Planning device, planning method and planning procedure

The maintenance plan of the electrolytic device is optimized through machine learning models, and the maintenance time is dynamically adjusted according to the equipment status and operating conditions, solving the problem of deterioration of the electrolytic device before regular maintenance, reducing operating costs and improving equipment reliability.

CN112805732BActive Publication Date: 2024-11-19ASAHI KASEI KOGYO KABUSHIKI KAISHA
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN201980066040.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Priority Date
2019-04-23
Filing Date
2019-10-09
Publication Date
2024-11-19
Estimated Expiration
2039-10-09

AI Technical Summary

Technical Problem

Existing electrolytic devices are prone to deterioration before regular maintenance operations, shortening the maintenance cycle will lead to an increase in operating costs.

Method used

It provides a planning device that uses machine learning to generate changes based on the equipment status and operating conditions, recommends postponing or early maintenance operations, and optimizes maintenance plans to reduce abnormal occurrence and unnecessary maintenance.

Benefits of technology

By dynamically adjusting maintenance plans, reduce equipment abnormalities, reduce operating costs, and improve equipment reliability and efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN112805732B_ABST
    Figure CN112805732B_ABST
Patent Text Reader

Abstract

There is a problem that when a device is regularly maintained, the device may deteriorate before the maintenance work, and if the maintenance work cycle is shortened, the cost will increase due to the unnecessary maintenance work. A planning device, a planning method, and a planning program are provided, wherein the planning device comprises: a state acquisition unit that acquires the state of the target device; a change plan generation unit that generates a change plan for the period of the maintenance work using a maintenance period model, wherein the maintenance period model outputs a change plan for the period of at least one maintenance work of the maintenance and replacement of the target device according to the state of the target device; and a change plan output unit that outputs the change plan.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a planning device, a planning method and a planning program. Background Art

[0002] Conventionally, there are known electrolysis devices that generate hydrogen by electrolyzing water or electrolyzing chlorine, hydrogen, and alkali metal hydroxide by electrolyzing an aqueous alkali metal chloride solution. Such electrolysis devices are regularly maintained to prevent degradation or failures caused by operation.

[0003] Problems to be solved by the invention

[0004] However, when the maintenance work of the device is performed regularly, there is a problem that the device may deteriorate before the maintenance work, and if the maintenance work cycle is shortened, the running cost of the device will increase due to the unnecessary maintenance work. Summary of the invention

[0005] In order to solve the above-mentioned problems, in a first aspect of the present invention, a planning device is provided. The planning device may include a state acquisition unit for acquiring the state of the target device. The planning device may include a change plan generation unit, which generates a change plan for the period of maintenance work using a maintenance period model, and the maintenance period model outputs a change plan for the period of at least one maintenance work of maintenance and replacement of the target device according to the state of the target device. The planning device may include a change plan output unit for outputting the change plan.

[0006] Furthermore, the maintenance period model may suggest whether the maintenance work scheduled after the preset first period can be changed in at least one of postponement and advancement based on the state of the target equipment acquired by the state acquisition unit. The change plan generation unit may suggest changing the cycle of the maintenance work on the condition that at least one of postponement and advancement of the maintenance work is recommended. The planning device may include a maintenance period model update unit that uses learning data to update the maintenance period model through learning, and the learning data includes the state of the target equipment and a target change plan for the period of performing the maintenance work. The planning device may include an operating status acquisition unit that acquires the operating status of the target equipment. The planning device may include an abnormality prediction generation unit that generates an abnormality prediction of the target equipment using an abnormality prediction model, and the abnormality prediction model predicts the occurrence of an abnormality of the target equipment based on the operating status of the target equipment. The planning device may include a maintenance plan generation unit that generates a maintenance plan for the target equipment based on the abnormality prediction of the target equipment generated by the abnormality prediction generation unit. The planning device may include a maintenance plan output unit that outputs the maintenance plan. The change plan generation unit may generate a change plan for the period of performing the maintenance work planned in the maintenance plan. The planning device may include an abnormality prediction model updating unit, which uses learning data and updates the abnormality prediction model through learning, wherein the learning data includes the operating status of the target device and the abnormality occurrence status of the target device. The planning device may include a maintenance plan model updating unit, which uses learning data and updates the maintenance plan model through learning, wherein the learning data includes the abnormality prediction of the target device and the maintenance plan of the target device to be the target. The maintenance plan generation unit may generate a maintenance plan for the target device using the maintenance plan model. The maintenance plan model may also generate a maintenance plan for the target device based on at least one of the skills, performance, and configuration of the operator performing the maintenance work. The target device may include an electrolysis device. The target device may include a hydrogen generation device that generates hydrogen by electrolysis.

[0007] In order to solve the above-mentioned problem, in a second embodiment of the present invention, a planning method is provided. A computer can obtain the state of an object device. The computer can use a maintenance period model to generate a change plan for the period of performing maintenance work, and the maintenance period model outputs a change plan for the period of at least one maintenance work of maintaining and replacing the object device according to the state of the object device. The computer can output the change plan.

[0008] In order to solve the above-mentioned problems, in a third aspect of the present invention, a planning program is provided. The planning program can be executed by a computer, so that the computer functions as a state acquisition unit that acquires the state of the target device. The planning program can be executed by a computer, so that the computer functions as a change plan generation unit that generates a change plan for the period of maintenance work using a maintenance period model, and the maintenance period model outputs a change plan for the period of at least one maintenance work of maintenance and replacement of the target device according to the state of the target device. The planning program can be executed by a computer, so that the computer functions as a change plan output unit that outputs the change plan.

[0009] In addition, the above summary of the invention does not list all the essential features of the present invention, and sub-combinations of these feature groups may also constitute other inventions. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 A system 10 according to this embodiment is shown.

[0011] Figure 2 A configuration example of the planning device 30 according to the present embodiment is shown.

[0012] Figure 3 An example of a maintenance plan generation flow of the planning device 30 according to the present embodiment is shown.

[0013] Figure 4 An example of a change proposal generation flow of the planning device 30 according to the present embodiment is shown.

[0014] Figure 5 An example of a computer 1900 that can embody the plurality of aspects of the present embodiment in whole or in part is shown. DETAILED DESCRIPTION

[0015] The present invention will be described below by way of the embodiments of the invention, but the following embodiments do not limit the invention according to the claims. Furthermore, the combination of features described in the embodiments is not necessarily required for the solution of the present invention.

[0016] Figure 1 A system 10 according to the present embodiment is shown. The system 10 generates a maintenance plan and changes the timing of maintenance work in the maintenance plan according to the state of the operating target equipment 20. The system 10 includes the target equipment 20, a maintenance management device 40, and a planning device 30.

[0017] The object device 20 is connected to the planning device 30. The object device 20 may be an electrolysis device or a system including an electrolysis device, and as an example, a hydrogen generating device that generates hydrogen by electrolysis or a system including a hydrogen generating device. The object device 20 is, for example, a hydrogen generating device that performs salt electrolysis or alkaline water electrolysis. As an example, a hydrogen generating device that performs salt electrolysis is a device that generates hydrogen by electrolysis of an aqueous solution such as an alkali metal chloride aqueous solution, and the device comprises: an anode chamber configured with an anode; a cathode chamber configured with a cathode; and a diaphragm such as an ion exchange membrane that separates the anode chamber and the cathode chamber. And, as an example, a hydrogen generating device that performs alkaline water electrolysis is a device that is configured with a diaphragm between the anode and the cathode, and generates hydrogen by electrolysis of an electrolyte such as a potassium hydroxide aqueous solution or a sodium hydroxide aqueous solution. The object device 20 is, for example, maintained to avoid abnormal operation or failure, and the abnormal operation includes a decrease in the amount of product generated per unit time generated by operation.

[0018] The maintenance management device 40 is connected to the planning device 30. The maintenance management device 40 may be a device owned by a company or operator who maintains and manages the target device 20. The operator who performs maintenance may input information on the state and maintenance of the target device 20 to the maintenance management device 40.

[0019] The planning device 30 generates and outputs a maintenance plan for the target device 20, and changes the state of the target device 20 in operation to determine the time for the maintenance work of the maintenance plan. The planning device 30 can use a model generated by machine learning to generate a maintenance plan and its change plan. The planning device 30 can output the generated maintenance plan and provide it to the maintenance management device 40, display it on the screen of the maintenance management device 40, etc. The planning device 30 can be a computer such as a personal computer, a tablet computer, a smart phone, a workstation, a server computer, or a general-purpose computer, or a computer system in which multiple computers are connected. The planning device 30 can generate a maintenance plan and a change plan by processing the CPU, GPU (Graphics Processing Unit) and / or TPU (Tensor Processing Unit) of the computer. In addition, the planning device 30 can perform various processes on the cloud provided by the server computer. The planning device 30 includes an acquisition unit 100, a storage unit 110, a learning unit 120, a generation unit 130, and an output unit 140.

