Maintenance assistance system, maintenance assistance method, and maintenance assistance program

Abstract

A maintenance assistance system (500) assists maintenance work performed by a maintenance worker of a device (30). An element unit estimation model acquisition unit (107) sets an element unit estimation model (108) for estimating the content of the maintenance work to an initial value of a maintenance work estimation model (105) for estimating the content of the maintenance work of the device (30), for each unit of device configuration elements. A maintenance work estimation unit (109) applies a maintenance work estimation model (105) to the abnormal feature amount, and estimates the maintenance work content presented to the maintenance worker. The maintenance job learning unit (104) repeatedly evaluates whether or not the maintenance job content is valid for the elimination of the abnormality, learns the maintenance job content for which the feature amount of the abnormality is valid, and updates the maintenance job estimation model (105).

Description

Technical Field

[0001] The present disclosure relates to a maintenance support system, a maintenance support method, and a maintenance support program for supporting a maintenance worker of equipment in performing maintenance work. Background Art

[0002] Among existing equipment maintenance support systems, there is a system that determines the status category of a machine when an abnormality occurs, extracts identical or similar cases from a case database, and displays the maintenance work content in descending order based on the credibility score. In the case database, the past maintenance work content, credibility, and machine status category are associated.

[0003] In addition, there is also an equipment maintenance support system that infers the content of maintenance work, such as disclosed in Patent Document 1. The equipment maintenance support system disclosed in Patent Document 1 is a system that learns machine operation information, machine phenomenon information, and information on maintenance work content, and infers the content of maintenance work based on the learning results when new machine operation information or machine phenomenon information is obtained.

[0004] Prior art literature

[0005] Patent Literature

[0006] Patent Document 1: Japanese Patent Application Publication No. 2020-154662 Summary of the invention

[0007] Problems to be solved by the invention

[0008] The existing equipment maintenance support system takes a certain equipment as an object, implements learning for determining the content of maintenance work, uses the learning results to determine the content of maintenance work for new events occurring in the object equipment, and prompts the maintenance operator. However, in the existing equipment maintenance support system, there is no mechanism for applying the information learned for a specific equipment to equipment of other specifications. In other words, even if the information learned for equipment of exactly the same specifications can be directly applied, it is necessary to re-learn the information required to determine the content of maintenance work for equipment of different specifications.

[0009] However, even if an event occurs in equipment with different specifications, if the physical cause is the same, it is sometimes considered that the same maintenance work content should be prompted and information learning is performed.

[0010] For example, when a new device is developed by using part of a certain device, the used part is exactly the same mechanism, so the content of the maintenance work can be determined by using the learning result.

[0011] More specifically, if the analysis result of the current waveform of a motor used in a certain part shows a specific characteristic value, it is determined that the structural component has deteriorated, and it can be determined as the maintenance work content that the component needs to be replaced. As long as the same part is used, the learning result can be used across devices.

[0012] Thus, in the existing equipment maintenance assistance system, learning is also performed separately for each device regarding events that can be processed as similar events. Therefore, there is the following problem: whenever a device is developed, the resources of the computer used for learning are prepared, and the calculation takes a long time depending on the scale of the learning object or the ability of the computer, and relearning is required. Furthermore, in the existing equipment maintenance assistance system, regarding events that can be processed as similar events as described above, even if the maintenance work content is determined by learning in a certain device, the result cannot be used in other devices. Therefore, there is the following problem: before learning is completed separately in each device, the maintenance work content cannot be determined for similar events. In addition, in the existing equipment maintenance assistance system, when credibility is given to the maintenance work content to be prompted, the credibility is also calculated separately for the events that can be processed as similar events as described above. Therefore, there is the following problem: compared with the case where learning is performed in multiple devices, the overall parameters of the data used to calculate the credibility are reduced, and the reliability of the credibility becomes lower.

[0013] The present disclosure aims to reduce the cost involved in learning the contents of maintenance work in a plurality of devices having different specifications and to improve the reliability of the contents of maintenance work presented to maintenance workers.

[0014] Means for solving problems

[0015] The maintenance assistance system disclosed in the present invention assists the maintenance operator in performing maintenance operations on equipment, wherein the maintenance assistance system comprises: an element unit inference model acquisition unit, which acquires structural elements of the equipment, namely, equipment structural elements, acquires an element unit inference model for estimating the content of the maintenance operation in units of the equipment structural elements, and sets the acquired element unit inference model as the initial value of the maintenance operation inference model for estimating the content of the maintenance operation of the equipment; a maintenance operation inference unit, which applies the maintenance operation inference model to the feature quantity of the abnormality extracted from the sensor data in the equipment, and estimates the content of the maintenance operation to be prompted to the maintenance operator; and a maintenance operation learning unit, which repeatedly evaluates whether the suggested maintenance operation content is effective for eliminating the abnormality based on the evaluation of the effectiveness of the suggested maintenance operation content, learns the maintenance operation content that is effective for the feature quantity of the abnormality, and updates the maintenance operation inference model.

[0016] Effects of the Invention

[0017] In the maintenance support device of the present disclosure, it is possible to reduce the cost involved in learning the content of maintenance work and to improve the reliability of the content of maintenance work presented to the maintenance worker. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a diagram showing a hardware configuration example of the maintenance support system according to the first embodiment.

[0019] Figure 2 This is a diagram showing a functional configuration example of the maintenance support system according to the first embodiment.

[0020] Figure 3 This is a diagram showing an example of constructing a maintenance work estimation model by the maintenance work learning unit according to the first embodiment.

[0021] Figure 4 This is a diagram showing an example of an input format of a device configuration by the device configuration input unit according to the first embodiment.

[0022] Figure 5 This is a diagram showing another example of the input format of the device configuration by the device configuration input unit according to the first embodiment.

[0023] Figure 6 This is a diagram showing an example in which the element unit estimation model acquisition unit of the first embodiment acquires an element unit estimation model in units of plant configuration elements.

[0024] Figure 7 This is a diagram showing a configuration example of an element unit estimation model according to the first embodiment.

[0025] Figure 8 This is a diagram showing the determination of the content of the maintenance work by the maintenance work estimation unit according to the first embodiment.

[0026] Fig. 9 This is a diagram showing a detailed configuration example of the maintenance work recording unit according to the first embodiment.

[0027] Fig.10 This is a flowchart showing an operation example of the maintenance support system according to the first embodiment.

[0028] Fig.11 This is a diagram showing a hardware configuration example of a maintenance support system according to a modified example of the first embodiment.

[0029] Fig.12 This is a diagram showing a functional configuration example of a maintenance support system according to the second embodiment.

[0030] Fig.13 This is a diagram showing a functional configuration example of a maintenance support system according to a third embodiment.

