JP2023014860A - Learning device, inference device and maintenance work support system - Google Patents

Abstract

To obtain a learning device capable of suppressing the occurrence of a revisit to a work site of a maintenance worker for maintenance work of a household apparatus caused by a report from a user.SOLUTION: A learning device 1 includes a data acquisition part 11 for acquiring report contents from a user about the maintenance of a household apparatus, and information of use objects in the maintenance work to the report contents, and a model generation part 12 for generating a learned model to infer necessary objects for the maintenance work of the household apparatus by using learning data created on the basis of the report contents and the information of the use objects.SELECTED DRAWING: Figure 2

Description

The present disclosure relates to a learning device, an inference device, and a maintenance work support system that realize support for maintenance work of housing equipment.

Maintenance work for housing equipment includes work that occurs at predetermined timing, such as periodic inspection of equipment, and work that occurs at unforeseen timing, for example, repair work for broken equipment.

In order to efficiently perform maintenance work that occurs at unforeseen timing, a worker in charge of maintenance work (hereafter referred to as a maintenance worker) has Therefore, it is important to properly select the necessary tools and parts. If you visit the work site with the necessary tools and parts in short supply, you will have to visit the work site again, which not only lowers work efficiency but also lowers customer service quality, such as losing trust from users. Connect.

Patent Document 1 discloses a method for inferring parts required for maintenance work based on learning information including information obtained from an air conditioner in which an abnormality has occurred and the procedure and content of maintenance work performed. There is described a maintenance work support device that generates a learned parts determination model, and uses the learned parts determination model to infer the parts required for maintenance work when maintenance work is newly required. According to the maintenance work support device described in Patent Literature 1, the number of dispatches of maintenance workers can be reduced.

JP 2020-154662 A

The maintenance work support device described in Patent Literature 1 can reduce the number of dispatches of maintenance workers by inferring necessary parts based on information that can be acquired from the equipment to be maintained. However, there is a problem that necessary parts and the like cannot be specified in the case of work that occurs in response to a report from a user who has discovered a device failure.

The present disclosure has been made in view of the above, and learning that can suppress the occurrence of re-visits to the work site by maintenance workers for maintenance work of housing equipment that occurs with reports from users. Aimed at obtaining a device.

In order to solve the above-described problems and achieve an object, a learning device according to the present disclosure acquires report content from a user regarding maintenance of housing equipment and information on items used in maintenance work for the report content. a data acquisition unit, a model generation unit that generates a trained model for inferring necessary items for maintenance work of housing equipment using learning data that is generated based on the report content and the information of the items to be used; Prepare.

According to the present disclosure, it is possible to obtain a learning device capable of suppressing the maintenance worker's revisit to the work site for the maintenance work of the housing equipment caused by the report from the user. play.

1 is a diagram showing a configuration example of a maintenance work support system according to an embodiment; FIG. Diagram showing a configuration example of a learning device Diagram for explaining a neural network Flowchart showing an example of the operation of the learning device Diagram showing a configuration example of an inference device Flowchart showing an example of the operation of the inference device Diagram showing an example of hardware that implements a learning device

A learning device, an inference device, and a maintenance work support system according to embodiments of the present disclosure will be described below in detail with reference to the drawings. Elements that are common in each drawing are denoted by the same reference numerals, and overlapping descriptions are omitted. In addition, in the present disclosure, a water heater installed in a house or the like will be exemplified and explained as the housing equipment, but the housing equipment is not limited to the water heater. For example, it goes without saying that the present disclosure can be applied to air conditioning equipment such as air conditioners, ventilation fans, bathroom heaters, and air purifiers, kitchen equipment such as IH cooking heaters and microwave ovens, lighting equipment such as recessed downlights, and the like.

Embodiment.
FIG. 1 is a diagram showing a configuration example of a maintenance work support system 100 according to an embodiment. A maintenance work support system 100 includes a learning device 1 , a storage device 2 , an inference device 3 and a terminal device 4 . A maintenance work support system 100 shown in FIG. 1 supports maintenance work for water heaters. to infer Tools, parts, secondary materials, measuring instruments, and the like that maintenance workers carry with them when they go out to work sites correspond to items necessary for maintenance work.