[0020] The acquisition unit 100 can be connected to the object device 20, the maintenance management device 40 and the storage unit 110, and obtain parameters for learning from the object device 20 and / or the maintenance management device 40. The acquisition unit 100 can acquire information and update it at predetermined intervals. The acquisition unit 100 can acquire information at approximately the same or different intervals according to the information to be acquired, and append or update them respectively. In addition, the acquisition unit 100 can be connected to a network and acquire data via the network. In the case where at least a part of the data to be acquired is stored in an external database, the acquisition unit 100 can access the database to acquire the data. The acquisition unit 100 provides the acquired data to the storage unit 110.

[0021] The storage unit 110 is connected to the learning unit 120 and the generating unit 130, and stores information acquired by the acquiring unit 100 and information generated by the generating unit 130. The storage unit 110 can store data processed by the planning device 30. The storage unit 110 can store intermediate data, calculation results, parameters, etc. calculated (or used) in the process of generating the maintenance plan and the change plan by the planning device 30. In addition, the storage unit 110 can respond to the request of each part in the planning device 30 and provide the stored data to the request source. As an example, the storage unit 110 responds to the request of the learning unit 120 and provides the stored data to the learning unit 120.

[0022] The learning unit 120 is connected to the generating unit 130. The learning unit 120 generates one or more learning models, and learns and updates the learning models. The learning unit 120 can learn the generated learning model based on the learning data stored in the storage unit 110. The learning unit 120 can perform reinforcement learning to update the learning model. The learning unit 120 provides the updated learning model to the generating unit 130.

[0023] The generation unit 130 is connected to the output unit 140. The generation unit 130 generates an abnormality prediction, a maintenance plan, and a change plan for the target device 20 based on the learning model updated by the learning unit 120. The generation unit 130 provides the generated maintenance plan and change plan for the target device 20 to the output unit 140. In addition, the generation unit 130 may store at least one of the generated abnormality prediction, maintenance plan, and change plan in the storage unit 110.

[0024] Here, the abnormal prediction of the object device 20 includes the result of predicting the occurrence of abnormal action of the object device 20 in the future. The abnormal prediction may include at least one of the occurrence date of the abnormal action in the future and the content of the abnormal action. The abnormal prediction, for example, includes at least one of the probability of abnormal action of the object device 20 in one or more future periods (for example, within a specified number of days, within a specified number of months, or within a specified number of years) and the period in which the probability of abnormal action of the object device 20 in the future exceeds a threshold. In addition, the abnormal action of the object device 20 means that the object device 20 does not operate normally. The abnormal action of the object device 20, for example, includes at least one of the following situations: the amount of product generated per unit time of the object device 20 decreases below the threshold; the object device 20 stops operating due to deterioration, failure, etc. of the parts of the object device 20; the temperature of the object device 20 exceeds the temperature threshold and becomes high temperature; and the temperature of the object device 20 is lower than the temperature threshold and becomes low temperature.

[0025] In the case where the object device 20 is a hydrogen generating device that performs electrolysis of common salt, the abnormal prediction of the object device 20 may include at least one of a voltage increase due to deterioration of the cathode or anode in the hydrogen generating device, a voltage change due to pinholes in the diaphragm separating the cathode chamber and the anode chamber, a decrease in current efficiency, a decrease in product purity, and a voltage increase and a decrease in current efficiency due to the deposition of impurities in the aqueous solution in the diaphragm.

[0026] Furthermore, in the case where the target device 20 is a hydrogen generating device that performs alkaline water electrolysis, the abnormality prediction of the target device 20 may include at least one of electrode degradation, short circuit, decreased cooling capacity, gas leakage, liquid leakage, poor adjustment valve, pipe blockage, poor pure water supply valve, electrolyte leakage, poor supply water purity, diaphragm rupture, adjustment valve degradation, increased cooling water temperature, abnormal differential pressure between the pressure of hydrogen generated on the anode side and the pressure of oxygen generated on the cathode side, increased mist amount in the rear stage of the electrolytic cell of the hydrogen generating device, and poor capture of mist in the rear stage of the electrolytic cell of the hydrogen generating device.

[0027] Furthermore, the maintenance plan includes a future plan for performing maintenance work on the target device 20. The maintenance plan includes, for example, a planned period for performing maintenance work on the target device 20, the content of the maintenance work, the devices used in the maintenance work, and at least one of the number, skills, performance, and configuration of operators performing the maintenance work. Furthermore, the maintenance work may include at least one of maintenance (e.g., repair, inspection, and maintenance of the target device 20) and replacement of the target device 20 or its parts.

[0028] Furthermore, the change proposal may be a proposal to change the period of maintenance work planned in the maintenance plan. The change proposal may suggest delaying or advancing the period of maintenance work. Furthermore, the change proposal may suggest changing the cycle of maintenance work. Furthermore, the change proposal may suggest changing the implementation period related to only part of the maintenance work planned in the maintenance plan. Furthermore, the change proposal may suggest changing the content of the maintenance work planned in the maintenance plan, the operator, the number of operators, or the configuration of operators, etc. together with the period of the maintenance work.

[0029] The output unit 140 is connected to the maintenance management device 40 , and outputs the maintenance plan and the modification plan generated by the generation unit 130 to the maintenance management device 40 .

[0030] According to the above-described planning device 30 of the present embodiment, the state of the target device 20 can be acquired and the maintenance plan can be changed according to the acquired state while generating and executing the maintenance plan of the target device 20. Next, a more specific configuration example of such a planning device 30 will be described.

[0031] Figure 2 FIG. 2 shows a configuration example of the planning device 30 according to the present embodiment. Figure 2 In the planning device 30, Figure 1 Elements of the planning device 30 according to the present embodiment shown in the figure that have substantially the same operation are denoted by the same reference numerals, and description thereof will be omitted.

[0032] The planning device 30 includes an operation status acquisition unit 200, an abnormality prediction model generation unit 210, an abnormality prediction model update unit 220, and an abnormality prediction generation unit 230. The planning device 30 predicts the occurrence of abnormality in the future target device 20. The planning device 30 includes a maintenance information acquisition unit 240, a maintenance plan model generation unit 250, a maintenance plan model update unit 260, a maintenance plan generation unit 270, and a maintenance plan output unit 280. The planning device 30 generates and outputs a maintenance plan for the future target device 20. The planning device 30 includes a state acquisition unit 290, a maintenance period model generation unit 300, a maintenance period model update unit 310, a change plan generation unit 320, and a change plan output unit 330. The planning device 30 generates and outputs a change plan for the maintenance plan. Here, the storage unit 110 stores the first factor, the second factor, and the third factor acquired by the acquisition unit 100.

[0033] The first factor (abnormality prediction factor) may include information that affects the occurrence of abnormalities in the target device 20. The first factor includes operating conditions such as the operating rate of the target device 20 before the target period of abnormality prediction. The first factor may include a history of abnormal actions such as degradation of the target device 20 that occurred in the past. The first factor includes, for example, the time of occurrence of abnormal actions such as degradation, the repair period, the operating rate of the target device 20 before and after the abnormal action occurred, and the content of the abnormal action. In addition, the first factor may include information such as the replacement period of parts recommended by the parts manufacturer of the target device 20, the usage time of the parts of the target device 20, or the time after the parts are installed on the target device 20. In addition, the first factor may include self-diagnosis results of the target device 20 obtained from sensors installed on the target device 20. The first factor may include parameters representing the operating status of the target device 20 such as the production efficiency of the target device 20.

[0034] When the target device 20 is a hydrogen generator that performs electrolysis of common salt, the first factor may include at least one of a voltage value, a voltage change, a current efficiency, and a purity of a product at a cathode and / or an anode in the hydrogen generator.

[0035] Furthermore, in the case where the object device 20 is a hydrogen generating device that performs alkaline water electrolysis, the first factor may include at least one of the voltage value of the cathode and / or anode in the hydrogen generating device (e.g., voltage change, etc.), the current value, the temperature, the pressure (the pressure of the hydrogen generated on the anode side, the pressure of the oxygen generated on the cathode side, or the differential pressure between the two), the electrolyte density, the purity of the product, the flow rate of the electrolyte, the instrument air pressure, the gas temperature, the electrolyte amount (e.g., the tank liquid level, etc.), and the pH of the subsequent stage of the electrolytic cell of the hydrogen generating device (e.g., a water seal or a cleaning tower, etc.).

