Maintenance plan management system, learning apparatus, and inference apparatus

Abstract

To easily formulate a maintenance plan in consideration of contract information.SOLUTION: A maintenance plan management system 1 includes: a data acquisition unit which acquires learning data including contract information and inspection information of a maintenance target, and modification processing information indicating the content of changes on information of a preset maintenance plan; a learning unit 112 which generates, based on the learning data, a trained model for inferring new modification processing information; and an inference unit 115 which infers, using the trained model for inferring new modification processing information, a new maintenance plan by inferring the new modification processing information.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present disclosure relates to a maintenance plan management system, a learning device, and an inference device. [Background technology]

[0002] There are known techniques for supporting efficient maintenance of equipment under a maintenance contract. For example, Patent Literature 1 discloses a technique for scheduling inspections by estimating recommended maintenance times for parts to be maintained based on inspection information acquired during inspection of the equipment to be maintained. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2015-88045 Summary of the Invention [Problem to be solved by the invention]

[0004] Regarding machine maintenance, a contract is concluded between a maintenance company and a machine user, and maintenance work such as inspections and part replacement is carried out based on the contents of the contract. In recent years, the appearance of new machine models and longer machine lifespans have led to a diversification of contract contents, contract periods, and the like. However, the technology in Patent Document 1 schedules inspection plans focusing only on inspection information including inspection history information and usage record information, and does not take into account the diversified contract information including the contract contents and contract periods. For this reason, there is a need for support for easily formulating maintenance plans that take contract information into account.

[0005] The present disclosure has been made in consideration of the above-mentioned problems, and aims to provide a maintenance plan management system, a learning device, and an inference device that can easily formulate maintenance plans that take contract information into consideration. [Means for solving the problem]

[0006] In order to achieve the above object, the maintenance plan management system according to the present disclosure includes a data acquisition unit that acquires learning data including contract information and inspection information for the maintenance target and change processing information that indicates changes to the information of a predetermined maintenance plan, a learning unit that generates a trained model for inferring new change processing information based on the learning data, and an inference unit that infers a new maintenance plan by inferring the new change processing information using the trained model for inferring the new change processing information. [Effects of the Invention]

[0007] According to the present disclosure, learning data including contract information and inspection information for a maintenance target and change processing information representing changes to information in a preset maintenance plan is acquired, a trained model for inferring new change processing information is generated based on the learning data, and the new change processing information is estimated using the trained model for inferring the new change processing information, thereby estimating a new maintenance plan. This makes it possible to easily formulate a maintenance plan that takes contract information into consideration. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1 is a diagram illustrating a configuration of an operating system including a maintenance plan management system according to an embodiment of the present disclosure. [Figure 2] FIG. 1 is a block diagram illustrating a configuration example of an operating system including a maintenance plan management system according to an embodiment of the present disclosure. [Figure 3] FIG. 1 is a diagram illustrating an example of inspection information according to an embodiment of the present disclosure. [Figure 4] FIG. 1 is a diagram illustrating an example of contract information according to an embodiment of the present disclosure. [Figure 5] FIG. 1 is a diagram illustrating an example of a hardware configuration of a maintenance plan management device according to an embodiment of the present disclosure. [Figure 6] FIG. 1 is a diagram illustrating a functional configuration of a learning unit according to an embodiment of the present disclosure. [Figure 7] 1 is a flowchart of a learning process according to an embodiment of the present disclosure; [Figure 8]FIG. 1 is a diagram illustrating the relationship between screens displayed on a display unit of a terminal device according to an embodiment of the present disclosure. [Figure 9] FIG. 10 is a diagram illustrating an example of a screen image displayed on a display unit of a terminal device according to an embodiment of the present disclosure. [Figure 10] FIG. 10 is a diagram illustrating an example of a screen image displayed on a display unit of a terminal device according to an embodiment of the present disclosure. [Figure 11] FIG. 10 is a diagram illustrating an example of a screen image displayed on a display unit of a terminal device according to an embodiment of the present disclosure. [Figure 12] FIG. 10 is a diagram illustrating an example of a screen image displayed on a display unit of a terminal device according to an embodiment of the present disclosure. [Figure 13] FIG. 1 is a diagram illustrating a functional configuration of an inference unit according to an embodiment of the present disclosure. [Figure 14] Flowchart of inference processing according to an embodiment of the present disclosure DETAILED DESCRIPTION OF THE INVENTION

[0009] (Embodiment) A maintenance schedule management system, a learning device, and an inference device according to embodiments of the present disclosure will be described in detail below with reference to the drawings. Note that identical or corresponding parts in the drawings are designated by the same reference numerals.

[0010] A maintenance schedule management system according to an embodiment of the present disclosure estimates and outputs a maintenance schedule, including information indicating a maintenance schedule, based on inspection information and contract information for equipment, machines, etc. to be maintained. Here, inspection information refers to information related to the equipment, machines, etc. to be maintained. Contract information refers to information based on the contents of a contract concluded with a customer regarding the maintenance of the equipment, machines, etc. to be maintained. FIG. 1 is a diagram illustrating the configuration of an operation system 100 including a maintenance schedule management system 1 according to an embodiment of the present disclosure. As illustrated in FIGS. 3 and 4, the inspection information includes information identifying the equipment, machines, etc. to be maintained and information about the customer who owns the equipment, machines, etc. to be maintained. The contract information includes information such as the contract start date, contract end date, and the number of inspections within the contract period. Details will be described later.