The learning device 1 performs machine learning based on the contents of the report from the user of the water heater and the results of maintenance work for the contents of the report, and generates a learned model for inferring necessary items when performing the maintenance work. do. The content of the report from the user includes complaints about the failure of the water heater, notification of abnormality detection, and the like. The maintenance work result includes information on tools, parts, measuring instruments, etc. used in the maintenance work. The details of the learning device 1 will be described later.

The storage device 2 includes a trained model storage unit 21 that receives and holds the trained model generated by the learning device 1 .

When maintenance work for the water heater newly occurs, the inference device 3 stores the learned model stored in the learned model storage unit 21 of the storage device 2 as necessary items for performing the new maintenance work. Make inferences using models. The details of the inference device 3 will be described later.

The terminal device 4 is, for example, a personal computer, and is used for inputting report contents from the user of the water heater, notifying the maintenance worker of the inference result by the inference device 3, and inputting the maintenance work result.

<Learning phase>
Next, a learning phase in which the learning device 1 performs machine learning to generate a learned model will be described.

FIG. 2 is a diagram showing a configuration example of the learning device 1. As shown in FIG. The learning device 1 includes a data acquisition unit 11 and a model generation unit 12 . Note that the learned model storage unit 21 shown in FIG. 2 is the learned model storage unit 21 included in the storage device 2 shown in FIG.

The data acquisition unit 11 acquires the information of the user's "details of complaint about failure" and the information of "actually used items" in the maintenance work performed based on the user's "details of complaint about failure", and creates learning data. do.

The ``failure complaint contents'' are, for example, ``no hot water'', ``hot water cannot be refilled'', ``remote controller does not work'', and furthermore, ``model of hot water heater'', ``model name of hot water heater''. , "manufacturing number of water heater", "time of purchase of water heater", "date and time of failure occurrence", "customer information", etc. are also included in "contents of failure complaint". "Customer information" is the user's name, address, and the like. Among the information examples included in the "details of complaint about failure", "no hot water", "cannot replenish hot water", and "remote control does not work" are information related to the state of the water heater. If the water heater has a function to notify the outside of the occurrence of an abnormality, for example, if the water heater has a function to notify an error code indicating an abnormal state, a function to notify an abnormal state by lighting or blinking a lamp, etc. The content of notification of the occurrence of an abnormality may be added to the information regarding the state of the water heater. "Hot water heater model", "Hot water heater model name", "Hot water heater manufacturing number", and "Hot water heater purchase time" are information related to the hot water heater. Note that the "failure complaint content" may include one or more of the above information examples. Further, in the present embodiment, for the sake of convenience, it is referred to as "contents of complaints about failure", but it includes not only complaints about failures of the water heater, but also general contents of reports from users regarding maintenance of the water heater, such as requests for inspection. That is, the "details of the complaint about failure" are the contents of the report from the user regarding the maintenance of the water heater.

"Actual items to be used" include, for example, "a pressure reducing valve that constitutes a water heater", a "piping removal tool", a "seal tape for piping", a "control board that constitutes a water heater", a "tester", and an "insulating tape," "dedicated remote controller for water heater," "connection tool for power supply wiring," and "sealant for caulking." In other words, the information on "actually used items" includes information on parts actually used in maintenance work, information on tools, and information on subsidiary materials.

The data acquisition unit 11 acquires each piece of information on the “contents of the complaint about failure” and the “actual product” via the terminal device 4 . For example, the information of the “details of complaint about failure” is input to the terminal device 4 by the person in charge who received the request for the maintenance work of the water heater, and is transmitted from the terminal device 4 to the learning device 1 . The information on the “actual product” is input to the terminal device 4 by, for example, the worker who performed the maintenance work corresponding to the “contents of complaint about failure”, and is transmitted from the terminal device 4 to the learning device 1 . After the information of the "details of the complaint about failure" is input, the terminal device 4 receives the information of the "actual item to be used" in the corresponding maintenance work. Information on "things to be used" is transmitted to the learning device 1. That is, the data acquisition unit 11 simultaneously acquires each of the information on the “details of the complaint about failure” and the “actual product” from the terminal device 4 .