[0036] As an example, at least one of the voltage value and the current value of the electrode in the hydrogen generator can be used as an abnormality prediction factor including at least one of electrode degradation and short circuit. The temperature of any structure of the hydrogen generator can be used as an abnormality prediction factor including at least one of electrode degradation and reduced cooling capacity. Pressure can be used as an abnormality prediction factor including at least one of gas leakage, liquid leakage, poor regulating valve and pipe clogging. Electrolyte density can be used as an abnormality prediction factor including at least one of poor pure water supply and electrolyte leakage. The purity of the product can be used as an abnormality prediction factor including at least one of poor supply water purity and diaphragm rupture. The flow rate of the electrolyte can be used as an abnormality prediction factor including at least one of gas leakage, liquid leakage, poor regulating valve and pipe clogging. The instrument air pressure can be used as an abnormality prediction factor including at least one of pipe clogging and regulating valve degradation. The gas temperature can be used as an abnormality prediction factor including at least one of reduced cooling capacity and increased cooling water temperature. The amount of electrolyte can be used as an abnormality prediction factor including at least one of pipe clogging, regulating valve degradation and differential pressure abnormality. The pH of the subsequent stage of the electrolytic cell of the hydrogen generating device can be used as an abnormality prediction factor including at least one of an increase in the amount of mist and poor capture.

[0037] The second factor (maintenance prediction factor) may include information on maintenance of the target device 20. The second factor may include an abnormality prediction generated by the abnormality prediction generation unit 230. In addition, the second factor may include a past maintenance plan of the target device 20. The second factor may include information on the configuration of operators who can perform maintenance work on the target device 20, devices that can perform maintenance work, and replacement parts of the target device 20. In addition, the second factor may include information such as the time, period, and content of maintenance work performed on the target device 20 in the past, and changes in the operating rate of the target device 20 caused by the maintenance work. In addition, the acquisition unit 100 may acquire prediction data for predicting abnormal operation of the target device 20 from the outside, etc., and store it in the storage unit 110 as information of the second factor. In this case, the prediction data may be data for predicting the next abnormal operation to occur in a period equivalent to the period from the past operation of the target device 20 to the occurrence of abnormal operation. In addition, the prediction data may be data using the history of abnormal operation acquired as a result of operating different target devices of the same type as the prediction data of the target device 20.

[0038] The third factor (state prediction factor) may include information about the state of the object device 20 received from the object device 20 or the maintenance management device 40. The third factor may include information such as the degree of wear, fatigue, and degradation of parts of the object device 20 obtained based on the inspection results and maintenance results of the object device 20. The third factor may include input from an operator during maintenance work. The third factor may include information on the amount of product generated per unit time (generation efficiency) of the object device 20 or the operating rate of the object device 20. The third factor may include a self-diagnosis result of the object device 20 obtained from a sensor installed on the object device 20. The third factor may include the value of a state register indicating the state of the object device 20, etc. In addition, the third factor may include a maintenance plan generated by the maintenance plan generation unit 270.

[0039] When the target device 20 is a hydrogen generator that performs electrolysis of common salt, the third factor may include at least one of a voltage value, a voltage change, a current efficiency, and a purity of a product at a cathode and / or an anode in the hydrogen generator.

[0040] Furthermore, in the case where the object device 20 is a hydrogen generating device that performs alkaline water electrolysis, the third factor may include at least one of the voltage value of the cathode and / or anode in the hydrogen generating device (e.g., voltage change, etc.), the current value, the temperature, the pressure (the pressure of the hydrogen generated on the anode side, the pressure of the oxygen generated on the cathode side, or the pressure difference between the two), the electrolyte density, the purity of the product, the flow rate of the electrolyte, the instrument air pressure, the gas temperature, the electrolyte amount (e.g., the tank liquid level, etc.), and the pH of the water seal or the cleaning tower of the subsequent stage of the electrolytic cell of the hydrogen generating device.

[0041] The information of the first factor, the second factor, and the third factor may be information of a time series at approximately a certain time. The information of the first factor, the second factor, and the third factor may be added or updated simultaneously with the passage of time. For example, the information of the first factor, the second factor, and the third factor may include information provided from an external device or the like.

[0042] The operating status acquisition unit 200 is connected to the storage unit 110, acquires the operating status (first factor) of the target device 20, and stores it in the storage unit 110. The operating status acquisition unit 200 may acquire the operating status from the target device 20 or a database of the manufacturer of the target device 20.

[0043] The abnormal prediction model generation unit 210 is connected to the abnormal prediction model update unit 220. The abnormal prediction model generation unit 210 generates an abnormal prediction model that predicts the occurrence of an abnormality of the object device 20 based on the operating status of the object device 20. The abnormal prediction model generation unit 210 can use the information before the predicted object period to generate the abnormal prediction model through a process called prior learning or offline learning. The abnormal prediction model generation unit 210 generates the abnormal prediction model using, for example, regression analysis, Bayesian reasoning, neural network, Gaussian mixture model, and hidden Markov model. In addition, as long as the abnormal prediction model uses, for example, LSTM (Long short-term memory), RNN (Recurrent Neural Network) and other models with storage, it is also possible to predict abnormal actions based on the time series of the first factor. The abnormal prediction model generation unit 210 provides the generated abnormal prediction model to the abnormal prediction model update unit 220.

[0044] The abnormal prediction model updating unit 220 is connected to the abnormal prediction generating unit 230. The abnormal prediction model updating unit 220 uses the learning data and updates the abnormal prediction model through learning, and the learning data includes the operating status of the object device 20 and the abnormal occurrence status of the object device 20. The abnormal prediction model updating unit 220 can, for example, update the abnormal prediction model through learning based on the value of the first factor in the past period and the abnormal occurrence status actually generated after the past period. The abnormal prediction model updating unit 220 can, for example, update to a new abnormal prediction model through learning every predetermined first update period or according to each abnormal occurrence that actually occurs. Instead of this, the abnormal prediction model updating unit 220 can also update the abnormal prediction model based on a plurality of conditions such as the condition that the learning is performed a predetermined number of times or the condition that the error difference generated by the learning is lower than a predetermined threshold.

[0045] The abnormal prediction model updating unit 220 can learn the abnormal prediction model through a process called adaptive learning or online learning. The abnormal prediction model updating unit 220 can learn the abnormal prediction model by, for example, performing reinforcement learning using an arbitrary machine learning model as a recognition model. By performing such machine learning, the abnormal prediction model updating unit 220 can take the first factor as input and predict the abnormal action corresponding to the first factor with an accuracy corresponding to the applied model.

[0046] Preferably, the abnormal prediction model updating unit 220 also uses information that is later in time than the information of the first factor used by the abnormal prediction model generating unit 210 in generating the abnormal prediction model for learning. The abnormal prediction model updating unit 220 uses the information of the first factor updated according to the abnormal action actually generated to learn the abnormal prediction model. The abnormal prediction model updating unit 220 can perform the learning of the abnormal prediction model according to the situation in which the information of the first factor is updated. The abnormal prediction model updating unit 220 can perform one or more learnings within the first update period. The abnormal prediction model updating unit 220 provides the updated abnormal prediction model to the abnormal prediction generating unit 230.

[0047] The abnormal prediction generation unit 230 is connected to the storage unit 110. The abnormal prediction generation unit 230 generates an abnormal prediction of the object device 20 using an abnormal prediction model. The abnormal prediction generation unit 230 predicts the occurrence of abnormal operation of the object device 20 within the future predetermined period, for example, at intervals of a predetermined period. The abnormal prediction generation unit 230 predicts the occurrence of abnormal operation using the abnormal prediction model and information of the first factor. The abnormal prediction generation unit 230 predicts the abnormal operation of the object device 20 by applying information of the first factor in the period immediately before the period in which the abnormal operation should be predicted to the abnormal prediction model, for example. The abnormal prediction generation unit 230 provides the prediction result to the storage unit 110 and stores it as a second factor. In addition, the abnormal prediction generation unit 230 may also directly provide the prediction result to the maintenance plan generation unit 270.

[0048] The maintenance information acquisition unit 240 is connected to the storage unit 110 and acquires information (second factor) about the maintenance of the target device 20. The maintenance information acquisition unit 240 may acquire the maintenance status from the database of the target device 20 or a carrier that maintains the target device 20. The maintenance information acquisition unit 240 acquires the information about the maintenance of the target device 20 and stores it in the storage unit 110.

[0049] The maintenance plan model generation unit 250 is connected to the maintenance plan model update unit 260. The maintenance plan model generation unit 250 can generate a maintenance plan model based on the first factor and the second factor. The maintenance plan model can be a model that generates a maintenance plan for the target device 20 by learning and based on at least one of the abnormality prediction of the target device 20, the skills, performance, and configuration of the operator who performs the maintenance work. The maintenance plan model generation unit 250 can generate the maintenance plan model by using past information and through a learning process called ex ante learning or offline learning.

