JP2021096590A - Recording apparatus, system including recording apparatus, control method of recording apparatus, and program - Google Patents

Abstract

To deliver multiple kinds of consumable supplies at a time in accordance with use state of a user.SOLUTION: A recording apparatus in an embodiment of the present invention having first transmission means which transmits an inference request for inferring a timing to deliver consumable supplies, is configured to determine whether it is necessary to deliver multiple kinds of consumable supplies at a time, or not, on the basis of an inference result notification which is generated with a learned model in accordance with the inference request.SELECTED DRAWING: Figure 15

Description

The present invention relates to a technique for an automatic consumable ordering system for automatically ordering consumables for a recording device.

Image recording devices such as printers, copiers, and facsimiles require consumables such as ink and paper. When the necessary consumables in the image recording device are exhausted or scarce, the user himself goes to the store and purchases the necessary consumables. In recent years, an image recording device has become able to connect to the Internet and communicate with a WEB server. As a service for image recording devices using the Internet, the image recording device notifies the WEB server of the remaining amount of consumables, and the WEB server automatically orders the consumables from the store based on the remaining amount information of the consumables. An automatic ordering system for consumables has appeared.

In the consumables automatic ordering system described in Patent Document 1, the image recording device transmits consumables information such as ink information and paper information to the management server. Then, when the remaining amount of consumables becomes equal to or less than a predetermined threshold value, the management server sends an order request to the store server. When the dealer server receives the order request, it ships the consumables to the user.

Japanese Unexamined Patent Publication No. 2018-136638

However, there is a problem that consumables such as ink and paper arrive at the user at short intervals depending on the frequency of use of the image recording device of the user and the print contents. When it is detected that the number of consumables such as ink and paper is low and the number of consumables other than the consumables that have been ordered is low immediately after the consumables have been ordered, the number of consumables is reduced again. Since it is necessary to deliver to the same user, consumables are ordered in quick succession. Therefore, it is troublesome for the user who receives the consumables because the plurality of types of consumables are delivered in quick succession at different timings.

Therefore, one embodiment of the present invention aims to enable simultaneous delivery of a plurality of types of consumables according to a user's usage situation.

One embodiment of the present invention is a recording device having a first transmission means for transmitting an inference request for inferring the timing of delivering consumables, and is created by using a trained model in response to the inference request. It is a recording device characterized in that it is determined whether or not simultaneous delivery of a plurality of types of the consumables is necessary based on the inference result notification.

According to one embodiment of the present invention, it is possible to simultaneously deliver a plurality of types of consumables according to the usage situation of the user.

It is a figure which shows the structure of the processing system 100 in one Embodiment of this invention. It is a block diagram which shows the structure of the cloud server 200, the consumables ordering server 260, and the edge server 300. It is an external view of the smartphone 500. It is an external view of the printer 600. It is a block diagram which shows the structure of the smartphone 500. It is a block diagram which shows the structure of a printer 600. It is a figure which shows the structure of the printing part 617 in a printer 600. It is a figure which shows the screen which notifies the ink-related error, etc. displayed on the operation panel 605. It is a figure which shows the software structure of the processing system 100 in 1st Embodiment. It is a conceptual diagram which shows the input / output structure at the time of using a learning model 252 and a trained model 352. It is a figure which shows the data structure of the print job 900. It is a figure which shows the data structure of job log information and printer log information. It is a figure which shows the data structure of a learning request. It is a figure which shows the data structure of the inference request and the inference result notification. It is a figure which shows the input / output structure of the learning model 252 at the time of learning in 1st Embodiment. It is a figure which shows the input / output structure of the trained model 352 at the time of inference in the 1st Embodiment. It is a flowchart which shows the processing content of a printer 600. It is a flowchart which shows the processing content of the edge server 300 in 1st Embodiment. It is a flowchart which shows the processing content of the cloud server 200 in 1st Embodiment. It is a flowchart which shows the processing content of the consumables ordering server 260 in 1st Embodiment. It is a figure which shows the software structure of the processing system 100 in 2nd Embodiment. It is a flowchart which shows the processing content of the edge server 300 in 2nd Embodiment. It is a flowchart which shows the processing content of the cloud server 200 in 2nd Embodiment. It is a figure which shows the input / output structure of the learning model 252 at the time of learning in 3rd Embodiment.

Hereinafter, embodiments of the present invention will be described in detail. The contents described below exemplify the forms to which the present invention can be applied, and the scope of the present invention is not limited to them.

<Processing system configuration>
FIG. 1 is a diagram showing a configuration of a processing system 100 according to an embodiment of the present invention. The processing system 100 includes a cloud server 200, a consumables ordering server 260, an edge server 300, and a device 400. The edge server 300 is connected to the device 400 via the local area network 102. Further, the edge server 300 and the device 400 are each connected to the Internet 104 via a router 103 installed in the local area network 102, and can transmit and receive data to and from the cloud server 200 on the Internet 104.

The device 400 includes various devices capable of connecting to a network. Examples of the device 400 include a smartphone 500, a printer 600, a client terminal 401 such as a personal computer or a workstation, and a digital camera 402. However, the device 400 is not limited to these devices, and may include home appliances such as refrigerators, televisions, and air conditioners. These devices 400 are connected to each other via the local area network 102, and can be connected to the Internet 104 via a router 103 installed in the local area network 102. In this specification, various devices that can be connected to a network are collectively referred to as a device 400.

Here, although the router 103 is shown as a device that connects the local area network 102 and the Internet 104, it is also possible for the router 103 to have a wireless LAN access point function that constitutes the local area network 102. In this case, in addition to connecting each device 400 to the router 103 via a wired LAN, the processing system can be configured to connect to an access point via a wireless LAN and participate in the local area network 102. For example, it is possible to configure the processing system so that the printer 600 and the client terminal 401 are connected by a wired LAN, and the smartphone 500 and the digital camera 402 are connected by a wireless LAN. Each device 400 and the edge server 300 can communicate with each other with the cloud server 200 and the consumables ordering server 260 via the Internet 104 connected via the router 103. The edge server 300 and each device 400 can communicate with each other via the local area network 102. Further, the devices 400 can also communicate with each other via the local area network 102. The smartphone 500 and the printer 600 can communicate with each other by short-range wireless communication 101. As the short-range wireless communication 101, it is conceivable to use wireless communication in accordance with the Bluetooth (registered trademark) standard or the NFC standard. The smartphone 500 is also connected to the mobile phone line network 105, and can communicate with the cloud server 200 and the consumables ordering server 260 via the mobile phone line network 105.

It should be noted that the configuration described above shows an example to which the present invention can be applied, and it is possible to obtain the effect of the present invention even if the configuration is different from this example. For example, although the router 103 has an access point function, the access point may be realized by a device different from the router 103. Further, the connection between the edge server 300 and each device 400 may use a connection means other than the local area network 102. For example, wireless communication such as LPWA, ZigBee, Bluetooth, short-range wireless communication other than wireless LAN, wired connection such as USB, infrared communication, etc. may be used. Further, the edge server 300 may be on the Internet 104 instead of on the local area network 102.

<Server configuration>
FIG. 2 is a block diagram showing the hardware configurations of the cloud server 200, the consumables ordering server 260, and the edge server 300. Here, the cloud server 200 will be described as the one that uses the common hardware configuration of these servers, and the description of the consumables ordering server 260 and the edge server 300 will be omitted. The cloud server 200 includes a main board 210 that controls the entire device, a network connection unit 201, and a hard disk unit 202. The CPU 211, which is a processor arranged on the main board 210, operates according to the control program stored in the program memory 213 connected via the internal bus 212 and based on the contents stored in the data memory 214. To do.

The CPU 211 controls the network connection unit 201 via the network control circuit 215 to connect to a network such as the Internet 104 or a local area network 102 and communicate with other devices. The CPU 211 can read and write data to and from the hard disk unit 202 connected via the hard disk control circuit 216. The hard disk unit 202 stores the operating system loaded in the program memory 213 and the control software of the server 200, and also stores various data. GPU 217 is mounted on the main board 210, and various arithmetic processes can be executed instead of the CPU 211. Since GPU 217 can perform efficient calculation by processing more data in parallel, when learning is performed multiple times using a learning model such as deep learning, it is possible to perform processing on GPU 217. It is valid. Therefore, in the present embodiment, the GPU 217 is used in addition to the CPU 211 for the processing by the learning unit 251 (see FIG. 9) described later. Specifically, when executing a learning program including a learning model, learning is performed by performing calculations in cooperation with the CPU 211 and the GPU 217. The process of the learning unit 251 may be executed by performing the calculation by either the CPU 211 or the GPU 217. Further, the processing of the inference unit 351 (see FIG. 9) may also be executed by using the GPU 217 in the same manner as the learning unit 251.

Although it has been described here that the edge server 300 and the cloud server 200 have a common configuration, the embodiment of the present invention is not limited to such a configuration. For example, the cloud server 200 may be equipped with the GPU 217, but the edge server 300 may not be equipped with the GPU 217, or the cloud server 200 and the edge server 300 may be configured to use the GPU 217 having different performances. Good. Further, although the consumables ordering server 260 and the cloud server 200 have been described here as having a common configuration, the embodiment of the present invention is not limited to such a configuration. For example, since the consumables ordering server 260 does not perform learning such as deep learning, the consumables ordering server 260 may not be equipped with the GPU 217.

<Appearance of smartphone>
FIG. 3 is an external view of the smartphone 500. A smartphone is a multifunctional mobile phone equipped with a camera, an internet browser, an email function, etc. in addition to the functions of a mobile phone. The short-range wireless communication unit 501 is a unit for performing short-range wireless communication, and can communicate with a short-range wireless communication unit of a communication partner within a predetermined distance. The wireless LAN unit 502 is a unit for connecting to and communicating with the local area network 102 using a wireless LAN, and is arranged in the device. The line connection unit 503 is a unit for connecting to a mobile phone line and performing communication, and is arranged in the device. The touch panel display 504 has both an LCD display mechanism and a touch panel operation mechanism, and is provided on the front surface of the smartphone 500. A typical operation method is to display a button-shaped operation part on the touch panel display 504 and issue an event in which the button is pressed by the user performing a touch operation on the touch panel display 504. The power button 505 is used when changing the power of the smartphone 500 from on to off or from off to on.

