Fault detection device for printer, printing system and fault detection method for printer

Abstract

To suppress lowering of quality of printing.SOLUTION: A fault detection device 100 of a printer 10 includes: a sound acquisition part 103 for acquiring operation sound S1 at the time of printing in the printer 10; a model generation part 105 for generating a learning model MD1 with the operation sound S1 acquired by the sound acquisition part 103 as teacher data, the operation sound S1 as input, and presence / absence of occurrence of a fault of the printer 10 as output, by machine learning; a framing part 107 for framing the operation sound S1 acquired by the sound acquisition part 103 at an interval of a predetermined frame shift, and thereby generating a plurality of frames FL1; and a learning model processing part 109 for inputting the plurality of frames FL1 generated by the framing part to the learning model MD1, and outputting whether or not the fault of the printer 10 has occurred.SELECTED DRAWING: Figure 5

Description

[Technical field]

[0001] The present invention relates to a printer fault detection device, a printing system, and a printer fault detection method. [Background technology]

[0002] For example, Patent Document 1 discloses a printer equipped with a platen on which a medium is supported and an ink head with nozzles formed therein for ejecting ink toward the medium supported by the platen. The ink head is mounted on a carriage. The carriage engages with a guide rail extending in the main scanning direction.

[0003] The printer is configured to eject ink from the ink head while the ink head moves together with the carriage in a main scanning direction, and the printer is also configured to move a medium supported by a platen in a sub-scanning direction perpendicular to the main scanning direction. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2004-196537 A Summary of the Invention [Problem to be solved by the invention]

[0005] However, in the above-mentioned printer, if an abnormality occurs in any of its components during printing, the printer may malfunction. If printing is continued while the printer is malfunctioning, the print quality may deteriorate.

[0006] The present invention has been made in consideration of these points, and its object is to provide a printer fault detection device, a printing system, and a printer fault detection method that can prevent a decrease in print quality caused by the occurrence of a printer fault. [Means for solving the problem]

[0007] The printer malfunction detection device according to the present invention includes a sound acquisition unit, a model generation unit, a framing processing unit, and a learning model processing unit. The sound acquisition unit acquires operating sounds of the printer when printing. The model generation unit uses the operating sounds acquired by the sound acquisition unit as teacher data, the operating sounds as input, and generates a learning model by machine learning, in which the input indicates whether a malfunction has occurred in the printer and the output indicates whether a malfunction has occurred in the printer. The framing processing unit creates multiple frames by performing framing processing on the operating sounds acquired by the sound acquisition unit at predetermined frame shift intervals. The learning model processing unit inputs the multiple frames created by the framing processing unit into the learning model, and outputs whether a malfunction has occurred in the printer.

[0008] The printer malfunction detection method according to the present invention includes a sound acquisition step, a model generation step, a framing processing step, and a learning model processing step. In the sound acquisition step, an operating sound of the printer during printing is acquired. In the model generation step, a learning model is generated by machine learning, in which the operating sound acquired in the sound acquisition step is used as training data, the operating sound is used as input, and the presence or absence of a malfunction of the printer is used as output. In the framing processing step, a plurality of frames are created by performing framing processing on the operating sound acquired in the sound acquisition step at a predetermined frame shift interval. In the learning model processing step, the plurality of frames created in the framing processing step are input to the learning model, and a signal indicating whether or not a malfunction has occurred in the printer is output.

[0009] For example, the operating sound of a printer when a malfunction occurs is considered to be different from the operating sound of a printer when a malfunction does not occur. Therefore, in the present invention, the operating sound of the printer is used as training data, and a learning model generated by machine learning is used to output whether or not a malfunction has occurred in the printer from the learning model. Therefore, it is possible to determine whether or not a malfunction has occurred in the printer from the operating sound of the printer by machine learning. Therefore, when a malfunction occurs in the printer, for example, printing can be stopped, and therefore it is possible to suppress a decrease in print quality caused by the malfunction of the printer. Effect of the Invention

[0010] According to the present invention, it is possible to provide a printer fault detection device, a printing system, and a printer fault detection method that can prevent a decrease in print quality caused by the occurrence of a printer fault. [Brief description of the drawings]

[0011] [Figure 1] 1 is a conceptual diagram illustrating a printing system according to an embodiment. [Diagram 2] FIG. [Diagram 3] 3 is a cross-sectional view of the printer taken along line III-III in FIG. 2. [Figure 4] FIG. 4 is a bottom view showing a schematic configuration of the bottom surface of the carriage and the ink head. [Diagram 5] 1 is a block diagram of a printing system according to an embodiment. [Figure 6] FIG. 1 is a diagram illustrating a learning model generated by machine learning. [Figure 7] FIG. 4 is a diagram showing types of operation sounds. [Figure 8] 1 is a graph showing normal operation sounds. [Figure 9] 11 is a graph showing malfunctioning operation sounds. [Figure 10] 4 is a flowchart showing a method for detecting a malfunction in a printer. [Figure 11] FIG. 11 is a diagram showing a first learning model in another embodiment. [Figure 12] FIG. 11 is a diagram showing a second learning model in another embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. Note that the embodiment described here is not intended to limit the present invention in particular. Also, the same reference numerals are used for members and parts that perform the same function, and duplicated descriptions are omitted or simplified as appropriate.