[0031] Fig.14 This is a diagram showing a functional configuration example of a maintenance support system according to a fourth embodiment. DETAILED DESCRIPTION

[0032] Below, the present embodiment is described using the accompanying drawings. In each figure, the same reference numerals are marked for the same or corresponding parts. In the description of the embodiment, the description of the same or corresponding parts is appropriately omitted or simplified. The arrows in the figure mainly represent data flow or processing flow. In addition, in the following figures, the relationship between the sizes of the various structural components is sometimes different from the actual situation. In addition, in the description of the embodiment, directions or positions such as top, bottom, left, right, front, back, front, and back are sometimes shown. These notations are records for the convenience of explanation and do not limit the configuration, direction, and direction of the device, appliance, or component.

[0033] Implementation Method 1

[0034] ***Description of the structure***

[0035] Figure 1 1 is a diagram showing a hardware configuration example of the maintenance support system 500 according to the present embodiment.

[0036] The maintenance support system 500 supports maintenance work performed by a maintenance operator of the equipment 30. The maintenance support system 500 derives the maintenance work content of the equipment 30 based on the information of the equipment 30 and presents it to the operator terminal 20 which is the terminal of the maintenance operator. The maintenance support system 500 is also called an equipment maintenance support system.

[0037] The maintenance support system 500 includes a maintenance work derivation device 10 and an operator terminal 20 .

[0038] The maintenance work derivation device 10 is also called a maintenance work content derivation device. The operator terminal 20 is also called a maintenance operator operation terminal.

[0039] The equipment 30 is an equipment to be subjected to maintenance work. The maintenance work recording device 40 is a device that records performance data of the maintenance work content.

[0040] The maintenance work derivation device 10 derives the maintenance work content of the equipment based on the information of the equipment 30 and the performance data of the maintenance work content recorded in the maintenance work recording device 40 , and transmits the data to the operator terminal 20 .

[0041] The operator terminal 20 displays the maintenance work content received from the maintenance work derivation device 10 on the display device 24 and communicates it to the maintenance operator.

[0042] The maintenance work derivation device 10 includes a processor 11 , a memory 12 , a storage 13 , a display device 14 , an operation interface 15 , a device information acquisition interface 16 , a work content acquisition interface 17 , and a communication interface 18 .

[0043] The processor 11 uses, for example, the following information to execute calculations for deriving the content of maintenance work to be presented to the maintenance worker.

[0044] Programs and data stored in the memory 13

[0045] The device setting parameter values ​​and device control parameter values, device status parameter values, sensor data, and operation logs stored in the memory 31 of the device 30 acquired via the device information acquisition interface 16 .

[0046] ·Performance data of maintenance work contents acquired through the work content acquisition interface 17

[0047] The memory 12 stores temporary data used in the calculation of the processor 11 .

[0048] The memory 13 stores a program representing a process for deriving the contents of maintenance work and accompanying data.

[0049] The display device 14 displays information necessary for the user of the maintenance work derivation device 10 to set, execute, and end such operations.

[0050] The operation interface 15 provides an interface for users of the maintenance work derivation device 10 , such as a person in charge of system operations, to perform operations such as setting, execution, and termination.

[0051] The device information acquisition interface 16 acquires the values ​​of control parameters and sensor data in the memory 31 of the device 30 .

[0052] The work content acquisition interface 17 acquires the performance data of the maintenance work content recorded by the maintenance work recording device 40 .

[0053] The communication interface 18 communicates with the operator terminal 20 . Specifically, the communication interface 18 transmits data on the maintenance work content derived by the calculation of the processor 11 to the operator terminal 20 .

[0054] The operator terminal 20 includes a processor 21 , a memory 22 , a storage 23 , a display device 24 , an operation interface 25 , and a communication interface 26 .

[0055] The processor 21 executes calculations for generating display contents using the program stored in the memory 23 and the data on the contents of the maintenance work received from the maintenance work deriving device 10 .

[0056] The memory 22 stores temporary data used in the calculation of the processor 21 .

[0057] The memory 23 stores a program representing a process for displaying the contents of maintenance work and accompanying data.

[0058] The display device 24 displays information necessary for setting, executing, and ending such operations or display contents related to the contents of the maintenance work generated by the processor 21 .

[0059] The operation interface 25 provides a maintenance operator with an interface for performing operations such as setting, execution, and termination.

[0060] The communication interface 26 communicates with the maintenance work derivation device 10. Specifically, the communication interface 26 receives data on the maintenance work content derived by the maintenance work derivation device 10.

[0061] In the device 30, the memory 31 stores the device setting parameter values ​​and control parameter values, sensor data, and log data. Generally speaking, the device 30 uses a control machine such as a PLC to control the action, and the memory 31 is equivalent to the memory of the control machine. PLC is the abbreviation of Programmable Logic Controller.

[0062] The maintenance work recording device 40 records the contents of the maintenance work performed by the maintenance worker. Specifically, the following performance data of the maintenance work contents are recorded.

[0063] Recording of images from cameras installed in or near the equipment or from the maintenance worker’s head-mounted camera

[0064] Recording of the physical or mental state of maintenance workers using wearable terminals

[0065] · Recording of characters and sounds input by maintenance workers using mobile terminals

[0066] · Operation history in the HMI (Human Machine Interface) of the equipment control terminal or display history of the manual display terminal

[0067] Figure 2 1 is a diagram showing a functional configuration example of the maintenance support system 500 according to the present embodiment.

[0068] The sensor data acquisition unit 101 acquires sensor data from the device 30 .

[0069] The abnormality sign detection unit 102 analyzes the sensor data and detects a sign of abnormality.

[0070] The feature quantity extraction unit 103 analyzes the sensor data and calculates the feature quantity. The analysis method includes statistics such as average, variance, maximum value and minimum value, differential conversion, calculus, peak detection, frequency analysis such as Fourier transform, or autocorrelation. For example, the analog waveform shown in the sensor data can be subjected to Fourier transform, and the feature quantity can be expressed as a multi-dimensional vector using representative frequency components.

[0071] The maintenance work learning unit 104 repeatedly evaluates whether the maintenance work content presented to the maintenance worker is effective in eliminating the abnormality based on the work results of the maintenance work and the evaluation of the effectiveness of the work. Thus, the maintenance work learning unit 104 learns effective maintenance work content and generates a maintenance work inference model.

[0072] Specifically, the maintenance work learning unit 104 learns the maintenance work content to be presented for the extracted feature amount based on the evaluation based on the maintenance work result acquired by the maintenance work result acquisition unit 111 , and constructs the maintenance work estimation model 105 .