<Appearance of printer>
FIG. 4 is an external view of the printer 600. Here, a multifunction printer (MFP) having a scanner function and other functions in addition to a print function will be described as a printer, but the embodiment of the present invention is not limited to the MFP. FIG. 4A is a perspective view showing the overall appearance of the printer 600. The document table 601 is a transparent glass-like table, which is used when a document is placed and read by a scanner. The platen pressure plate 602 is a cover for pressing the document against the platen so that the document does not float when scanning with a scanner, and for preventing outside light from entering the scanner unit. The printing paper insertion slot 603 is an insertion slot for loading various sizes of paper. The paper set here is conveyed to the printing unit one by one, and after the desired printing is performed, the paper is ejected from the printing paper ejection port 604.

FIG. 4B shows an external view of the upper surface of the printer 600. An operation panel 605 and a short-range wireless communication unit 606 are arranged above the platen pressure plate 602. The short-range wireless communication unit 606 is a unit for performing short-range wireless communication, and can communicate with a short-range wireless communication unit of a communication partner within a predetermined distance. The wireless LAN antenna 607 is an antenna for connecting to and communicating with the local area network 102 using a wireless LAN.

<Smartphone configuration>
FIG. 5 is a block diagram showing the configuration of the smartphone 500. The smartphone 500 includes a main board 510 that controls the entire device, a wireless LAN unit 502, a short-range wireless communication unit 501, a line connection unit 503, and a touch panel display 504. The CPU 511, which is a processor arranged on the main board 510, has a control program stored in a program memory 513 (specifically ROM) connected via an internal bus 512 and a data memory 514 (specifically, ROM). Operates according to the contents of RAM).

The CPU 511 controls the wireless LAN unit 502 via the wireless LAN control circuit 515 to perform wireless LAN communication with other communication terminal devices. Further, the CPU 511 controls the short-range wireless communication unit 501 via the short-range wireless communication control circuit 516 to detect a connection with another short-range wireless communication terminal or with another short-range wireless communication terminal. Data can be sent and received between. Further, the CPU 511 controls the line connection unit 503 via the line control circuit 517 to connect to the mobile phone line network 105 and perform a call and data transmission / reception.

By controlling the operation unit control circuit 518, the CPU 511 can display a desired display on the touch panel display 504 and accept an operation from the user. The CPU 511 can control the camera 519 to take an image, and stores the acquired image in the image memory 520 in the data memory 514. In addition to the image acquired by imaging, the CPU 511 stores an image acquired from the outside through a mobile phone line, a local area network 102, a short-range wireless communication 101, or the like in the image memory 520, or conversely transmits the image to the outside. It is also possible to do.

The non-volatile memory 521 is composed of a flash memory or the like, and the non-volatile memory 521 stores data to be retained even after the power is changed from on to off. Examples of such data include telephone directory data, various communication connection information, device information connected in the past, image data to be retained, application software for realizing various functions on the smartphone 500, and the like. ..

<Printer configuration>
FIG. 6 is a block diagram showing the configuration of the printer 600. The printer 600 includes a main board 610 that controls the entire device, a wireless LAN unit 608, a short-range wireless communication unit 606, and an operation panel 605. The CPU 611, which is a processor arranged on the main board 610, has a control program stored in a program memory 613 (specifically ROM) connected via an internal bus 612 and a data memory 614 (specifically, ROM). Operates according to the contents of RAM). The CPU 611 controls the scanner unit 615 to read the original, and stores the image data acquired by the reading in the image memory 616 in the data memory 614. Further, the CPU 611 can control the printing unit 617 to form (that is, print) an ink image on a recording medium such as paper based on the image data stored in the image memory 616.

The CPU 611 controls the wireless LAN unit 608 via the wireless LAN control unit 618 to perform wireless LAN communication with other communication terminal devices. Further, the CPU 611 controls the short-range wireless communication unit 606 via the short-range wireless communication control circuit 619 to detect a connection with another short-range wireless communication terminal or to connect with another short-range wireless communication terminal. Data can be sent and received between.

By controlling the operation unit control circuit 620, the CPU 611 can display the status of the printer 600 and the function selection menu on the operation panel 605, and can accept operations from the user. The operation panel 605 is provided with a backlight, and the CPU 611 can control the lighting and extinguishing of the backlight via the operation unit control circuit 620. When the backlight is turned off, the display on the operation panel 605 becomes difficult to see, but the power consumption of the printer 600 can be suppressed.

<Structure of printing unit>
FIG. 7 is a diagram showing a configuration of a printing unit 617 in the printer 600. An ink cartridge 632 having nozzle rows of black, cyan, magenta, and yellow is mounted on the carriage 631 that reciprocates on the guide shaft 630 in the directions of arrows A and B. The ink cartridge 632 is composed of a recording head and an ink tank that supplies ink to the recording head, and these are integrated. In the embodiment described here, the recording head and the ink tank are integrated, but the recording head and the ink tank may be separated from each other. In the following, the four colors of cyan, magenta, yellow, and black will be represented by one letter each as C, M, Y, and K.

By applying a drive pulse to the ink cartridge 632 while moving the carriage 631 by the carriage motor 634 connected to the carriage 631 via the belt 633, a print recording operation for one scan is performed. Further, a sensor 635 is provided at a position on the carriage 631 alongside the ink cartridge 632. By monitoring the recording medium 636 using the sensor 635 while moving the carriage 631, the result of printing and recording by the ink cartridge 632 can be read. The recording medium 636 is fed by the transfer roller 638 driven by the transfer motor 637 in a specified amount in the direction of arrow C for each print recording operation for one scan of the carriage 631, and by repeating this, one page of the recording medium 636 is displayed. The minute print recording operation is completed.

At a position deviated from the recording medium 636 in the direction of arrow A, a recovery mechanism 639 for keeping the state of the ink cartridge 632 in good condition is provided so as to face the ink cartridge 632. After the ink cartridge 632 moves the carriage 631 to a position facing the recovery mechanism 639, the recovery mechanism 639 is driven. As a result, it is possible to perform a cleaning operation such as removing clogging of the ink cartridge 632 and removing color mixing between colors. As the cleaning operation, in addition to the standard cleaning operation, a plurality of types of cleaning operations such as strong cleaning that sucks the ink in the nozzle of the ink cartridge 632 more strongly can be provided. Further, when the recording operation is not performed for a long time, the recovery mechanism 639 is brought into contact with the ink cartridge 632 to cap the nozzle portion of the ink cartridge 632 to prevent drying and solidification of the ink. This operation is referred to as "capping (operation)". These cleaning operations and capping operations are automatically executed in accordance with the operation control of the printer 600 to control the print quality to be maintained appropriately. Even if such control is performed, if the ink is faint or disturbed, the user can select and execute the maintenance operation from the operation panel.

Further, the printer 600 can detect the remaining amount of ink in the ink cartridge 632. The detection method may be various. For example, there is one in which a sensor is provided in the ink tank to directly measure the remaining amount of ink. As this sensor, an electrode or an optical sensor is used. In the method using electrodes for the sensor, the ink remaining amount is estimated by providing two electrodes in contact with the ink inside the ink tank and measuring the resistance value between the electrodes that changes according to the ink remaining amount. On the other hand, in the method using an optical sensor, an optical sensor consisting of a light source and a light receiving part is provided in the vicinity of the ink tank, light is incident on the ink in the ink tank, and the light changes depending on the remaining amount of ink and the presence or absence of ink. Detects reflected or transmitted light. As a result, the remaining amount of ink and the presence or absence of ink are detected. These detection methods using a sensor are called "sensor method". However, in the sensor method, accurate remaining amount detection may not be possible due to external factors such as electrical noise interfering with the sensor output value or the light receiving portion receiving ambient light.

Further, as a method different from the above-mentioned sensor method, the ink consumption amount is indirectly calculated by counting the number of times the ink droplets are ejected and multiplying the pre-estimated amount of ink per ejection and the number of times of ejection. It may be estimated to. This method is called a "dot count method". In the dot count method, the amount of ink per ejection differs depending on the variation in the ejection amount for each recording device and the usage environment (specifically, temperature and humidity) of the recording device, and the amount of ink changes due to ink evaporation. Since it is not supported, it may not be possible to accurately detect the remaining amount of ink. As described above, since each method has advantages and disadvantages, a plurality of remaining amount detection methods may be used in combination in order to complement each other. The specific configuration of the ink remaining amount detection will be described below.

On the lateral side of the recovery mechanism 639, an optical unit 640 including an infrared LED (light emitting element) and a phototransistor (light receiving element) constituting a reflective sensor for detecting the remaining amount of ink is provided. These light emitting elements and light receiving elements are attached so as to be arranged along the conveying direction (arrow C direction) of the recording paper. When the ink cartridge 632 is mounted on the carriage 631 and moves to the right from the position shown in FIG. 6, the ink cartridge 632 comes to be located on the optical unit 640. By doing so, the optical unit 640 can inject light toward the bottom surface of the ink cartridge 632 and detect the state of the ink cartridge 632 such as the remaining amount of ink.

Further, the printing unit 617 is provided with a dot counter (not shown) for counting the number of ink droplets ejected from the ink cartridge 632. By providing this dot counter, it is possible to count both the ink droplets ejected to form an image and the ink droplets ejected during the preliminary ejection performed as part of the recovery process. It is possible. Further, in the recording device of the present embodiment, the ink consumption amount in each ink tank in various suction recovery processes is experimentally obtained in advance, and the weight consumption when the suction recovery process is executed is managed for each ink tank. Can be done. As a result, the amount of ink consumed can be calculated by adding the value obtained by multiplying the number of dots counted by the above-mentioned dot counter by the weight of the ejected droplets and the weight consumed for the above-mentioned suction recovery process. Is possible. The recording device in the present embodiment has a configuration having both an optical unit 640 and a dot counter (not shown), but may have a configuration in which only one of them is provided.

<GUI that displays ink-related errors>
FIG. 8 shows a GUI for notifying the user of an ink-related error (specifically, an error caused by the remaining amount of ink) generated in the printer 600, specifically, an error notification screen displayed on the operation panel 605. It is a figure which shows an example. The printer 600 in the present embodiment selectively displays one of the three types of ink-related error notification screens according to the remaining amount of ink.