[0013] <Embodiment> FIG. 1 is a conceptual diagram showing a printing system 1 according to the present embodiment. In the present embodiment, the printing system 1 includes a printer 10, a terminal 70, and a malfunction detection device (hereinafter simply referred to as a "malfunction detection device") 100 for the printer 10. In the printing system 1, a malfunction of the printer 10 is detected by the malfunction detection device 100, and when a malfunction of the printer 10 is detected, the terminal 70 is notified that a malfunction of the printer 10 has been detected. In FIG. 1, the number of the printer 10, the terminal 70, and the malfunction detection device 100 is one each. However, the number of the printer 10, the terminal 70, and the malfunction detection device 100 may be more than one each. In addition, the number of the printers 10 and the terminals 70 may be the same or different. Below, the printer 10, the terminal 70, and the malfunction detection device 100 will be described in that order.

[0014] FIG. 2 is a front view showing the printer 10 according to the present embodiment. FIG. 3 is a cross-sectional view of the printer 10 taken along line II-II in FIG. 2. FIG. 4 is a bottom view showing a schematic configuration of the bottom of the carriage 17 and the ink head 40 of the printer 10. FIG. 5 is a block diagram of the printing system 1 according to the present embodiment. In the following description of the printer 10, the left, right, top, and bottom refer to the left, right, top, and bottom as seen from a user in front of the printer 10. The side approaching the user from the printer 10 is the front, and the side approaching the printer 10 from the user is the rear. The symbols F, Rr, L, R, U, and D in the drawings respectively indicate the front, rear, left, right, top, and bottom of the printer 10. The symbol Y in the drawings indicates the main scanning direction. In this embodiment, the main scanning direction Y is the left-right direction. The symbol X in the drawings indicates the sub-scanning direction. In this embodiment, the sub-scanning direction X is the front-back direction, and is a direction that intersects (here, perpendicular to) the main scanning direction Y in a plan view. However, these directions are merely defined for the convenience of explanation, and do not limit the installation mode of the printer 10 or the present invention in any way.

[0015] The printer 10 is an inkjet printer, or a so-called inkjet printer. However, the printing method of the printer 10 is not limited to the inkjet method, and may be, for example, a thermal printer or a laser printer. In this embodiment, the printer 10 is a roll-to-roll type printer that moves the medium 5 in the sub-scanning direction X relative to a support table 13 (see FIG. 3 ) described below. However, the printer 10 may also be, for example, a flatbed type printer that moves the medium 5 together with the support table 13 in the sub-scanning direction X.

[0016] The printer 10 prints on a medium 5 as shown in Fig. 3. The medium 5 is, for example, a roll of recording paper, so-called roll paper. However, the medium 5 is not limited to roll of recording paper. For example, the medium 5 may be paper such as plain paper or inkjet printing paper, or may be a sheet or film made of resin such as polyvinyl chloride or polyester, a plate material, a fabric such as a woven fabric or a nonwoven fabric, or other medium.

[0017] As shown in FIG. 2, the printer 10 includes a printer body 11, a support stand 13, a medium moving mechanism 20, a guide rail 15, a carriage 17, an ink head 40 (see FIG. 4), a head moving mechanism 30, and a control device 60.

[0018] 2, the printer body 11 has a casing that extends in the main scanning direction Y. The printer body 11 is supported by legs 12. The legs 12 are provided on the bottom surface of the printer body 11 and extend downward from the bottom surface of the printer body 11.

[0019] 3, the support table 13 supports the medium 5. The upper surface of the support table 13 extends in the main scanning direction Y and the sub-scanning direction X. Here, the medium 5 is placed on the upper surface of the support table 13. Printing on the medium 5 is performed on the support table 13. The support table 13 is a so-called platen.

[0020] The medium moving mechanism 20 is a mechanism that moves the medium 5 in the sub-scanning direction X relative to the ink head 40. Here, the medium moving mechanism 20 moves the medium 5 supported by the support base 13 in the sub-scanning direction X. The configuration of the medium moving mechanism 20 is not particularly limited.

[0021] In this embodiment, the medium moving mechanism 20 includes a pinch roller 21, a grit roller 22, and a feed motor 23. The pinch roller 21 is provided above the support base 13 and below the guide rail 15, and presses down the medium 5 from above. The pinch roller 21 is disposed behind the carriage 17 in a plan view. The grit roller 22 is provided on the support base 13, and is a member having a cylindrical outer periphery. The grit roller 22 is embedded in the support base 13 with its upper surface exposed. The grit roller 22 faces the pinch roller 21. The feed motor 23 is connected to the grit roller 22.