[0073] As a first method related to evaluation, if it is determined based on the maintenance work result that the maintenance work content derived by the maintenance work inference unit 109 is helpful to restore the equipment to a normal state, the maintenance work content is considered to be effective for the extracted feature quantity and a high evaluation is given. As a second method, if it is determined based on the maintenance work result that the maintenance work content recorded by the maintenance work recording unit 112 is helpful to restore the equipment to a normal state, the maintenance work content is considered to be effective for the extracted feature quantity and a high evaluation is given. By repeatedly performing this process, it is possible to construct a maintenance work inference model 105 for determining appropriate maintenance work content for the feature quantity extracted from the sensor data.

[0074] Figure 3 1 is a diagram showing an example of constructing a maintenance work estimation model 105 by the maintenance work learning unit 104 according to the present embodiment.

[0075] exist Figure 3 , an example is shown in which three feature quantities a, b, and c are extracted from certain sensor data to construct a maintenance work estimation model 105 for determining the content of the maintenance work.

[0076] As described above, the maintenance work estimation model 105 is a model for determining appropriate maintenance work content for the feature amount extracted from the sensor data.

[0077] Figure 4 1 is a diagram showing an example of a format for inputting a device configuration by the device configuration input unit 106 according to the present embodiment.

[0078] Figure 5FIG. 1 is a diagram showing another example of the input format of the device configuration by the device configuration input unit 106 according to the present embodiment.

[0079] The equipment configuration input unit 106 receives input of equipment configuration from a user of the maintenance support system 500. The user of the maintenance support system 500 is a person such as a person in charge of system operations.

[0080] As input form, for example, use Figure 4 As shown, a graphical description using an engineering tool is provided. Alternatively, as an input form, for example Figure 5 The description shown is based on a structured formal language such as XML.

[0081] Figure 6 1 is a diagram showing an example in which the element unit estimation model acquisition unit 107 of the present embodiment acquires the element unit estimation model 108 in units of plant configuration elements.

[0082] The element unit estimation model 108 is stored in a storage unit included in the maintenance support system 500. The function of the storage unit included in the maintenance support system 500 is realized by the memory 12 or the storage 13, or by the memory 12 and the storage 13.

[0083] First, the element unit estimation model acquisition unit 107 acquires the equipment configuration of the equipment to be maintained from the system user via the equipment configuration input unit 106. Then, the element unit estimation model acquisition unit 107 acquires the element unit estimation model 108 in units of equipment configuration elements based on the elements of the equipment configuration.

[0084] Specifically, the element unit estimation model acquisition unit 107 identifies a reusable element unit estimation model 108 in units of equipment structural elements in the construction of the maintenance work estimation model 105, and acquires the element unit estimation model 108 for the same equipment structural element. The element unit estimation model 108 is a maintenance work estimation model learned in units of equipment structural elements.

[0085] Then, the element unit estimation model acquisition unit 107 stores the acquired element unit estimation model 108 of the equipment configuration element unit as an initial model in the maintenance work estimation model 105. Figure 6 As shown, the equipment structural element unit is composed of equipment, parts of the equipment and components of the parts.

[0086] exist Figure 6 In the example below, the following status is indicated.

[0087] Regarding the type of device, there are device 1 type, device 1 ' type, device 2 type, ... For example, there are sometimes a plurality of individuals (products) of device 1 type.

[0088] Regarding the type of part, there are part A type, part B type, part C type, ... For example, there may be a plurality of individuals (products) of part A type.

[0089] Regarding the types of components, there are component a type, component b type, component c type, ... For example, there may be a plurality of individuals (products) of component a type.

[0090] Here, an ID for identifying the type or type of a device is referred to as a device category ID. An ID for identifying the type or type of a part is referred to as a part category ID. In addition, an ID for identifying the type or type of a component is referred to as a component category ID. ID is an abbreviation of Identifier.

[0091] For example, the element unit inference model acquisition unit 107 acquires a maintenance work inference model in units of equipment, parts, or parts for elements with the same equipment category ID, part category ID, or component category ID, and sets it as an element of the maintenance work inference model 105. Alternatively, instead of using the category IDs of equipment, parts, and parts, the user may directly specify the model to be used in the element unit inference model 108 for a specific equipment, part, or part. In addition, it may be configured so that when the element unit inference model acquisition unit 107 acquires an element unit inference model of a certain equipment structure element unit according to the association of equipment, parts, and parts in the element unit inference model 108, the associated element unit inference model is also acquired.

[0092] use Figure 6 , and explain it in more detail.

[0093] For example, in the element unit inference model 108, the device unit inference model 701, the part unit inference model 702, and the component unit inference model 703 are pre-stored for the device structure "device 1" of the learned device. Furthermore, if "part A+part B" is included in the newly learned device structure, the element unit inference model acquisition unit 107 acquires the "part unit inference model 702" of "part A+part B" from the "element unit inference model 108". Similarly, if "component a" is included in the newly learned device structure, the element unit inference model acquisition unit 107 acquires the "component unit inference model 703" of "component a" from the "element unit inference model 108".

[0094] Figure 7 It is a diagram showing a configuration example of the element unit estimation model 108 according to the present embodiment.

[0095] like Figure 7 As shown, the element unit estimation model 108 stores the learned maintenance work estimation model in equipment configuration element units (equipment, part, component).

[0096] The equipment unit estimation model 701, the location unit estimation model 702, and the component unit estimation model 703 are all maintenance work estimation models 105, and are the following models.

[0097] The equipment unit inference model 701 is a model that determines the content of maintenance work using sensor data of the entire equipment across multiple locations.

[0098] The location unit inference model 702 is a model that determines the content of maintenance work using the associated sensor data under the location

[0099] The component-based inference model 703 is a model that determines the content of maintenance work using only the sensor data associated with a single component.

[0100] Furthermore, the equipment unit estimation model 701, the part unit estimation model 702, and the component unit estimation model 703 are associated with each other according to the equipment structure. That is, according to the equipment structure of one or more parts constituting a certain equipment and one or more parts constituting each part, a specific equipment unit estimation model 701 is associated with one or more specific part unit estimation models 702, and a specific part unit estimation model 702 is associated with one or more specific component unit estimation models 703.

[0101] In addition, the unit of equipment structure element does not have to be Figure 7 The three hierarchical structures shown may have a hierarchical structure in which a plurality of devices constitute one device, or a plurality of parts constitute one part.

[0102] Figure 8 1 and 2 are diagrams showing the determination of the content of maintenance work by the maintenance work estimation unit 109 according to the present embodiment.

[0103] The maintenance work inference unit 109 applies the maintenance work inference model 105 to the feature quantity of the abnormality extracted from the sensor data to infer the content of the maintenance work. When a new abnormality sign is detected, the maintenance work inference unit 109 applies the maintenance work inference model 105 to the feature quantity extracted by the feature quantity extraction unit 103 based on the sensor data obtained from the equipment to determine the content of the maintenance work.