FIG. 8A shows an ink low warning screen 650. The ink claw warning screen 650 is a screen for notifying the user that the ink is running low when the remaining amount of ink falls below a certain threshold value (for example, 20% or less). The ink claw warning screen 650 is displayed to prompt the user to prepare a replacement ink cartridge 632 because the ink is running low. The user who visually recognizes the ink low warning screen 650 can erase the ink low warning screen 650 and continue printing by pressing the OK button 651 or replacing the ink cartridge 632.

FIG. 8B shows an inkless error screen 660. The out-of-ink error screen 660 indicates that the ink is running out when the remaining amount of ink is further reduced than when the ink low warning screen 650 is displayed and becomes below a certain threshold value (for example, 10% or less). This is a screen for notifying the user. Since this screen is running out of ink, the output image may be faint, or the recording head may have an abnormality such as ejection failure if the recording operation is performed without ink. Therefore, it is displayed to prompt the user to replace the ink cartridge 632. The user who visually recognizes the no-ink error screen 660 can erase the no-ink error screen 660 and continue printing by pressing the OK button 661 or replacing the ink cartridge 632. If the print job, copy job, or maintenance job is executed again after printing is continued, this error screen may be displayed again. By displaying the out-of-ink error screen 660 again, the user can be repeatedly notified that the ink needs to be replaced.

FIG. 8C shows an ink-free error screen 670. The no-ink error screen 670 indicates that the ink has run out when the remaining amount of ink is further reduced than when the no-ink error screen 660 is displayed and becomes below a certain threshold value (for example, 1% or less). This is a screen for notifying the user. Since this screen runs out of ink, the output image may be faint, or if the recording operation is performed without ink, an abnormality such as ejection failure may occur in the recording head. It is displayed to prompt the user to replace the ink cartridge 632. The user who visually recognizes the ink completeness error screen 670 can erase the ink completeness error screen 670 and continue printing by replacing the ink cartridge 632.

In the example shown in FIG. 8, the information for specifying which ink of CMYK the error information is related to is not shown, but the error information may be displayed by specifying the ink color. Alternatively, in the procedure for replacing the ink cartridge 632, information on the ink color in which the error has occurred may be presented to the user.

<Software configuration>
FIG. 9 is a diagram showing a software configuration of the processing system 100 (specifically, a printing system). Note that, in FIG. 9, among the software configurations of the processing system 100, only those related to the learning and inference processing in the present embodiment are described, and other software modules are omitted. For example, the operating system running on each device or server, various middleware, maintenance applications, etc. are omitted.

The cloud server 200 has a learning data generation unit 250, a learning unit 251 and a learning model 252. The learning data generation unit 250 is a module that generates learning data that can be processed by the learning unit 251 based on the data received from the outside. The learning data is a set of the input data X of the learning unit 251 and the teacher data T indicating the correct answer of the learning result (see FIG. 10). The learning unit 251 is a program module that executes learning for the learning model 252 based on the learning data received from the learning data generation unit 250. The learning model 252 accumulates the learning results performed by the learning unit 251. Here, an example of realizing the learning model 252 as a neural network will be described. By correcting the weighting parameters between each node of the neural network to appropriate ones, the input data can be classified and the evaluation value can be determined. The accumulated learning model 252 is distributed to the edge server 300 as a trained model, and is used for inference processing in the edge server 300.

The edge server 300 has a data collection / provision unit 350, an inference data generation unit 353, an inference unit 351 and a trained model 352. The data collection / provision unit 350 is a module that receives data from the device 400 and collects and stores log information. Further, the data collection / provision unit 350 passes these data to the inference data generation unit 353, or transmits the cloud server 200 as a data group to be used for learning. The inference data generation unit 353 is a module that generates inference data that can be processed by the inference unit 351 based on the data passed from the data collection / provision unit 350. The inference data is the input data X of the inference unit 351. The input data X is data in the same format as the input data X used in the learning unit 251 but the data used for inference is always the latest job data. The inference unit 351 is a program module that executes inference using the trained model 352 based on the inference data passed from the inference data generation unit 353 and returns the result to the device 400. The data sent from the device 400 to the edge server 300 is the data that becomes the input data X of the inference unit 351. The trained model 352 is used for inference performed by the edge server 300. It is assumed that the trained model 352 is also realized as a neural network like the training model 252. However, as described later, the trained model 352 may be the same as the learning model 252, or a part of the learning model 252 may be extracted and used. As the learned model 352, the learning model 252 accumulated and distributed by the cloud server 200 is stored. The trained model 352 may be the one in which the entire learning model 252 is distributed, or the training model 252 is distributed by extracting only a part necessary for inference by the edge server 300. There may be.

The device 400 has an application unit 450 and a data transmission / reception unit 451. The application unit 450 is a module that realizes various functions executed by the device 400, and is a module that uses a learning / inference mechanism by machine learning. The data transmission / reception unit 451 is a module that requests the edge server 300 to learn or infer. At the time of learning, the data transmission / reception unit 451 transmits the data used for learning to the data collection / provision unit 350 of the edge server 300 at the request of the application unit 450. At the time of inference, the data transmission / reception unit 451 transmits the data used for inference to the edge server 300 at the request of the application unit 450, receives the inference result, and returns it to the application unit 450.

In the above description, the learning model 252 trained by the cloud server 200 is distributed to the edge server 300 as the trained model 352 and used for inference, but the embodiment of the present invention is limited to this mode. It is not something that is done. Where to execute learning and inference in the cloud server 200, the edge server 300, and the device 400, respectively, may be determined according to the allocation of hardware resources, the amount of calculation, and the amount of data communication. Specifically, for example, the device 400 may hold the trained model 352 by itself and make inferences by the trained model 352 by itself. Alternatively, the system may be configured to dynamically change according to the allocation of these resources, the amount of calculation, and the increase / decrease in the amount of data communication. When the subject performing the learning and the inference are different, the inference side can reduce the capacity of the logic used only for the inference and the trained model 352, or can be configured to execute at higher speed.

<Learning model>
FIG. 10 is a conceptual diagram showing an input / output structure when the learning model 252 and the trained model 352 are used.

FIG. 10A shows the relationship between the learning model 252 and its input / output data at the time of learning. The input data X (801) is the data of the input layer of the learning model 252. Details of the input data X in this embodiment will be described later. As a result of recognizing the input data X using the learning model 252 which is a machine learning model, the output data Y (803) is output. At the time of learning, the teacher data T (802) is given as the correct answer data of the recognition result of the input data X. Therefore, by giving the output data Y and the teacher data T to the loss function 804, the amount of deviation L from the correct answer of the recognition result L. (805) is obtained. The coupling weighting coefficient between the nodes of the neural network in the training model 252 is updated so that the deviation amount L becomes smaller for a large number of training data. The error backpropagation method is a method of adjusting the coupling weighting coefficient between the nodes of each neural network so that the above error becomes small. Specific algorithms for machine learning include the nearest neighbor method, the naive Bayes method, a decision tree, and a support vector machine. In addition, deep learning (deep learning) in which features and coupling weighting coefficients for learning are generated by themselves using a neural network can also be mentioned. The present embodiment can be applied by using any of these algorithms that can be appropriately used.

FIG. 10B shows the relationship between the trained model 352 and its input / output data at the time of inference. The input data X (801) is the data of the input layer of the trained model 352. Details of the input data X in this embodiment will be described later. The output data Y (803) is output as a result of recognizing the input data X using the trained model 352 which is a machine learning model. At the time of inference, this output data Y is used as the inference result. Although the trained model 352 at the time of inference has been described as having a neural network equivalent to the learning model 252 at the time of training, it is also possible to prepare a trained model 352 in which only the parts necessary for inference are extracted. it can. This makes it possible to reduce the amount of data in the trained model 352 and shorten the neural network processing time at the time of inference. An embodiment in which the present invention is applied will be described on the premise of the configuration described above.

[First Embodiment]
In the first embodiment, a case where ink is consumed by printing, copying, maintenance, etc., the possibility that automatic ink ordering is required is estimated by machine learning, and ink is automatically ordered at an appropriate timing will be described. ..

<Print job data structure>
FIG. 11 is a diagram showing a data structure of the print job 900. The print job 900 is data transmitted to the printer 600 in order to instruct the printer 600 to execute printing from a device 400 such as a smartphone 500 or a client terminal 401. This data structure is also used to transmit information about a print job to the cloud server 200 and the edge server 300 as input data for learning.

FIG. 11A shows the overall structure of the print job. The data structure of the print job 900 includes a print job header unit 910 and a print job data unit 920. Information about the source and destination of the print job is stored in the print job header section 910. The print job data unit 920 stores information necessary for executing the print job.

FIG. 11B shows the details of the print job header section 910. The transmission date and time 911 indicates the date and time when the print job was transmitted from the transmission source. As the date and time information, information indicating the year, month, day, hour, minute, and second is used, but if necessary, time information in finer units such as milliseconds may be included. The sender identifier 912 is an identifier that uniquely identifies the sender user. This corresponds to the user's login name for the source device. The sender name 913 is an arbitrary character string used for a purpose such as displaying the name of the sender user. The source device identifier 914 is an identifier that uniquely identifies the source terminal device. IP address, MAC address, etc. correspond. The source device name 915 is an arbitrary character string used for a purpose such as displaying the name of the source device. The print job identifier 916 is an identifier for uniquely identifying the print job 900. By appropriately combining the print job identifier 916, the source device identifier 914, and the sender identifier 912, it is possible to identify one of a huge number of print jobs 900 collected in the cloud server 200. .. The destination printer identifier 917 is an identifier that uniquely identifies the destination printer 600 to which the print job 900 is transmitted. That is, the printer 600 that executes a print job and prints is shown.

FIG. 11C shows the details of the print job data unit 920. The document name 921 is an arbitrary character string indicating the name of the document to be printed by the print job. The created application name 922 is an arbitrary character string indicating the name of the application that created the print target document. The print paper size 923 is setting information regarding the paper size to be printed. Examples of the paper size include sizes used for document printing such as A3, A4, A5, B4, and B5, sizes used for photo printing such as L size and 2L size, and sizes of postcards and envelopes. The paper size is specified by a method such as allowing the user to select from the paper sizes supported by the printer 600 to be used. The print paper type 924 is setting information regarding the type of paper to be printed. Various types of paper such as plain paper, photo paper, and kraft paper can be mentioned. The paper type is specified by a method such as allowing the user to select from the paper types supported by the printer 600 to be used. The print quality setting 925 is setting information regarding the quality of the printed matter. You can specify that you want to print at high speed instead of degrading the quality, or you can print at high quality even if it takes some printing time. The number of copies setting 926 is setting information regarding the number of copies of the printed matter. By using the number of copies setting 926, it is possible to send one print target document data to print a plurality of copies. The double-sided printing setting 927 is setting information related to double-sided printing. By using the double-sided printing setting 927, it is possible to specify that the document to be printed is arranged on one side of the printing paper for printing, or arranged on both sides of the printing paper for printing. It should be noted that the above content is an example, and it is not always necessary to include all of the above items, and it is possible to include information other than the above in the print job. The print target document data 928 is data indicating the contents to be printed. In the case of printing a document, it may be text data or image data of each page constituting the document, or in the case of photo printing, it may be image data.