[0022] In addition, although two pinch rollers 21 are arranged in Fig. 2, in reality, a plurality of (e.g., three or more) pinch rollers 21 are arranged side by side in the main scanning direction Y. Furthermore, although two grit rollers 22 are also arranged in Fig. 2, in reality, a plurality of grit rollers 22 are arranged side by side in the main scanning direction Y so as to be disposed below the pinch rollers 21. The plurality of grit rollers 22 are provided and connected to, for example, a shaft (not shown) extending in the main scanning direction Y. A feed motor 23 is connected to one of the plurality of grit rollers 22, or to the above-mentioned shaft.

[0023] Here, when the feed motor 23 is driven with the medium 5 sandwiched between the pinch roller 21 and the grit roller 22, the grit roller 22 rotates. This causes the medium 5 supported by the support base 13 to move in the sub-scanning direction X.

[0024] The guide rail 15 is disposed above the support base 13. The guide rail 15 is disposed parallel to the upper surface of the support base 13, and extends in the main scanning direction Y as shown in Fig. 2. A carriage 17 is engaged with the guide rail 15. The carriage 17 is slidably provided on the guide rail 15, and configured to be movable in the main scanning direction Y.

[0025] As shown in FIG. 3, the ink head 40 ejects ink. The ink head 40 is provided on the carriage 17. The ink head 40 is supported by the carriage 17 so that the bottom surface of the ink head 40 is exposed downward. The number of ink heads 40 is not particularly limited. In this embodiment, as shown in FIG. 4, the number of ink heads 40 is four. The four ink heads 40 are arranged side by side in the main scanning direction Y. Each ink head 40 has a nozzle surface 45. The nozzle surface 45 constitutes the bottom surface of the ink head 40. Nozzles 46 that eject ink are formed on each nozzle surface 45. A plurality of nozzles 46 are formed side by side in the sub-scanning direction X. Here, a row of a plurality of nozzles 46 is referred to as a nozzle row 48. The number of nozzle rows 48 for one ink head 40 is two. However, the number of nozzle rows 48 for one ink head 40 is not particularly limited, and may be one or may be three or more.

[0026] The ink ejected from the ink head 40 (here, the nozzles 46) is, for example, one of a process color ink and a special color ink. Here, the process color ink includes, for example, cyan ink, magenta ink, yellow ink, and black ink. The special color ink is an ink of a color other than the process color ink. The special color ink includes, for example, white ink, clear ink, gloss ink, primer ink, fluorescent ink, metallic ink, orange ink, red ink, violet ink, blue ink, and green ink. Here, one color of ink is ejected from one nozzle row 48. Note that there is no limitation on the material of the ink, and various materials that have been conventionally used as ink materials for inkjet printers and the like can be used. The ink may be, for example, a solvent-based pigment ink or a water-based pigment ink. Alternatively, the ink may be a water-based dye ink or an ultraviolet-curable ink that is cured by receiving ultraviolet light.

[0027] 2, the head moving mechanism 30 is a mechanism that moves the carriage 17 and the ink head 40 (see FIG. 4) in the main scanning direction Y relative to the medium 5 supported by the support base 13. Here, the head moving mechanism 30 moves the carriage 17 and the ink head 40 in the main scanning direction Y. Note that the configuration of the head moving mechanism 30 is not particularly limited.

[0028] In this embodiment, the head moving mechanism 30 includes left and right pulleys 31a and 31b, a belt 32, and a scan motor 33. The left pulley 31a is provided around the left end of the guide rail 15. The right pulley 31b is provided around the right end of the guide rail 15. The belt 32 is, for example, an endless belt, and is wound around the left and right pulleys 31a and 31b. The carriage 17 is attached and fixed to the belt 32. The scan motor 33 is connected to the right pulley 31b. Here, the right pulley 31b rotates when the scan motor 33 is driven, and the belt 32 runs. As a result, the carriage 17 and the ink head 40 move in the main scanning direction Y along the guide rail 15.

[0029] 2, in this embodiment, an operation panel 50 is provided on the right end of the printer body 11 of the printer 10. The operation panel 50 is provided with a display screen 51 that displays the status of the printer 10, and operation keys 52 that are operated by the user.

[0030] The control device 60 is a device that performs control related to printing, etc. The configuration of the control device 60 is not particularly limited. The control device 60 is, for example, a microcomputer. The hardware configuration of the microcomputer is not particularly limited. The control device 60 includes, for example, an I / F, a CPU, a ROM, and a RAM. The control device 60 is provided inside the printer body 11. However, the control device 60 may be realized by a computer or the like that is installed outside the printer body 11. In this case, the control device 60 is connected to a control board (not shown) of the printer 10 via wire or wirelessly so as to be able to communicate with it.