[0011] The operating system 100 includes an operating device 30 that is the target of maintenance, and a maintenance plan management system 1 that estimates and outputs a maintenance plan. The operating device 30 is the operating device for which a maintenance plan is estimated by the maintenance plan management system 1. The operating device 30 stores various information related to its own operation in its own memory. The operating device 30 includes an operation information management unit 31 that manages operation information that indicates its own operating status.

[0012] The maintenance plan management system 1 includes a maintenance plan management device 10 that estimates and outputs a maintenance plan, and a terminal device 40 that is used by a person in charge of setting the maintenance plan, a person in charge, etc. The terminal device 40 includes a display unit 41 that displays various information, etc. The maintenance plan management device 10 and the terminal device 40 are each configured as a personal computer, for example. Furthermore, the maintenance plan management device 10, the terminal device 40, and the operating device 30 are connected via a network NW.

[0013] 2, the maintenance plan management device 10 includes a calculation processing unit 11 that executes processing to acquire or generate various data, and a storage unit 12 that stores data necessary for the processing performed by the calculation processing unit 11. When the maintenance plan management device 10 is started, the operation information management unit 31 transmits operation information including various information indicating its own operation status to the maintenance plan management device 10 via the network NW. For example, the operation information management unit 31 acquires and accumulates information detected by a temperature sensor, a vibration sensor, etc. attached to the operating device 30. The operation information management unit 31 transmits the accumulated operation information to the maintenance plan management device 10 via the network NW at a preset timing, for example, as operation information data indicating the daily operation status of the operating device 30.

[0014] The calculation processing unit 11 includes, for example, a CPU (Central Processing Unit). By executing a program stored in the storage unit 12, the calculation processing unit 11 functions as an operation information acquisition unit 110 that acquires operation information from the operating device 30, a maintenance plan estimation unit 111 that estimates a maintenance plan for the operating device 30, and an output unit 118 that outputs the maintenance plan.

[0015] The storage unit 12 includes, for example, a nonvolatile semiconductor memory including a flash memory and an EPROM (Erasable Programmable Read Only Memory).The storage unit 12 may also include a nonvolatile memory including a magnetic disk, a flexible disk, an optical disk, a compact disk, a minidisk, and a DVD (Digital Versatile Disc).

[0016] The storage unit 12 includes an inspection information storage unit 121 that stores inspection information, a contract information storage unit 122 that stores contract information, a maintenance plan information storage unit 123 that stores maintenance plan information, and a trained model storage unit 124 that stores the trained model generated by the maintenance plan estimation unit 111. The storage unit 12 also stores a program executed by the calculation processing unit 11. The storage unit 12 also stores software, applications, programs, data, parameters, etc. that are required to execute various processes in this embodiment.

[0017] The inspection information storage unit 121 stores inspection information of the operating device 30. In this embodiment, as illustrated in Fig. 3, the inspection information includes device information that identifies the operating device 30 to be inspected, information on the customer who owns the operating device 30, etc. The inspection information includes operation information that indicates the operation record and operation status, unit number information that indicates information that identifies the operating device 30, customer information that indicates information that identifies the customer, work history information that indicates the history of unscheduled work that is not based on the details of the contract with the customer, and sales history information that indicates sales performance.

[0018] The operation information includes the cumulative operation time of the operating device 30, operation information data indicating the operation status, the usage time of each part, etc. As described above, the operation information is collected and accumulated by the operating device 30 itself, and transmitted to the maintenance plan management device 10 at a preset timing.

[0019] The unit information represents information that identifies the operating device 30. The unit information includes the unit number of the operating device 30, the time when the device started operation, the unit ID, the manufacturing date of the operating device 30, information about the components of the operating device 30, etc. The customer information includes the company name, address, contact information of the person in charge, desired day of the week for inspection, desired time for inspection, inquiry history, etc. The unit information and customer information mainly consist of information that can be collected at the time of introducing the operating device 30.

[0020] In this embodiment, the work history information represents history information on unscheduled inspection work. Specifically, the work history information represents history information on unexpected inspection work not based on the details of a contract with a customer. For example, the work history information represents history information on unexpected inspection work performed when a malfunction occurs in the operating device 30. The work history information includes information such as the previous maintenance work date, which indicates the time when the last unscheduled inspection work was performed, the previous maintenance work content, which indicates the content of the previous inspection work, the previous person in charge, which indicates the person in charge of the previous inspection work, the past work result, which indicates the results of the past work, and the past work content, which indicates the content of the past work. Unlike maintenance, the sales history information includes information such as customer visit history, which indicates the number of visits the maintenance plan manager or person in charge made to customers who own the operating device 30, and consumable sales history, which indicates the number of consumable sales made by the maintenance plan manager or person in charge. The work history information, sales history information, inquiry history, etc. are updated by the maintenance plan manager or person in charge entering information from the terminal device 40 each time an action is taken. Note that the inspection information may include information other than those listed above.

[0021] The contract information storage unit 122 shown in FIG. 2 stores contract information data indicating the details of the maintenance contract for the operating device 30, as exemplified in the contract information table of FIG. 4. The contract information includes contract period information indicating the start date, end date, etc. of the contract between the maintenance company and the owner of the operating device 30; contract implementation item information indicating the contract ID, the number of scheduled inspections, etc.; contract implementation history information indicating the inspection date, the inspection implementation details, etc.; past contract IDs; and past contract conclusion information indicating the results of past contract conclusions. In this embodiment, the inspection types covered are "engineer dispatch inspection," which refers to inspection performed by an engineer in charge of maintenance work dispatched to the site, and "remote inspection," which refers to inspection performed without an engineer visiting the site. Note that maintenance work includes inspection, monitoring, repair, part replacement, etc., but this embodiment focuses on inspection work.