[0104] For example, Figure 8As shown, the maintenance work content is determined according to the distance such as the Euclidean distance or the Mahalanobis distance, based on the maintenance work content associated with the feature quantity existing near the feature quantity of the sensor data of the equipment when the new abnormality sign is detected. For example, the maintenance work content of the feature quantity closest to the equipment is selected, or the maintenance work content is selected by giving priority to the feature quantity with the closest distance in order of multiple feature quantities, or the maintenance work content with credibility is calculated by giving weights corresponding to the distance to multiple feature quantities located within a certain distance.

[0105] The maintenance work prompting unit 110 prompts the maintenance work content determined by the maintenance work inference unit 109 to the maintenance operator. When there are multiple maintenance work contents to be prompted, the maintenance work prompting unit 110 prompts so that the maintenance operator can select the maintenance work content to be viewed. At this time, if the maintenance work content is given priority or credibility, the maintenance work prompting unit 110 prompts it at the same time when the maintenance work content is selected. Alternatively, the maintenance work prompting unit 110 may prompt the maintenance work content in order from the maintenance work content with high priority or credibility, without the maintenance operator selecting the maintenance work content to be viewed.

[0106] The maintenance work result acquisition unit 111 acquires the result related to the elimination of abnormal signs through the maintenance work performed by the maintenance worker. As a first method related to the judgment of the result, when the characteristic quantity of the sensor data returns from the value indicating the abnormal sign to the same value as the normal state, it is judged that the abnormal sign has been eliminated. On the contrary, if the characteristic quantity of the sensor data is still the value indicating the abnormal sign, it is judged that the abnormal sign cannot be eliminated. As a second method, when the device status parameter value or the information of the action log obtained from the device changes from the information indicating the abnormality to the information indicating the normal state, it is judged that the abnormal sign has been eliminated. On the contrary, when the device status parameter value or the information of the action log does not change from the information indicating the abnormality, it is judged that the abnormal sign cannot be eliminated.

[0107] Fig. 9 It is a diagram showing a detailed configuration example of the maintenance work recording unit 112 according to the present embodiment.

[0108] The maintenance work recording unit 112 records the content of the maintenance work performed by the maintenance worker. The maintenance work recording unit 112 includes a recording object selection and editing unit 901 , a sensor information acquisition unit 902 , a device operation history acquisition unit 903 , and a worker input acquisition unit 904 .

[0109] The recording object selection and editing unit 901 accepts selection of information determined to be useful by the maintenance operator after the maintenance work is completed, with respect to the information acquired by the sensor information acquisition unit 902, the equipment operation history acquisition unit 903, and the operator input acquisition unit 904. The recording object selection and editing unit 901 also accepts editing such as cutting out of the interval designation of the dynamic image or sound, and designating the range of the operation history or manual.

[0110] The sensor information acquisition unit 902 acquires sensor information such as images from a camera installed in or near the equipment or a head-mounted camera of a maintenance worker, and information on the physical state or mental state of a wearable terminal worn by the maintenance worker.

[0111] The device operation history acquisition unit 903 acquires a history such as a change history of device setting parameters or an operation history of the HMI or the like in the device control terminal.

[0112] The operator input acquisition unit 904 acquires input such as characters and voice input by a maintenance operator using a mobile terminal.

[0113] In addition, you can also target Figure 1 The functional configuration of the maintenance support system 500 of the present embodiment may be arbitrarily configured within a range without contradiction among the equipment 30, the maintenance work derivation device 10, the operator terminal 20, and the maintenance work recording device 40.

[0114] ***Description of the action***

[0115] Next, the operation of the maintenance support system 500 of this embodiment will be described. The operation procedure of the maintenance support system 500 corresponds to the maintenance support method. In addition, the program that realizes the operation of the maintenance support system 500 corresponds to the maintenance support program.

[0116] Fig.10 1 is a flowchart showing an operation example of the maintenance support system 500 according to the present embodiment.

[0117] First, before the device is operated, the process of step S101 is performed.

[0118] In step S101 , the element-based estimation model acquisition unit 107 acquires an available maintenance work estimation model for the equipment configuration element from the element-based estimation model 108 based on the equipment configuration information held by the equipment configuration input unit 106 , and sets it as the initial maintenance work estimation model 105 .

[0119] Specifically, the element unit estimation model acquisition unit 107 acquires the structural elements of the equipment, that is, the equipment structural elements, and acquires the element unit estimation model 108 for estimating the content of the maintenance work in units of the equipment structural elements. The element unit estimation model acquisition unit 107 sets the acquired element unit estimation model 108 as the initial value of the maintenance work estimation model 105 for estimating the content of the maintenance work of the equipment.

[0120] The equipment structural elements include, for example, the type of equipment, the type of parts constituting the equipment, and the type of components constituting the parts. In the storage unit, element unit estimation models 108 are stored respectively according to the type of equipment, the type of parts, and the type of components. The element unit estimation model acquisition unit 107 acquires the element unit estimation model 108 respectively according to the type of equipment, the type of parts, and the type of components in the equipment.

[0121] More specifically, it is as follows.

[0122] Before new learning is performed, i.e., before the equipment is operated, the element unit inference model acquisition unit 107 determines whether there is a match between the equipment structure of the newly operated equipment and the equipment structure of the previously learned equipment. The element unit inference model acquisition unit 107 determines whether there is a match based on the category ID of the equipment structure element such as equipment, part, or component.

[0123] If there is a matching element, the element unit inference model acquisition unit 107 acquires the element unit inference model 108 associated with the matching element and sets it as the initial maintenance work inference model 105. At this time, the element unit inference model 108 acquired by the element unit inference model acquisition unit 107 is reused as the initial maintenance work inference model 105.

[0124] If there is no matching element, the element-based estimation model acquisition unit 107 does not set the initial maintenance work estimation model 105 .

[0125] Next, after the device is operated, the processing after step S102 is performed.

[0126] In step S102 , the abnormality sign detection unit 102 analyzes the sensor data to detect abnormality.

[0127] In step S103 , the feature quantity extraction unit 103 extracts feature quantities from the sensor data.

[0128] In step S104, the maintenance work estimation unit 109 receives the feature amount, applies the maintenance work estimation model 105 to the feature amount, and specifies the content of the maintenance work.

[0129] In step S105, the maintenance operation prompting unit 110 receives the maintenance operation content and prompts the maintenance operator. At this time, the maintenance operation content with the highest priority or credibility can also be mechanically prompted. In addition, the maintenance operator can also accept the selection request for multiple maintenance operation contents assigned with priority or credibility to perform prompts.

[0130] In step S106 , the maintenance worker performs the maintenance work according to the maintenance work content presented by the maintenance work presentation unit 110 .

[0131] In step S107 , the maintenance work result acquisition unit 111 confirms that the abnormality sign has been eliminated.

[0132] If the abnormal sign can be eliminated, proceed to step S108.

[0133] If the abnormal sign cannot be eliminated, go to step S109.