[0031] 5, in this embodiment, the control device 60 is communicatively connected to the medium moving mechanism 20 (more specifically, the feed motor 23), the head moving mechanism 30 (more specifically, the scan motor 33), the ink head 40, and the operation panel 50. The control device 60 controls the medium moving mechanism 20, the head moving mechanism 30, the ink head 40, and the operation panel 50.

[0032] The printer 10 according to the present embodiment has been described above. Next, the terminal 70 shown in FIG. 1 will be described. The terminal 70 is used by a user. The terminal 70 may be used to instruct the printer 10 to print, or may be used to operate the printer 10. In the present embodiment, the terminal 70 is communicably connected to the control device 60 (see FIG. 5) of the printer 10. As shown in FIG. 1, for example, the terminal 70 may be communicably connected to the printer 10 via the Internet 200. The terminal 70 is realized, for example, by a desktop or laptop personal computer used by a user. The terminal 70 may be realized by a dedicated computer or a general-purpose computer. The terminal 70 may also be a portable terminal such as a so-called smart fan or tablet terminal.

[0033] As shown in Fig. 5, the terminal 70 includes a display screen 71, an operation means 72, and a terminal control device 73. For example, information about the printer 10 communicably connected to the terminal 70 is displayed on the display screen 71. The operation means 72 is operated by a user. For example, the user operates the operation means 72 to select information on the display screen 71 or to switch the information displayed on the display screen 71. The operation means 72 is realized by a keyboard, a mouse, a touch panel, or the like.

[0034] The terminal control device 73 is, for example, a microcomputer. The terminal control device 73 includes, for example, an I / F, a CPU, a ROM, and a RAM. The terminal control device 73 is electrically connected to the display screen 71 and the operation means 72. In this embodiment, a dedicated application for operating the printer 10 is installed in the terminal control device 73. The user starts the application via the operation means 72 to display it on the display screen 71. This makes it possible to use the terminal 70 to instruct the printer 10 to perform printing, etc.

[0035] Next, a description will be given of the fault detection device 100 shown in Fig. 1. The fault detection device 100 is a device that detects faults in the printer 10. Here, a "fault in the printer 10" refers to an operation of the printer 10 that causes a decrease in print quality, an operation of the printer 10 when a component breaks down, or an operation of the printer 10 that is a precursor to a fault before the printer 10 breaks down.

[0036] For example, the printing system 1 is realized by a so-called cloud computing environment. In this case, the defect detection device 100 is a cloud server provided in the cloud computing environment. However, the printing system 1 may be realized by a so-called client-server environment. In this case, the defect detection device 100 is a server provided in the client-server environment. Also, the printing system 1 may be realized in a so-called stand-alone environment. In this case, the defect detection device 100 is realized by a personal computer. The defect detection device 100 may be realized by a dedicated computer or a general-purpose computer. As shown in FIG. 1, the defect detection device 100 is communicably connected to the printer 10 (more specifically, the control device 60 (see FIG. 5)) and the terminal 70 (more specifically, the terminal control device 73 (see FIG. 5)). Here, the defect detection device 100 is communicably connected to the printer 10 and the terminal 70 via the Internet 200.

[0037] Fig. 6 is a diagram showing a learning model MD1 generated by machine learning. In this embodiment, as shown in Fig. 6, the malfunction detection device 100 acquires an operation sound S1 of the printer 10 when the printer 10 prints, and determines whether or not a malfunction has occurred in the printer 10 by machine learning using the acquired operation sound S1. Then, when it is determined by machine learning that a malfunction has occurred in the printer 10, the malfunction detection device 100 notifies the terminal 70.

[0038] In this embodiment, in order to realize a series of controls for determining whether or not a malfunction occurs in the printer 10, the printer 10 is provided with a recorder 55 as shown in FIG. 5. The recorder 55 records the operation sound S1 (see FIG. 6) of the printer 10 during printing. Here, the operation sound S1 of the printer 10 refers to the sound of the motor driving during printing, or the sound of members such as the carriage 17 sliding. FIG. 7 is a diagram showing the types of operation sound S1. In this embodiment, as shown in FIG. 7, the types of operation sound S1 of the printer 10 include at least a first motor sound S21, a second motor sound S22, and a sliding sound S23.

[0039] The first motor sound S21 is the motor sound of the scan motor 33 (see FIG. 2) of the head moving mechanism 30 when the scan motor 33 is driven. The first motor sound S21 is the sound generated when the scan motor 33 is driven and the ink head 40 is moving in the main scanning direction Y. The second motor sound S22 is the motor sound of the feed motor 23 (see FIG. 3) of the medium moving mechanism 20 when the feed motor 23 is driven. The second motor sound S22 is the sound generated when the feed motor 23 is driven and the medium 5 supported by the support base 13 is moving in the sub-scanning direction X.