<Data structure of log information>
FIG. 12 is a diagram showing a data structure of job log information and printer log information. The data structure of the job log information is used to send the information of the job executed by the printer 600 to the edge server 300. The data structure of the printer log information is used for the printer 600 to send information such as ink information to the edge server 300 when the ink information is updated due to job execution, ink cartridge replacement, or the like. The job log information and printer log information sent to the edge server 300 are used as input data at the time of inference. Further, the job log information and the printer log information are also used as input data for learning on the cloud server 200.

FIG. 12A shows the contents of the print job log information 940, which is the job log information at the time of printing execution. The print job log information 940 includes the following information. The log identifier 941 is an identifier that uniquely identifies the job log information. The job type information 942 indicates the job type of the job log information, and the job type information 942 in the print job log information 940 indicates that the job type is "print". The job start date and time 943 is information indicating the start date and time of the print job, and the job end date and time 944 is information indicating the end date and time information of the print job. The printer identifier 945 is an identifier that uniquely identifies the printer 600. The print job header information 946 is information on the contents of the print job header section 910 of the executed print job 900, and the print job data information 947 is information on the contents of the print job data section 920 of the executed print job 900. Here, all the contents of the print job header section 910 and the print job data section 920 are stored in the job log information, but in order to reduce the capacity of the job log information, the job log information is limited to the necessary information. It may be configured to be stored in. In particular, since the print target document data 928 of the print job data unit 920 has a large capacity, it may not be included in the job log information. The number of color pages 948 is information on the number of pages of the color pages printed as a result of the printer 600 executing the print job 900. The number of monochrome pages 949 is information on the number of pages of monochrome pages printed as a result of the printer 600 executing the print job 900. The ink usage amount 950 is information on the amount of ink used for each color used for printing as a result of the printer 600 executing the print job 900. The value of the ink usage amount 950 may be calculated according to a predetermined mathematical formula by counting the number of ink dots ejected from the ink cartridge at the time of printing.

FIG. 12B shows the contents of the copy job log information 960, which is the job log information at the time of copy execution. The copy job log information 960 includes the following information. The log identifier 961 is an identifier that uniquely identifies the job log information. The job type information 962 indicates the job type of the job log information, and the job type information 962 in the copy job log information 960 indicates that the job type is "copy". The job start date and time 963 is information indicating the start date and time of the copy operation, and the job end date and time 964 is information indicating the end date and time of the copy operation. The printer identifier 965 is an identifier that uniquely identifies the printer 600. The paper size 966 is setting information regarding the paper size of the printing paper used at the time of copying. The paper type 967 is setting information regarding the paper type of the printing paper used at the time of copying. The print quality setting 968 is setting information regarding the quality of the printed matter used at the time of copying. The number of copies setting 969 is setting information regarding the number of copies used for copy execution. The double-sided printing setting 970 is setting information related to double-sided printing used at the time of copying. The number of color pages 971 is information on the number of pages of the color pages printed as a result of executing the copy operation. The number of monochrome pages 972 is information on the number of pages of monochrome pages printed as a result of executing the copy operation. The ink usage amount 973 is information on the ink amount for each color used as a result of executing the copy operation.

FIG. 12C shows the contents of the maintenance job log information 980, which is the job log information at the time of maintenance execution. The maintenance job log information 980 includes the following information. The log identifier 981 is an identifier that uniquely identifies the job log information. The job type information 982 indicates the job type of the job log information, and the job type information 982 in the maintenance job log information 980 indicates that the job type is "maintenance". The job start date and time 983 is information on the start date and time of the maintenance operation, and the job end date and time 984 is information on the end date and time of the maintenance operation. The printer identifier 985 is an identifier that uniquely identifies the printer 600. The maintenance type 986 indicates whether the executed maintenance operation is executed by the user selecting the maintenance execution from the operation panel, or is automatically executed instead of the user operation. The cleaning type 987 indicates the type of cleaning executed as the content of maintenance, and the possible values include each value indicating standard cleaning, strong cleaning, capping, uncapping, and the like.

FIG. 12D shows the contents of the printer log information 990, which is various information of the printer 600. The printer log information 990 includes the following information. The log identifier 991 is an identifier that uniquely identifies the job log information. The ink remaining amount (K / C / M / Y) 992 is ink remaining amount information for each ink color acquired by using the optical unit 640 or the dot counter. The final ink replacement date and time (K / C / M / Y) 993 is date and time information for replacing the ink cartridge 632 for each ink color. The usage environment 994 is information on what kind of environment the user is using the printer 600. Examples of such information include the number of users, the purpose of use (office / home, etc.), the network environment (Internet / intranet / USB, etc.) entered by the user using the operation panel 605 when the printer is installed. The user information 995 is, for example, information such as country information, age, and gender input by the user using the operation panel 605 when the printer is installed. The printer identifier 996 is an identifier that uniquely identifies the printer 600.

<Data structure of learning request>
FIG. 13 is a diagram showing a data structure of a learning request. This data structure is used when a learning request is made from the printer 600 to the edge server 300 and the cloud server 200. The learning in the present embodiment is performed when the ink information is updated by the user executing a job, replacing the ink cartridge, or the like with respect to the printer 600. Therefore, the learning request 700 includes the printer identifier 701, the log identifier 702, the presence / absence information 703 of the ink completeness error, the last ink replacement date / time (K / C / M / Y) 704, the log history information 705, and the delivery information 807. Is done. For example, when the ink information is updated with the print job, the log identifier 702 includes two log identifiers, the print job log information 940 and the printer log information 990. Further, when the ink information is updated by the ink exchange, the log identifier 702 includes the log identifier of the printer log information 990.

The presence / absence information 703 of the ink completeness error is information indicating whether or not the ink completeness error has occurred as the state of the printer 600 at the time of the learning request. When there is an ink completeness error, the ink completeness error screen 670 is displayed on the operation panel 605. Further, the last ink replacement date and time (K / C / M / Y) 704 is information on the date and time when the ink cartridge 632 was replaced for each ink color, and has the same contents as the last ink replacement date and time 993 included in the printer log information 990. is there.

The log history information 705 is used as information for acquiring the operation history from the previous learning to the time when the next learning is required, such as when reducing the learning frequency on the server in order to reduce the server load. .. For this information, it is assumed that the data group of the log identifier that specifies the log information generated between the time when the previous learning is executed and the time when the next learning is executed is passed as a table. Instead of using a table to hold the log identifier, it may be configured to store the entire log information. If it is difficult to retain and store all the log history information 705, it may be configured to store only the log history within a certain period of time. Further, when learning every time the ink information is updated, the log history information 705 is unnecessary. The delivery information 807 is information such as a delivery address and a delivery means. The user who uses the delivery system inputs information such as the delivery address at the time of service use registration, and the input information is stored in the edge server 300 or the cloud server 200.

<Data structure of inference request and inference result notification>
FIG. 14 is a diagram showing a data structure of an inference request and an inference result notification. This data structure is used when an inference request is made from the printer 600 to the edge server 300 and the cloud server 200, and the result is notified to the consumables ordering server 260. The inference request in the present embodiment is information indicating whether or not automatic ordering of ink is required when the ink information is updated by the user executing a job or exchanging an ink cartridge for the printer 600. Is done to get.

FIG. 14A shows the data structure of the inference request 710 sent from the printer 600 to the edge server 300 and the cloud server 200. The inference request 710 includes a printer identifier 711, a log identifier 712, and log history information 713. For example, when the ink information is updated with the copy job, the log identifier 712 includes two log identifiers, the copy job log information 960 and the printer log information 990. Further, when the ink information is updated by the ink exchange, the log identifier 712 includes the log identifier of the printer log information 990.

The log history information 713 is used as information for acquiring the operation history from the previous inference to the time when the next inference is required, such as when reducing the inference frequency on the server in order to reduce the server load. .. This information shall be passed as a table of log identifier data groups that specify log information that occurred between the time when the previous inference was executed and the time when the next inference was executed. Instead of using a table to hold the log identifier, it may be configured to store the entire log information. If it is difficult to retain and store all the log history information 713, it may be configured to store only the log history within a certain period of time. Further, when inferring each time the ink information is updated, the log history information 713 is unnecessary.

FIG. 14B shows the data structure of the inference result notification 720 that notifies the consumables ordering server 260 of the result of inference performed by the edge server 300 and the cloud server 200. In the inference result notification 720, in addition to the printer identifier 721 and the log identifier 722, the simultaneous delivery ink information 723 is stored as the information obtained as a result of inferring the simultaneous delivery ink. The simultaneous delivery ink information 723 is specifically information on ink to be delivered at the same time as a result of inference.

<I / O structure of learning model during learning>
FIG. 15 is a diagram showing an input / output structure of the learning model 252 at the time of learning in the present embodiment. As shown in the figure, the input data X is generated from the log information 806 stored in the server, the log identifier 702 and the log history information 705 included in the learning request 700, and the delivery information 807 passed by the learning request 700. To use. The delivery information 807 is information such as a delivery address and a delivery means. The user who uses the delivery system inputs information such as the delivery address at the time of service use registration, and the input information is stored in the edge server 300 or the cloud server 200.

In order to generate the input data X, the log information 806 stored in the edge server 300 is extracted as the log information specified by the log identifier 702 as described later. Further, as necessary, the log information 806 stored in the edge server 300 is similarly extracted as the log information specified in the log history information 705. The learning is performed in order to predict the tendency of the ink remaining amount of the printer 600 and to predict the ink to be delivered at the same time. When executing the learning performed in step S1903 (see FIG. 19) described later, it is conceivable to appropriately derive and use the parameters listed below.