[0134] If the abnormality sign can be eliminated, then in step S108 , the maintenance work learning unit 104 regards the maintenance work content presented with respect to the feature amount this time as effective and gives a high evaluation.

[0135] If the abnormality sign cannot be eliminated, then in step S109 , the maintenance work learning unit 104 regards the maintenance work content presented for the feature amount this time as invalid and sets a low evaluation.

[0136] Then, in step S110, if there is a maintenance work content with the next highest priority or reliability, the maintenance work presentation unit 110 mechanically selects the maintenance work content, or accepts a selection request from the maintenance worker to select a maintenance work content and presents it to the maintenance worker.

[0137] If there is no maintenance work content to be presented, then in step S111 , the maintenance work recording unit 112 associates the current maintenance work content with the current feature amount.

[0138] Furthermore, in step S112 , the maintenance work learning unit 104 updates the maintenance work estimation model 105 based on the results of steps S108 , S109 , and S111 .

[0139] As described above, in the maintenance support system 500 of the present embodiment, when there is a coincidence in the device structure with the device learned in the past, the element unit inference model is obtained from the "element unit inference model 108" for the element, and it is set as the initial model of the "maintenance work inference model 105". In addition, the maintenance support system 500 updates the "maintenance work inference model 105" based on actual performance during the operation of the device. In addition, when there is no coincidence in the device structure with the device learned in the past, there is no initial model of the "maintenance work inference model 105", and the maintenance support system 500 builds the "maintenance work inference model 105" from scratch based on actual performance during the operation of the device.

[0140] Here, a specific example will be used to describe how the generation of the maintenance work estimation model 105 is made more efficient by setting the element unit estimation model 108 acquired by the element unit estimation model acquisition unit 107 as the initial maintenance work estimation model 105 .

[0141] Here, the maintenance work content of "if the analysis result of the current waveform of the motor used in a certain part shows a specific characteristic value, it is determined that the structural component has deteriorated and the component is replaced" is taken as an example.

[0142] The maintenance work content such as "replace components for the current waveform of the motor" is determined (step S104), and a prompt is given to the maintenance worker (S105). If the maintenance worker who receives the prompt performs the maintenance work according to the prompt and succeeds, the credibility of the prompt content is improved (step S108). For example, if the process is repeated 10 times, the maintenance work content such as "replace components for the current waveform of the motor" becomes a reasonable prompt content accordingly. Assume that this process is repeated 10 times in a certain device A, and a maintenance work inference model with improved credibility of the prompt content is generated. By using the maintenance work inference model in device B of the same category ID, in device B, as a process from the 11th time onwards, the credibility of the prompt content can be cumulatively improved. If the process is further repeated 10 times in device B, a high-credibility maintenance work inference model with a total of 20 processes can be generated. In this way, when the maintenance work inference model is not used between devices, if the process has not been carried out 20 times in device A and 20 times in device B, a maintenance work inference model with the same high degree of credibility cannot be generated.

[0143] ***Description of the Effects of the Embodiment***

[0144] As described above, in the maintenance support system of the present embodiment, the following characteristic processes are performed.

[0145] (A) Before the equipment is put into operation, the element unit inference model acquisition unit acquires the equipment structural elements of the equipment to be maintained, and based on the element unit inference model, acquires the learned inference model with the same equipment structural element unit, and sets it as the initial maintenance work inference model.

[0146] (B) When an abnormality is detected by the abnormality sign detection unit, a maintenance work estimation model generated efficiently is applied to a feature amount of the abnormality to estimate the content of the maintenance work.

[0147] (C) The maintenance work learning unit repeatedly evaluates whether the maintenance work content is effective in eliminating the abnormality based on the maintenance work result obtained by the maintenance work result acquisition unit, learns effective maintenance work content, and generates a maintenance work inference model.

[0148] According to the maintenance support system of this embodiment, the process (A) to (C) above can reduce the cost involved in the generation of the model. In particular, the process (A) above can improve the credibility of the maintenance work content presented to the maintenance operator compared to the learning in a single device. Even if it is a single device, a high-credibility model can be generated if it takes a long time. However, if a high-credibility model can be used from the beginning, the success rate of the maintenance work is improved, and the cost of the maintenance work itself can be reduced.

[0149] As described above, in the maintenance support system of this embodiment, the maintenance work content is shown to the maintenance operator of the equipment, thereby making the maintenance work more efficient, and further, the information required to determine the maintenance work content can be shared among multiple equipments with different specifications. Therefore, according to the maintenance support system of this embodiment, the cost involved in learning the maintenance work content can be reduced, and the credibility of the maintenance work content shown to the maintenance operator can be improved.

[0150] In addition, the maintenance support system of this embodiment provides the maintenance operator with the maintenance work content to be performed. Therefore, the maintenance support system of this embodiment has the following effect: the time spent on maintenance work can be shortened, excluding the time the maintenance operator himself spends on researching the maintenance work content.

[0151] Furthermore, in the maintenance support system of the present embodiment, the evaluation of the effectiveness of the maintenance work content for the characteristic quantity is cumulatively updated according to the success or failure of the maintenance work. Therefore, according to the maintenance support system of the present embodiment, the following effect is achieved: as the system is continuously used, the reliability of the priority or credibility of the maintenance work content presented to the maintenance operator can be continuously improved.

[0152] Furthermore, in the maintenance support system of this embodiment, the element unit inference model can be stored in the structural unit of the equipment. Therefore, in the case of newly developing equipment, the learning results of the equipment that is the source of derivative development or the equipment that uses the same part / component can be reused. Therefore, according to the maintenance support system of this embodiment, the effect of reducing the computer resources and time spent on the construction of the maintenance work inference model is achieved.

[0153] Furthermore, the maintenance support system of this embodiment has the effect that even for abnormal signs that have not been learned in a newly developed device, the content of maintenance work can be determined using the learning results of a device that is a derivative development source or a device that uses the same part / component.

[0154] Furthermore, according to the maintenance support system of this embodiment, learning is performed based on more actual results than when learning is performed only in one newly developed device, thereby achieving an effect of improving the reliability of the priority or credibility of the maintenance work content presented to the maintenance operator.

[0155] ***Description of Hardware Configuration Example***

[0156] use Figure 1 The hardware configurations of the maintenance work derivation device 10 and the operator terminal 20 of the present embodiment will be described. The maintenance work derivation device 10 and the operator terminal 20 are referred to as devices of the maintenance support system.

[0157] Each device of the maintenance support system is a computer. Each device of the maintenance support system has a processor and other hardware such as a memory, a storage, various interfaces, and a display device. The processor is connected to other hardware via a signal line to control the other hardware.

[0158] For example, the functions of each device of the maintenance support system are realized by software.

[0159] The processor is a device that executes the maintenance support program. The maintenance support program is a program that realizes the functions of each device of the maintenance support system.

