Facility Maintenance Support System and Facility Maintenance Support Method

Abstract

To provide a facility maintenance support system capable of referring to and using knowledge of past maintenance activities to make a maintenance coping plan so that even an unskilled person can adequately cope with maintenance.SOLUTION: A facility maintenance support system according to an embodiment comprises a data relation part, a fault prediction part and a facility maintenance history search part. The data relation part relates facility monitor data showing activity states of facilities gathered by monitoring the facilities, to facility maintenance history data showing maintenance details of the facilities recorded when the facilities are maintained. The fault prediction part uses a prediction model generated through machine learning using the related data to predict maintenance details of facilities to be maintained from the facility monitor data at an arbitrary point of time. The facility maintenance history search part uses the related data to search for facility maintenance history data having relevancy to the facility monitor data to which the whole or part of the facility monitor data used when the fault prediction part predicts the maintenance details of the facilities matches.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] Embodiments of the present invention relate to a facility maintenance support system and a facility maintenance support method.

Background Art

[0002] There is a support technology for not only detecting abnormalities in facilities but also supporting maintenance corresponding to the abnormalities. For example, in an abnormality detection and diagnosis method that detects an abnormality or its sign in a facility and diagnoses the facility, while detecting an abnormality with a plurality of sensors and converting it into a keyword indicating the type of abnormality, a keyword is extracted from the maintenance history to generate a diagnosis model, and an abnormality detection and diagnosis method for diagnosing the measures to be taken has been proposed.

[0003] In this abnormality detection and diagnosis method, a keyword described in a maintenance history data group is extracted, and a correspondence relationship between an abnormality content keyword and a maintenance response content keyword is used as a diagnosis model. Based on abnormality detection targeting the output of a multi-dimensional sensor attached to the facility, an abnormality keyword is specified, and by obtaining a maintenance response content keyword using the diagnosis model, the user is shown the measures to be taken for the occurred abnormality.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, in the above-described abnormality detection and diagnosis method, the output data of the multi-dimensional sensor attached to the facility and the facility maintenance history data are not associated, and the diagnosis model is generated only from the maintenance history data. Therefore, there is a problem that diagnosis cannot be performed in the case of an abnormality not described in the maintenance history data.

[0006] Also, since what is obtained through the diagnostic model is keywords rather than the maintenance history data itself, it was not possible to formulate maintenance response policies while referring to the maintenance history documents, which are the raw voices of on-site maintenance staff.

[0007] One embodiment of the present invention provides a facility maintenance support system and a facility maintenance support method that can be used to formulate maintenance response policies with reference to the knowledge of past maintenance activities, enabling accurate maintenance responses even by non-experts.

Means for Solving the Problems

[0008] According to the embodiment, the facility maintenance support system includes a data association unit, a failure prediction unit, and a facility maintenance history search unit. The data association unit associates facility monitoring data indicating the operating status of the facility collected by monitoring the facility with facility maintenance history data indicating the maintenance details of the facility recorded during the implementation of facility maintenance, from among the facility monitoring data and the facility maintenance history data, for the period when a maintenance case of the facility occurred or a period retroactively from the occurrence of the maintenance case. The failure prediction unit predicts the maintenance details of the facility to be implemented at an arbitrary point in time using a prediction model generated by machine learning using the data associated by the data association unit, which predicts the maintenance details of the facility to be implemented with the facility monitoring data as input. The facility maintenance history search unit searches for facility maintenance history data having relevance to facility monitoring data that matches all or part of the facility monitoring data used when predicting the maintenance details of the facility by the failure prediction unit, using the data associated by the data association unit. The failure prediction unit outputs input items emphasized by the prediction model in predicting the maintenance content of the equipment. The equipment maintenance history search unit searches for equipment maintenance history data having relevance to equipment monitoring data in which all or part of the input items match. The equipment maintenance support system includes a display unit. The display unit preferentially displays, as high-priority cases, data in which the values of the input items emphasized in predicting the maintenance content of the equipment match among the one or more pieces of retrieved equipment maintenance history data.