Note that the data acquisition unit 11 selects a text representing the content of the complaint from the user as the “details of the complaint about failure” (for example, “The hot water heater (model name XX) has stopped supplying hot water since around XX hours on month x day”). etc.) is acquired as it is, and information (date and time, model name, details of failure, etc.) to be included in the learning data may be extracted from the acquired text. The terminal device 4 instead of the data acquisition unit 11 may perform the process of extracting information to be included in the learning data from the text.

The model generation unit 12 generates learning data created based on a combination of each piece of information that constitutes the “details of the complaint about failure” output from the data acquisition unit 11 and each piece of information that constitutes the “actual product”. By using it, the relationship between the "contents of complaints about failures" from the user and the "actual items to be used" in the maintenance work corresponding to the "contents of complaints about failures" is learned. That is, the model generation unit 12 generates a learned model for inferring necessary items such as parts, tools, and secondary materials required for maintenance work corresponding to the "details of complaints about failures" from the user, based on the learning data. to generate

Here, the learning data is data in which the user's "contents of complaint about failure" and the "actual product" are associated with each other. It is data that associates each piece of information that constitutes a "use product".

The learning device 1 is used to learn necessary items for maintenance work corresponding to the user's complaint about failure of the water heater. It may be configured to operate as a device separate from the maintenance work support system for water heaters. Also, the learning device 1 may exist on a cloud server.

The learning algorithm used by the model generation unit 12 can be known algorithms such as supervised learning, unsupervised learning, and reinforcement learning. As an example, a case where a neural network is applied will be described.

The model generator 12 learns, for example, what is required for maintenance work corresponding to the user's "details of complaints about failure" by so-called supervised learning according to a neural network model. Here, supervised learning refers to a method in which input and result (label) data sets are given to the learning device 1 to learn features in the learning data, and the result is inferred from the input.

A neural network consists of an input layer made up of multiple neurons, an intermediate layer (hidden layer) made up of multiple neurons, and an output layer made up of multiple neurons. The intermediate layer may be one layer, or two or more layers.

FIG. 3 is a diagram for explaining a neural network. For example, in the case of a three-layer neural network as shown in FIG. Y (Y1-Y2) is input, and the result is further multiplied by weight W2 (w21-w26) and output from output layer Z (Z1-Z3). This output result varies depending on the values of weight W1 and weight W2.

The neural network used in the model generation unit 12 of the learning device 1 according to the present embodiment is created based on the combination of the user's "details of complaint about failure" and the "actual product" acquired by the data acquisition unit 11. In accordance with the learning data obtained, so-called supervised learning is performed to learn the necessary items for maintenance work corresponding to the user's "details of complaints about failures."

For example, it is assumed that the "details of complaint about failure" include information indicating the state of "no hot water".

In such a case, if the cause of the "no hot water" state is a failure of the pressure reducing valve built into the hot water heater, the items required for maintenance work are "the pressure reducing valve that constitutes the water heater, Piping Removal Tool, Piping Sealing Tape". In other words, during maintenance work in this case, it is necessary to remove the pressure reducing valve and repair the failure, so a pipe removal tool is required. are required, so these are also included in the necessary items.

On the other hand, if the cause of the "no hot water" condition is freezing of the piping inside the water heater due to a drop in temperature, and simply no hot water or cold water, then the configuration of the necessary items described above is insufficient. becomes. In such a case, a factor related to the decrease in outside air temperature is added to the learning data. By doing this, if there is a "defect complaint" that "hot water does not come out when the outside temperature is low", there is a possibility that the pipes are simply frozen in addition to the pressure reducing valve, and the hot water heater The model generation unit 12 can learn that the internal pipe built in the is broken due to freezing, repair to replace this internal pipe is required, and the internal pipe is required as a replacement part. Factors relating to the drop in outside air temperature may be the date of occurrence of the failure, the date of discovery of the failure, or the like. If the date of occurrence of the failure or the date of discovery of the failure is known, it is possible to determine whether the failure occurred during a period when the possibility of freezing of the piping is high. It is possible to realize a learned model that can infer whether or not preparation of tools, parts, secondary materials, etc.) is necessary.