[0022] Contract period information includes information indicating the contract start date, contract end date, and if there is a change, the date of the contract change and the reason for the change. The contract start date and end date are set when the contract is concluded. Information regarding changes is updated each time. Contract implementation information includes the contract ID, the number of scheduled inspections based on the contract contents during the contract period, the scheduled number of engineer dispatch inspections, the scheduled number of remote inspections, and information indicating the contract plan and its contents.

[0023] The contract implementation history information includes information indicating the types of inspections that have been performed and the number of times they have been performed, information indicating the types of inspections that have been scheduled within the contract period but have not been performed and the number of times they have been performed, the scheduled date of the next inspection, the inspection details, confirmation of the next inspection schedule, the status of the inspection indicating whether the inspection has been completed, the engineer in charge of the inspection, and history information on maintenance responses performed during the contract period. In this embodiment, "engineer dispatch inspection" can be subdivided into inspection type P, inspection type Q, inspection type R, etc. Similarly, "remote inspection" can be subdivided into inspection type P, inspection type Q, inspection type R, etc. The status of inspection implementation includes statuses such as "unarraged" where the inspection schedule has not been confirmed, "arranged" where the inspection schedule has been confirmed, and "inspection completed" where the inspection has been completed.

[0024] The past contract conclusion information includes information indicating a past contract ID if there is a past contract, and information on past contract conclusion records indicating the records of contracts that have been concluded in the past for the same operating device 30. Note that the contract information may include information other than that listed above.

[0025] The maintenance plan information storage unit 123 shown in FIG. 2 stores the maintenance plan information estimated by the maintenance plan estimation unit 111. In this embodiment, a tentative maintenance plan is stored in advance in the maintenance plan information storage unit 123. The maintenance plan information storage unit 123 also stores schedule information of engineers indicating whether or not there is free time in the engineers' schedules. Furthermore, the maintenance plan information storage unit 123 stores information indicating the content of changes to the maintenance plan. Specifically, the maintenance plan information storage unit 123 stores information indicating the content to be replaced with a new maintenance plan. For example, When a tentative maintenance plan is replaced with a new maintenance plan, information indicating candidate dates that can become new dates for performing maintenance within the contract period is stored in the maintenance plan information storage unit 123. Note that information indicating the content of the replacement with the new maintenance plan will hereinafter be referred to as "change processing information." Also, an example of engineer schedule information is shown in FIG. 12, but details will be described later. Also, an operating program for the maintenance plan estimation unit 111 to adjust the maintenance plan schedule is stored in the maintenance plan information storage unit 123.

[0026] The trained model storage unit 124 stores the trained model generated by the learning unit 112 in the maintenance schedule estimation unit 111.

[0027] The various data in the storage unit 12 of the maintenance plan management device 10 may be stored in a cloud-based server external to the maintenance plan management device 10.

[0028] The operation information acquisition unit 110 of the calculation processing unit 11 acquires the above-mentioned operation information from the operating device 30 and stores it in the inspection information storage unit 121 of the storage unit 12. The operation information includes, for example, the accumulated operation time, operation information data, and part usage period shown in FIG. 3 .

[0029] The maintenance plan estimation unit 111 includes a learning unit 112 that generates a trained model, and an inference unit 115 that estimates a maintenance plan using the trained model.

[0030] The learning unit 112 uses learning data including inspection information, contract information, and information representing the maintenance plan to learn an action-value function for determining a new maintenance plan. Learning is performed using known algorithms such as reinforcement learning and unsupervised learning. The learned action-value function is stored in the learned model storage unit 124 as a learned model.

[0031] The inference unit 115 uses the trained model generated by the learning unit 112 and stored in the trained model storage unit 124 to infer a new maintenance plan.

[0032] The output unit 118 outputs information on the new maintenance plan inferred by the inference unit 115 to the terminal device 40.

[0033] Next, an example of the hardware configuration of the maintenance plan management device 10 will be described with reference to Fig. 5. The maintenance plan management device 10 in Fig. 5 is realized by a computer such as a personal computer or a microcontroller.

[0034] The maintenance plan management device 10 comprises a processor 1001 that executes an operation program that is a program for the operation of the maintenance plan management device 10, a memory 1002 that serves as the main storage area, an interface 1003 that realizes the communication function of the maintenance plan management device 10, and a secondary storage device 1004 that stores the operation program for executing processing, all of which are connected to each other via a bus 1000.

[0035] The processor 1001 is, for example, a CPU (Central Processing Unit). The processor 1001 reads an operating program stored in a secondary storage device 1004 into a memory 1002 and executes the program, thereby realizing each function of the maintenance plan management device 10.

[0036] The memory 1002 is a main storage device configured, for example, by a RAM (Random Access Memory). The memory 1002 stores the operating program that the processor 1001 reads from the secondary storage device 1004. The memory 1002 also functions as a work memory when the processor 1001 executes the operating program.

[0037] The interface 1003 is an I / O (Input / Output) interface such as a serial port, a USB (Universal Serial Bus) port, a network interface, etc. The interface 1003 realizes the communication function of the maintenance plan management device 10.