[0050] The maintenance plan model generation unit 250 generates a maintenance plan model by, for example, performing reinforcement learning using any machine learning model as a recognition model, such as regression analysis, Bayesian reasoning, neural network, Gaussian mixture model, and hidden Markov model. Furthermore, as long as LSTM, RNN, and other models with storage are used as the maintenance plan model, the maintenance plan of the target device 20 can also be predicted based on the time series of the second factor. The maintenance plan model generation unit 250 provides the generated maintenance plan model to the maintenance plan model update unit 260.

[0051] The maintenance plan model updating unit 260 is connected to the maintenance plan generating unit 270. The maintenance plan model updating unit 260 updates the maintenance plan model by learning using learning data, the learning data including the abnormality prediction of the target device 20 and the maintenance plan of the target device 20. Here, the maintenance plan of the target device 20 may be an ideal maintenance plan derived from past actual data. As an example, in the case where an abnormal operation of the target device 20 occurs before the maintenance work of the period set according to the abnormality prediction of the target device 20 in the past, the maintenance plan of the target device 20 is a maintenance plan that sets the day or several days before the occurrence of the abnormal operation as the period of the maintenance work. The maintenance plan model updating unit 260 may also use other second factors to update the maintenance plan generating model by learning.

[0052] The maintenance plan model updating unit 260 may update the maintenance plan model to a new maintenance plan model after learning, for example, every second update period that is set in advance. Alternatively, the maintenance plan model updating unit 260 may update the maintenance plan model based on a plurality of conditions, such as a condition that learning has been performed a predetermined number of times or a condition that an error difference generated by learning is lower than a predetermined threshold.

[0053] The maintenance plan model updating unit 260 can learn the maintenance plan model by a process called adaptive learning or online learning. The maintenance plan model updating unit 260 learns the maintenance plan generation model by, for example, performing reinforcement learning using an arbitrary machine learning model as a recognition model. By performing such machine learning, the maintenance plan model updating unit 260 can use the second factor as an input and predict a value corresponding to the second factor with an accuracy corresponding to the applied model.

[0054] Preferably, the maintenance plan model updating unit 260 also learns using information later in time than the information used by the maintenance plan model generating unit 250 to generate the maintenance plan model. For example, the maintenance plan model updating unit 260 learns the maintenance plan model using information on the second factor updated by actual maintenance work on the target machine 20.

[0055] The maintenance plan model updating unit 260 may learn the maintenance plan model according to the updated information of the second factor. The maintenance plan model updating unit 260 may perform learning once or multiple times during the second update period. The maintenance plan model updating unit 260 may provide the updated maintenance plan model to the maintenance plan generating unit 270.

[0056] The maintenance plan generation unit 270 is connected to the maintenance plan output unit 280. The maintenance plan generation unit 270 generates a maintenance plan for the target device 20 based on the abnormality prediction of the target device 20 generated by the abnormality prediction generation unit 230. The maintenance plan generation unit 270 may generate a maintenance plan for the target device 20 using a maintenance plan model. The maintenance plan generation unit 270 may generate a maintenance plan for the target device 20 in the target period based on the value of the second factor including the abnormality prediction of the target device 20 in the target period.

[0057] The maintenance plan generation unit 270 generates a maintenance plan for a future predetermined period, for example, every predetermined period. The maintenance plan generation unit 270 generates a maintenance plan by applying information of a second factor of a period before the future predetermined period is about to begin to a maintenance plan model. The maintenance plan generation unit 270 generates a maintenance plan for a period of several days or more than ten days, a week or several weeks, a month or several months, a year or several years, for example. As an example, the maintenance plan generation unit 270 generates a maintenance plan for N days. The maintenance plan generation unit 270 provides the generated maintenance plan to the maintenance plan output unit 280. The maintenance plan generation unit 270 may also provide the generated maintenance plan to the storage unit 110 to store it as a third factor.

[0058] The maintenance plan output unit 280 is connected to the target machine 20 . The maintenance plan output unit 280 outputs the maintenance plan generated by the maintenance plan generating unit 270 to the maintenance management device 40 .

[0059] The state acquisition unit 290 is connected to the storage unit 110 to acquire the state (third factor) of the target device 20. The state acquisition unit 290 may acquire the state information of the target device 20 from the target device 20 or a database of a carrier or the like that maintains the target device 20. The state acquisition unit 290 acquires information on the state of the target device 20 and stores it in the storage unit 110 as the third factor.

[0060] The maintenance period model generation unit 300 is connected to the maintenance period model update unit 310. The maintenance period model generation unit 300 generates a maintenance period model based on the third factor. The maintenance period model may be a model that outputs a change plan for the period of at least one maintenance operation of the maintenance and replacement of the target device 20 by learning and based on the state of the target device 20. For example, the maintenance period model is a model that recommends whether the maintenance operation scheduled after the preset first period can be changed in at least one of a postponement and an advance according to the state of the target device 20 acquired by the state acquisition unit 290.

[0061] The maintenance period model generation unit 300 can generate a maintenance period model by using past information and a learning process called prior learning or offline learning. The maintenance period model generation unit 300 generates a maintenance period model by, for example, performing reinforcement learning using any of the following machine learning models as a recognition model, such as regression analysis, Bayesian reasoning, neural network, Gaussian mixture model, and hidden Markov model. Furthermore, as long as the maintenance period model uses, for example, LSTM, RNN, and a model with other storage, it is also possible to predict the maintenance period of the target device 20 based on the time series of the third factor. The maintenance period model generation unit 300 provides the generated maintenance period model to the maintenance period model update unit 310.

[0062] The maintenance period model updating unit 310 is connected to the change plan generating unit 320. The maintenance period model updating unit 310 updates the maintenance period model by learning using the learning data, wherein the learning data includes the state of the object device 20 and the target change plan for the period of maintenance work. Here, the target change plan for the period of maintenance work can be an ideal change plan derived from past actual data. As an example, in the case where an abnormal operation of the object device 20 occurs before the maintenance operation period changed according to the change plan generated according to the state of the object device 20 in the past, the target change plan becomes a change plan for changing the period of maintenance work to the day before or a few days before the abnormal operation occurs. In addition, in the case where the operator or the like determines that the maintenance operation of the period changed according to the change plan generated according to the state of the object device 20 in the past is not necessary because the degradation of the object device 20 is small, the target change plan becomes a change plan for changing the period of maintenance work to a date later than the changed period (for example, one day or a few days later). The target change plan can be derived from at least one of the first factor, the second factor, and the third factor.

[0063] The maintenance period model updating unit 310 may update the maintenance period model to a new maintenance period model after learning, for example, every third update period that is set in advance. Alternatively, the maintenance period model updating unit 310 may update the maintenance period model based on a plurality of conditions, such as a condition that learning has been performed a predetermined number of times or a condition that an error difference generated by learning is lower than a predetermined threshold.

[0064] The maintenance period model updating unit 310 can learn the maintenance period model by a process called adaptive learning or online learning. The maintenance period model updating unit 310 learns the maintenance period model by, for example, performing reinforcement learning using an arbitrary machine learning model as a recognition model. By performing such machine learning, the maintenance period model updating unit 310 can take the third factor as input and predict a value corresponding to the third factor with an accuracy corresponding to the applied model.

[0065] Preferably, the maintenance time model updating unit 310 also learns using information later in time than the information of the third factor used by the maintenance time model generating unit 300 to generate the maintenance time model. The maintenance time model updating unit 310 learns the maintenance time model using information of the third factor updated by actual maintenance work of the target device 20 or the like.

[0066] The maintenance period model updating unit 310 may perform learning of the maintenance period model according to the updated information of the third factor. The maintenance period model updating unit 310 performs learning once or multiple times within the third update period of the maintenance period model updating unit 310. The maintenance period model updating unit 310 provides the updated maintenance period model to the modification proposal generating unit 320.

[0067] The change proposal generation unit 320 is connected to the change proposal output unit 330. The change proposal generation unit 320 generates a change proposal for the period of maintenance work of the target equipment 20 using the maintenance period model. The change proposal generation unit 320 may suggest changing the cycle of maintenance work on the condition that at least one of delaying and advancing the maintenance work is suggested. The change proposal generation unit 320 may generate a change proposal for the period of maintenance work based on the value of the third factor stored in the storage unit 110.