Brief Description of the Drawings

[0009]

Figure 1

Figure 2A

Figure 2B

Figure 2C

Figure 2D

Figure 2E

Figure 2F

Figure 2G

Figure 2H

Figure 2I

Figure 3A

Figure 3B

Figure 3C

Figure 3D

Mode for Carrying Out the Invention

[0010] Hereinafter, embodiments will be described with reference to the drawings. FIG. 1 is a diagram showing a configuration example of the facility maintenance support system 1 according to the embodiment. FIG. 1 also shows external systems related to the operation and maintenance of facilities that are coordinated with the facility maintenance support system 1. Here, as external systems related to the operation and maintenance of facilities, for example, a facility management system 5, a facility monitoring system 6, a facility maintenance information system 7, and a facility environment information system 8 are assumed. The facility maintenance support system 1 and the external systems are information processing systems constructed by, for example, a computer and a storage device. The facility maintenance support system 1 and the external systems realize various data processing units by a processor of the computer executing various programs. The various data processing units may be realized as hardware such as an electric circuit.

[0011] The facility management system 5 is a system that manages facility ledger data 51 in which the type of each facility to be maintained, the installation location, the model of the devices that make up the facility, the start time of operation, etc. are registered for all facilities subject to maintenance management. An example of the facility ledger data 51 is shown in FIG. 2A.

[0012] As shown in FIG. 2A, here, it is assumed that the facility ledger data 51 includes a "facility name" field, a "direction" field, a "equipment name" field, a "equipment type" field, a "section" field, etc. Also, it is assumed that the values of the three fields of the "facility name" field, the "direction" field, and the "equipment name" field among these uniquely indicate the facility to be maintained. Also, it is assumed that the value of the "section" field uniquely indicates the surrounding environment where the facility to be maintained is installed.

[0013] The equipment monitoring system 6 constantly monitors one or more equipment to be monitored 3, acquires and records the operation records of the equipment, the sensor time series values of the operation status, the abnormal detection reports coming from the equipment, etc., and notifies the necessary information to the monitoring staff in the maintenance department. Although various forms of internal structures are assumed, here, as an example, an equipment monitoring system 6 is assumed that has an equipment status change data extraction unit 61 that receives status change report information from the equipment to be monitored 3, and equipment monitoring history data 62 that stores and manages the history of the extracted equipment status change data. An example of the equipment monitoring history data 62 is shown in FIG. 2B.

[0014] As shown in FIG. 2B, here, it is assumed that the equipment monitoring history data 62 includes fields such as "date and time", "facility name", "direction", "equipment name", "status", etc. The equipment monitoring history data 62 includes three fields that can uniquely identify the equipment to be conservatively managed.

[0015] The equipment maintenance information system 7 is a system that records information such as the individual information of the target equipment, the date and time of maintenance response, and the content of maintenance implementation as the maintenance response history performed by the maintenance worker 9 in the maintenance department 2 on the equipment, and can be referred to as needed. An example of the equipment maintenance history data 71 managed by the equipment maintenance information system 7 is shown in FIG. 2C.

[0016] As shown in FIG. 2C, here, it is assumed that the equipment maintenance history data 71 includes fields such as "facility name", "direction", "equipment name", "date and time", "fault location", "processing content", etc. The equipment maintenance history data 71 includes three fields that can uniquely identify the equipment to be conservatively managed.

[0017] The equipment environment information system 8 is a system that records data notified from the environment information collection equipment 4 which collects time-series data of meteorological information and other state fluctuations regarding the surrounding environment where the equipment is installed, and can be referred to as necessary. Fig. 2D shows an example of the equipment environment history data 81 managed by the equipment environment information system 8. Here, meteorological data is cited as an example of the equipment environment history data 81.

[0018] As shown in Fig. 2D, here, it is assumed that the equipment environment history data 81 includes a "date and time" field, a "section" field, a "temperature" field, a "weather" field, a "precipitation amount" field, and the like. The equipment environment history data 81 includes a "section" field that can uniquely indicate the surrounding environment where the equipment to be subject to maintenance management is installed.