As another example, it is assumed that the "description of complaint about failure" includes information indicating the state that "remote control does not work".

In such a case, if the cause of the "remote control does not work" state is a failure of the dedicated remote control for the hot water heater itself, the items required for maintenance work are "the dedicated remote control for the hot water heater and the connection of the power supply wiring." Tools, sealants for caulking”. In other words, in maintenance work in this case, it is necessary to remove and replace the remote control, so a connection tool for connecting and disconnecting the power supply wiring and, for example, a remote control installed in the bathroom are water-tightly sealed against the wall. Since it is necessary to peel off the caulking and apply caulking again, a sealant for caulking (auxiliary material) is essential, so these are also included in the necessary items.

On the other hand, if the "remote control does not work" state is caused by a failure of the control board on the water heater main body side, the configuration of the necessary items described above is insufficient. In such a case, the necessary items are "the control board of the water heater, the tester, and the insulating tape."

In other words, the neural network learns by adjusting the weights W1 and W2 so that the result output from the output layer after inputting the user's "failure complaint content" to the input layer approaches the "actual product". do.

The model generating unit 12 generates and outputs a learned model by executing the learning as described above.

The learned model storage unit 21 stores the learned model output from the model generation unit 12 .

Next, the operation of the learning device 1 to generate a trained model will be described with reference to FIG. FIG. 4 is a flow chart showing an example of the operation of the learning device 1. As shown in FIG.

The learning device 1 first acquires learning data (step S11). Specifically, the data acquisition unit 11 acquires the user's "details of the complaint about failure" and the "actual product" and associates them to create learning data. Although the user's "contents of the complaint about failure" and the "actually used item" are acquired at the same time, it is sufficient if the user's "contents of complaint about failure" and the "actually used item" can be acquired in association with each other. The “details of the complaint about failure” and the “actual product” may be obtained at different timings.

The learning device 1 then performs a learning process (step S12). Specifically, the model generation unit 12 generates a so-called supervised model according to the learning data created based on the combination of the user's "failure complaint content" and the "actual product" acquired by the data acquisition unit 11. Through learning, necessary items for maintenance work corresponding to the user's "details of complaints about failure" are learned, and a learned model is generated.

The learning device 1 then stores the learned model in the learned model storage unit 21 (step S13). Specifically, the model generation unit 12 outputs the learned model generated in step S12, and the learned model storage unit 21 stores the learned model.

Here, as described above, the “details of the complaint about failure” in the learning data may include the model name information of the water heater. When the model name information is included in the learning data, it is possible to associate the model name information of the water heater to be maintained with the "actual product" for learning. Data corresponding to the difference in the tendency of maintenance work corresponding to the difference in the model name (for example, there are many repairs and replacements of part A for one model name, but there are many repairs and replacements of part B for another model name). can be accumulated, and more accurate learning becomes possible.

<Utilization phase>
Next, a utilization phase in which the inference device 3 infers necessary items for maintenance work using the learned model generated by the learning device 1 will be described.

FIG. 5 is a diagram showing a configuration example of the inference device 3. As shown in FIG. The inference device 3 includes a data acquisition unit 31 and an inference unit 32 . 5 is the learned model storage unit 21 included in the storage device 2 shown in FIG. 1, and is the same as the learned model storage unit 21 shown in FIG. FIG. 5 also shows a terminal device 4 as an example of an output destination of the inference result by the inference device 3 . The terminal device 4 includes an information notification unit 41 that notifies a maintenance worker or the like of the inference result.

The data acquisition unit 31 acquires the information of the user's "details of complaint about failure". The information of the user's “failure complaint content” acquired by the data acquisition unit 31 is the same as the user’s “failure complaint content” information acquired by the data acquisition unit 11 of the learning device 1 .

The inference unit 32 uses the learned model stored in the learned model storage unit 21 to infer what is required for maintenance work corresponding to the user's "details of complaint about failure". That is, the inference unit 32 inputs the information of the user's "failure complaint content" acquired by the data acquisition unit 31 to the learned model, and causes the user's "failure complaint content" to infer necessary items for maintenance work, Necessary items for maintenance work, which are inferred from the information of the "contents of the complaint about failure", are obtained as inference results.