[0160] A processor is an IC that performs computational processing. Specific examples of processors are CPU, DSP, and GPU. IC is the abbreviation of Integrated Circuit. CPU is the abbreviation of Central Processing Unit. DSP is the abbreviation of Digital Signal Processor. GPU is the abbreviation of Graphics Processing Unit.

[0161] Memory is a storage device that temporarily stores data. Specific examples of memory are SRAM or DRAM. SRAM is the abbreviation of Static Random Access Memory. DRAM is the abbreviation of Dynamic Random Access Memory.

[0162] The memory is a storage device for storing data. A specific example of the memory is HDD. In addition, the memory may also be a mobile storage medium such as SD (registered trademark) memory card, CF, NAND flash memory, floppy disk, optical disk, high-density disk, Blu-ray (registered trademark) disk, DVD. In addition, HDD is the abbreviation of Hard Disk Drive. SD (registered trademark) is the abbreviation of Secure Digital. CF is the abbreviation of CompactFlash (registered trademark). DVD is the abbreviation of Digital Versatile Disk.

[0163] The input / output interface is an interface for connecting input / output devices. As a specific example, the input / output interface is a port of USB or HDMI (registered trademark). USB is the abbreviation of Universal Serial Bus. HDMI (registered trademark) is the abbreviation of High-Definition Multimedia Interface.

[0164] The communication interface is an interface for communicating with an external device. As a specific example, the communication interface is an Ethernet (registered trademark) port or a device for wireless communication.

[0165] The maintenance assistance program is executed in each device of the maintenance assistance system. The maintenance assistance program is read into the processor and executed by the processor. In the memory, not only the maintenance assistance program but also the OS is stored. OS is the abbreviation of Operating System. The processor executes the maintenance assistance program while executing the OS. The maintenance assistance program and the OS can also be stored in the memory. The maintenance assistance program and the OS stored in the memory are loaded into the memory and executed by the processor. In addition, part or all of the maintenance assistance program can also be embedded in the OS.

[0166] Each device of the maintenance support system may include a plurality of processors instead of the processor. These plurality of processors share the responsibility of executing the maintenance support program. Each processor is a device that executes the maintenance support program in the same manner as the processor.

[0167] Data, information, signal values, and variable values ​​utilized, processed, or output by the maintenance assistance program are stored in memory, storage, or registers or cache within a processor.

[0168] The "part" of each part of each device of the maintenance support system may be rewritten as "circuit", "process", "step", "processing" or "line". The maintenance support program causes the computer to execute each process after the "part" of each part of each device of the maintenance support system is rewritten as "processing". The "processing" of each process of each device of the maintenance support system may be rewritten as "program", "program product", "computer-readable storage medium storing a program" or "computer-readable recording medium recording a program". In addition, the maintenance support method is a method implemented by executing the maintenance support program by each device of the maintenance support system.

[0169] The maintenance support program may be provided by being stored in a computer-readable recording medium. In addition, the maintenance support program may be provided as a program product.

[0170] ***Other structures***

[0171] <Modification>

[0172] In the present embodiment, the functions of each part of each device of the maintenance support system are realized by software. As a modification, the functions of each part of each device of the maintenance support system may also be realized by hardware.

[0173] Specifically, each device of the maintenance support system includes electronic circuits 11 a and 21 a instead of the processors 11 and 21 .

[0174] Fig.11 1 is a diagram showing a hardware configuration example of a maintenance support system 500 according to a modification of the present embodiment.

[0175] The electronic circuit is a dedicated electronic circuit that realizes the functions of each part of each device of the maintenance support system. Specifically, the electronic circuit is a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, a logic IC, a GA, an ASIC or an FPGA. GA is the abbreviation of Gate Array. ASIC is the abbreviation of Application Specific Integrated Circuit. FPGA is the abbreviation of Field-Programmable Gate Array.

[0176] The functions of each part of each device of the maintenance support system may be realized by one electronic circuit or may be realized by distributing them across a plurality of electronic circuits.

[0177] As another variation, some functions of each part of each device of the maintenance support system may be implemented by electronic circuits, and the remaining functions may be implemented by software. In addition, some or all functions of each part of each device of the maintenance support system may also be implemented by firmware.

[0178] The processor and the electronic circuit are also referred to as processing circuits. That is, the functions of each part of each device of the maintenance support system are realized by the processing circuits.

[0179] Implementation Method 2

[0180] In this embodiment, differences from Embodiment 1 and additions to Embodiment 1 are mainly described.

[0181] In this embodiment, the same reference numerals are given to components having the same functions as those in Embodiment 1, and description thereof will be omitted.

[0182] In the first embodiment, the maintenance work estimation model is acquired from the element unit estimation model 108 for each element of the equipment configuration. In the present embodiment, a method of registering the element unit maintenance work estimation model learned in the target equipment in the element unit estimation model 108 will be described.

[0183] Fig.12 1 is a diagram showing a functional configuration example of the maintenance support system 500 according to the present embodiment.

[0184] In this embodiment, in addition to the functional configuration of the first embodiment, an element unit estimation model registration unit 201 is provided.

[0185] exist Fig.12 In the example, the element unit estimation model registration unit 201 registers the maintenance work estimation model 105 , which is a result of learning in the equipment, in the element unit estimation model 108 in association with the equipment configuration acquired by the equipment configuration input unit 106 .

[0186] The association between the maintenance work inference model and the equipment structure uses the category ID of the equipment, part or component. If a maintenance work inference model of the equipment, part or component with the same category ID has already been registered in the element unit inference model 108, the model is mechanically overwritten. Alternatively, the user may be allowed to select whether to overwrite.

[0187] Furthermore, the registration of the maintenance work estimation model may be performed mechanically after a certain period of time has passed, or may be performed by the user at any timing.

[0188] As described above, in the maintenance support system of this embodiment, the model learned in the equipment is mechanically registered as the element unit estimation model. Therefore, according to the maintenance support system of this embodiment, the labor and time of manually constructing the element unit estimation model are not required, and the effect of mechanical expansion is achieved.

[0189] Furthermore, in the maintenance support system of the present embodiment, a "maintenance work estimation model" learned in a certain device is associated with elements of the device and registered as a "learned estimation model", so that it can be reused in other devices of the same type.

[0190] Implementation 3

[0191] In this embodiment, differences from Embodiment 2 and additions to Embodiment 2 are mainly described.

[0192] In this embodiment, the same reference numerals are given to components having the same functions as those in Embodiments 1 and 2, and description thereof will be omitted.

[0193] In the second embodiment, the component-based maintenance work estimation model learned in one device is registered in the component-based estimation model 108. In the present embodiment, a method of registering component-based maintenance work estimation models learned in a plurality of devices in the component-based estimation model 108 is described.