Printer identifier. Since the printer identifiers 945, 965, 985, and 996 are stored in each log information, they can be acquired and included in the input data X. By including the printer identifier in the input data X, learning suitable for the printer 600 specified by the included printer identifier can be performed. Further, by not including the printer identifier in the input data X, it is possible to make a plurality of printers learn common characteristics.

Elapsed time since the last printing or maintenance was completed. This can be calculated as follows. From the log information of the printer 600, the print job log information 940, the copy job log information 960, or the maintenance job log information 980 is acquired as the information of the last executed job. Finally, using the information of the job end time (944, 964, 984) stored in the acquired print job log information 940, copy job log information 960, or maintenance job log information 980, the printing or maintenance is finally completed. Derived the elapsed time from. Specifically, the difference between the job end time and the current time may be calculated.

The number of prints and maintenance performed since the last ink change. This can be calculated as follows. The latest printer log information 990 is extracted from the log information of the printer 600. Then, the final ink replacement date and time 993 is acquired, and all the print job log information 940, copy job log information 960, and maintenance job log information 980 performed after the acquired final ink replacement date and time 993 are extracted. At the time of this extraction, the job end times 944, 964, 984 and the final ink replacement date and time 993 are compared. Then, the number of times of printing and maintenance is obtained by counting the number of the extracted print job log information 940, copy job log information 960, and maintenance job log information 980, respectively. Further, since the information is stored for each ink color (specifically, K / C / M / Y) as the final ink replacement date and time 993, the number of times may be calculated for each ink color.

The number of printed pages since the last ink change. This can be calculated as follows. The latest printer log 990 is extracted from the log information of the printer 600. Then, the final ink exchange date and time 993 is acquired, and all the print job log information 940 and the copy job log information 960 performed after the acquired final ink exchange date and time 993 are extracted. At the time of this extraction, the job end times 944 and 964 are compared with the last ink replacement date and time 993. Then, the number of print pages is calculated by calculating the sum of the values of the number of color pages 948 and 971 and the number of monochrome pages 949 and 972 stored in the extracted print job log information 940 and copy job log information 960. Ask. The number of color pages and the number of monochrome pages may be calculated and used separately. Further, since the information is stored for each ink color as the last ink replacement date and time 993, the number of pages may be calculated for each ink color.

The amount of ink used for printing since the last ink change. This can be calculated as follows. The latest printer log 990 is extracted from the log information of the printer 600. Then, the final ink exchange date and time 993 is acquired, and all the print job log information 940 and the copy job log information 960 performed after the acquired final ink exchange date and time 993 are extracted. At the time of this extraction, the job end times 944 and 964 are compared with the last ink replacement date and time 993. Then, the sum of the ink usage amounts 950 and 973 stored in the extracted print job log information 940 and copy job log information 960 is calculated. Further, since the information is stored for each ink color as the final ink replacement date and time 993 and the ink usage amounts 950 and 973, the ink usage amount may be calculated for each ink color.

Ink density in the final print. This can be calculated as follows. The print job log information 940 of the last print job or the copy job log information 960 of the last copy job is acquired. The value of the ink usage amount 950 stored in the acquired print job log information 940 or the value of the ink usage amount 973 included in the acquired copy job log information 960 is defined as the ink usage amount. At this time, the sum of the value of the number of color pages 948 stored in the print job log information 940 and the value of the number of monochrome pages 949, or the value of the number of color pages 971 included in the copy job log information 960 and the monochrome page. The sum with the value of the number of pages 972 is defined as the number of pages. The ink usage density per page is calculated by dividing the ink usage by the number of pages. Although the ink usage density in the last printing is calculated here, the ink usage density may be calculated by accumulating all the printings performed after the last ink replacement.

Printing frequency (number of times) since the last ink change. This can be calculated as follows. The latest printer log information 990 is extracted from the log information of the printer 600. Then, the last ink exchange date and time 993 is acquired, and the difference between the time indicated by the acquired last ink exchange date and time 993 and the current time is calculated. Then, the calculated difference is divided by the above-mentioned "number of printings and maintenances performed since the last ink replacement". The printing frequency in a longer span is calculated by dividing the cumulative number of prints and maintenance performed after the printer 600 is installed, not after the last ink replacement, by the period after the printer 600 is installed. You can also do it. In addition, by periodically calculating the printing frequency (number of times) since the last ink replacement and learning in chronological order, it is possible to estimate the transition of the user's usage frequency from the viewpoint of chronological order. In addition, based on this estimation, the time when the ink runs out is predicted, and the ink delivery timing is estimated.

Printing frequency (number of printed pages) since the last ink change. This can be calculated as follows. The latest printer log information 990 is extracted from the log information of the printer 600. Then, the last ink exchange date and time 993 is acquired, and the difference between the time indicated by the acquired last ink exchange date and time 993 and the current time is calculated. Then, the calculated difference is divided by the above-mentioned "number of printed pages performed since the last ink replacement". It is also possible to calculate the printing frequency in a longer span by dividing the cumulative number of all printed pages performed after the printer 600 is installed, not after the last ink replacement, by the period after the printer 600 is installed. it can. In addition, by periodically calculating the printing frequency (number of printed pages) since the last ink replacement and learning in chronological order, it is possible to estimate the transition of the user's usage frequency from the viewpoint of chronological order. In addition, based on this estimation, the time when the ink runs out is predicted, and the ink delivery timing is estimated.

Printed content (ink usage) since the last ink change. This can be calculated as follows. The latest printer log information 990 is extracted from the log information of the printer 600. Then, the last ink exchange date and time 993 is acquired, and the difference between the time indicated by the acquired last ink exchange date and time 993 and the current time is calculated. Then, the calculated difference is used to divide the above-mentioned "ink usage amount used for printing performed after the last ink replacement". By dividing the cumulative total of all ink usage performed after the printer 600 is installed, not after the last ink replacement, by the period after the printer 600 is installed, the print content (ink usage) over a longer span. Can also be calculated. In addition, by periodically calculating the print content (ink usage) since the last ink change and learning in chronological order, it is also possible to estimate the transition of the user's usage content from the time series perspective. it can. The transition of the user's usage content is, for example, a tendency that monochrome printing is more common, but color printing is more common. In addition, based on this estimation, the time when the ink runs out is predicted, and the ink delivery timing is estimated.

Shipping information. The delivery information 807 is information input by the user when the printer 600 is registered in the automatic ordering system, and is stored in the store server. The delivery information 807 is also replicated to the edge server 300 or the cloud server 300, which can be acquired and included in the input data X. The delivery information 807 includes information on the distance to the delivery destination, the delivery method, and the delivery time. The delivery method includes delivery by car or delivery by airplane. In addition, the delivery time information considers factors that suddenly affect the delivery time, such as the time to send a New Year's card at the end of the year. For example, in December, the amount and period of ink that can be delivered at the same time can be changed by adjusting the parameters of machine learning because the delivery of ink increases due to the printing of New Year's cards. Due to these factors, the time it takes for consumables to reach the user varies. Therefore, by including the delivery information in the input data X, it is possible to perform learning suitable for the usage environment and time of each user.

The parameters that can be used as the input data X to the learning model are not limited to the above example. Some or all of the parameters stored in the log information may be used as they are as input data X to the learning model. This also makes it possible to reflect unknown parameters in the control by machine learning, for example, as a factor of whether or not an ink-related error occurs.

Further, the teacher data T is generated based on the presence / absence information 703 of the ink completeness error stored in the learning request 700, the last ink exchange date / time (K / C / M / Y) 704, and the delivery information 807. A delivery cost of 707 is used. If the ink completeness error occurs, it means that the ink delivery was slow, and the inference of the past ink delivery timing was incorrect. In addition, by acquiring information on the actual ink shipping date and time from the consumables ordering server 260 and calculating the difference from the last ink replacement date and time, from the time when new ink arrives at the user's hand until it is actually attached to the printer 600. You can see the period of. If this period is long, it means that the ink delivery was early, and the inference of the past ink delivery timing was incorrect. If you only want to deliver one ink, you can learn the optimum delivery timing of each ink and deliver it, but if you deliver multiple types of ink, the additional ink may not be delivered at the optimum timing. large. Therefore, regarding the teacher data T used for optimizing the deviation amount L used in learning, in addition to the presence / absence information 703 of ink completeness error and the last ink replacement date / time (K / C / M / Y) 704, the delivery cost 707 needs to be considered. Specifically, the delivery cost 707 is a labor cost and a fuel cost calculated based on the delivery information 807 held by the edge server 300 or the cloud server 200, so it is necessary to learn to make it as small as possible. .. A threshold value of the delivery cost 707 is set, and whether the simultaneous ink delivery is successful or unsuccessful is determined based on whether or not the threshold value is exceeded. If the delivery cost 707 exceeds this threshold when each ink is sent multiple times at the optimum timing without sending each ink at the same time, the delivery cost 707 is more important than the optimum delivery timing, and the weight of the parameter of the delivery cost 707 is increased. Learn in the direction of increasing. The threshold value of the delivery cost 707 may be set to a value unique to each user. By using these three teacher data, learning is performed in consideration of the balance between ink delivery timing and delivery cost for each user. In some cases, the satisfaction level of the users delivered in quick succession may be quantitatively calculated and added to the teacher data. The user's satisfaction is obtained by having the user send feedback at the time of ink delivery. That is, the output when the input data X generated based on the log identifier 702, the delivery information 807, and the log history information 705 as needed is input to the learning model 252, and the value derived from the above-mentioned teacher data T. Learning is performed so that the deviation amount L of is minimized. Then, as a result, the learning model 252 is updated.

<I / O structure of trained model at the time of inference>
FIG. 16 is a diagram showing an input / output structure of the trained model 352 at the time of inference in the present embodiment. At the time of inference, the necessary log history information is extracted from the log information 806 stored in the edge server 300 and used by using the printer identifier 711 passed in the inference request 710. From the extracted log history information, necessary parameters are acquired and used as input data X as in the case of learning. The inference is performed to predict the tendency of the ink remaining amount of the printer 600 and to predict the simultaneous delivery ink. Then, when executing the inference performed in step S1807 (see FIG. 18) described later, it is conceivable to calculate and use the above-mentioned parameters as in the case of learning. The output data Y obtained as a result of the inference by the trained model 352 becomes the inference result notification 720. The inference result notification 720 notifies the consumables ordering server 260 of the simultaneous delivery ink information 723.