In the present embodiment, the inference device 3 uses the learned model generated by the learning device 1 that constitutes the maintenance work support system 100 to output the items required for maintenance work. A trained model generated outside the system 100 may be obtained and used to output maintenance work requirements.

Next, with reference to FIG. 6, the operation of inferring necessary items for maintenance work by the inference device 3 will be described. FIG. 6 is a flow chart showing an example of the operation of the inference device 3. As shown in FIG.

The inference device 3 first acquires data used for inference of necessary items for maintenance work (step S21). Specifically, the data acquisition unit 31 acquires the information of the user's "details of complaint about failure".

The inference device 3 then inputs the data acquired in step S21 to the learned model stored in the learned model storage unit 21 (step S22) to obtain an inference result. That is, the inference unit 32 inputs the information of the user's "failure complaint content" acquired by the data acquisition unit 31 to the learned model stored in the learned model storage unit 21, and accordingly, the learned model Acquire the information of necessary items for maintenance work corresponding to the user's "details of complaint about failure" output from.

The inference device 3 then outputs the inference result (step S23), and provides the maintenance worker with information on necessary items for maintenance work (step S24). That is, the inference unit 32 outputs the inference result obtained by executing step S22 to the terminal device 4, and the information notification unit 41 notifies the maintenance worker of the items required for the maintenance work.

Here, the information notification unit 41 may notify the inference result output from the inference device 3 as it is, or may process the information created based on the inference result and notify the maintenance worker. For example, the information notification unit 41 acquires and holds information on tools, parts, and sub-materials that the maintenance worker always carries with him/herself in advance, and stores the necessary items indicated by the inference result and the information that the maintenance worker always carries with him/her. The tools, parts, and sub-materials that are available may be compared, and the result of the comparison, for example, missing items that are not carried by the maintenance worker, may be reported.

As a result, maintenance workers can carry out appropriate work on-site and respond in a single visit, thereby improving the quality of customer service. In other words, in the case of repairing a water heater by going to the site, there are cases such as "I have the parts, but I can't complete the repair because I don't have the tools, secondary materials, or measuring instruments." Other than that, there are also problems with piping and wiring, so if you don't repair this part as well, you can't confirm whether the repair is complete or not." It doesn't matter if it's a direct internal part of the hot water heater, or if it's caused by the wiring or piping connected to the hot water heater, it's only natural that you'll fix it because you called for repair. There are situations where you have no choice but to respond from a customer service point of view."

Note that the inference result by the inference device 3 is not always correct. Therefore, the information reporting unit 41 may report information that supplements the inference result together with the inference result.

As an example, in the case where the above-mentioned ``content of complaint about failure'' is ``no hot water'', the probability that the cause is the failure of the pressure reducing valve built into the water heater is based on past maintenance work results. There is a high probability of over 80%, while other internal parts of the water heater, such as the pump, have a 10% probability of failure, a mixing valve failure of 5%, and other internal piping and Consider the case where the probability of component failure is 5%. In this case, the information notifying unit 41 indicates that items (tools, parts, auxiliary materials, etc.) for coping with the failure of the pressure reducing valve built into the water heater are used with a probability of 80% or more, and that the failure of the pump 10% chance of using things to deal with mixing valve failure, 5% chance of using things to deal with other internal piping and component failures The inference result of the inference device 3 may be notified together with the fact that the object for the purpose is used with a probability of 5%.

The record data of past maintenance work may be stored in the storage device 2 described above, or may be stored in the terminal device 4 . Further, the calculation of the ratio of each cause of the symptom "no hot water" may be performed by the information notification unit 41 of the terminal device 4 or another processing unit (not shown), or may be performed inside the inference device 3. It is good also as a structure to carry out. The above ratio may be reported in association with the inference result.

In addition, it is not necessary to notify all necessary items, and for example, a certain range (fixed number) of items with the highest probability of use (hereinafter referred to as use probability) may be notified.