[0068] The change proposal generation unit 320 generates a change proposal for a future preset period, for example, every preset period. The change proposal generation unit 320 generates a change proposal by applying information of a third factor of a period immediately before the future preset period to a maintenance period model, for example. The change proposal generation unit 320 may generate a plurality of change proposals for respectively changing a plurality of periods of maintenance work in the maintenance plan. The change proposal generation unit 320 outputs the generated change proposal to the change proposal output unit 330.

[0069] The modification plan output unit 330 is connected to the target device 20. The modification plan output unit 330 outputs the modification plan generated by the modification plan generating unit 320 to the maintenance management device 40.

[0070] The planning device 30 according to the present embodiment described above generates a maintenance plan for the target equipment 20 using a model generated by learning, and changes the timing of maintenance work in the maintenance plan according to the current state of the target equipment 20. Next, the operation of the planning device 30 will be described.

[0071] Figure 3 An example of a maintenance plan generation flow of the planning device 30 according to the present embodiment is shown.

[0072] The acquisition unit 100 acquires information of the first factor and the second factor that are past trends related to the operating status of the target device 20 and the maintenance of the target device 20 (S310). The acquisition unit 100 acquires information of the first factor and the second factor from time t0 to time t1, for example. The acquisition unit 100 stores the acquired information of the first factor and the second factor in the storage unit 110. In addition, the acquisition unit 100 may directly provide the information of the first factor and the second factor to the learning unit 120 and the generating unit 130.

[0073] Next, the learning unit 120 generates a learning model (S320). The learning unit 120 generates a learning model based on the values ​​of the first factor and the second factor during the period from time t0 to time t1. For example, the abnormality prediction model generation unit 210 generates an abnormality prediction model using the value of the first factor during the period from time t0 to time t1. The maintenance plan model generation unit 250 generates a maintenance plan model using the value of the second factor during the period from time t0 to time t1.

[0074] Furthermore, the abnormality prediction model generation unit 210 and the maintenance plan model generation unit 250 may use hypothetical data based on the physical model of the target device 20 as prediction data, and compare the prediction data with actual data obtained in the past operation of the target device 20, thereby generating a maintenance plan model and an abnormality prediction model. For example, the abnormality prediction model generation unit 210 and the maintenance plan model generation unit 250 perform reinforcement learning to generate a model in such a way that the error between the prediction data and the target data derived from the past actual data is minimized (for example, 0) or is less than a preset value.

[0075] As an example, the abnormal prediction model generation unit 210 and the maintenance plan model generation unit 250 use the period of M days from time t0 to time t1 as the assumed prediction period. In addition, M days can be, for example, a period of several days, more than ten days, a week, or several weeks. Moreover, the abnormal prediction model generation unit 210 and the maintenance plan model generation unit 250 perform reinforcement learning in a manner that minimizes the error between the following two parts: the prediction result of the prediction period generated based on the values ​​of the first factor and the second factor of the period before the prediction period from time t0 to time t1; and the actual data or hypothetical data of the prediction period.

[0076] In addition, the learning model may be generated by such a learning unit 120 before the planning device 30 acquires actual data of the target device 20 as the target device 20 operates.

[0077] Next, the learning unit 120 performs adaptive learning on the generated learning model (S330). Here, the acquisition unit 100 may also acquire information on the first factor and the second factor. For example, the acquisition unit 100 acquires information on the first factor and the second factor from time t2 to time t3. In addition, the period from time t2 to time t3 is set to the period after the period from time t0 to time t1. The learning unit 120 may perform adaptive learning using the information on the first factor and the second factor newly acquired by the acquisition unit 100.

[0078] For example, the abnormal prediction model updating unit 220 performs adaptive learning on the abnormal prediction model according to the value of the first factor. The abnormal prediction model updating unit 220 may use at least one of the operating status of the object device 20 during the period from time t2 to time t3 and the abnormal occurrence status of the object device 20 to perform adaptive learning on the abnormal prediction model. The abnormal prediction model updating unit 220 may perform reinforcement learning in the following manner: the result obtained by using the abnormal prediction model to predict the abnormal action of the object device 20 during the period from time t2 to time t3 is consistent with the acquired operating status or abnormal occurrence status of the object device 20 during the period from time t2 to time t3.

[0079] As an example, the abnormal prediction model updating unit 220 uses the M-day period from time t2 to time t3 as an imaginary prediction period. In addition, M days can be, for example, a period of several days or more than ten days, a week or several weeks, a month or several months, a year or several years. As an example, the abnormal prediction model updating unit 220 performs reinforcement learning in a manner in which the error between the following two parts is the minimum error (for example, 0) or less than a preset value, and the two parts are: the prediction result of the prediction period generated according to the value of the first factor of the period before the prediction period in the period from time t2 to time t3; and the actual data of the prediction period.

[0080] Furthermore, the maintenance plan model updating unit 260 may adaptively learn the maintenance plan model according to the values ​​of the first factor and the second factor. For example, the maintenance plan model updating unit 260 may learn the maintenance plan model using learning data, wherein the learning data includes an abnormality prediction of the object device 20 during the period from time t2 to time t3 and a maintenance plan of the object device 20 that should be the target. The maintenance plan model updating unit 260 may perform reinforcement learning in a manner such that the error between the following two parts is minimized (e.g., 0) or is less than a preset value, wherein the two parts are: a result obtained by predicting the maintenance status (e.g., the period of maintenance work, etc.) of the object device 20 during the period from time t2 to time t3 using the maintenance plan model; and actual data obtained during the period from time t2 to time t3 (or a target value derived from the actual data).

[0081] As an example, the maintenance plan model updating unit 260 uses the M-day period in the period from time t2 to time t3 as a hypothetical prediction period. In addition, M days can be, for example, a period of several days or more than ten days, a week or several weeks, a month or several months, a year or several years. Moreover, the maintenance plan model updating unit 260 performs reinforcement learning in a manner in which the error between the following two parts is minimized (for example, 0) or is less than a preset value, and the two parts are: the prediction result of the maintenance status in the prediction period generated based on the first factor and the value of the second factor in the period before the prediction period in the period from time t2 to time t3; and the actual data of the prediction period (or the target value derived from the actual data).

[0082] Next, the learning unit 120 updates the learning model (S340). The learning unit 120 can update the learning model at predetermined intervals. For example, after starting the adaptive learning, the learning unit 120 continues the adaptive learning for an initial update period required for the update, then performs the initial update of the learning model, and then repeatedly updates it at regular intervals. Here, it is preferred that the initial update period is N days or more as a period planned in the maintenance plan to be generated (or a period from the output of the maintenance plan to the initial maintenance work, etc.). In addition, the regular period of repeated updates can be several hours, more than ten hours, one day, dozens of hours, or several days, etc.

[0083] For example, the abnormality prediction model updating unit 220 updates the abnormality prediction model every first update period after the initial update period. Also, the maintenance plan model updating unit 260 updates the maintenance plan model every second update period after the initial update period. As an example, the first update period and the second update period are one day, one month, or one year.

[0084] Next, the abnormal prediction generation unit 230 generates an abnormal prediction of the object device 20 using the updated abnormal prediction model (S350). For example, the abnormal prediction generation unit 230 uses the updated abnormal prediction model and the value of the first factor to predict the occurrence of abnormal action of the object device 20 from time t4 to time t5. In addition, the period from time t4 to time t5 is a period after the period from time t2 to time t3, and can be a future period at the prediction time point. As an example, the abnormal prediction generation unit 230 applies the value of the first factor of N days acquired by the acquisition unit 100 during the initial update period to the abnormal prediction model to predict the occurrence of abnormalities within N days after the initial update period. The abnormal prediction generation unit 230 can provide the generated abnormal prediction to the storage unit 110 and store it as a second factor.

[0085] Next, the maintenance plan generation unit 270 generates a maintenance plan for the target device 20 using the updated learning model (S360). The maintenance plan generation unit 270 may apply the value of the second factor including the abnormal prediction generated by the abnormal prediction generation unit 230 to the updated maintenance plan model to generate a maintenance plan from time t4 to time t5. As an example, the maintenance plan generation unit 270 applies the value of the second factor for N days acquired by the acquisition unit 100 during the initial update period to the maintenance plan model to generate a maintenance plan for N days after the initial update period.

[0086] The maintenance plan generation unit 270 may generate a maintenance plan from time t4 to time t5 in such a manner that maintenance work is performed on a date before the time point at which the abnormal operation is predicted to occur according to the abnormal prediction generated by the abnormal prediction generation unit 230. The maintenance plan generation unit 270 may generate a maintenance plan based on the type or scale of the abnormal operation predicted according to the abnormal prediction, and the maintenance plan sets at least one of the content of the maintenance work, the operator, the skills of the operator, the number of operators, the period of the maintenance work, and the replacement parts.