[0040] 2, when the ink head 40 is moved in the main scanning direction Y by the head moving mechanism 30, the carriage 17 moves in the main scanning direction Y while sliding against the guide rail 15. The sliding sound S23 in FIG. 7 is a sound generated when the carriage 17 slides against the guide rail 15.

[0041] The position where the recorder 55 is disposed is not particularly limited. The recorder 55 is provided, for example, in the printer body 11 (see FIG. 2). The recorder 55 may be provided near the scan motor 33, near the feed motor 23, or in the carriage 17. The number of recorders 55 is not particularly limited, and may be one or more. When there are multiple recorders 55, for example, a first recorder that mainly records the first motor sound S21 of the scan motor 33, a second recorder that mainly records the second motor sound S22 of the feed motor 23, and a third recorder that mainly records the sliding sound S23 of the carriage 17 may be provided.

[0042] The type of recorder 55 is not particularly limited as long as it can record the operation sound S1 of the printer 10. In this embodiment, the recorder 55 has a microphone, although not shown. The recorder 55 is also communicably connected to the control device 60 of the printer 10. The operation sound S1 recorded by the recorder 55 (in other words, the operation sound S1 picked up by the microphone of the recorder 55) is sent to the control device 60.

[0043] In this embodiment, the control device 60 of the printer 10 includes a printer storage unit 61, an acquisition unit 63, and a communication unit 65. The terminal control device 73 of the terminal 70 includes a terminal storage unit 75, a receiving unit 77, and a display unit 79. The fault detection device 100 includes a storage unit 101, a sound acquisition unit 103, a model generation unit 105, a framing processing unit 107, a learning model processing unit 109, and a notification unit 111. Each unit of the control device 60, each unit of the terminal control device 73, and each unit of the fault detection device 100 may be realized by one or a plurality of processors, or may be realized by a circuit.

[0044] In this embodiment, as described above, a malfunction of the printer 10 is detected by machine learning (e.g., deep learning). In order to detect a malfunction of the printer 10 by machine learning, a learning model MD1 in FIG. 6 is generated by machine learning. This learning model MD1 is generated by the model generation unit 105 of the malfunction detection device 100 in FIG. 5. As shown in FIG. 6, the model generation unit 105 generates the learning model MD1 by machine learning using the operating sound S1 of the printer 10 as teacher data. Note that in this embodiment, the model generation unit 105 embodies a model generation process of the malfunction detection method for the printer 10.

[0045] The operation sound S1 of the printer 10 is acquired by the sound acquisition unit 103 of FIG. 5. The sound acquisition unit 103 acquires the operation sound S1 when the printer 10 prints. In this embodiment, the sound acquisition unit 103 acquires the operation sound S1 from the control device 60 of the printer 10. Here, in the printer 10, the acquisition unit 63 of FIG. 5 acquires the operation sound S1 recorded by the recorder 55 when printing. The operation sound S1 acquired by the acquisition unit 63 may be stored in, for example, the printer storage unit 61 (see FIG. 5) of the printer 10. The communication unit 65 of the printer 10 transmits the operation sound S1 acquired by the acquisition unit 63 to the malfunction detection device 100. In this way, the sound acquisition unit 103 of the malfunction detection device 100 can acquire the operation sound S1 transmitted by the communication unit 65 of the printer 10.

[0046] In this embodiment, as shown in FIG. 6, the learning model MD1 is a model in which the input is the operation sound S1 of the printer 10 and the output is the presence or absence of a malfunction of the printer 10. As described above, the model generation unit 105 generates the learning model MD1 by machine learning using the operation sound S1 of the printer 10 as training data. Here, the operation sound S1 includes a malfunction operation sound S11 and a normal operation sound S12. The malfunction operation sound S11 is the operation sound S1 of the printer 10 when a malfunction occurs in the printer 10. The malfunction operation sound S11 also includes the operation sound S1 of the printer 10 when a malfunction is expected to occur in the printer 10. The normal operation sound S12 is the operation sound S1 of the printer 10 when no malfunction occurs in the printer 10 and is the operation sound S1 when the printer 10 is operating normally.

[0047] In this embodiment, depending on the location where the printer 10 is installed, environmental sound S13 (see FIG. 9) may be generated. The environmental sound S13 is a sound generated from other than the printer 10, for example, a location where the printer 10 is installed or a sound generated from the surroundings of that location. The environmental sound S13 includes, for example, a sound generated when a person moves or carries an object, and a mechanical sound generated from a machine other than the printer 10. Here, the operating sound S1 acquired by the sound acquisition unit 103 may be mixed with the environmental sound S13. In other words, the malfunction operating sound S11 and the normal operating sound S12 may be mixed with the environmental sound S13.

[0048] 6, in this embodiment, the teacher data used to generate the learning model MD1 includes malfunction sound S11 and normal operation sound S12 as the operation sound S1 of the printer 10. Furthermore, the teacher data includes environmental sound S13 that is mixed with the operation sound S1.