Since the learning parameter is determined with the case where the delivery cost 707 exceeds the threshold value as a failure, it is determined by the trained model 352 which of the optimum delivery timing and the delivery cost 707 should be prioritized. For example, it is estimated that the ink low warning screen 650 is displayed within one week from the display of the ink low warning screen 650 of the first ink (referred to as ink 1) at the delivery timing for the user who takes one week or more for delivery. Consider the case where there is ink (referred to as ink 2). In this case, in order to reduce the cost of re-delivering after one week, even if the ink low warning screen 650 of ink 2 is not displayed, the ink 1 is delivered at the same time one week ahead of schedule. Since it is estimated that the ink 2 arrives after the ink low warning screen 650 is displayed even if the simultaneous delivery is performed, it is possible to prevent the ink inventory from continuing to increase in the user's hand. In this way, the number of days to be targeted for simultaneous delivery is determined according to the usage frequency and usage amount of the user. Further, even if the inks have the same color, a plurality of inks of the same color can be delivered at the same time in the case of a user whose previous usage frequency and usage amount are very large and the ink low warning screen 650 is displayed in a few days. The factor that determines the delivery timing may be an error shown in FIG. 8 other than the ink low warning, or a case where the remaining amount of ink becomes equal to or less than the threshold value.

<Printer processing details>
FIG. 17 is a flowchart showing the processing contents of the printer 600. The program for executing this process is stored in the program memory 613 of the printer 600. When the power of the printer 600 is turned on, after the initialization operation is executed, the CPU 611 of the printer 600 executes this process every time a job or an instruction is received from the outside to the printer 600. Note that FIG. 17 describes the processes related to the present embodiment, and the other processes are not shown.

In step S1701, the CPU 611 determines whether the processing of the print job, the copy job, or the maintenance job, which is a job whose ink information may be updated, has been completed. If the determination result of this step is true, the process proceeds to S1703. On the other hand, if the determination result of this step is false, the process proceeds to S1702. Hereinafter, "step S-" will be abbreviated as "S-".

In S1702, the CPU 611 determines whether the ink cartridge 632 of the printer 600 has been replaced. If the determination result of this step is true, the process proceeds to S1704. On the other hand, if the determination result of this step is false, a series of processes is terminated.

In S1703, the CPU 611 transmits job log information including information such as ink consumption of a print job, a copy job, or a maintenance job to the edge server 300. Further, the CPU 611 transmits the printer log information 990 including the information such as the ink remaining amount after the job is completed to the edge server 300. For example, when the maintenance job is executed, the CPU 611 transmits the maintenance job log information 980 to the edge server 300 as the job log information.

In S1704, the CPU 611 transmits printer log information 990 including information such as the remaining amount of ink after replacing the ink cartridge 632 to the edge server 300.

Subsequently, since the ink information has been updated, the CPU 611 transmits the learning request 700 to the edge server 300 in S1705 in order to verify the previous inference result.

Subsequently, since the ink information has been updated, the CPU 611 transmits the inference request 710 to the edge server 300 in S1706 in order to know the possibility that the automatic ordering of ink is required.

Although the mode in which the printer 600 transmits the inference request has been described here, the embodiment of the present invention is not limited to this. Since the information required for inference has already been transmitted to the edge server 300 as job log information or printer log information, the edge server 300 may execute inference when receiving the job log information or printer log information.

Further, although the mode in which the printer 600 transmits the learning request has been described here, the embodiment of the present invention is not limited to this. Of the teacher data T required for learning, the printer log information includes the same information as the last ink replacement date and time (K / C / M / Y) 704. On the other hand, since the presence / absence information 703 of the ink completeness error is not included in the printer log information, the edge server 300 can change the job log information and the printer log information by changing the information so that the information is included in the job log information and the printer log information. Learning may be performed when it is received.

<Processing content of edge server>
FIG. 18 is a flowchart showing the processing contents of the edge server 300 in the present embodiment. The program for executing this process is stored in the hard disk unit 202 of the edge server 300. When the edge server 300 is started, after the initialization operation is executed, the CPU 211 of the edge server 300 executes this process every time it receives a request from the outside to the edge server 300. Note that FIG. 18 describes the processes related to the present embodiment, and the other processes are not shown.

In S1801, the CPU 211 determines whether or not the inference request 710 transmitted by the printer 600 has been received. If the determination result of this step is true, the process proceeds to S1805. On the other hand, if the determination result of this step is false, the process proceeds to S1802.

First, the case where the reception of the inference request 710 is detected (when YES in S1801) will be described. In this case, in S1805, the CPU 211 extracts the log information necessary for performing the requested inference from the area (log information storage area) in which the log information is stored in the hard disk unit 202. Specifically, the CPU 211 extracts the log information in which the printer identifier 711 and the printer identifier match in the inference request 710, and then uses the log identifier 712 and the log history information 713 to perform the inference this time. Extract the log information that becomes. If it is difficult to extract and process all the log information, it may be configured to extract only the log information within a certain period.

After extracting the necessary log information in S1805, in S1806, the CPU 211 generates inference data (specifically, input data X for the trained model 352) based on the extracted log information.

In S1807, the CPU 211 executes inference based on the inference data generated in S1806. Specifically, the inference data is sent to the inference unit 351 and input to the trained model 352 to execute the inference.

In S1808, the CPU 211 creates an inference result notification 720 based on the inference result of S1807, and transmits the created inference result notification 720 to the consumables ordering server 260.

Subsequently, the case where the reception of the inference request 710 is not detected (the case of NO in S1801) will be described. In this case, in S1802, the CPU 211 determines whether or not some log information has been received from the printer 600. If the determination result of this step is true, the process proceeds to S1809. On the other hand, if the determination result of this step is false, the process proceeds to S1803.

In S1809, the CPU 211 stores the received log information as log information for the corresponding printer in the log information recording area provided in the hard disk unit 202 in the edge server 300.

In S1803, the CPU 211 determines whether or not the learning request 700 has been received from the printer 600. If the determination result of this step is true, the process proceeds to S1810. On the other hand, if the determination result of this step is false, the process proceeds to S1804.

In S1810, the CPU 211 extracts the log information necessary for performing the requested learning from the log information storage area in the hard disk unit 202. Specifically, the CPU 211 extracts the log information in which the printer identifier 701 and the printer identifier match the printer identifier stored in the learning request 700, and then uses the log identifier 702 and the log history information 705 to use the log identifier 702 and the log history information 705 to be the target of the current learning. Extract the log information that becomes.

After extracting the necessary log information in S1810, in S1811, the CPU 211 transmits the learning request 700 to which the extracted log information is added to the cloud server 200.

In S1804, the CPU 211 determines whether or not the trained model delivered by the cloud server 200 has been received. If the determination result of this step is true, the process proceeds to S1812. On the other hand, if the determination result of this step is false, a series of processes is terminated.

In S1812, the CPU 211 reflects the received trained model information in the trained model 352 of the edge server 300, and stores the reflected trained model 352. As a result, the updated trained model 352 will be used for subsequent inference requests.

<Processing content of cloud server>
FIG. 19 is a flowchart showing the processing contents of the cloud server 200. The program for executing this process is stored in the hard disk unit 202 of the cloud server 200. When the cloud server 200 is started, after the initialization operation is executed, the CPU 211 of the cloud server 200 executes this process every time it receives a request from the outside to the cloud server 200. Note that FIG. 19 describes the processes related to the present embodiment, and the other processes are not shown.

In S1901, the CPU 211 determines whether or not the learning request 700 from the outside of the cloud server 200 has been received. If the determination result of this step is true, the process proceeds to S1902. On the other hand, if the determination result of this step is false, a series of processes is terminated.

In S1902, the received learning request 700 is sent to the learning data generation unit 250 to generate learning data. Specifically, the received learning request 700 includes the extracted log information added by the edge server 300, and the learning data generation unit 250 uses the input data for the learning model 252 based on this information. Generate X. Further, the learning data generation unit 250 generates the teacher data T based on the presence / absence information 703 of the ink completeness error and the last ink exchange date / time (K / C / M / Y) 704 stored in the learning request 700. Generate.

In S1903, the CPU 211 executes learning based on the learning data generated in S1902. Specifically, the learning data is sent to the learning unit 251 to execute the learning for the learning model 252.

In S1904, the CPU 211 reflects the learning result of S1903 in the learning model 252 and accumulates it. When the accumulation of learning results is completed, the process proceeds to S1905.

In S1905, the CPU 211 generates a trained model 352 based on the learning model 252, and distributes the generated trained model 352, that is, transmits it to the edge server 300. Here, it is explained that the cloud server 200 executes learning every time a learning request is received. However, since it takes a processing time to execute learning, the received learning requests are accumulated and the learning request is executed. It may be configured to execute learning asynchronously with reception. Further, although the form in which the trained model is delivered to the edge server 300 each time the learning is performed has been described, when a large amount of learning is performed on the cloud server 200, it is not necessary to deliver the trained model each time the learning is performed. Good. That is, the trained model may be delivered periodically, or the trained model may be delivered in response to a delivery request sent from the edge server 300 to the Clyde server 200.

<Processing contents of consumables ordering server>
FIG. 20 is a flowchart showing the processing contents of the consumables ordering server 260. The program for executing this process is stored in the hard disk unit 202 of the consumables ordering server 260. When the consumables ordering server 260 is started, after the initialization operation is executed, the CPU 211 of the consumables ordering server 260 executes this process every time it receives a request from the outside to the consumables ordering server 260. Note that FIG. 20 describes the processes related to the present embodiment, and the other processes are not shown.

In S2001, the CPU 211 determines whether or not the inference result notification 720 from the outside of the consumables ordering server 260 has been received. If the determination result of this step is true, the process proceeds to S2002. On the other hand, if the determination result of this step is false, a series of processes is terminated.

In S2002, the CPU 211 determines whether ink delivery is necessary by using the simultaneous delivery ink information 723 stored in the received inference result notification 720, that is, the delivery date and time information for each ink color as the inference result. For example, the inferred delivery date and time may be compared with the current date and time, and if the inferred delivery date and time exceeds the current date and time, ink delivery may be considered necessary. If the determination result of this step is true, the process proceeds to S2003, while if the determination result is false, a series of processes is terminated.