As a result, even if the necessary items for maintenance work obtained by reasoning are listed and displayed on a display device, etc., the importance of each necessary item cannot be conveyed to the maintenance worker, but the probability of using the necessary item is high. By sequentially providing information, the maintenance worker can select necessary items based on the probability of use.

Further, the terminal device 4 may include a notification range setting section that is means for setting a range (hereinafter referred to as a notification range) in which the information notification section 41 performs the above notification. For example, if the ``content of the complaint'' is ``no hot water'', the cause is a failure of the pressure reducing valve built into the water heater with a probability of 40%. Yes, 10% failure probability of pump, 8% failure probability of mixing valve, 8% failure probability of relief valve, 6% failure probability of heat exchanger, 4% failure probability of water level sensor, pressure Sensor failure probability is 4%, internal pipe failure probability is 4%, internal joint failure probability is 3%, hot water storage tank failure probability is 2%, water supply and drainage socket failure probability is 2%, drain plug failure probability is 2%, 2% probability of failure of three-way valve, 2% probability of failure of sterilization unit, 1% probability of failure of drain plug, 1% probability of failure of four-way valve, 1% probability of failure of built-in rubber hose. If there is, even if all of these causes are associated with the failure probability and output, the amount of information is too large and contains a lot of noise, which will confuse the maintenance worker. The above may be set so that information with a low failure probability is not output.

In this way, when extracting information about "necessary items" based on the "probability of failure", even if too much "necessary item" information is obtained, the receiving maintenance worker will be confused. By providing the setting function, it becomes possible for the maintenance worker to adjust the notification range as appropriate, thereby adjusting the amount of information on the "necessary items", thereby enhancing convenience.

In addition, when the same user makes a plurality of complaints regarding the same water heater within a predetermined short period of time, the notification range is expanded more than before the occurrence of the same. good too.

For example, if the first "details of complaint" is "no hot water", the cause is a failure of the pressure reducing valve built into the water heater, which is 40% of the past results. Below, for example, the pump failure probability is 10%, the mixing valve failure probability is 8%, the relief valve failure probability is 8%, the heat exchanger failure probability is 6%, and the water level sensor failure probability. is 4%, pressure sensor failure probability is 4%, internal pipe failure probability is 4%, internal joint failure probability is 3%, hot water storage tank failure probability is 2%, water supply and drainage socket failure probability is 2%, drainage 2% failure probability of the faucet, 2% failure probability of the three-way valve, 2% failure probability of the sterilization unit, 1% failure probability of the drain plug, 1% failure probability of the four-way valve, failure of the built-in rubber hose. Let the probability be 1%. At this time, the notification range is set to a failure probability of 10% or more. If there is a similar "defect complaint" from , it is unlikely that the part that was just replaced will break down again in a short period of time, even compared to other users (who have used the part for a long time). , it is assumed that the possibility of failure of the same parts as the previous time should be expanded and handled. Therefore, the alarm range is changed to a failure probability of 5% or more, and parts are prepared for mixing valves with a failure probability of 8%, relief valves with a failure probability of 8%, and heat exchangers with a failure probability of 6%. You should do it like this.

The fact that similar complaints have occurred multiple times in a short period of time from the same user means that it is assumed that the appropriate "necessary items" could not be selected in past maintenance work. The quality of customer service can be improved by urging customers to provide services including "necessary items", that is, "necessary items" that have a relatively low usage probability.

In addition, by including the model name information of the water heater to be maintained in the learning data, it is possible to learn the tendency of repair response corresponding to the failure tendency of each model name, so that more accurate learning and maintenance work can be performed. It is possible to infer the necessities of

For example, even if it is the same water heater, model name A, model name C, and model name F tend to have many failure repair replacements of part a and part b, and a special tool G is required for repair. For model names B and D, there is a tendency for part d to be repaired and replaced frequently, and special tools are not required for repairs. , the inference result of the necessary items for maintenance work reflects the tendency of each type name, and the accuracy rate of the necessary items for maintenance work is improved.