[0019] The external system may exist separately, or may be a composite function system that comprehensively manages information including the equipment maintenance support system 1. However, the data recorded in the external system is related to each other by some data items (techniques for associating known data such as a key data item, a group of data items serving as a composite key, or an association of recording times allowing a certain time difference), like the "facility name" field, "direction" field, "equipment name" field, and "section" field exemplified in Figs. 2A, 2B, 2C, and 2D, and can be handled in an associated manner.

[0020] The equipment maintenance support system 1 receives status monitoring data of equipment and operating environment and past maintenance history data from some or all of a separately operating equipment management system 5, equipment monitoring system 6, equipment maintenance information system 7, and equipment environment information system 8, and predicts the failure location when the equipment monitoring system 6 detects an abnormality and presents the maintenance history in past similar situations to the maintenance department 2. Here, for past monitoring data, as shown in FIG. 1, it is referred to as equipment monitoring history data 62, equipment environment history data 81, etc., while for real-time monitoring data, it may be referred to as equipment monitoring data, equipment environment data, etc. (excluding "history"). Furthermore, equipment monitoring data and equipment environment data may be referred to as equipment status change data.

[0021] As a result, in the maintenance department 2, it is possible to predict the failure content before going to the failure site, confirm the maintenance content in past similar failures, formulate a response policy, select the equipment to bring to the failure site, and determine an efficient inspection and repair order. Or, even when the equipment monitoring system 6 has not detected a clear occurrence of an abnormality, predictions are periodically executed from the latest data of external systems, it is detected that the occurrence status of the status monitoring data is different from normal, and by presenting to the maintenance department examples where an abnormal state occurred after similar monitoring data content occurred, it is possible to give notice to the maintenance worker 9 even when the equipment monitoring system 6 has not explicitly caught the abnormality and utilize it for improving the efficiency of maintenance work.

[0022] Next, the configuration of the equipment maintenance support system 1 will be described. As shown in FIG. 1, the equipment maintenance support system 1 includes a data association unit 11, a prediction model learning unit 12, a failure prediction model 13, a failure prediction unit 14, a search condition generation unit 15, an equipment maintenance history search unit 16, a condition input / display unit 17, an association data management unit 18, and a prediction input data association unit 19.

[0023] The data association unit 11 associates the data obtained from the facility management system 5, the facility monitoring system 6, the facility maintenance information system 7, and the facility environment information system 8, and stores the associated data (associated data 18A) in the associated data management unit 18. An example of the associated data 18A is shown in FIG. 2E.

[0024] As shown in FIG. 2E, here, it is assumed that the associated data 18A includes an "ID" field, a "key item" field, a "ledger part" field, a "monitoring part" field, an "environment part" field, a "maintenance history part" field, and the like.

[0025] The "key item" field includes a "facility name" field, a "direction" field, a "equipment name" field, and a "date and time" field. The "ledger part" field includes a "equipment type" field. The "monitoring part" field includes a "status" field. The "environment part" field includes a "section" field, a "temperature" field, a "weather" field, and a "precipitation amount" field. The "maintenance history part" field includes a "failure location" field and a "processing content" field.

[0026] Association is performed by associating, for each case of the facility maintenance history data 71, the facility master data 51 such as the installation year and model of the target facility where the maintenance occurred, the facility monitoring history data 62 during the time point when the maintenance occurred or a period retroactively from the time point when the maintenance occurred, and the environmental data around the facility (facility environment history data 81). As a specific example, it will be described using the facility master data 51 shown in FIG. 2A, the facility monitoring history data 62 shown in FIG. 2B, the facility maintenance history data 71 shown in FIG. 2C, the facility environment history data 81 shown in FIG. 2D, and the association data 18A shown in FIG. 2E. The data in FIGS. 2A, 2B, 2C, and 2D can be associated with each other using some of the key items in FIG. 2E as keys. As a result, as shown in FIG. 2E, for the maintenance response result, it is possible to associate and confirm in what monitoring state the maintenance was required, where the target facility was installed, and what type of device it was. In this example, it is an example of association between data that exactly matches in date and time. However, for example, one or more monitoring events or sensor data values reported from a certain time before the occurrence date and time recorded in the maintenance history to the occurrence date and time may be associated.