[0194] The maintenance support system 500 of the present embodiment includes a learning inference execution unit 200 in each of the plurality of devices, and the learning inference execution unit 200 includes an element unit inference model acquisition unit 107, a maintenance operation inference unit 109, and a maintenance operation learning unit 104. In addition, the maintenance support system 500 of the present embodiment includes an element unit inference model storage unit 300, and the learning inference execution units 200 of the plurality of devices can refer to the element unit inference model storage unit 300 in a manner of sharing the element unit inference model 108.

[0195] Fig.13 1 is a diagram showing a functional configuration example of the maintenance support system 500 according to the present embodiment.

[0196] exist Fig.13 In the embodiment, the learning and inference execution unit 200 has functions other than the element unit inference model 108 in the maintenance support system 500. The learning and inference execution unit 200 is provided in each of the plurality of devices.

[0197] The element-based estimation model storage unit 300 has the function of the element-based estimation model 108 in the maintenance support system 500. In the second embodiment, one element-based estimation model 108 is provided in one maintenance support system 500. In the present embodiment, a plurality of learning and estimation execution units 200 share one element-based estimation model 108.

[0198] In this embodiment, when a learning inference execution unit 200 set up for a certain device registers the maintenance work inference model of the structural element unit learned in the device into the element unit inference model 108, the learning result can be utilized in the learning inference execution unit 200 set up for other devices.

[0199] The learning inference execution unit 200 and the element unit inference model storage unit 300 are assumed to be able to operate in different devices as hardware. In addition, if the devices are connected using a network such as the Internet or LAN, the learning results can be shared beyond the physical distance. LAN is the abbreviation of Local Area Network.

[0200] As described above, in the maintenance support system of the present embodiment, the learning results are shared via the network, that is, the element unit inference model is shared by multiple learning inference execution units. Therefore, according to the maintenance support system of the present embodiment, the following effect is achieved: by learning based on the actual performance of multiple devices, the reliability of the priority or credibility of the maintenance work content presented to the maintenance operator can be improved.

[0201] In the maintenance support system of this embodiment, when there are multiple devices of the same type across factories, etc., each learning result can be mechanically aggregated. For example, assume that the same type of equipment is operating in factory A and factory B, and the "maintenance work inference model 105" is updated in the system of each factory, and the updated model is reflected in the "element unit inference model 108". In this way, the credibility of the model can be improved based on the operating performance of the equipment of the two factories. More specifically, the following process can be cited.

[0202] (1) In the systems of Factory A and Factory B, a "maintenance work estimation model 105" is constructed respectively ( Fig.10 Step S101).

[0203] (2) In the factory A, a sign of abnormality is detected, maintenance is performed, and the "maintenance work estimation model 105" is updated.

[0204] (3) In the system of plant A, the "element unit estimation model 108" is overwritten (updated) with the "maintenance work estimation model 105".

[0205] (4) Factory B is notified that the “element unit estimation model 108” has been overwritten (updated) in Factory A.

[0206] (5) In the system of the plant B, the element unit estimation model is again obtained from the "element unit estimation model 108" for the element, and is set as the initial model of the "maintenance work estimation model 105".

[0207] As described above, in the maintenance support system of this embodiment, when the "maintenance work estimation model 105" is updated in a certain factory, the "element unit estimation model 108" is overwritten (updated) as a result. By using the overwritten (updated) "element unit estimation model 108" again as the initial model in other factories, the learning results of the systems of multiple factories can be aggregated.

[0208] Furthermore, the effects of the maintenance support system according to the present embodiment will be described using another specific example.

[0209] As another specific example, it is assumed that each of devices A to J of the same type has the learning inference execution unit of the present embodiment.

[0210] When a fault occurs in device A, a maintenance operation such as "replace components for the current waveform of the motor" is performed once, and the credibility of the prompt content in the maintenance operation inference model is updated. Furthermore, when a fault also occurs in each of devices B to J, the credibility of the maintenance operation inference model is updated in the same way. As a result, the maintenance operation inference model can derive highly reliable prompt content in the same manner as when the process has been repeated 10 times. Assuming that the fault occurs once within a period T1, devices A to J generate models separately. In order to obtain the same credibility, T1×10 time is required in each device. However, in the present embodiment, it can be generated in T1 time.

[0211] Implementation 4

[0212] In this embodiment, differences from Embodiment 1 and additions to Embodiment 1 are mainly described.

[0213] In this embodiment, the same reference numerals are given to components having the same functions as those in Embodiments 1 to 3, and description thereof will be omitted.

[0214] In the first embodiment, it is determined whether the abnormality sign has been eliminated based on the maintenance work result, and the result is used to evaluate the content of the performed maintenance work. In this embodiment, a method in which a maintenance worker evaluates the content of the performed maintenance work is described.

[0215] Fig.141 is a diagram showing a functional configuration example of the maintenance support system 500 according to the present embodiment.

[0216] In this embodiment, in addition to the functional configuration of the first embodiment, a maintenance work evaluation input unit 401 is provided.

[0217] exist Fig.14 The maintenance work evaluation input unit 401 allows the maintenance worker to input an evaluation of the effectiveness of the maintenance work content presented by the maintenance work presentation unit 110 in eliminating the abnormality. The maintenance work evaluation input unit 401 receives the maintenance worker's evaluation of the maintenance work content and submits it to the maintenance work learning unit 104.

[0218] In an embodiment, if Fig.10 As shown in step S108 and step S109, the maintenance operation learning unit 104 evaluates the maintenance operation content provided according to whether the abnormality sign is successfully eliminated. In this embodiment, the evaluation result of the maintenance operation content input by the maintenance operator is used. Specifically, the evaluation of the maintenance operation content that the maintenance operator assigns a high evaluation is increased, and the evaluation of the maintenance operation content that the maintenance operator assigns a low evaluation is reduced. In addition, it can also be combined with the evaluation method of embodiment 1.

[0219] For example, after evaluating the suggested maintenance work content based on the success or failure of eliminating the abnormal signs in the same way as in Implementation 1, the maintenance work content that the maintenance worker assigns a high evaluation may have its evaluation further improved, while the maintenance work content that the maintenance worker assigns a low evaluation may have its evaluation lowered.

[0220] As described above, in the maintenance support system of the present embodiment, the maintenance operator gives an evaluation to the content of the maintenance work that has been performed. Therefore, according to the maintenance support system of the present embodiment, the following effects are achieved: the content of the maintenance work can also be evaluated based on the criterion of the ease of handling by a person, and as a result of learning, the content of the maintenance work that is easy for a person to handle can be given priority.

[0221] Furthermore, when a plurality of maintenance work contents are presented to the maintenance operator, not necessarily only the maintenance work contents performed immediately before the abnormality sign is eliminated are effective. In the maintenance support system of this embodiment, in this case, the maintenance work contents that the maintenance operator determines to be useful can also be selected from the plurality of maintenance work contents that have been performed. Thus, according to the maintenance support system of this embodiment, the maintenance work contents that are helpful in eliminating the abnormality sign can be evaluated more highly.