[0049] FIG. 8 is a graph showing the normal operation sound S12. FIG. 9 is a graph showing the malfunction sound S11. In FIG. 8 and FIG. 9, the horizontal axis indicates the frame number after the framing process, and the vertical axis indicates the frequency. For example, as shown in FIG. 8, the frequency of the normal operation sound S12 is a predetermined reference frequency NF1 or around the reference frequency NF1, and can be almost constant at the reference frequency NF1. However, as shown in FIG. 9, there is a timing when the frequency of the malfunction sound S11 temporarily becomes higher than the reference malfunction frequency NF2. A frequency higher than the reference frequency NF1 is set for this reference malfunction frequency NF2. The reference frequency NF1 and the reference malfunction frequency NF2 are stored in advance in the storage unit 101 of FIG. 5. In the example of FIG. 9, in the malfunction sound S11, a malfunction occurs in the printer 10 at the timing T1. Therefore, since the frequency of the malfunction sound S11 is higher than the reference malfunction frequency NF2 at the timing T1, it is determined that a malfunction occurs in the printer 10 when the malfunction sound S11 is recorded.

[0050] In addition, the normal operation sound S12 in FIG. 8 and the malfunction operation sound S11 in FIG. 9 are mixed with the environmental sound S13. In the normal operation sound S12 in FIG. 8 and the malfunction operation sound S11 in FIG. 9, the environmental sound S13 is generated at timing T2. Here, as shown in FIG. 9, the environmental sound S13 is lower in frequency than the malfunction operation sound S11 at timing T1 when a malfunction occurs in the printer 10. The environmental sound S13 is, for example, higher than the reference frequency NF1 and lower than the reference malfunction frequency NF2. Here, the reference malfunction frequency NF2 is set to be higher than the environmental sound S13 and lower than the frequency when a malfunction occurs in the printer 10. In this embodiment, even if the operation sound S1 of the printer 10 is mixed with the environmental sound S13, by adjusting the reference malfunction frequency NF2, it is possible to make it difficult to erroneously detect that a malfunction occurs in the printer 10 from the environmental sound S13, even if a malfunction does not occur in the printer 10.

[0051] However, a filtering process may be performed on the operation sound S1 mixed with the environmental sound S13 to remove the environmental sound S13 from the operation sound S1. In this case, the operation sound S1 from which the environmental sound S13 has been removed is added to the teacher data when generating the learning model MD1.

[0052] In this embodiment, each time an operation sound S1 is acquired by the sound acquisition unit 103 in Fig. 5, the operation sound S1 is sequentially added to the teacher data. When an operation sound S1 is added to the teacher data, the learning model MD1 that has already been trained is subjected to machine learning based on the teacher data including the added operation sound S1, thereby sequentially updating the learning model MD1. As the number of operation sounds S1 included in the teacher data increases, the accuracy of the output of the presence or absence of a malfunction of the printer 10 improves.

[0053] Next, a method for detecting a malfunction of the printer 10 while the printer 10 is printing (i.e., a control procedure for detecting that a malfunction has occurred in the printer 10) will be described with reference to the flowchart in Fig. 10. Step S101 in Fig. 10 is executed while the printer 10 is printing, for example, as shown in Fig. 2, when the feed motor 23 of the medium moving mechanism 20 and the scan motor 33 of the head moving mechanism 30 are appropriately driven to appropriately move the medium 5 supported by the support base 13 in the sub-scanning direction X, or when the carriage 17 is appropriately moving in the main scanning direction Y relative to the guide rail 15.

[0054] In step S101 of FIG. 10, a sound acquisition process is executed. In the sound acquisition process, an operation sound S1 generated during printing by the printer 10 is acquired. The sound acquisition process is embodied by the sound acquisition unit 103 of FIG. 5. The sound acquisition unit 103 is configured to acquire the operation sound S1 from the printer 10 at a predetermined interval. Here, while the printer 10 is printing, the acquisition unit 63 of the printer 10 acquires the operation sound S1 recorded by the recorder 55 (see FIG. 5). For example, the acquisition unit 63 acquires the operation sound S1 from the recorder 55 at a predetermined interval. Each operation sound S1 acquired by the acquisition unit 63 is saved in one file and stored in the printer storage unit 61 of FIG. 5. The recording time of one operation sound S1 saved in the file is, for example, the same length as the above-mentioned predetermined interval. Here, at the timing when the operation sound S1 is acquired by the acquisition unit 63, the communication unit 65 of the printer 10 (see FIG. 5) transmits the operation sound S1 acquired by the acquisition unit 63 to the malfunction detection device 100. The sound acquisition unit 103 of the malfunction detection device 100 acquires the operation sound S1 transmitted by the communication unit 65. The operation sound S1 of the printer 10 acquired by the sound acquisition unit 103 is stored in the storage unit 101 of FIG.