[0027] Also, when there are facility monitoring history data 62 and facility environment history data 81 even for the date and time or facilities where there are no maintenance records in the facility maintenance history data 71, association data 18A with the values of the non-existing data item parts set as NULL values or the like may be created. In FIG. 2E, examples of association data where there are no values in the data items of the facility maintenance history data 71 are shown for ID109 and ID110.

[0028] In the case of the association data 18A shown in FIG. 2E, the facility inventory data 51 is associated with other data based on the values of the "facility name" field, "direction" field, and "equipment name" field, and the value of the "equipment type" field is stored in the "inventory part" field of the association data 18A. The equipment monitoring history data 62 is associated with other data based on the values of the "date and time" field, "facility name" field, "direction" field, and "equipment name" field, and the value of the "status" field is stored in the "monitoring part" field of the association data 18A. The equipment maintenance history data 71 is associated with other data based on the values of the "facility name" field, "direction" field, "equipment name" field, and "date and time" field, and the values of the "fault location" field and "processing details" field are stored in the "maintenance history part" field of the association data 18A. The equipment environment history data 81 is associated with other data based on the values of the "date and time" field and "section" field, and the values of the "section" field, "temperature" field, "weather" field, and "precipitation amount" field are stored in the "maintenance history part" field of the association data 18A.

[0029] Note that the "key item" field of the association data 18A does not include the "section" field used for the association of the equipment environment history data 81, but the value of the "section" field is derived from the "section" field of the facility inventory data 51 that is associated with the equipment monitoring history data 62 and the equipment maintenance history data 71 based on the values of the "facility name" field, "direction" field, and "equipment name" field.

[0030] The prediction model learning unit 12 performs machine learning using the association data 18A stored in the association data management unit 18. It takes as input observable information that changes moment by moment, such as equipment monitoring data and equipment environment data, and outputs, as predicted values, information necessary for maintenance response actions, such as the failure location, failure time, and equipment required for maintenance, estimated therefrom. It generates a prediction model (failure prediction model 13). For example, in the case of Fig. 2E, a machine learning model is generated that estimates the failure location in the maintenance history part from all or part of the data items in the key item, ledger part, monitoring part, and environment part, either as they are or as processed multi-dimensional features. Regarding the learning method of the prediction model, many known machine learning techniques have been devised, and in the present invention, any known technique is used for learning the prediction model. Note that it is assumed that the prediction model learning unit re-learns the prediction model using the latest association data 18A in the association data management unit 18 at an arbitrary timing according to an arbitrarily determined operation policy. The arbitrary timing may be, for example, a timing at regular intervals such as every year, or a timing when the accuracy of the prediction result decreases as the trend of the data content changes.

[0031] The failure prediction model 13 refers to the failure prediction model generated by the prediction model learning unit 12. Although it is named "failure prediction model", the content to be predicted is not limited to the content of failures. By changing the machine learning method performed by the prediction model learning unit 12, it also includes cases where it is a model that predicts a state where the operation is unstable although no failure has occurred as an abnormal state, or an abnormal value, risk value, etc. Also, the failure prediction model 13 may hold a plurality of models with different learning conditions. For example, in Fig. 2E, data regarding equipment named "TN ventilation" and equipment named "ETC" is recorded as the equipment name. A prediction model for the TN ventilation equipment may be generated from the data regarding the TN ventilation equipment, and a prediction model for the ETC equipment may be generated from the data regarding the ETC equipment, and two prediction models may be prepared.

[0032] The prediction input data association unit 19 generates data that serves as input when the failure prediction unit 14 executes a prediction. Although the processing content is similar to that of the data association unit 11, since the prediction is to be made in the situation before maintenance is performed, data corresponding to the equipment maintenance history data 71 is not included, and data obtained from the equipment management system 5, the equipment monitoring system 6, and the equipment environment information system 8 is associated. Fig. 2F shows an output example of the prediction input data association unit 19.