[0038] The secondary storage device 1004 is, for example, a flash memory, a hard disk drive (HDD), or a solid state drive (SSD). The secondary storage device 1004 stores the operating program executed by the processor 1001, the inspection information exemplified in Fig. 3, the contract information exemplified in Fig. 4, the engineer schedule information exemplified in Fig. 12, information representing the maintenance plan estimated by the maintenance plan estimation unit 111 and change processing information, and the trained model generated by the learning unit 112 in the maintenance plan estimation unit 111.

[0039] 6 is a diagram showing the functional configuration of the learning unit 112 included in the maintenance plan management device 10. The learning unit 112 includes a data acquisition unit 113 and a model generation unit 114. Learning is performed using known algorithms such as reinforcement learning and unsupervised learning, but reinforcement learning will be taken as an example for explanation.

[0040] The data acquisition unit 113 acquires, as learning data, the change processing information stored in the maintenance plan information storage unit 123 as action A, and the inspection information stored in the inspection information storage unit 121 and the contract information stored in the contract information storage unit 122 as state S.

[0041] The model generation unit 114 learns change processing information that will be the optimal action A based on learning data that includes change processing information as action A and inspection information and contract information as state S. That is, a learned model is generated that infers change processing information that will be the optimal action A from the contract information and inspection information that correspond to state S. A new maintenance plan is inferred from the inferred change processing information.

[0042] In reinforcement learning used by the model generation unit 114, an agent (subject of action) in a certain environment observes the current state (environmental parameters) and decides on the action to be taken. The environment changes dynamically depending on the agent's actions, and the agent is given a reward according to the changes in the environment. The agent repeats this process and learns the course of action that will obtain the most reward through a series of actions. Known representative methods of reinforcement learning include Q-learning and TD-learning. For example, in the case of Q-learning, the general update formula for the action value function Q(s, a) is expressed by Equation 1.

[0043]

number

[0044] In equation 1, s t represents the state of the environment at time t, and a t represents the action at time t. Action a t Therefore, the state is s t+1 Changes to r t+1 represents the reward obtained by the change in state, γ represents the discount rate, and α represents the learning coefficient. Note that γ is in the range of 0<γ≦1, and α is in the range of 0<α≦1. If action A at time t is action a t and state S becomes state s t and the state s at time t t Best Practices in a t Learn.

[0045] According to the update formula expressed by Equation 1, if the action value Q of action a with the highest Q value at time t+1 is greater than the action value Q of action a executed at time t, the action value Q is increased, and vice versa. In other words, the action value function Q(s, a) is updated with the aim of bringing the action value Q of action a at time t closer to the best action value at time t+1. As a result, the best action value in a certain environment is propagated sequentially to the action value in the previous environment.

[0046] As shown in FIG. 6, the model generation unit 114 includes a reward calculation unit 114a and a function update unit 114b.

[0047] The reward calculation unit 114a calculates a reward based on the change process information corresponding to the action A and the inspection information and contract information corresponding to the state S. For example, the reward calculation unit 114a calculates an inspection interval, which indicates the interval between inspections, based on the change process information. The reward calculation unit 114a calculates a reward r based on the inspection interval as a reward criterion. For example, the reward calculation unit 114a calculates the standard deviation of the inspection intervals within the contract period. If there is no significant variation in the inspection intervals within the contract period, in other words, if the standard deviation of the inspection intervals is smaller than a predetermined threshold, the reward calculation unit 114a increases the reward r. For example, a reward of "1" is awarded. On the other hand, if the standard deviation of the inspection intervals within the contract period is equal to or greater than a predetermined threshold, the reward calculation unit 114a reduces the reward r. For example, a reward of "-1" is awarded. The standard deviation is an example of a dispersion according to the present disclosure.

[0048] The function update unit 114b updates a function for determining optimal change processing information in accordance with the reward calculated by the reward calculation unit 114a, and outputs the updated function to the learned model storage unit 124. For example, in the case of Q-learning, the function update unit 114b updates the action value function Q(s t ,a t ) is output to the trained model storage unit 124 as a function for calculating optimal change processing information.

[0049] The learning process is repeated as described above. The learned model storage unit 124 stores the action-value function Q(s t ,a t ), i.e., stores the trained model.

[0050] Next, the learning process performed by the learning unit 112 will be described with reference to Fig. 7. Fig. 7 is a flowchart of the learning process performed by the learning unit 112.

[0051] The data acquisition unit 113 acquires, as learning data, change process information corresponding to the action A, and contract information and inspection information corresponding to the state S (step S101).

[0052] The model generation unit 114 calculates the reward based on the change process information corresponding to the action A and the contract information and inspection information corresponding to the state S (step S102). Specifically, the reward calculation unit 114a acquires the change process information corresponding to the action A and the contract information and inspection information corresponding to the state S, and calculates the inspection interval representing the interval at which inspections are performed. The reward calculation unit 114a determines whether to increase or decrease the reward based on the inspection interval within a predetermined contract period (step S102).

[0053] If the reward calculation unit 114a determines that the reward should be increased (step S102: Yes), it increases the reward in step S103. On the other hand, if the reward calculation unit 114a determines that the reward should be decreased (step S102: No), it decreases the reward in step S104.

[0054] The function update unit 114b updates the action value function Q(s) represented by Equation 1 stored in the trained model storage unit 124 based on the reward calculated by the reward calculation unit 114a. t ,a t ) is updated (step S105).

[0055] The learning unit 112 repeatedly executes the above steps S101 to S105 at times t, t+1, . . . The learned model storage unit 124 stores the generated action value function Q(s t ,a t ) is stored as a trained model.