[0087] Furthermore, the maintenance plan generation unit 270 may generate maintenance plans for each of the plurality of target devices 20. When the plurality of target devices 20 are substantially the same, the maintenance plan generation unit 270 may generate substantially the same maintenance plans for each of the plurality of target devices 20. Furthermore, the maintenance plan generation unit 270 may generate different maintenance plans for each of the target devices 20 of different types, target devices 20 purchased at different times, target devices 20 of different manufacturers, or a combination thereof.

[0088] In this case, the maintenance plan model generation unit 250 may generate a plurality of maintenance plan models corresponding to each of the plurality of target devices 20 or each combination of the plurality of target devices 20. Furthermore, the maintenance plan model update unit 260 may learn and update the plurality of maintenance plan generation models respectively.

[0089] The output unit 140 outputs the maintenance plan generated by the maintenance plan generating unit 270 ( S370 ). Thus, a service provider or the like who performs maintenance work can maintain the target device 20 according to the maintenance plan received by the maintenance management device 40 .

[0090] When the planning device 30 continues to generate the maintenance plan after outputting the maintenance plan or after the period from time t4 to time t5 has passed (S380: No), the process returns to S330, and the learning unit 120 performs adaptive learning on the learning model. In this case, the acquisition unit 100 sequentially acquires information on the first factor and the second factor that change due to the operation of the target device 20 during the period from time t4 to time t5, and sequentially stores them in the storage unit 110. That is, the planning device 30 includes the information during the period from time t4 to time t5 in the past information, and sets the target period to be predicted to the period after the period from time t4 to time t5.

[0091] Then, the planning device 30 repeatedly performs adaptive learning of the model, updates the model as a certain period passes, generates and outputs a maintenance plan. In this way, the planning device 30 according to the present embodiment can continuously update the learning model and continuously output the maintenance plan of the target device 20 by repeatedly generating a maintenance plan for the target period of the target device 20 and operating and maintaining the target period.

[0092] In the above operation flow of the planning device 30, an example is described in which the planning device 30 is operated in a time series in the order of time t0 to t5. Here, each period may be a period that is continuous in time.

[0093] The planning device 30 according to the present embodiment can predict abnormal operation of the target equipment 20 through learning and create an appropriate maintenance plan. Next, generation of a modification plan for modifying the maintenance plan of the target equipment 20 will be described.

[0094] Figure 4 An example of a change proposal generation flow of the planning device 30 according to the present embodiment is shown.

[0095] The acquisition unit 100 acquires information of the third factor that is a past trend related to the state of the target device 20 (S410). For example, the acquisition unit 100 acquires information of the third factor from time t10 to time t11. The acquisition unit 100 stores the acquired information of the third factor in the storage unit 110. In addition, the acquisition unit 100 may directly provide the information of the third factor to the learning unit 120 and the generation unit 130.

[0096] Next, the maintenance time model generation unit 300 generates a maintenance time model (S420). The maintenance time model generation unit 300 generates a learning model based on the value of the third factor during the period from time t10 to time t11.

[0097] Furthermore, the maintenance period model generation unit 300 may generate a maintenance period model by using hypothetical data based on the physical model of the target device 20 as prediction data and comparing the prediction data with actual data acquired in the past operation of the target device 20. For example, the maintenance period model generation unit 300 performs reinforcement learning and generates a model in such a manner that the error between the prediction data and the target data derived from the past actual data is minimized (for example, 0) or is smaller than a preset value.

[0098] As an example, the maintenance period model generation unit 300 uses the M-day period from time t10 to time t11 as a hypothetical prediction period. In addition, M days may be, for example, a period of several days or more than ten days, a week or several weeks, a year or several years. Furthermore, the maintenance period model generation unit 300 performs reinforcement learning in a manner that minimizes the error between the following two parts: the prediction result of the prediction period generated based on the value of the third factor of the period before the prediction period from time t10 to time t11; and the actual data of the prediction period.

[0099] Next, the maintenance period model updating unit 310 performs adaptive learning on the generated maintenance period model (S430). Here, the acquisition unit 100 may also acquire information on the third factor. The acquisition unit 100 acquires information on the third factor from time t12 to time t13, for example. In addition, it is assumed that the period from time t12 to time t13 is set to a period later than the period from time t10 to time t11. The maintenance period model updating unit 310 may perform adaptive learning using the information on the third factor newly acquired by the acquisition unit 100.

[0100] For example, the maintenance period model updating unit 310 may learn the maintenance period model using learning data including the state of the target device 20 acquired during the period from time t12 to time t13 and the target change plan for the period of maintenance work. The maintenance period model updating unit 310 may perform reinforcement learning in such a way that the error between the following two parts is minimized (e.g., 0) or is less than a preset value, the two parts being: the result obtained by using the maintenance period model to predict the period during which maintenance of the target device 20 is required during the period from time t12 to time t13; and the actual data acquired during the period from time t12 to time t13 (or the target value derived from the actual data).

[0101] As an example, the maintenance period model updating unit 310 uses the period of M days from time t12 to time t13 as a hypothetical prediction period. In addition, M days can be, for example, a period of several days or more than ten days, a week or several weeks, a month or several months, a year or several years. Moreover, the maintenance period model updating unit 310 performs reinforcement learning in a manner in which the error between the following two parts is minimized (for example, 0) or is less than a preset value, and the two parts are: the prediction result of the period of maintenance required during the prediction period generated based on the value of the third factor of the period before the prediction period from time t12 to time t13; and the actual data of the prediction period (or the target value derived from the actual data).

[0102] Next, the maintenance period model updating unit 310 updates the maintenance period model (S440). The maintenance period model updating unit 310 can update the maintenance period model at preset intervals. For example, after starting the adaptive learning, the maintenance period model updating unit 310 continues the adaptive learning for an initial update period required for the update, then performs the initial update of the learning model, and then repeatedly updates at regular intervals. Here, it is preferred that the initial update period is more than N days, which is the period from the generation of the maintenance plan to the first maintenance work of the maintenance plan. In addition, the regular period of repeated updates can be a few hours, more than ten hours, a day, dozens of hours, or a few days.

[0103] For example, the maintenance period model updating unit 310 updates the maintenance period model every third update period after the initial update period. The first update period, the second update period, and the third update period may be different periods, or may be substantially the same period instead. As an example, the third update period is one day, one month, or one year.

[0104] Next, the change proposal generation unit 320 generates a change proposal for the period of the maintenance work planned in the maintenance plan of the target device 20 using the updated maintenance period model (S460). The change proposal generation unit 320 may apply the state of the target device 20 from time t12 to time t13, the state of the target device 20 from the last maintenance work to the present, and / or the state of the target device 20 at the current time point included in the third factor to the updated maintenance period model, and generate a change proposal for the period of the maintenance work scheduled to be at time t14 after the preset first period. For example, time t14 may be a period after the period from time t12 to time t13.

[0105] The modification proposal generation unit 320 may generate a modification proposal between consecutive maintenance operations in the maintenance plan. The modification proposal generation unit 320 may generate a modification proposal for the time of the latest maintenance operation in the maintenance plan.

[0106] For example, if it is predicted that the abnormal operation of the target device 20 will occur before the scheduled time t14 of the maintenance operation, the modification plan generation unit 320 can generate a modification plan for the maintenance operation to be performed in advance by performing the maintenance operation at the predicted abnormal operation time point or a date before the time t14. Also, if it is predicted that the abnormal operation of the target device 20 will occur after the time t14, the modification plan generation unit 320 can generate a modification plan for the maintenance operation to be performed in advance by performing the maintenance operation at a date before the predicted abnormal operation time point and after the time t14.

[0107] The change proposal output unit 330 outputs the change proposal generated by the change proposal generation unit 320 ( S370 ). This allows a business operator who owns the target device 20 or a business operator who performs maintenance work on the target device 20 to maintain the target device 20 according to the maintenance plan changed according to the change proposal.

[0108] When the planning device 30 continues to generate a change proposal after outputting the change proposal (S470: No), the process returns to S430, and the learning unit 120 performs adaptive learning on the learning model. In this case, the acquisition unit 100 acquires information on the third factor that changes due to the operation of the target device 20 and stores it in the storage unit 110.

[0109] The planning device 30 of the present embodiment described above can change the maintenance plan of the target device 20 maintained according to the maintenance plan according to the current state of the target device 20, thereby avoiding abnormal operation and unnecessary maintenance, and reducing the operating cost of the target device 20.