[0033] In the example shown in Fig. 2F, for the single alarm of "unable to open or close" detected by the equipment monitoring system 6 at "2021 / 4 / 1 1:11", an association with the data of the equipment management system 5 and the equipment environment information system 8 is performed. However, when multiple monitoring alarms occur at the same time or within a certain period of time, they may be aggregated as a combined alarm and associated.

[0034] The failure prediction unit 14 uses the failure prediction model 13 to execute prediction processing using the input data generated by the prediction input data association unit 19 each time equipment monitoring data is added in the equipment monitoring system 6, or each time data with a specific keyword is added in the equipment monitoring system 6, or periodically, and outputs the prediction result and a list of the data items and the degrees of importance (hereinafter referred to as the "list of important feature quantities") that were emphasized when deriving the prediction result. Fig. 2G shows an example of the output data of the failure prediction unit 14.

[0035] In the example shown in Fig. 2G, "opening and closing part" is output as the prediction result, and a value "0.9" indicating the confidence level for the prediction result in the range of 0 to 1 is output, and as the list of important feature quantities, the names of the top three feature quantities with strong influence degrees and the influence degree values for the determination of the prediction result are output.

[0036] For the generation of a list of important feature quantities, generally, the use of technologies such as "explanation of AI" is assumed, and for example, technologies such as SHAP, which is a known technology, are used. When using SHAP, it is possible to output the feature quantities and influence degree values that the entire prediction model emphasizes. In addition, for each prediction result for one input, it is also possible to output the important feature quantities and influence degree values according to the input content. Either can be used. By utilizing the above technologies to generate a list of important items, it is possible to dynamically change the data items that should be important feature quantities in reflection of the machine learning results without declaratively defining as rules the data items that experts should emphasize in prediction. More appropriate case search can be expected in the equipment maintenance history search unit 16 described later.

[0037] Regarding the prediction results and the data items emphasized, a plurality of results may be output. For example, when using a machine learning algorithm such as a decision tree or SVM that can calculate the confidence level for a predicted value, a plurality of prediction results can be output in descending order of confidence level.

[0038] The search condition generation unit 15 uses the prediction results and the list of data items emphasized at the time of prediction output from the failure prediction unit 14 to generate search conditions necessary for executing a search in the equipment maintenance history search unit 16. For example, when the output of the failure prediction unit 14 is the content shown in FIG. 2G, instead of setting all input data items at the time of prediction by the failure prediction unit 14 as matching conditions, a priority condition setting is generated such that the matching priority of the search conditions is in the order of "state", "weather", and "precipitation amount". In addition, for each data item with a high matching priority, the value of the influence degree may be interpreted as the importance degree, and the importance degree value may be converted to set a weight value. The method of converting the importance degree into a weight is not limited. For example, a search condition is generated using the importance degree values themselves assigned to the three items output as important feature quantities in the example of FIG. 2G as weights, and search means such as calculating a comprehensive matching degree is used, where the search result with a higher total weight value of the data items with matching values is ranked higher. In this case, for data items that take continuous values such as "precipitation amount", even if they do not match exactly, a process such as reducing the matching degree according to the difference in values may be added to calculate the comprehensive matching degree.

[0039] Note that as priority conditions, it is also possible to set mandatory conditions that are higher than the first place. For example, in addition to the output of the failure prediction unit 14 shown in FIG. 2G, the data of FIG. 2F used as input may also be output, and it is conceivable to make the equipment name a mandatory condition. In this case, the output of the equipment maintenance history search unit 16 will only be the history related to the ETC equipment, and it is possible to prevent the search for the history of unrelated equipment.

[0040] In addition to the above conditions, the presence or absence of agreement in the prediction results may be added as a priority condition. For example, in this embodiment, the prediction result is the predicted value of the failure location. However, taking the prediction result as the first priority condition, a high overall agreement degree may be given to the association data 18A in which the "opening / closing part", which is the predicted value in FIG. 2G, is the failure location.

[0041] The equipment maintenance history search unit 16 performs a search of the association data management unit 18 using the conditions created by the search condition generation unit 15. The search execution must perform a search that at least matches the values of the input data of the failure prediction unit 14 for the data items that are mandatory conditions in the priority condition setting generated by the search condition generation unit 15. For the list of cases that match the mandatory conditions, which conditions are met for the conditions other than the mandatory conditions are output as the list of priority condition matching items. FIG. 2H shows an example of the output of the equipment maintenance history search unit 16.