In this embodiment, the case where supervised learning is applied to the learning algorithm used by the model generation unit 12 of the learning device 1 has been described, but the present invention is not limited to this. In addition to supervised learning, it is also possible to apply reinforcement learning, unsupervised learning, semi-supervised learning, and the like as learning algorithms.

In addition, the model generator 12 may learn necessary items for maintenance work corresponding to the user's "details of complaints about failure" according to learning data created for a plurality of maintenance work support systems 100 . Note that the model generation unit 12 may acquire learning data from a plurality of maintenance work support systems 100 used in the same area, or may acquire learning data from a plurality of maintenance work support systems 100 operating independently in different areas. The collected data for learning may be used to learn what is required for maintenance work corresponding to the user's "details of complaints about failures." It is also possible to add or remove the maintenance work support system 100 that collects learning data from the target on the way. Furthermore, the learning device 1 that has learned the maintenance work requirements corresponding to the user's "failure complaint content" with respect to a certain maintenance work support system 100 is applied to a different maintenance work support system 100, and the other maintenance work support system 100 is applied to the maintenance work support system 100. Regarding the work support system 100, it is also possible to re-learn and update the necessary items for maintenance work corresponding to the user's "details of complaints about failure".

In addition, as the learning algorithm used in the model generation unit 12, deep learning that learns to extract the feature amount itself can be used, and other known methods such as genetic programming, functional logic programming, Machine learning may be performed according to support vector machines and the like.

Next, hardware that implements the learning device 1 will be described.

FIG. 7 is a diagram showing an example of hardware that implements the learning device 1. As shown in FIG. The learning device 1 can be implemented by the processor 91, memory 92 and interface circuit 93 shown in FIG.

The processor 91 is a CPU (Central Processing Unit, central processing unit, processor, arithmetic unit, microprocessor, microcomputer, DSP (Digital Signal Processor)), system LSI (Large Scale Integration), or the like. The memory 92 is a RAM (Random Access Memory), a ROM (Read Only Memory), an EPROM (Erasable Programmable Read Only Memory), an EEPROM (registered trademark) (Electrically Erasable Programmable Read Only Memory), a hard disk drive, or the like. The interface circuit 93 is a circuit for exchanging data between the learning device 1 and external devices such as the storage device 2 and the terminal device 4 .

The data acquisition unit 11 and the model generation unit 12 of the learning device 1 are implemented by the processor 91 executing programs for operating as these units. Programs for operating as the data acquisition unit 11 and the model generation unit 12 are stored in advance in the memory 92 . The processor 91 operates as the data acquisition unit 11 and the model generation unit 12 by reading the above program from the memory 92 and executing it.

It is assumed that programs for operating as the data acquisition unit 11 and the model generation unit 12 are pre-stored in the memory 92, but the present invention is not limited to this. The above program is written in a recording medium such as a CD (Compact Disc)-ROM, a DVD (Digital Versatile Disc)-ROM, and is supplied to the user of the learning device 1, and the user installs the above program in the memory 92. may be in the form In this case, the hardware that implements the learning device 1 further includes a reading device for reading the program from the recording medium. It is also possible to connect a reading device to the interface circuit 93 and install the program.

Although an example of hardware that implements the learning device 1 has been described, the inference device 3 can also be implemented with similar hardware. That is, the processor 91 reads out the program from the memory 92 storing the program for operating as the data acquisition unit 31 and the inference unit 32 of the inference device 3 and executes the program, so that the data acquisition unit 31 and the inference unit of the inference device 3 32 can be realized.

As described above, the maintenance work support system 100 according to this embodiment includes the learning device 1 and the reasoning device 3 . The learning device 1 provides information on the status of the housing equipment, which is included in the contents of reports from the users of the equipment such as water heaters, information on the equipment, and information on the items actually used in the maintenance work corresponding to the contents of the report. Based on this, learn the necessary items for maintenance work corresponding to the content of the report. The inference device 3 uses the learned model generated by the learning device 1 to infer the necessary items for maintenance work corresponding to the content of the report, when receiving a report on the housing equipment from the user of the housing equipment. As a result, the maintenance worker can appropriately prepare necessary items when visiting the site for maintenance work, and can suppress the occurrence of revisits to the site.