[0110] In addition, the planning device 30 may not include at least one of the operating status acquisition unit 200, the abnormal prediction model generation unit 210, the abnormal prediction model update unit 220, the abnormal prediction generation unit 230, the maintenance information acquisition unit 240, the maintenance plan model generation unit 250, the maintenance plan model update unit 260, the maintenance plan generation unit 270, and the maintenance plan output unit 280. In this case, the planning device 30 may generate a change proposal for changing the period of maintenance work for a maintenance plan input from an external source such as a manufacturer of the target device 20 or a maintenance plan set in advance.

[0111] Furthermore, the abnormality prediction model generation unit 210, the abnormality prediction model update unit 220, the abnormality prediction generation unit 230, the maintenance plan model generation unit 250, the maintenance plan model update unit 260, the maintenance plan generation unit 270, the maintenance period model generation unit 300, the maintenance period model update unit 310 and the change proposal generation unit 320 can use any one of the first factor, the second factor and the third factor stored in the storage unit 110 in order to generate a model, update a model, generate an abnormality prediction, generate a maintenance plan or generate a change proposal.

[0112] Furthermore, the planning device 30 may output the plan and the modification proposal to the maintenance management device 40 and display the plan and the modification proposal on the screen of the planning device 30 instead of outputting the plan and the modification proposal to the maintenance management device 40 .

[0113] Various embodiments of the present invention may be described with reference to flowcharts and block diagrams, where a block may represent: (1) a stage of a process for performing an operation, or (2) a portion of a device having the function of performing an operation. Specific stages and portions may be implemented by dedicated circuits, programmable circuits provided together with computer-readable instructions stored on a computer-readable medium, and / or processors provided together with computer-readable instructions stored on a computer-readable medium. Dedicated circuits may include digital and / or analog hardware circuits, and may also include integrated circuits (ICs) and / or discrete circuits. Programmable circuits may include reconfigurable hardware circuits, which may include logical AND, logical OR, logical XOR, logical NAND, logical NOR, and other logical operations, flip-flops, registers, field programmable gate arrays (FPGAs), programmable logic arrays (PLAs), and other memory elements.

[0114] The computer-readable medium may include any tangible device capable of storing commands executed by an appropriate device, with the result that the computer-readable medium having commands stored in the device includes a product that includes commands that can be executed in order to create a unit for performing the operations specified in the flowchart or block diagram. As an example of a computer-readable medium, electronic storage media, magnetic storage media, optical storage media, electromagnetic storage media, semiconductor storage media, etc. may be included. As a more specific example of a computer-readable medium, a floppy disk (registered trademark), a magnetic disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an electrically erasable programmable read-only memory (EEPROM), a static random access memory (SRAM), a compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a Blu-ray (RTM) disk, a memory stick, an integrated circuit card, etc. may also be included.

[0115] The computer-readable commands may include any code in source code and object code described by any combination of one or more programming languages, wherein the one or more programming languages ​​include assembly commands, instruction set architecture (ISA) commands, machine commands, machine-related commands, microcodes, firmware commands, state setting data, or object-oriented programming languages ​​such as Smalltalk, JAVA (registered trademark), C++, Python, and "C" programming language, or existing procedural programming languages ​​such as the same programming language.

[0116] The computer readable instructions can be provided to a processor or programmable circuit of a general-purpose computer, a special-purpose computer, or other programmable data processing device via a local or local area network (LAN), the Internet, or other wide area network (WAN), and the computer readable instructions are executed to create a unit for performing the operations specified in the flowchart or block diagram. Examples of processors include computer processors, processing units, microprocessors, digital signal processors, controllers, microcontrollers, etc.

[0117] Figure 5 An example of a computer 1900 that can embody multiple aspects of the present invention in whole or in part is shown. The program installed in the computer 1900 can cause the computer 1900 to function as an operation associated with the device involved in the embodiment of the present invention or one or more parts of the device; or can cause the computer 1900 to perform the operation or the one or more parts; and / or can cause the computer 1900 to perform a process involved in the embodiment of the present invention or a stage of performing the process. Such a program can be executed by the CPU 2000 in order to cause the computer 1900 to perform specific operations associated with some or all of the blocks in the flowcharts and block diagrams described in this specification.

[0118] The computer 1900 involved in this embodiment comprises: a CPU peripheral unit, which has a CPU 2000, a RAM 2020, a graphics controller 2075 and a display device 2080 interconnected via a main controller 2082; an input-output unit, which has a communication interface 2030, a hard disk drive 2040 and a DVD drive 2060 connected to the main controller 2082 via an input-output controller 2084; and a conventional (regacy) input-output unit, which has a ROM 2010, a flash drive 2050 and an input-output chip 2070 connected to the input-output controller 2084.

[0119] The main controller 2082 connects the RAM 2020 with the CPU 2000 and the graphic controller 2075, which access the RAM 2020 at a high transfer rate. The CPU 2000 operates according to the programs stored in the ROM 2010 and the RAM 2020 and controls each part. The graphic controller 2075 acquires the image data generated by the CPU 2000 and the like in the frame buffer provided in the RAM 2020, and displays it on the display device 2080. Instead of this, the graphic controller 2075 may include a frame buffer storing the image data generated by the CPU 2000 and the like.

[0120] The input / output controller 2084 connects the main controller 2082 with the communication interface 2030, the hard disk drive 2040, and the DVD drive 2060, which are relatively high-speed input / output devices. The communication interface 2030 communicates with other devices via a network in a wired or wireless manner. In addition, the communication interface functions as hardware for communication. The hard disk drive 2040 stores programs and data used by the CPU 2000 in the computer 1900. The DVD drive 2060 reads programs or data from the DVD 2095 and provides them to the hard disk drive 2040 via the RAM 2020.

[0121] In addition, the input / output controller 2084 is connected to relatively low-speed input / output devices such as the ROM 2010, the flash drive 2050, and the input / output chip 2070. The ROM 2010 stores a startup program executed when the computer 1900 is started and / or a program that depends on the hardware of the computer 1900. The flash drive 2050 reads a program or data from the flash memory 2090 and provides it to the hard disk drive 2040 via the RAM 2020. The input / output chip 2070 connects the flash drive 2050 to the input / output controller 2084, and connects various input / output devices to the input / output controller 2084 via, for example, a parallel port, a serial port, a keyboard port, a mouse port, etc.

[0122] The program provided to the hard disk drive 2040 via the RAM 2020 is stored in a storage medium such as the flash memory 2090, DVD 2095 or an IC card, and is provided by the user. The program is read out from the storage medium, installed in the hard disk drive 2040 in the computer 1900 via the RAM 2020, and executed by the CPU 2000. The information processing described in these programs is read into the computer 1900 to realize the cooperation between the software and the various types of hardware resources described above. The device or method can be constituted by realizing the operation or processing of information as the computer 1900 is used.

[0123] As an example, when the computer 1900 communicates with an external device, the CPU 2000 executes a communication program loaded on the RAM 2020, and instructs the communication interface 2030 to perform communication processing according to the processing content described in the communication program. The communication interface 2030 reads the transmission data stored in the transmission buffer area set on the storage device such as the RAM 2020, the hard disk drive 2040, the flash memory 2090 or the DVD 2095 under the control of the CPU 2000, and sends it to the network, or writes the received data received from the network to the receiving buffer area set on the storage device. In this way, the communication interface 2030 can transmit and receive data with the storage device through the DMA (Direct Memory Access) method. Instead of this, the CPU 2000 can also read data from the storage device or the communication interface 2030 as the transmission source, and write the data to the communication interface 2030 or the storage device as the transmission destination, thereby transmitting the transmission and receiving data.

[0124] Furthermore, CPU 2000 reads all or a required portion of files or databases stored in external storage devices such as hard disk drive 2040, DVD drive 2060 (DVD 2095), flash drive 2050 (flash memory 2090) into RAM 2020 through DMA transmission, and performs various processes on the data on RAM 2020. Then, CPU 2000 rewrites the processed data into the external storage device through DMA transmission. In such a process, RAM 2020 can be regarded as a device for temporarily storing the contents of the external storage device, so in this embodiment, RAM 2020 and external storage devices are collectively referred to as memory, storage unit or storage device, etc.

[0125] Various information such as various programs, data, tables, databases in this embodiment are stored in such storage devices and become information processing objects. In addition, CPU 2000 can also store a part of RAM 2020 in a cache and read and write on the cache. Even in such a way, the cache is also responsible for a part of the RAM 2020 function, so in this embodiment, except for the situation of distinguishing the representation, the cache is also included in the RAM 2020, the memory and / or the storage device.