[0056] The learning unit 112 in this embodiment is configured to store the learned model in a learned model storage unit 124 provided outside the learning unit 112, but the learned model storage unit 124 may also be provided inside the learning unit 112.

[0057] Next, various processing flows in this embodiment will be explained using Figures 8 to 12. The explanation will be made in order, assuming that a person in charge of or in charge of a maintenance plan searches for inspection schedules, formulates a maintenance plan, etc. on the terminal device 40. The person in charge of or in charge of a maintenance plan will hereinafter also be simply referred to as a "user."

[0058] When a maintenance contract is concluded, the date and type of inspection to be carried out within the contract period will be tentatively set in advance, and then the official inspection schedule will be determined after taking into consideration customer requests, inspection information, and contract information.

[0059] 8 shows a relationship diagram between screens that are displayed when a user searches for a maintenance schedule, plans a new maintenance schedule, etc. In response to an operation or input by the user, the maintenance schedule estimation unit 111 outputs information shown in screens A to D, which will be described later.

[0060] Screen A is a screen where the user searches for cases for which a maintenance schedule has already been set. Note that cases for which a maintenance schedule has already been set include cases that were provisionally set when the contract was signed. Screen B is a screen where the user extracts a list of cases that match the conditions entered by the user on Screen A and sorts them by priority in order to plan a maintenance schedule for cases for which an official maintenance schedule has not yet been decided.

[0061] Screen C is a screen for displaying detailed information for each scheduled maintenance case and determining the maintenance schedule, including the schedule. Screen D is a pop-up screen showing schedules for each engineer, and the user refers to information about the engineer as reference information when determining the maintenance schedule on screen C. In response to the information input by the user, the maintenance plan estimation unit 111 acquires information stored in the inspection information storage unit 121, contract information storage unit 122, maintenance plan information storage unit 123, and trained model storage unit 124. After the information input by the user on screens B, C, and D and processing by the calculation processing unit 11, the maintenance plan estimation unit 111 estimates a new maintenance plan. The output unit 118 outputs the new maintenance plan to the terminal device 40. When a new maintenance plan is determined, the information in the maintenance plan information storage unit 123 is updated.

[0062] Screen A shown in FIG. 9 is an example of a screen displayed when a user searches for a project for which a maintenance schedule has been set. Screen A's components include a condition input field 701, which indicates a field for inputting conditions; a detail field 702, which indicates a field for selecting detailed information for each unit or project; and a schedule 703 for each engineer for the input year and month. The condition input field 701 includes selection fields for confirming the schedule, creating a new schedule, and changing the schedule; a target month and year input field for displaying the schedule; and an optional field for selecting the target engineer, and the user inputs each of these fields. In the case of confirming the schedule, the work is completed on this screen. When the maintenance plan management device 10 is started, the maintenance plan estimation unit 111 outputs the condition input field 701 and the detail field 702 to the terminal device 40. The terminal device 40 displays the condition input field 701 and the detail field 702 on the display unit 41. For example, suppose the user selects "Check Schedule" in the condition input item 701, inputs August of a certain year as the year and month for which the user wants to check the schedule, and inputs "X, Y, Z" as the target engineers for whom the user wants to check the schedule. Based on these input items, the maintenance plan estimation unit 111 outputs schedules 703 for engineers X, Y, and Z from the schedule information of the engineers stored in the maintenance plan information storage unit 123.

[0063] When formulating a new maintenance plan, the user selects the information for each unit in the detailed item 702. In this case, the maintenance plan estimation unit 111 outputs the information shown on screen B, which is stored in the maintenance plan information storage unit 123, to the terminal device 40 based on information input by the user. When changing the maintenance schedule for a specific project, the user selects the project detailed information in the detailed item 702. In this case, the maintenance plan estimation unit 111 outputs the information shown on screen C, which is stored in the maintenance plan information storage unit 123, to the terminal device 40 based on information input by the user. Note that an input screen for the unit ID, contract ID, etc. may also be provided. Furthermore, a selection item for the information for each engineer may be provided, for example, in the detailed item 702. In the event of an emergency, if the user wants to assign work to an engineer's available schedule by predetermining the date, time, and person in charge, the user may select the information for each engineer. In this case, the maintenance plan estimation unit 111 outputs the information shown on screen D, which is stored in the maintenance plan information storage unit 123, to the terminal device 40 based on information input by the user.

[0064] The components of screen B shown in FIG. 10 include a condition input field 711, which represents an item for inputting conditions, and a detailed information confirmation field 712, which represents detailed information for each contract ID or unit ID that is output according to the input conditions. In the condition input field 711, the user inputs the display year and month, inspection plan status, and sorting conditions for the inspection list. The maintenance schedule estimation unit 111 selects information corresponding to the user's input information from information including the inspection information stored in the inspection information storage unit 121 and the contract information stored in the contract information storage unit 122 and outputs it to the terminal device 40. Based on the output information, the terminal device 40 displays detailed information for each contract ID, for example, in the detailed information confirmation field 712. Note that in FIG. 10, the next scheduled inspection month is displayed in the detailed information confirmation field 712, but the next scheduled inspection month has not yet been determined. The sorting conditions include projects whose contracts are about to expire, projects whose inspections were scheduled in the previous month but have not yet been inspected, important customers, such as VIP customers, and customers requiring careful attention.