[0055] Next, in step S103 of FIG. 10, a framing process is executed. In the framing process, the operation sound S1 acquired in the sound acquisition process is subjected to framing process at intervals of a predetermined frame shift, thereby dividing the operation sound S1 into a plurality of frames FL1. Note that the frames FL1 divided by the framing process are numbered. Here, the framing process is embodied by the framing processor 107 of the fault detection device 100 of FIG. 5. The framing processor 107 executes framing process at intervals of a predetermined frame shift on the operation sound S1 acquired by the sound acquisition unit 103. As a result, the operation sound S1 is divided into a plurality of frames FL1. Here, the predetermined frame shift is stored in advance in the storage unit 101. The specific value of the predetermined frame shift is not particularly limited, but is appropriately set according to the recording time of the operation sound S1 and the processing speed of the fault detection device 100.

[0056] Next, in step S105 of Fig. 10, a learning model processing step is executed. In the learning model processing step, the multiple frames FL1 created in the framing processing step are input to the learning model MD1, and a process is performed to output a result of whether or not a malfunction has occurred in the printer 10. Here, the learning model processing step is embodied by the learning model processing unit 109 of Fig. 5. The learning model processing unit 109 inputs the multiple frames FL1 after framing processing has been performed on the operation sound S1 to the learning model MD1 (here, the learned learning model MD1), and outputs a result of whether or not a malfunction has occurred in the printer 10.

[0057] In this embodiment, the learning model MD1 outputs a result of whether or not a malfunction has occurred in the printer 10 for each frame FL1. For example, if the frequency of the operation sound S1 in the frame FL1 is higher than the reference malfunction frequency NF2 (see FIG. 9), a result that a malfunction has occurred in the printer 10 is output for that frame FL1. The learning model processing unit 109 then determines that a malfunction has occurred in the printer 10 when the ratio of frames FL1 output by the learning model MD1 indicating that the printer 10 is malfunctioning to the entire plurality of frames FL1 after the framing processing by the framing processing unit 107 is equal to or greater than a predetermined determination ratio. That is, the learning model processing unit 109 determines that a malfunction has occurred in the printer 10 based on the operation sound S1 when the ratio of frames FL1 output by the learning model MD1 indicating that the printer 10 is malfunctioning to all frames FL1 after the framing processing for the operation sound S1 is equal to or greater than a predetermined determination ratio.

[0058] In this embodiment, regardless of the result of machine learning using the learning model MD1 in step S105 of FIG. 10 (here, the output result as to whether or not a malfunction has occurred in the printer 10), the operating sound S1 acquired by the sound acquisition unit 103 is sequentially added to the teacher data used in machine learning for the learning model MD1.

[0059] Next, in step S107 of Fig. 10, it is determined whether the output result by the learning model MD1 in step S105 indicates that a malfunction has occurred in the printer 10 or not. If it is determined that the output result by the learning model MD1 indicates that no malfunction has occurred in the printer 10, the flowchart in Fig. 10 is terminated. In this case, when the next operating sound S1 of the printer 10 is transmitted from the printer 10 to the malfunction detection device 100, step S101 in Fig. 10 is executed anew.

[0060] On the other hand, if it is determined in step S107 of FIG. 10 that the output result by the learning model MD1 in step S105 is an output indicating that a malfunction has occurred in the printer 10, the process proceeds to step S109 of FIG. 10. In step S109, a notification process is executed. In the notification process, the user is notified that a malfunction has occurred in the printer 10. Here, the notification process is embodied by the notification unit 111 of FIG. 5. When the learning model processing unit 109 determines that a malfunction has occurred in the printer 10, the notification unit 111 notifies the user that a malfunction has occurred in the printer 10.

[0061] The specific method of notification by the notification unit 111 is not particularly limited. In this embodiment, the notification unit 111 notifies the terminal 70 (see FIG. 5) that a malfunction has occurred in the printer 10, thereby notifying the user via the terminal 70. For example, the notification unit 111 transmits a malfunction occurrence signal to the terminal 70. At this time, an identifier for identifying the printer 10 in which the malfunction has occurred may be transmitted to the terminal 70 together with the malfunction occurrence signal. The receiving unit 77 of the terminal 70 in FIG. 5 receives the malfunction occurrence signal transmitted from the notification unit 111 and the identifier of the printer 10 in which the malfunction has occurred. Thereafter, the display unit 79 of the terminal 70 in FIG. 5 displays a message on the display screen 71 indicating that a malfunction has occurred in the printer 10. A template of this message is stored in the terminal storage unit 75 in advance. Here, when the receiving unit 77 receives the identifier, the display unit 79 may display information on the printer 10 identified based on the identifier on the display screen 71. The user can know that a malfunction has occurred in the printer 10 by looking at the message displayed on the display screen 71.