[0222] In the above embodiments 1 to 4, the parts of the devices of the maintenance support system are described as independent functional blocks. However, the structure of the maintenance support device may not be the structure of the above embodiment. The functional blocks of the maintenance support device can achieve the functions described in the above embodiment, and may be any structure. In addition, the maintenance support device may not be a single device, but a system composed of multiple devices.

[0223] Furthermore, a plurality of parts of Embodiments 1 to 4 may be implemented in combination. Alternatively, one part of these embodiments may be implemented. In addition, these embodiments may be implemented in combination as a whole or in part.

[0224] That is, in Embodiments 1 to 4, it is possible to freely combine the respective embodiments, to modify any constituent element in the respective embodiments, or to omit any constituent element in the respective embodiments.

[0225] In addition, the above-mentioned embodiments are essentially preferred illustrations and are not intended to limit the scope of the present disclosure, the scope of the application of the present disclosure, and the scope of the use of the present disclosure. The above-mentioned embodiments can be variously changed as needed. For example, the steps described using a flowchart or a timing diagram can also be appropriately changed.

[0226] Description of symbols

[0227] 10: Maintenance operation export device; 11, 21: Processor; 11a, 21a: Electronic circuit; 12, 22: Memory; 13, 23: Storage; 14, 24: Display device; 15, 25: Operation interface; 16: Equipment information acquisition interface; 17: Operation content acquisition interface; 18, 26: Communication interface; 20: Operator terminal; 30: Equipment; 31: Memory; 40: Maintenance operation recording device; 101: Sensor data acquisition unit; 102: Abnormal sign detection unit; 103: Feature extraction unit; 104: Maintenance operation learning unit; 105: Maintenance operation inference model; 106: Equipment structure input unit; 107: Element unit inference model Model acquisition unit; 108: Element unit inference model; 109: Maintenance operation inference unit; 110: Maintenance operation prompt unit; 111: Maintenance operation result acquisition unit; 112: Maintenance operation recording unit; 200: Learning inference execution unit; 201: Element unit inference model registration unit; 300: Element unit inference model storage unit; 401: Maintenance operation evaluation input unit; 500: Maintenance support system; 701: Equipment unit inference model; 702: Part unit inference model; 703: Component unit inference model; 901: Record object selection editing unit; 902: Sensor information acquisition unit; 903: Equipment operation history acquisition unit; 904: Operator input acquisition unit.

Claims

1. A maintenance support system that supports a maintenance operator of equipment in performing maintenance work, wherein: The maintenance assistance system comprises: an element unit inference model acquisition unit, which acquires a structural element of the equipment, namely, an equipment structural element, acquires an element unit inference model for estimating the content of the maintenance work in units of the equipment structural element, and sets the acquired element unit inference model as an initial value of the maintenance work inference model for estimating the content of the maintenance work of the equipment; a maintenance work estimation unit that applies the maintenance work estimation model to a feature quantity of an abnormality extracted from sensor data in the equipment to estimate the content of maintenance work to be presented to a maintenance worker; and The maintenance work learning unit repeatedly evaluates whether the suggested maintenance work content is effective in eliminating the abnormality based on the evaluation of the effectiveness of the suggested maintenance work content, learns the maintenance work content that is effective for the feature quantity of the abnormality, and updates the maintenance work inference model.

2. The maintenance support system according to claim 1, wherein: The equipment structural elements include the category of the equipment, the category of the parts constituting the equipment, and the category of the components constituting the parts. The maintenance assistance system has a storage unit that stores the element unit inference model according to the category of the equipment, the category of the parts, and the category of the components.

3. The maintenance support system according to claim 2, wherein: The element unit estimation model acquisition unit acquires the element unit estimation model in accordance with the type of the equipment, the type of the part, and the type of the component in the equipment.

4. The maintenance support system according to claim 1, wherein: The maintenance support system includes a maintenance work presenting unit that presents the maintenance work content estimated by the maintenance work estimating unit to the maintenance worker.

5. The maintenance support system according to any one of claims 1 to 4, wherein: The maintenance support system includes an element unit estimation model registration unit that records the maintenance work estimation model updated by the maintenance work learning unit in the element unit estimation model in association with the equipment configuration element.

6. The maintenance support system according to claim 5, wherein: The maintenance support system includes a learning and inference execution unit in each of the plurality of devices, and the learning and inference execution unit includes the element unit inference model acquisition unit, the maintenance work inference unit, and the maintenance work learning unit. The maintenance support system includes an element-based estimation model storage unit, and the learning estimation execution unit of each of the plurality of devices can refer to the element-based estimation model storage unit so as to share the element-based estimation model.

7. The maintenance support system according to any one of claims 1 to 6, wherein: The maintenance support system includes a maintenance work evaluation input unit that causes the maintenance worker to input the effectiveness evaluation of whether the presented maintenance work content is effective in eliminating the abnormality.

8. A maintenance assistance method for a maintenance assistance system, the maintenance assistance system assisting a maintenance operator of equipment in performing maintenance work, wherein: The computer obtains a structural element of the device, namely, a device structural element, obtains a component unit inference model for estimating the content of the maintenance work in units of the device structural element, sets the obtained component unit inference model as an initial value of the maintenance work inference model for estimating the content of the maintenance work of the device, The computer applies the maintenance work inference model to the feature quantity of the abnormality extracted from the sensor data in the equipment to estimate the content of the maintenance work to be presented to the maintenance worker. The computer repeatedly evaluates whether the suggested maintenance work content is effective for eliminating the abnormality based on the evaluation of the effectiveness of the suggested maintenance work content, learns the maintenance work content that is effective for the feature amount of the abnormality, and updates the maintenance work inference model.

9. A maintenance assistance program for a maintenance assistance system for assisting a maintenance operator of equipment in performing maintenance work, wherein: The maintenance assistance program causes the computer to execute the following processing: An element unit inference model acquisition process is performed to acquire a structural element of the equipment, namely, an equipment structural element, acquire an element unit inference model for estimating the content of the maintenance work in units of the equipment structural element, and set the acquired element unit inference model as an initial value of the maintenance work inference model for estimating the content of the maintenance work of the equipment; Maintenance work inference processing, applying the maintenance work inference model to the feature quantity of the abnormality extracted from the sensor data in the equipment, and estimating the content of the maintenance work to be presented to the maintenance worker; as well as The maintenance work learning process repeatedly evaluates whether the suggested maintenance work content is effective in eliminating the abnormality based on the evaluation of the effectiveness of the suggested maintenance work content, learns the maintenance work content that is effective for the characteristic quantity of the abnormality, and updates the maintenance work inference model.

Images