[0065] In response to the output of the maintenance plan estimation unit 111, the detailed information confirmation item 712 displays, for each contract ID, information such as the machine number, machine operation date, company name, address, estimated work time, scheduled number of contract inspections, remaining number of inspections for each inspection type, date of the last engineer dispatch inspection, scheduled month of the next inspection, contract expiration month, customer requests, etc. The estimated work time indicates the standard time required for work and can be set for each inspection type.

[0066] The estimated work time is set for each inspection type in units such as half a day, one day, or two days. In the example of Figure 9, Engineer X's schedules for August 21st and 22nd are both marked as "Inspection Scheduled." If the person in charge arranges an inspection type that requires two days of work time, and the inspection is allocated to the 21st, the schedule for the 22nd will automatically also be marked as "Inspection Scheduled." Depending on the arrangement by the person in charge, for example, the maintenance plan estimation unit 111 outputs "Inspection Scheduled" information. In response to the output of the maintenance plan estimation unit 111, the schedule is displayed on the calendar screen shown in Figure 9. Engineer Y's schedule for August 21st is marked as "Inspection Scheduled (Second Inspection Possible)." This is displayed when an inspection schedule with an estimated work time of half a day is allocated, indicating that another inspection schedule can be allocated on the same day. Engineer Z's schedule for August 25th is marked as "Inspection Scheduled," which indicates that an inspection schedule with an estimated work time of one day is allocated and is reflected in calendar format.

[0067] Suppose a user selects one or more items from the inspection schedule list displayed in the detailed information confirmation field 712 on screen B shown in FIG. 10, such as an item with a contract ID of 00001. For example, the user operates the mouse on screen B to move the cursor to the field with the contract ID of 00001 and click it. In response to the user's selection, the maintenance plan estimation unit 111 infers the next scheduled inspection date of August 2023, which corresponds to new change processing information, based on information including the inspection information stored in the inspection information storage unit 121 and the contract information stored in the contract information storage unit 122, and the trained model stored in the trained model storage unit 124. For contract ID 00001, the last engineer dispatch date is August 2022, the number of remaining inspections for inspection P is two, and the contract expiration month is May 2024. Based on this information, the maintenance plan estimation unit 111 infers that the next inspection date will be August 2023, taking into account, for example, the inspection interval. At this point, the maintenance plan estimation unit 111 outputs information indicating that the next scheduled inspection month in the detailed information confirmation item 712 for the project corresponding to contract ID 00001 is August 2023 to the terminal device 40. As illustrated in Fig. 11, the maintenance plan estimation unit 111 outputs information shown on screen C for the project corresponding to contract ID 00001 to the terminal device 40. Note that although the inspection date is displayed in Fig. 11, the inspection date has not been determined at this point. If the user specifies a plurality of items, the inspection schedules may be calculated and set in order according to the priority.

[0068] 11 includes a detailed information confirmation item 721 and engineer availability information 722. The maintenance schedule estimation unit 111 determines an appropriate inspection schedule by outputting the schedule information for August 2023, which is the next scheduled inspection month obtained by inference, and the engineer availability information 722. Specifically, the maintenance schedule estimation unit 111 outputs the customer's wishes and the engineer availability information for August 2023, and, for example, sets August 26, 2023 as the inspection date and designates Engineer X as the engineer in charge. At this point, the maintenance schedule estimation unit 111 causes the terminal device 40 to output information that the inspection date in the detailed information confirmation item 721 is August 26, 2023.

[0069] If there are no problems with the engineers' available man-hours, the inspection schedule arrangement is completed, and the maintenance schedule planning work ends here. If there are sufficient engineers available and it is possible to plan all inspection schedules, the maintenance plan is output to the screen, the information in the maintenance plan information storage unit 123 is updated, and the user's maintenance schedule planning work ends. If there are still cases for which inspection schedules have not been arranged, the user can, for example, return to screen B from screen C and click to select the unarranged contract ID, and the maintenance plan estimation unit 111 will continue the inspection schedule planning process. It is possible that there are insufficient available engineer man-hours for the number of scheduled inspection cases, in which case the work described below will proceed.

[0070] If there are insufficient engineer slots for the number of scheduled inspections, for example, screen B may display "Expected shortage of engineer slots" or screen C may display "Insufficient engineer slots, no vacant slots." After the user confirms the shortage of engineer slots, when processing is performed by the maintenance plan estimation unit 111, all maintenance plans are adjusted to be able to be arranged within the range according to the engineers' available man-hours, and a schedule proposal is created. After that, it is possible to check the details on screen C and screen D as necessary. In principle, arrangements are made according to the priority of each project determined on screen B.

[0071] When determining the inspection schedule, person in charge, and type of inspection, the following conditions are incorporated, for example: Regarding the inspection schedule, only inspections with higher priority are scheduled for the originally scheduled month, and inspections with lower priority are rescheduled to the following month or later. Furthermore, the timing of inspection is determined based on the operational information of the operating equipment 30 and periods when it is prone to breakdowns based on past performance. Regarding the type of inspection, remote inspections are prioritized during the first year of operation of the equipment, when the contract begins, when there is a low possibility of breakdowns. During the customer's busy season, remote inspections are used instead to reduce commitments or machine downtime. Once a year, an engineer is dispatched to the customer's location to perform maintenance. Furthermore, the timing of engineer dispatch inspections and remote inspections is reversed after taking each condition into consideration. Note that these conditions are merely examples, and any other conditions may be used.