In S2003, the CPU 211 determines whether or not there is simultaneous delivery ink based on the simultaneous delivery ink information 723. When the judgment result of this step is true (in other words, when it is judged from the viewpoint of delivery cost 707 that the ink that should be originally delivered and the ink different from this should be delivered at the same time as a result of inference) , S2004. On the other hand, if the determination result of this step is false, the process proceeds to S2005.

In S2004, the CPU 211 adds an ink different from the ink that should be originally delivered (an ink that should be delivered at the same time as the ink that should be delivered and is referred to as "simultaneous delivery ink") to the ink to be delivered. ..

In S2005, the CPU 211 executes an ink delivery process for the ink to be delivered. For example, when there is no simultaneous delivery ink, only the ink specified on the ink low warning screen 650 is delivered. The factor for determining the delivery timing may be an error shown in FIG. 8 other than the ink low warning, or a case where the remaining amount of ink becomes equal to or less than a predetermined threshold value. Since the ink delivery process is a process executed by using an existing delivery system, detailed description thereof will be omitted. For example, ink may be delivered based on the user information registered in the consumables ordering server 260, or the consumables ordering server 260 acquires user information from a user information server (not shown) to deliver ink. You may go.

<Effect of this embodiment>
According to this embodiment, it is possible to deliver a plurality of types of ink at an appropriate timing based on a prediction result of the possibility that ink is consumed by printing, copying, maintenance, etc. and automatic ordering of ink is required. ..

[Second Embodiment]
In the first embodiment, the configuration in which the edge server 300 uses the trained model 352 based on the learning model 252 trained by the cloud server 200 for inference has been described. In such a configuration, if there are a plurality of edge servers 300, the same learning model 252 trained by the cloud server 200 is reflected in the trained models 352 of each of the plurality of edge servers 300 and used. However, there are variations in the manufacturing of ink cartridges, and the characteristics also vary depending on the environment such as ambient temperature and humidity. Therefore, the timing at which automatic ink ordering is required may differ for each individual printer 600. In addition, since the intended use of the printed matter differs depending on the user, the criteria for determining how much the printed matter is allowed to be blurred or disturbed may differ depending on the user.

Therefore, in the present embodiment, when there are a plurality of printers 600, a case of inferring the necessity of automatic ordering of inks different for each printer 600 will be described. Specifically, the general-purpose learning model 253 learned by the cloud server 200 is reflected in the individual learning model 356 of the edge server 300, and the edge server 300 also performs unique learning. The edge server 300 performs learning only on the learning request sent to itself, and reflects the learning result only on its own individual learning model 356. This makes it possible to perform learning depending on the characteristics unique to the device group in the range managed by the edge server 300, the characteristics of each user to be used, the printer to be used, and the like.

<Software configuration>
FIG. 21 is a diagram showing a software configuration of the processing system 100 according to the present embodiment. The cloud server 200 has a learning data generation unit 250, a learning unit 251 and a general-purpose learning model 253. The learning data generation unit 250 is a module that generates learning data that can be processed by the learning unit 251 based on the data received from the outside. The learning data is a set of the input data X of the learning unit 251 and the teacher data T indicating the correct answer of the learning result. The learning unit 251 is a program module that executes learning for the general-purpose learning model 253 based on the learning data received from the learning data generation unit 250. The general-purpose learning model 253 accumulates the results of learning performed by the learning unit 251. Here, an example of realizing the general-purpose learning model 253 as a neural network will be described. By correcting the weighting parameters between each node of the neural network to appropriate ones, the input data can be classified and the evaluation value can be determined. The accumulated general-purpose learning model 253 is distributed to the edge server 300 as a general-purpose learned model, and is used for updating the individual learning model 356 in the edge server 300.

The edge server 300 has a data collection / provision unit 350, a learning / inference data generation unit 354, a learning unit 355, an individual learning model 356, and an inference unit 351. The data collection / provision unit 350 is a module that transmits the data received from the device 400 and the data collected by the edge server 300 to the cloud server 200 as a data group to be used for learning. Further, these data are also sent to the learning / inference data generation unit 354 as a data group for the edge server 300 itself to use for learning. The learning / inference data generation unit 354 is a module that generates learning data that can be processed by the learning unit 355 based on the data received from the outside. The learning data is a set of the input data X of the learning unit 355 and the teacher data T indicating the correct answer of the learning result. The learning unit 355 is a program module that executes learning for the individual learning model 356 based on the learning data received from the learning / inference data generation unit 354. The individual learning model 356 accumulates the results of learning performed by the learning unit 355. Further, when the edge server 300 receives the general-purpose trained model learned and distributed by the cloud server 200, the difference between the received general-purpose trained model and the individual learning model 356 is reflected in the neural network of the individual learning model 356. Let me. As a result, it becomes possible to execute inference using a trained model that reflects both the general-purpose learning result performed by the cloud server 200 and the individual learning result performed by the edge server 300 itself. The inference unit 351 is a program module that executes inference using the individual learning model 356 based on the data sent from the device 400 and returns the result to the device 400. The data sent from the device 400 to the edge server 300 is the data that becomes the input data X of the inference unit 351.

The device 400 has an application unit 450 and a data transmission / reception unit 451. The application unit 450 is a module that realizes various functions executed by the device 400, and is a module that uses a learning / inference mechanism by machine learning. The data transmission / reception unit 451 is a module that requests the edge server 300 to learn or infer. At the time of learning, the data transmission / reception unit 451 transmits the data used for learning to the data collection / provision unit 350 of the edge server 300 at the request of the application unit 450. At the time of inference, the data transmission / reception unit 451 transmits the data used for inference to the edge server 300 at the request of the application unit 450, receives the inference result, and returns it to the application unit 450. As shown in FIGS. 9 and 21, the configuration of the device 400 in this embodiment is the same as that in the first embodiment, and the processing of the device 400 in this embodiment is also in the first embodiment (see FIG. 17). Since it is the same as the above, the description regarding the processing of the device 400 will be omitted below. Further, since the processing of the consumables ordering server 260 is the same as that of the first embodiment, the description thereof will be omitted.

<Processing content of edge server>
FIG. 22 is a flowchart showing the processing contents of the edge server 300 in the present embodiment. The program for executing this process is stored in the hard disk unit 202 of the edge server 300. When the edge server 300 is started, after the initialization operation is executed, the CPU 211 of the edge server 300 executes this process every time it receives a request from the outside to the edge server 300. Note that FIG. 22 describes the processes related to the present embodiment, and the other processes are not shown.

In S2201, the CPU 211 determines whether or not the inference request 710 transmitted by the printer 600 has been received. If the determination result of this step is true, the process proceeds to S2205. On the other hand, if the determination result of this step is false, the process proceeds to S2202.

First, the case where the reception of the inference request 710 is detected (when YES in S2201) will be described. In this case, in S2205, the CPU 211 extracts the log information necessary for performing the requested inference from the log information storage area in the hard disk unit 202. Specifically, the CPU 211 extracts the log information in which the printer identifier 711 and the printer identifier match in the inference request 710, and then uses the log identifier 712 and the log history information 713 to perform the inference this time. Extract the log information that becomes. If it is difficult to extract and process all the log information, it may be configured to extract only the log information within a certain period.

After extracting the necessary log information in S2205, in S2206, the CPU 211 generates the input data X for the individual learning model 356 based on the extracted log information. Here, in order to perform learning depending on the individual printer 600, it can be configured to generate the input data X including the information of the printer identifier.

In S2207, the CPU 211 executes inference based on the inference data generated in S2206. Specifically, the inference data is sent to the inference unit 351 and input to the individual learning model 356 to execute the inference.

In S2208, the CPU 211 creates an inference result notification 720 based on the inference result of S2207, and transmits the created inference result notification 720 to the consumables ordering server 260.

Subsequently, the case where the reception of the inference request 710 is not detected (the case of NO in S2201) will be described. In this case, in S2202, the CPU 211 determines whether or not some log information has been received from the printer 600. If the determination result of this step is true, the process proceeds to S2209. On the other hand, if the determination result of this step is false, the process proceeds to S2203.

In S2209, the CPU 211 stores the received log information as log information for the corresponding printer in the log information recording area provided in the hard disk unit 202 in the edge server 300.

In S2203, the CPU 211 determines whether or not the learning request 700 has been received from the printer 600. If the determination result of this step is true, the process proceeds to S2210. On the other hand, if the determination result of this step is false, the process proceeds to S2204.

In S2210, the CPU 211 extracts the log information necessary for performing the requested learning from the log information storage area in the hard disk unit 202. Specifically, the CPU 211 extracts the log information to be the target of the current learning by using the log history information 705 stored in the learning request 700.

After extracting the necessary log information in S22110, the learning / inference data generation unit 354 generates learning data in S2211. Specifically, the learning / inference data generation unit 354 generates the input data X for the individual learning model 356 based on the log information extracted in step S2210. Further, the learning / inference data generation unit 354 generates the teacher data T based on the received ink-related error presence / absence information 703.

In S2212, the CPU 211 executes learning based on the learning data generated in S2211. Specifically, the learning data is sent to the learning unit 355, and learning for the individual learning model 356 is executed.

In S2213, the CPU 211 reflects the learning result of S2212 in the individual learning model 356 and accumulates it. By the way, the learning request 700 transmitted from the printer 600 and received by the edge server 300 is used by the edge server 300 for learning to its own individual learning model 356, and to the cloud server 200 as a learning request to the general-purpose learning model 253. Need to pass.

Therefore, in the following S2214, the CPU 211 adds the log information extracted in S2210 and transmits the received learning request 700 to the cloud server 200.

In S2204, the CPU 211 determines whether or not the general-purpose learning model distributed by the cloud server 200 has been received. If the determination result of this step is true, the process proceeds to S2215. On the other hand, if the determination result of this step is false, a series of processes is terminated.

In S2215, the CPU 211 stores the received general-purpose learning model.

In S2216, the CPU 211 reflects the difference between the general-purpose learning (learned) model information stored in S2215 and the individual learning model 356 in the individual learning model 356, and stores the reflected individual learning model 356. As a result, the updated individual learning model 356 will be used for subsequent inference requests.

<Processing content of cloud server>
FIG. 23 is a flowchart showing the processing contents of the cloud server 200. The program for executing this process is stored in the hard disk unit 202 of the cloud server 200. When the cloud server 200 is started, after the initialization operation is executed, the CPU 211 of the cloud server 200 executes this process every time it receives a request from the outside to the cloud server 200. Note that FIG. 23 describes the processes related to the present embodiment, and the other processes are not shown.