The configuration shown in the above embodiment is an example, and can be combined with another known technique, and part of the configuration can be omitted or changed without departing from the scope of the invention. It is possible.

For example, in the above embodiment, the maintenance work support system 100 includes the terminal device 4, and the learning device 1 and the reasoning device 3 acquire necessary data via the terminal device 4. It is also possible to adopt a configuration in which data is acquired without going through. In addition, the terminal device 4 notifies the inference result of the inference device 3 to the outside. It is also possible to configure Further, it is possible to adopt a configuration in which the learning device 1 or the inference device 3 is provided with the above-described learned model storage unit 21 and the storage device 2 is omitted.

1 learning device, 2 storage device, 3 inference device, 4 terminal device, 11, 31 data acquisition unit, 12 model generation unit, 21 learned model storage unit, 32 inference unit, 41 information notification unit, 100 maintenance work support system.

Claims (19)

a data acquisition unit that acquires report content from a user regarding maintenance of housing equipment and information on items used in maintenance work in response to the report content;
a model generation unit that generates a trained model for inferring necessary items for maintenance work of housing equipment, using learning data created based on the report content and the information on the items to be used;
A learning device comprising: The report content includes information on housing equipment and information on the status of housing equipment,
2. The learning device according to claim 1, wherein: The information on the state of the housing equipment includes an error code indicating an abnormal state of the housing equipment.
3. The learning device according to claim 2, wherein: The information about the housing equipment includes model name information of the housing equipment,
4. The learning device according to claim 2 or 3, characterized in that: The information on the state of the housing equipment includes the date of occurrence of the failure or the date of discovery of the failure of the equipment.
5. The learning device according to any one of claims 2 to 4, characterized in that: The items to be used are the component parts of the housing equipment replaced in the repair of the housing equipment, the tools used in the repair, the parts directly connected to the housing equipment used in the repair, and the parts used in the repair. At least one of secondary materials and measuring instruments used in the repair,
6. The learning device according to any one of claims 1 to 5, characterized in that: The housing equipment is a water heater,
7. The learning device according to any one of claims 1 to 6, characterized in that: a data acquisition unit that acquires the contents of reports from users regarding maintenance of housing equipment;
Report content acquired by the data acquisition unit using a learned model generated by a learning device that performs machine learning using report content from a user regarding maintenance of housing equipment and maintenance work results corresponding to the report content an inference unit for inferring maintenance work requirements corresponding to
An inference device characterized by comprising: The report content includes information on housing equipment and information on the status of housing equipment,
9. The reasoning apparatus according to claim 8, characterized by: The information on the state of the housing equipment includes an error code indicating an abnormal state of the housing equipment.
10. The reasoning apparatus according to claim 9, characterized by: The information about the housing equipment includes model name information of the housing equipment,
11. The reasoning device according to claim 9, characterized by: The information on the state of the housing equipment includes the date of occurrence of the failure or the date of discovery of the failure of the equipment.
12. The reasoning apparatus according to claim 9, characterized by: an information notification unit that notifies a maintenance worker of an inference result by the inference unit;
13. A reasoning apparatus according to any one of claims 8 to 12, comprising: The information notification unit notifies necessary items that are not items that are always carried by the maintenance worker, among the necessary items for maintenance work included in the inference result.
14. The reasoning apparatus according to claim 13, characterized by: The information notification unit notifies the inference result and the usage probability of the necessary items for maintenance work included in the inference result.
15. The reasoning device according to claim 13 or 14, characterized by: The information notification unit notifies necessary items included in a certain range with a high usage probability among the necessary items for maintenance work included in the inference result.
16. The reasoning apparatus according to any one of claims 13 to 15, characterized by: a notification range setting unit for setting the certain range in which the information notification unit notifies necessary items;
17. A reasoning apparatus according to claim 16, comprising: The notification range setting unit performs setting to widen the certain range when the data acquisition unit acquires the same report content from the same user multiple times within a predetermined period,
18. The reasoning apparatus according to claim 17, characterized by: a learning device according to any one of claims 1 to 7;
A reasoning device according to any one of claims 8 to 18;
A maintenance work support system comprising:

Images