[0126] Furthermore, the CPU 2000 performs various processing specified by the command sequence of the program, including various operations, information processing, conditional judgment, information retrieval / replacement, etc. described in this embodiment, on the data read from the RAM 2020, and rewrites it into the RAM 2020. For example, when performing conditional judgment, the CPU 2000 judges whether the various variables shown in this embodiment are greater than, less than, greater than, less than, or equal to other variables or constants, and when the conditions are met (or not met), it branches to a different command sequence or calls a subroutine.

[0127] Furthermore, the CPU 2000 can search for information stored in a file or database in the storage device. For example, when the storage device stores a plurality of entries in which the attribute values ​​of the second attribute are respectively associated with the attribute values ​​of the first attribute, the CPU 2000 searches for an entry that matches the condition for specifying the attribute value of the first attribute from the plurality of entries stored in the storage device, and reads the attribute value of the second attribute stored in the entry, thereby being able to obtain the attribute value of the second attribute associated with the first attribute that satisfies the specified condition.

[0128] Furthermore, when a plurality of elements are listed in the description of the embodiments, elements other than the listed elements may be used. For example, when it is stated that "X uses A, B, and C to perform Y", X may use D in addition to A, B, and C to perform Y.

[0129] The present invention has been described above using the embodiments, but the technical scope of the present invention is not limited to the scope described in the above embodiments. It should be clear to those skilled in the art that various changes or improvements can be made to the above embodiments. It is clear from the claims that the method to which such changes or improvements are made can also be included in the technical scope of the present invention.

[0130] It should be noted that the execution order of each processing such as actions, processes, steps and stages in the devices, systems, programs and methods shown in the claims, specifications and drawings can be implemented in any order as long as there is no special indication of "before...", "prior to...", etc., and the output of the previous processing is not used for the subsequent processing. Regarding the action flow in the claims, specifications and drawings, "first," "next," etc. are used for convenience, but it does not mean that it must be implemented in this order.

[0131] Explanation of symbols

[0132] 10 System

[0133] 20 Target devices

[0134] 30 Planning Devices

[0135] 40 Maintenance management device

[0136] 100 Acquisition Department

[0137] 110 Storage

[0138] 120 Learning Department

[0139] 130 Generation Department

[0140] 140 Output

[0141] 200 Operation status acquisition unit

[0142] 210 Abnormal Prediction Model Generation Department

[0143] 220 Abnormal Prediction Model Update Department

[0144] 230 Abnormal prediction generation unit

[0145] 240 Maintenance Information Acquisition Department

[0146] 250 Maintenance plan model generation department

[0147] 260 Maintenance plan model update department

[0148] 270 Maintenance plan generation department

[0149] 280 Maintenance plan output unit

[0150] 290 Status Acquisition Unit

[0151] 300 Maintenance period model generation department

[0152] 310 Maintenance period model update department

[0153] 320 Change Plan Generation Department

[0154] 330 Change Plan Output Department

[0155] 1900 Computer

[0156] 2000 CPU

[0157] 2010 ROM

[0158] 2020 RAM

[0159] 2030 Communication Interface

[0160] 2040 Hard Drive

[0161] 2050 Flash Drive

[0162] 2060 DVD drive

[0163] 2070 Input and Output Chip

[0164] 2075 Graphics Controller

[0165] 2080 display device

[0166] 2082 Main Controller

[0167] 2084 Input and Output Controller

[0168] 2090 Flash Memory

[0169] 2095 DVD

Claims

1. A planning device, comprising: a status acquisition unit that acquires the status of the target device; a maintenance period model generating unit that uses hypothetical data based on a physical model of the target equipment as prediction data, performs reinforcement learning in such a way that an error between the prediction data and target data derived from past actual data is minimized or is smaller than a preset value, and generates a maintenance period model; a change plan generating unit that generates a change plan for a period of performing maintenance work using the maintenance period model, wherein the maintenance period model outputs a change plan for a period of performing at least one of maintenance and replacement of the target device according to a state of the target device; a change proposal output unit configured to output the change proposal; as well as The maintenance period model updating unit updates the maintenance period model by learning using learning data at predetermined intervals, the learning data including a state of the target equipment and a target change plan for a period of performing the maintenance work.

2. The planning device according to claim 1, wherein: The maintenance period model suggests whether the maintenance work scheduled after a preset first period can be changed by at least one of delaying and advancing the maintenance work, based on the state of the target equipment acquired by the state acquisition unit.

3. The planning device according to claim 2, wherein: The change proposal generating unit recommends changing the cycle of the maintenance work on the condition that at least one of delaying and advancing the maintenance work is recommended.

4. The planning device according to claim 1, wherein: The planning device also has: an operating status acquisition unit configured to acquire an operating status of the target device; an abnormality prediction generating unit that generates an abnormality prediction of the target device using an abnormality prediction model, wherein the abnormality prediction model predicts the occurrence of an abnormality of the target device based on an operating condition of the target device; a maintenance plan generating unit configured to generate a maintenance plan for the target equipment based on the abnormality prediction for the target equipment generated by the abnormality prediction generating unit; as well as A maintenance plan output unit outputs the maintenance plan.

5. The planning device according to claim 4, wherein: The change proposal generating unit generates a change proposal for performing a time period of the maintenance work planned in the maintenance plan.

6. A planning device according to claim 4 or 5, wherein: The planning device further includes an abnormality prediction model updating unit that updates the abnormality prediction model through learning using learning data including an operating status of the target device and an abnormality occurrence status of the target device.

7. The planning device according to claim 4 or 5, wherein: The planning device further includes a maintenance plan model updating unit that updates the maintenance plan model by learning using learning data, wherein the learning data includes an abnormality prediction of the target equipment and a maintenance plan for the target equipment to be targeted. The maintenance plan generating unit generates a maintenance plan for the target equipment using the maintenance plan model.

8. The planning device according to claim 7, wherein: The maintenance plan model further generates a maintenance plan for the target equipment based on at least one of the skills, performance, and configuration of an operator who performs the maintenance work.

9. The planning device according to claim 1, wherein: The target equipment includes an electrolysis device.

10. The planning device according to claim 1, wherein: The target equipment includes a hydrogen generation device that generates hydrogen by electrolysis.

11. A planning method, wherein: The computer obtains the status of the target device. Using hypothetical data based on the physical model of the target equipment as prediction data, reinforcement learning is performed in such a way that the error between the prediction data and the target data derived from the past actual data is minimized or smaller than a preset value to generate a maintenance period model. The computer generates a change plan for the period of maintenance work using the maintenance period model, wherein the maintenance period model outputs a change plan for the period of at least one of maintenance and replacement of the target device according to the state of the target device, The computer outputs the change proposal, The computer updates the maintenance period model by learning using learning data including a state of the target equipment and a target change plan for the period of performing the maintenance work at predetermined intervals.

12. A recording medium storing a planning program executed by a computer, wherein: The planning program enables the computer to function as the following five parts, which are: a status acquisition unit that acquires the status of the target device; a maintenance period model generating unit that uses hypothetical data based on a physical model of the target equipment as prediction data, performs reinforcement learning in such a way that an error between the prediction data and target data derived from past actual data is minimized or is smaller than a preset value, and generates a maintenance period model; a change plan generating unit that generates a change plan for a period of performing maintenance work using the maintenance period model, wherein the maintenance period model outputs a change plan for a period of performing at least one of maintenance and replacement of the target device according to a state of the target device; a change proposal output unit configured to output the change proposal; as well as The maintenance period model updating unit updates the maintenance period model by learning using learning data at predetermined intervals, the learning data including a state of the target equipment and a target change plan for a period of performing the maintenance work.

13. A computer program product comprising a planning program executed by a computer, wherein: The planning program enables the computer to function as the following five parts, which are: a status acquisition unit that acquires the status of the target device; a maintenance period model generating unit that uses hypothetical data based on a physical model of the target equipment as prediction data, performs reinforcement learning in such a way that an error between the prediction data and target data derived from past actual data is minimized or is smaller than a preset value, and generates a maintenance period model; a change plan generating unit for generating a change plan for a period of performing maintenance work using a maintenance period model, wherein the maintenance period model outputs a change plan for a period of performing at least one of maintenance and replacement of the target device according to a state of the target device; a change proposal output unit configured to output the change proposal; as well as The maintenance period model updating unit updates the maintenance period model by learning using learning data at predetermined intervals, the learning data including a state of the target equipment and a target change plan for a period of performing the maintenance work.

Citation Information

Patent Citations

  • Maintenance support device for plant equipment

    JP2004240642A

  • Maintenance management of a machine

    US20080133178A1

  • Anomaly Detection / Diagnostic Method and Anomaly Detection / Diagnostic System

    US20140195184A1

  • Monitored machine performance as a maintenance predictor

    US20170344897A1

  • Information processing device, information processing method, and storage medium

    US20180136616A1