[0042] FIG. 2H shows an example in which, with the equipment name as a mandatory condition, the value obtained by multiplying the importance value by the presence or absence of agreement for the priority conditions is taken as the agreement degree, and the total agreement degree of the agreement degrees of each important feature amount is taken as the overall agreement degree.

[0043] The condition input / display unit 17 outputs on the screen the prediction result output by the failure prediction unit 14 and the search result output by the equipment maintenance history search unit 16. In addition, the person in charge of the maintenance department 2 performs condition input for adjusting the display content, such as changing the display order or changing the priority order of the search conditions. When displaying the equipment maintenance history search results, they are listed and displayed in descending order of the overall matching degree output by the equipment maintenance history search unit 16. When displaying the listed results, the content of the list of priority condition matching items may be displayed together to show for which data items each of the listed equipment maintenance cases is prioritized and listed. Furthermore, it is possible to select a specific priority condition from the list of priority condition matching items and sort them in descending order of the matching degree of the said priority condition. Fig. 2I shows an example of the display screen of the condition input / display unit 17.

[0044] In the example shown in Fig. 2I, using the content of Fig. 2F, the observation information at the time of prediction execution is displayed (a1). Also, using the content of Fig. 2G, the content of the prediction result is displayed (a2).

[0045] Also, for condition input, a pull-down menu (a3) for selecting a sort key is displayed. With this pull-down menu, a sort key can be selected from each column of the output of the equipment maintenance history search unit 16 shown in Fig. 2H. The output of the equipment maintenance history search unit 16 displayed on the display screen changes the display order according to the selection of the sort key. In the example of Fig. 2I, only one sort key is selected, but it is also possible to select multiple sort keys, or data items other than the feature quantities displayed in the prioritized items can be used as mandatory items and added to the mandatory conditions in the equipment maintenance history search unit 16 to execute the search.

[0046] So far, the person in charge of the maintenance department 2 has judged the situation from the display content of the equipment monitoring system 6, recalled past similar maintenance cases from their memory to plan the setup for failure maintenance, prepared the equipment required for the planned response, and then gone to the site.

[0047] In the facility maintenance support system 1 of the present embodiment, reference information for formulating a setup for failure maintenance response is presented by the operations of the respective parts described above. Thus, it becomes possible to formulate an efficient maintenance work setup by checking the reference information without relying on memory, and even a person in charge with little experience can make appropriate judgments.

[0048] Figure 3A is a flowchart showing the processing flow of the data association unit 11.

[0049] The data association unit 11 inputs the facility monitoring history data 62 from the facility monitoring system 6 (S101). The data association unit 11 inputs the facility maintenance history data 71 from the facility maintenance information system 7 (S102). The data association unit 11 inputs the facility ledger data 51 from the facility management system 5 (S103). Further, the data association unit 11 inputs the facility environment history data 81 from the facility environment information system 8 (S104). Note that the input of various data is not limited to being performed in the order shown in Figure 3A. The input of various data may be performed in parallel.

[0050] The data association unit 11 extracts key item values for identifying individual facilities from the various input data (S105). The data association unit 11 associates the various data (facility monitoring history data 62, facility maintenance history data 71, facility environment history data 81, facility ledger data 51) according to the key items (S106). The data association unit 11 outputs the associated data 18A in which the various data are associated (S107).

[0051] Figure 3B is a flowchart showing the processing flow of the prediction model learning unit 12.

[0052] The prediction model learning unit 12 inputs the association data 18A managed by the association data management unit 18 (S201). The prediction model learning unit 12 performs machine learning for creating the failure prediction model 13 using the input association data 18A as learning data (S202). The prediction model learning unit 12 outputs the failure prediction model 13 created by the machine learning (S203).

[0053] Figure 3C is a flowchart showing the processing flow of the prediction input data association unit 19 and the failure prediction unit 14.