In S2301, the CPU 211 determines whether or not the learning request 700 from the outside of the cloud server 200 has been received. If the determination result of this step is true, the process proceeds to S2302. On the other hand, if the determination result of this step is false, a series of processes is terminated.

In S2302, the received learning request 700 is sent to the learning data generation unit 250 to generate learning data. Specifically, the received learning request 700 includes the extracted log information added by the edge server 300, and the learning data generation unit 250 uses the input data for the learning model 252 based on this information. Generate X. Here, in order to eliminate individual-dependent learning of the printer 600, the input data X can be generated by removing the information of the printer identifier. Further, the learning data generation unit 250 is a teacher based on the presence / absence information 703 of the ink completeness error, the last ink exchange date / time (K / C / M / Y) 704, and the delivery information 807 stored in the learning request 700. Generate data T.

In S2303, the CPU 211 executes learning based on the learning data generated in S2302. Specifically, the learning data is sent to the learning unit 251 to execute learning for the general-purpose learning model 253.

In S2304, the CPU 211 reflects the learning result of S2303 in the general-purpose learning model 253 and accumulates it. When the accumulation of learning results is completed, the process proceeds to S2305.

In S2305, the CPU 211 transmits the generated general-purpose learning model 253 to the edge server 300.

<Effect of this embodiment>
According to this embodiment, it is possible to perform learning depending on the unique characteristics different for each individual printer 600 and the characteristics of the user who uses the printer 600. The learning depending on the unique characteristics is not limited to the individual printer 600, and the learning depending on the group can be performed by using other grouping information. Examples of grouping include information about the company or organization to which the device or user belongs, information about the country or region where the printer 600 is installed and operating, and business format information according to the business content in which the device is used. .. These can be obtained by analyzing the usage environment 994 and the user information 995 included in the printer log information 990 acquired by the edge server 300.

[Third Embodiment]
In the first embodiment and the second embodiment, a configuration (see FIG. 15) in which information on the presence or absence of an ink completeness error as shown in FIG. 8C is learned using the teacher data T has been described. , The embodiment of the present invention is not limited to this.

Regarding ink-related errors, the errors that can occur differ depending on the configuration of the printer 600. For example, a printer that does not have a highly expressive panel such as the operation panel 605 and expresses an ink error only by blinking or lighting an LED cannot express a plurality of ink-related errors and cannot support the display of an ink low warning. In this case, since the number of errors that can be expressed is two, it is necessary to change the threshold value of the ink remaining amount used for determining each error display. Also, depending on the printer configuration, you may want to add a new error, and even in that case, you need to set a new ink level threshold.

Therefore, instead of learning using the information on the presence or absence of the ink completeness error as the teacher data T, the information on whether or not the ink remaining amount specified for each printer has been reached is used as the teacher data T for learning. It may be configured. The remaining amount of ink that the printer 600 wants to learn as teacher data T is registered in the edge server 300 in advance. Further, the ink remaining amount may be registered in the edge server 300 for each ink color.

FIG. 24 is a diagram showing an input / output structure of the learning model 252 at the time of learning in the present embodiment. Since the same as the first embodiment (see FIG. 15) except for the ink remaining amount 706 shown in the figure, the description thereof will be omitted. For the teacher data T802, the ink remaining amount information 706 stored in the learning request 700, the last ink replacement date and time) (K / C / M / Y) 704, and the delivery cost 707 are used. Then, the output when the input data X generated based on the log identifier 702 and the log history information 705 as needed is input to the learning model 252, and the values derived from the teacher data T (706, 704, and 707). Learning is performed so that the deviation amount L from the above is minimized. As a result, the learning model 252 is updated.

[Fourth Embodiment]
In the first embodiment and the second embodiment, when the ink information may be updated, the printer log information including the ink remaining amount information is transmitted to the edge server 300 in S1703 or S1704 (see FIG. 17). I'm sending. However, the embodiments of the present invention are not limited to this.

When the edge server 300 or the cloud server 200 needs the ink remaining amount information when performing inference or learning, the necessary information may be acquired by directly communicating with the printer 600. As a result, when there is a possibility that the ink information is updated, there is no data communication for transmitting the printer log information between the printer 600 and the edge server 300, and the communication load of the printer 600 and the edge server 300 is reduced. it can.

[Other Embodiments]
In the above embodiment, ink has been described as an example of consumables used in the printer, but the present invention is not limited to this, and other recording materials such as toner or a recording medium such as paper may be used. In addition, the treatment of the above embodiment can be applied to various consumables.

The present invention can also be realized by supplying a program that realizes one or more functions of the above-described embodiment to a system or a device via a network or a storage medium, and the computer of the system or the device reads and executes the program. is there. A computer may have one or more processors or circuits and may include multiple separate computers or a network of separate processors or circuits to read and execute computer executable instructions.

The processor or circuit may include a central processing unit (CPU), a microprocessing unit (MPU), a graphics processing unit (GPU), an application specific integrated circuit (ASIC), a field programmable gateway (FPGA). Also, the processor or circuit may include a digital signal processor (DSP), a data flow processor (DFP), or a neural processing unit (NPU).

The storage medium can also be referred to as a non-transitory computer-readable medium. The storage medium may also include one or more hard disks (HD), random access memory (RAM), read-only memory (ROM), and storage devices for distributed computing systems. The storage medium may also include an optical disc (eg, a compact disc (CD), a digital versatile disc (DVD), or a Blu-ray disc (BD, registered trademark)), a flash memory device, and a memory card. That is, the present invention supplies a program that realizes one or more functions of the above-described embodiment to a system or device via a network or storage medium, and one or more processors in the computer of the system or device reads and executes the program. It can also be realized by the processing to be performed. It can also be realized by a circuit (for example, ASIC) that realizes one or more functions. The present invention may be configured by appropriately combining the elements of the above-described embodiments.

Further, in the above-described embodiment, automatic ordering for ink for consumables has been described, but the present invention is not limited thereto. For example, it may be various types of paper for printing or consumables such as a maintenance cartridge that absorbs waste ink. When these consumables are used, the ink remaining amount information in FIG. 12D becomes the paper remaining amount information and the maintenance cartridge remaining amount information.

451 Data transmitter / receiver 600 Printer 710 Inference request 720 Inference result notification

Claims (19)

A recording device having a first transmission means for transmitting an inference request for inferring the timing of delivering consumables.
A recording device characterized in that it is determined whether or not simultaneous delivery of a plurality of types of the consumables is necessary based on an inference result notification created by using a trained model in response to the inference request. The recording device according to claim 1, wherein the inference request includes a first identifier for identifying a recording device and a second identifier for identifying a log. The recording device according to claim 2, wherein the inference request further includes log history information. The recording device according to any one of claims 1 to 3, wherein the trained model is acquired by machine learning using teacher data including at least information on the remaining amount of the consumables. The consumable is ink
The teacher data includes information on the presence or absence of an ink completeness error indicating whether or not an ink completeness error has occurred in the recording device as the information, information on the date and time when the ink cartridge was last replaced, and delivery cost. The recording device according to claim 4, wherein the information of the above is included. The consumable is ink
The recording device according to claim 4, wherein the teacher data includes ink remaining amount information as the information, information on the date and time when the ink cartridge was last replaced, and information on the delivery cost. The recording device according to any one of claims 1 to 6, wherein the consumable includes ink, toner, paper, and a maintenance cartridge. A system including the recording device according to any one of claims 4 to 6 and a first server.
The first server is
Based on the inference request transmitted by the first transmission means, data regarding the operation performed by the recording device, the remaining amount of the consumables, and the settings in the recording device and the job are collected as the log information of the recording device. Data collection means and
An inference data generation means that generates the first input data necessary for inference based on the data collected by the data collection means, and an inference data generation means.
An inference means for making inferences by inputting the first input data into the trained model,
A system characterized by having. 8. The inference result notification acquired when the inference is performed by the inference means includes information indicating whether or not a plurality of types of the consumables are to be delivered at the same time, according to claim 8. The system described. When the print job, copy job, or maintenance job is completed in the recording device, or when the ink cartridge is replaced, the first transmission means learns to update the inference request and the trained model. The system according to claim 8 or 9, wherein the request is transmitted to the first server. The learning request includes a first identifier for identifying the recording device, a second identifier for identifying the log, and an ink completeness error indicating whether or not an ink completeness error has occurred in the recording device. The system according to claim 10, further comprising information on the presence or absence of an ink cartridge, information on the date and time when the ink cartridge was last replaced, log history information, and delivery information. Has a second server,
The second server is
A learning data generation means that generates learning data that is a set of a second input data generated based on log information and the teacher data based on the learning request.
A first learning means for accumulating a learning model for updating the learned model and a part or all of the accumulated learning model as a result of learning by the first learning means are transmitted to the first server. Second transmission means to
10. The system according to claim 10 or 11. The first learning means corrects the learning model so that the amount of deviation between the output data obtained as a result of recognizing the second input data using the learning model and the teacher data is minimized. 12. The system according to claim 12. The first server further has a second learning means for performing learning depending on an individual recording device.
The first learning means of the second server performs general-purpose learning independent of a recording device, and performs general-purpose learning.
The system according to claim 12 or 13, wherein the learning result of the first learning means and the learning result of the second learning means are reflected in the trained model. When the inference is performed by the inference means, the first server directly communicates with the recording device to acquire the log information of the recording device.
Any one of claims 12 to 14, wherein the second server directly communicates with the recording device to acquire log information of the recording device when learning is performed by the first learning means. The system described in the section. Has a third server,
The system according to any one of claims 12 to 15, wherein the first server transmits the inference result notification created by the inference performed by the inference means to the third server. The third server is
The receiving means for receiving the inference result notification and
Based on the inference result notification received by the receiving means, the first determining means for determining whether the consumables need to be delivered, and the first determining means.
Based on the inference result notification, a second determination means for determining whether simultaneous delivery of a plurality of types of the consumables is necessary, and
If the determination result by the second determination means is true, an additional means for adding consumables to be delivered at the same time to the delivery target, and
16. The system according to claim 16. A method of controlling a recording device having a step of transmitting an inference request for inferring the timing of delivering consumables.
A control method characterized in that it is determined whether or not simultaneous delivery of a plurality of types of the consumables is necessary based on an inference result notification created by using a trained model in response to the inference request. A program for causing a computer to perform the method according to claim 18.

Images