[0072] Screen D shown in FIG. 12 displays schedule information for each engineer. For example, while the user is viewing screen C, the maintenance plan estimation unit 111 may output the schedule information for each engineer to the terminal device 40. The terminal device 40 may display the schedule information as a pop-up screen on the display unit 41. For example, when the user wants to check information related to other engineers in the maintenance schedule planning process, the user can check it as needed. As shown in FIG. 12, the user's input fields are the display year and month and the target engineer's name. For example, the user can check the schedule status of engineers for each schedule within their jurisdiction in a list format on screen 731, which displays vacation, internal work, and available slots numerically; screen 732, which displays each engineer's schedule; and screen 733, which displays a detailed list of each engineer's inspection schedule. Screen D may also display the skills of each engineer for each inspection type they can handle. Note that while the user started the operation from screen A, the operation may also be performed from screen B. The user may start the operation from screen B, for example, when no new engineer dispatch slots are secured, i.e., when the available man-hours of engineers are not reduced. Examples of cases in which an engineer's available man-hours do not decrease include changing inspection type P to inspection type Q, or changing the scheduled work time from one day to half a day. Alternatively, the user may be prompted to select the screen to operate when starting the operation. The screen display method, display processing, display form, etc. may be arbitrary as long as they achieve the effects of the present disclosure.

[0073] 13 is a diagram showing the functional configuration of the inference unit 115 included in the maintenance plan management device 10. The inference unit 115 includes a data acquisition unit 116 and a maintenance plan inference unit 117. The inference unit 115 corresponds to a part of the maintenance plan estimation unit 111.

[0074] The data acquisition unit 116 acquires inspection information and contract information corresponding to state S, and engineer schedule information. Specifically, the data acquisition unit 116 acquires the information shown on screen B in FIG. 10 and the information shown on screen D in FIG. 12. The maintenance plan inference unit 117 infers a new maintenance plan within the contract period using the trained model. Specifically, the maintenance plan inference unit 117 infers new change processing information and formulates candidate maintenance plans by inputting the inspection information and contract information corresponding to state S, among the information acquired by the data acquisition unit 116, into the trained model. The maintenance plan inference unit 117 formulates a new maintenance plan by outputting candidate maintenance plans and engineer schedule information. Specifically, the maintenance plan inference unit 117 inputs the information shown on screen B in FIG. 10 into the trained model and infers the next scheduled inspection month, which will be a candidate maintenance plan. The maintenance plan inference unit 117 outputs the next scheduled inspection month obtained by inference, the information shown in screen C in Figure 11, and the information shown in screen D in Figure 12, and thereby formulates information that corresponds to a new maintenance plan, including the next scheduled inspection date and the name of the engineer in charge.

[0075] The maintenance plan inference unit 117 stores the new maintenance plan in the maintenance plan information storage unit 123. Specifically, the maintenance plan inference unit 117 stores information including the formulated inspection date and the name of the engineer in charge in the maintenance plan information storage unit 123. The output unit 118 outputs the new maintenance plan to the terminal device 40.

[0076] Next, a process for obtaining a new maintenance plan within the contract period using the inference unit 115 will be described with reference to Fig. 14. This process is started, for example, by specifying one or more items on the inspection schedule list displayed on screen B.

[0077] The data acquisition unit 116 acquires data related to the specified project, i.e., data related to the project for which a maintenance plan is to be formulated (step S201). The data acquisition unit 116 acquires inspection information and contract information corresponding to the state S, and engineer schedule information stored in the maintenance plan information storage unit 123. Specifically, the data acquisition unit 116 acquires the information shown on screen B in FIG. 10 and the information shown on screen D in FIG. 12.

[0078] The maintenance plan inference unit 117 inputs the inspection information and contract information corresponding to the state S into the learned model stored in the learned model storage unit 124 (step S202), and infers new change processing information. As described above, the change processing information is information indicating the content to be replaced by a new maintenance plan. By obtaining the new change processing information, a candidate maintenance plan is formulated. The maintenance plan inference unit 117 formulates a new maintenance plan by outputting the candidate maintenance plan and engineer schedule information. The new maintenance plan is saved in the maintenance plan information storage unit 123.

[0079] Specifically, the maintenance plan inference unit 117 infers the next scheduled inspection month, which corresponds to new change processing information, by inputting the information shown on screen B in Fig. 10 into the trained model. The maintenance plan inference unit 117 formulates the next scheduled inspection date, which corresponds to a new maintenance plan, by outputting the information on the next scheduled inspection month obtained by inference, the information shown on screen C in Fig. 11, and the information shown on screen D in Fig. 12. Information including the next scheduled inspection date, inspection type, and the name of the engineer in charge, which corresponds to the new maintenance plan, is saved in the maintenance plan information storage unit 123.

[0080] The output unit 118 outputs the information on the new maintenance plan stored in the maintenance plan information storage unit 123 to the terminal device 40 (step S203).

[0081] The terminal device 40 displays the output new maintenance plan information for the contract period on the display unit 41 (step S204).

[0082] According to this embodiment, a maintenance plan can be formulated taking into account not only inspection information but also contract information, including the number of scheduled inspections within the contract period, and a new maintenance plan can be clearly indicated. This makes it possible to support the easy formulation of a maintenance plan that takes contract information into account.

[0083] (Variation) In the embodiment, a new maintenance plan is output using a trained model trained by the model generation unit 114 of the maintenance plan management device 10, but a trained model may be obtained from another maintenance plan management device, and a new maintenance plan may be output to the terminal device 40 based on this trained model.