[0054] The prediction input data association unit 19 inputs equipment status change data (equipment monitoring data) from the equipment monitoring system 6 (S301). The prediction input data association unit 19 inputs equipment monitoring history data 62 (for example, from a certain period back) from the equipment monitoring system 6 according to the type of the failure prediction model 13 (S302).

[0055] The prediction input data association unit 19 extracts key item values for identifying the equipment individual from the input equipment monitoring history data 62 (S303). The prediction input data association unit 19 inputs equipment ledger data 51 corresponding to the key items from the equipment management system 5 (S304). Also, the prediction input data association unit 19 inputs equipment environment history data 81 corresponding to the key items from the equipment environment information system 8 (S305).

[0056] The prediction input data association unit 19 associates various data (equipment monitoring history data 62, equipment environment history data 81, equipment ledger data 51) according to the key items (S306). The failure prediction unit 14 executes failure prediction using the associated various data (S307). The failure prediction unit 14 outputs the prediction result and the importance degree by input item (S308).

[0057] Figure 3D is a flowchart showing the processing flow of the search condition generation unit 15 and the equipment maintenance history search unit 16.

[0058] The search condition generation unit 15 inputs the display target conditions from the user (of the maintenance department 2) via the condition input / display unit 17 (S401). Note that the input of the display target conditions is not mandatory (optional). If there is no input of the display target conditions, the default display target conditions are applied.

[0059] The search condition generation unit 15 inputs the prediction results output from the failure prediction unit 14 and the importance levels by input item (S402). The search condition generation unit 15 generates search conditions for the association data 18A managed by the association data management unit 18 based on the input display target conditions, prediction results, and importance levels by input item (S403).

[0060] The equipment maintenance history search unit 16 executes a search for the association data 18A using the search conditions generated by the search condition generation unit 15 (S404). The equipment maintenance history search unit 16 sorts the search results of the association data 18A based on the importance levels by input item (S405). If there are no display target conditions in S401, this process is skipped (optional). The equipment maintenance history search unit 16 outputs the search results of the association data 18A, which are ordered by the overall matching degree as shown in, for example, FIG. 2H (S406). The search results of the association data 18A output by the equipment maintenance history search unit 16 are presented to the user via the condition input / display unit 17.

[0061] As described above, the facility maintenance support system 1 of the present embodiment generates a failure prediction model 13 that predicts a part that requires maintenance with the facility monitoring history data 62 as an input, from the associated data 18A that collects and accumulates by associating the facility monitoring data from the time when an abnormality occurs (a state where maintenance is required) or a certain period before the occurrence of the abnormality until the occurrence of the abnormality, and the maintenance history data at the time of the occurrence of the abnormality. When new data is added to the facility monitoring history data 62, it predicts and presents a part that requires maintenance, or even if there is no new data added to the facility monitoring data 62, it predicts and presents the occurrence of an abnormality from the occurrence status of the facility monitoring data 62 that is different from the normal operation state at regular intervals. In addition, by searching for the maintenance history in which an abnormality occurred in a similar situation from the facility maintenance history data 71 in which the maintenance of the part was performed and presenting it to the maintenance worker 9, it can be used as a reference for formulating a maintenance response policy with reference to the knowledge of past maintenance activities, enabling accurate maintenance responses even for non-experts.

[0062] Also, when searching for and presenting the maintenance history, the facility maintenance support system 1 of the present embodiment ranks and displays the search results based on the similarity of the values of the feature quantities emphasized by the failure prediction model 13, so as to preferentially present maintenance cases in which the data items emphasized during prediction are similar among the multiple data items included in the facility monitoring history data 62 of each facility. As a result, even when there are a large number of search results, it can be used as a reference for formulating a maintenance response policy by checking in order from the maintenance history that occurred in a more similar situation.

[0063] Furthermore, the facility maintenance support system 1 of the present embodiment searches for past monitoring data in which the feature quantities emphasized by the failure prediction model 13 are similar during inference execution, and presents the probability of maintenance occurring in the past in a similar situation, thereby determining the degree of relevance between the situation indicated by the observed monitoring data and the maintenance target and using it as a reference for formulating a maintenance response policy.

