] As a result, for example, in the choice estimation model in Figure 4, if it is known from the information I13 on the known inter-item dependency that there is a high interdependence relationship between "Phenomenon 1," "Cause 3," and "Countermeasure 1," the probability P 32 ef and P 21 ac By correcting the choice inference model so that the probability value of becomes high, the choice inference model can be configured with higher accuracy.

[0033] According to the second embodiment, in order to reduce the man-hours required for creating maintenance records, the dependency relationships between maintenance record items are input and the option estimation model is re-learned. After that, based on the dependency relationships and past maintenance records, keywords are extracted from the free description of the device error code and failure state input by the maintenance worker, and appropriate options for the maintenance record items: "phenomenon," "cause," and "action" are estimated. In this way, by using the dependency relationships between maintenance record items, appropriate options can be estimated with higher accuracy. EXAMPLES

[0034] Example 3 shown in FIG. 6 represents a configuration in which the option estimation means 12 displays and outputs estimation results O1a, O1b, and O1c for each of the items "phenomenon," "cause," and "measure," and further allows selection results I2a, I2b, and I2c to be input for each of these.

[0035] According to the third embodiment, in order to present an appropriate response procedure to reduce the man-hours for maintenance work, keywords are extracted from the free description of the device error code and failure state entered by the maintenance worker M1, the "phenomenon" of the maintenance record is inferred, the "cause" is inferred based on the confirmation and correction of the "phenomenon" by the maintenance worker M1, and then the "measure" is inferred based on the confirmation and correction of the "cause" by the maintenance worker M1. In this way, a more useful response procedure can be presented by re-estimating the "cause" and "measure" based on the facts discovered during the maintenance work. EXAMPLES

[0036] In Example 4, a response when a new, unexperienced event is observed will be described. Here, a case is assumed in which there is no corresponding option inferred item on the user interface UI1 in FIG.

[0037] The display screen in Figure 2 has options for the presumed phenomenon, such as data corruption, function outage, line stop, and display not possible, but in this case, none of these apply to the events experienced by maintenance worker M1. In this case, maintenance worker M1 selects "Other" (free description) in this column, and writes the detected data such as the device status in free description format, which becomes input I2.

[0038] Fig. 7 shows an example of a configuration in which a cluster label setting unit 1B is further added to the maintenance record input support device 1 for this purpose. The cluster label setting unit 1B may be configured within the maintenance record input support device 1, but here shows an example in which it is configured in a separate computer device. The user interface UI3 of the cluster label setting unit 1B is accessible by the domain expert M2.

[0039] In the cluster label setting unit 1B in Fig. 7, data on the target device and its malfunction, which is written in free text format by the maintenance worker M1 when a new unexperienced phenomenon is observed, is extracted from the maintenance record database DB, and clustered by the clustering means 15. As a result, the unexperienced data is not included in the existing clusters, but is classified into a new cluster. It is necessary to define an appropriate cluster name for the new cluster, and the label name estimation means 16 assigns a name that is deemed appropriate, and presents the cluster and its label name (O2) to the domain expert M2 via the user interface UI3.

[0040] The domain expert M2 refers to the presented cluster and its label name (O2) and provides an input Ie (cluster label correction result) that he / she deems more appropriate or accepts the proposal to the clustering label name correction means 17 via the user interface UI3. As a result, a new cluster including a new, unexperienced phenomenon is defined in the maintenance record database DB, and this cluster contains data about the target device and its malfunction, written in free text format by the maintenance worker M1.

[0041] Here, naming and setting a new cluster in the maintenance record database DB means, for example, that a new phenomenon, Phenomenon 4, has been added when three types of phenomena, Phenomenon 1, Phenomenon 2, and Phenomenon 3, are defined in the option estimation model of Fig. 4. It also means that model modifications are made in relation to Phenomenon 4, such as reviewing the probability between Phenomenon 4 and Cause 1 and Cause 2, adding a new cause, and further reviewing the probability between Phenomenon 4 and countermeasures, and adding a new countermeasure.

[0042] According to the fourth embodiment, in order to input "phenomenon," "cause," and "measure" that are not available in the options, if there are no options, free-text input is accepted, and when a certain amount of maintenance records is accumulated, the input free-text is clustered and a label name is assigned to each cluster. By presenting the label names to the maintenance manager (domain expert) M2, accepting corrections to the label names and clustering results, and adding the corrected label names as new options, it becomes possible to create a maintenance record even when there are no appropriate options, and to add appropriate options based on information from the maintenance site.

[0043] The maintenance record input support device according to the present invention in the above embodiment can be applied to situations where a maintenance worker supports on-site maintenance work, or where a call center operator supports customer service. [Explanation of symbols]

[0044] 1: Maintenance record input support device 1A: Model performance improvement section 1B: Cluster label setting section 10: Maintenance record creation method 11: Method of constructing and updating option estimation model 12: Choice estimation method 13: Keyword extraction method 14: Means of inputting dependencies 15: Clustering methods 16: Label name estimation method 17: Clustering label name correction method UI1, UI2, UI3: User Interface

Claims

1. The system comprises a keyword extraction means for inputting a device status of a device to be maintained and extracting a keyword representing the device status, an option estimation means for estimating options for maintenance items using the keywords and presenting them to the outside, a maintenance record creation means for performing maintenance based on the options presented to the outside, creating a maintenance record from the maintenance results and the device status of the device to be maintained input by a maintenance technician, and storing the maintenance record in a maintenance record database, and an option estimation model construction and update means for generating an option estimation model by learning using the maintenance records, The option inference means uses the keyword to refer to the option inference model, infers options for maintenance items, and presents the options to an external device.

2. The maintenance record input support device according to claim 1, A maintenance record input support device, characterized in that the input of the device status of the maintenance target device is made from any of an error code generated by the target device, image information of the target device, and free-form description of the failure status information of the target device.

3. 3. The maintenance record input support device according to claim 1, 2. A maintenance record input support device, comprising: a maintenance item selection section for selecting an abnormality in the target device, the abnormality being detected, the cause of the abnormality, and a countermeasure for the abnormality.

4. A maintenance record input support device as claimed in claim 1, The option estimation model is a probabilistic model that shows the probability between each item of abnormality phenomenon, cause, and countermeasure in the target equipment, and the maintenance record input support device is characterized in that the probability is modified according to known inter-item dependency relationships.

5. A maintenance record input support device as claimed in claim 1, A maintenance record input support device characterized in that, when the options presented externally do not match the device status of the device to be maintained, the device status of the device to be maintained is input by free description, a new cluster is generated by clustering with existing maintenance records, and the option estimation model is reconstructed by learning including the maintenance records of the newly named cluster.

Images