[0084] In addition, in this embodiment, a case where reinforcement learning is applied to the learning algorithm used by the maintenance plan inference unit 117 has been described, but the present invention is not limited to this. As for the learning algorithm, other than reinforcement learning, supervised learning, unsupervised learning, semi-supervised learning, or the like can also be applied.

[0085] Furthermore, the learning algorithm used in the model generation unit 114 may be deep learning, which learns to extract the features themselves, or machine learning may be performed according to other known methods, such as neural networks, genetic programming, inductive logic programming, support vector machines, etc.

[0086] The learning unit 112 and the inference unit 115 may be connected to the maintenance plan management device 10 via a network, for example, and may be separate devices from the maintenance plan management device 10. The learning unit 112 and the inference unit 115 may also exist on a cloud server.

[0087] The model generation unit 114 may also learn a new maintenance plan using learning data acquired from multiple maintenance plan management devices 10. The model generation unit 114 may acquire learning data from multiple maintenance plan management devices 10 used in the same area, or may learn a recommended maintenance plan for the contract period using learning data collected from multiple maintenance plan management devices 10 operating independently in different areas. It is also possible to add or remove a maintenance plan management device 10 that collects learning data to or from the target at any time. Furthermore, the learning unit 112 that has learned a new maintenance plan for a certain maintenance plan management device 10 may be applied to another maintenance plan management device 10, and the new maintenance plan for the other maintenance plan management device 10 may be re-learned and updated.

[0088] In the embodiment, the remuneration calculation unit 114a increases or decreases the remuneration r depending on whether the standard deviation of the inspection intervals is equal to or greater than a predetermined threshold, but the present disclosure is not limited to this. The remuneration calculation unit 114a may increase or decrease the remuneration r depending on the variance of the inspection intervals.

[0089] Although the preferred embodiments have been described in detail above, the present invention is not limited to the above-described embodiments, and various modifications and substitutions can be made to the above-described embodiments without departing from the scope of the claims.

[0090] Various aspects of the present disclosure are summarized below as appendices.

[0091] (Appendix 1) a data acquisition unit that acquires learning data including contract information and inspection information of a maintenance target, and change processing information that indicates changes to information on a preset maintenance plan; a learning unit that generates a learned model for inferring new change processing information based on the learning data; an inference unit that infers new change process information using a trained model for inferring new change process information, thereby inferring a new maintenance plan; A maintenance plan management system comprising: (Appendix 2) the inference unit infers a new maintenance plan based on the inferred new change process information and schedule information of an engineer who will perform maintenance on the maintenance target. 1. A maintenance plan management system as described in Appendix 1. (Appendix 3) The learning unit calculates a dispersion of an execution interval representing an interval at which the maintenance is performed, and generates the trained model based on whether the calculated dispersion is equal to or greater than a predetermined threshold. 3. The maintenance plan management system of claim 1 or 2. (Appendix 4) a data acquisition unit that acquires learning data including contract information and inspection information of a maintenance target, and change processing information that indicates changes to information on a preset maintenance plan; a model generation unit that generates a trained model for inferring new change processing information based on the training data; A learning device comprising: (Appendix 5) a data acquisition unit that acquires learning data including contract information and inspection information of a maintenance target; an inference unit that infers a new maintenance plan using a trained model for inferring change processing information that represents changes to information on a preset maintenance plan; An inference device comprising: [Explanation of symbols]

[0092] 1 Maintenance plan management system, 10 Maintenance plan management device, 11 Calculation processing unit, 12 Memory unit, 30 Operating device, 31 Operation information management unit, 40 Terminal device, 41 Display unit, 100 Operating system, 110 Operation information acquisition unit, 111 Maintenance plan estimation unit, 112 Learning unit, 113, 116 Data acquisition unit, 114 Model generation unit, 114a Remuneration calculation unit, 114b Function update unit, 115 Inference unit, 117 Maintenance plan inference unit, 118 Output unit, 121 Inspection information storage unit, 122 Contract information storage unit, 123 Maintenance plan information storage unit, 124 Learned model storage unit, 701, 711 Condition input item, 702 Detailed item, 703 Schedule, 712, 721 Detailed information confirmation item, 722 Engineer availability schedule information, 731, 732, 733 Screen, 1000 bus, 1001 processor, 1002 memory, 1003 interface, 1004 secondary storage device.

Claims

1. a data acquisition unit that acquires learning data including contract information and inspection information of a maintenance target, and change processing information that indicates changes to information on a preset maintenance plan; a learning unit that generates a learned model for inferring new change processing information based on the learning data; an inference unit that infers new change process information using a trained model for inferring new change process information, thereby inferring a new maintenance plan; A maintenance plan management system comprising:

2. the inference unit infers a new maintenance plan based on the inferred new change process information and schedule information of an engineer who will perform maintenance on the maintenance target. The maintenance plan management system according to claim 1 .

3. The learning unit calculates a dispersion of an execution interval representing an interval at which the maintenance is performed, and generates the trained model based on whether the calculated dispersion is equal to or greater than a predetermined threshold. The maintenance plan management system according to claim 1 or 2.

4. a data acquisition unit that acquires learning data including contract information and inspection information of a maintenance target, and change processing information that indicates changes to information on a preset maintenance plan; a model generation unit that generates a trained model for inferring new change processing information based on the training data; A learning device comprising:

5. a data acquisition unit that acquires learning data including contract information and inspection information of a maintenance target; an inference unit that infers a new maintenance plan using a trained model for inferring change processing information that represents changes to information on a preset maintenance plan; An inference device comprising:

Images