[0064] Note that the above failure prediction model 13 is not limited to a model that predicts parts requiring maintenance. The same effect can be obtained in the case of a model that estimates the period until a failure occurs, a model that estimates whether the target facility is in an abnormal operating state compared to normal, and the like.

[0065] Although some embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the gist of the invention. These embodiments and their modifications are included in the scope and gist of the invention, and are included in the invention described in the claims and its equivalent scope.

Explanation of Reference Numerals

[0066] 1... Facility maintenance support system, 5... Facility management system, 6... Facility monitoring system, 7... Facility maintenance information system, 8... Facility environment information system, 11... Data association unit, 12... Prediction model learning unit, 13... Failure prediction model, 14... Failure prediction unit, 15... Search condition generation unit, 16... Facility maintenance history search unit, 17... Condition input / display unit, 18... Association data management unit, 18A... Association data, 19... Prediction input data association unit, 51... Facility ledger data, 61... Facility state change data extraction unit, 62... Facility monitoring history data, 71... Facility maintenance history data, 81... Facility environment history data.

Claims

1. A data association unit that associates equipment monitoring data indicating the operating status of the equipment collected by monitoring the equipment and equipment maintenance history data indicating the maintenance details of the equipment recorded when the equipment is maintained, from among the equipment monitoring data during the period when a maintenance case of the equipment occurs or a period retroactively from the occurrence of the maintenance case, and the equipment maintenance history data of the maintenance case; A failure prediction unit that predicts the maintenance details of the equipment to be performed from the equipment monitoring data at an arbitrary point in time, using a prediction model that predicts the maintenance details of the equipment to be performed with the equipment monitoring data as an input, generated by machine learning using the data associated by the data association unit; An equipment maintenance history search unit that searches for equipment maintenance history data having relevance to the equipment monitoring data that all or part of the equipment monitoring data used when predicting the maintenance details of the equipment by the failure prediction unit matches, using the data associated by the data association unit; Comprising: The failure prediction unit outputs input items emphasized by the prediction model in predicting the maintenance details of the equipment; The equipment maintenance history search unit searches for equipment maintenance history data having relevance to the equipment monitoring data that matches all or part of the input items; A display unit that preferentially displays, as high-priority cases, data in which the values of the input items emphasized in predicting the maintenance details of the equipment among the one or more pieces of equipment maintenance history data retrieved match; An equipment maintenance support system.

2. The failure prediction unit outputs the priority order of the input items; The display unit adds information indicating which of the input items the data matches to each of the one or more pieces of equipment maintenance history data retrieved, and displays the one or more pieces of equipment maintenance history data retrieved in the order determined based on the information; The equipment maintenance support system according to Claim 1.

3. The prediction model predicts at least one of a part of the equipment that requires maintenance, a period until a failure occurs in the equipment, and an abnormal operating state of the equipment. The equipment maintenance support system according to Claim 1 or 2.

4. An equipment maintenance support method executed by a computer, From among the equipment monitoring data indicating the operating status of the equipment monitored and collected, and the equipment maintenance history data indicating the maintenance details of the equipment recorded when the maintenance of the equipment is carried out, associate the equipment monitoring data during the period when the maintenance case of the equipment occurs or a period retroactively from the occurrence of the maintenance case, and the equipment maintenance history data of the maintenance case. Using a prediction model that predicts the maintenance details of the equipment to be carried out based on the equipment monitoring data at an arbitrary point in time, which is generated by machine learning using the associated data, predict the maintenance details of the equipment to be carried out from the equipment monitoring data at any point in time. Using the associated data, search for the equipment maintenance history data having relevance to the equipment monitoring data in which all or part of the equipment monitoring data used at the time of predicting the maintenance details of the equipment matches. The predicting includes outputting input items emphasized by the prediction model in predicting the maintenance details of the equipment. The searching includes searching for the equipment maintenance history data having relevance to the equipment monitoring data in which all or part of the input items match. Among the one or more pieces of the searched equipment maintenance history data, preferentially display as high-priority cases the data in which the values of the input items emphasized in predicting the maintenance details of the equipment match. Equipment maintenance support method.

Images