[0062] As described above, in this embodiment, as shown in FIG. 1, the printing system 1 includes the printer 10 and the fault detection device 100. As shown in FIG. 5, the fault detection device 100 includes the sound acquisition unit 103, the model generation unit 105, the framing processing unit 107, and the learning model processing unit 109. The sound acquisition unit 103 acquires the operation sound S1 during printing of the printer 10 as in step S101 of FIG. 10. The model generation unit 105 uses the operation sound S1 acquired by the sound acquisition unit 103 as teacher data, and generates a learning model MD1 by machine learning, in which the input is the operation sound S1 and the output is the presence or absence of a fault in the printer 10, as shown in FIG. 6. The framing processing unit 107 creates multiple frames FL1 by performing framing processing on the operation sound S1 acquired by the sound acquisition unit 103 at predetermined frame shift intervals as in step S103 of FIG. 10. The learning model processing unit 109 inputs a plurality of frames FL1 created by the framing processing unit 107 into the learning model MD1 as in step S105 of FIG. 10, and outputs whether or not a malfunction has occurred in the printer 10. Here, for example, the operation sound S1 of the printer 10 when a malfunction has occurred (for example, the malfunction operation sound S11 (see FIG. 6)) is considered to be different from the operation sound S1 of the printer 10 when no malfunction has occurred (for example, the normal operation sound S12 (see FIG. 6)). For example, the malfunction operation sound S11 has a higher frequency (for example, the maximum frequency) than the normal operation sound S12. Therefore, in this embodiment, the operation sound S1 of the printer 10 is used as teacher data, and the learning model MD1 generated by machine learning is used to output a result of whether or not a malfunction has occurred in the printer 10 from the learning model MD1. Therefore, it is possible to determine whether or not a malfunction has occurred in the printer 10 from the operation sound S1 of the printer 10 by machine learning. Therefore, when a malfunction has occurred in the printer 10, for example, printing can be stopped, and therefore it is possible to suppress a decrease in print quality caused by the malfunction of the printer 10.

[0063] In this embodiment, as shown in FIG. 6, the operation sound S1 includes a malfunction operation sound S11 when a malfunction occurs in the printer 10, and a normal operation sound S12 when no malfunction occurs in the printer 10. As shown in FIG. 6, the model generation unit 105 in FIG. 5 generates a learning model MD1 using the malfunction operation sound S11 and the normal operation sound S12 as training data. As a result, the training data is accumulated with the operation sound S1 when a malfunction occurs in the printer 10. Therefore, even if various malfunctions occur in the printer 10, machine learning using the learning model MD1 makes it easier to detect the occurrence of a malfunction in the printer 10 from the operation sound S1.

[0064] In this embodiment, the learning model processing unit 109 in Fig. 5 determines that a malfunction has occurred in the printer 10 when the proportion of frames FL1 output by the learning model MD1 indicating that the printer 10 is malfunctioning is equal to or greater than a predetermined determination proportion, for a plurality of frames FL1. In this way, when there are many frames FL1 output indicating that the printer 10 is malfunctioning, it is highly likely that the printer 10 is malfunctioning continuously. Therefore, it is possible to detect that a malfunction has occurred continuously in the printer 10 based on the operating sound S1.

[0065] In this embodiment, the model generation unit 105 in Fig. 5 generates a learning model MD1 using the operation sound S1 mixed with the environmental sound S13 (see Fig. 9) as training data. In this way, by using the operation sound S1 and the environmental sound S13 mixed with the operation sound S1 as training data, it is possible to generate a learning model MD1 that takes the environmental sound S13 into consideration. Therefore, even if the operation sound S1 is mixed with the environmental sound S13, it is possible to detect that a malfunction has occurred in the printer 10.

[0066] In this embodiment, as shown in Fig. 5, the malfunction detection device 100 includes a notification unit 111. When the learning model processing unit 109 determines that a malfunction has occurred in the printer 10, the notification unit 111 notifies the user that a malfunction has occurred in the printer 10, as in step S109 in Fig. 10. This allows the user to know that a malfunction has occurred in the printer 10. Therefore, the user can take appropriate measures against the malfunction that has occurred in the printer 10, and it is possible to prevent the printer 10 from continuing to print when a malfunction has occurred in the printer 10.

[0067] In this embodiment, as shown in FIG. 5, the printer 10 includes a recorder 55 that records the operation sound S1 of the printer 10, and a control device 60. The control device 60 includes an acquisition unit 63 and a communication unit 65. The acquisition unit 63 acquires the operation sound S1 recorded by the recorder 55. The communication unit 65 transmits the operation sound S1 acquired by the acquisition unit 63 to the malfunction detection device 100. The sound acquisition unit 103 of the malfunction detection device 100 acquires the operation sound S1 transmitted by the communication unit 65 of the printer 10. This allows the sound acquisition unit 103 to acquire the operation sound S1 recorded by the recorder 55 of the printer 10 from the printer 10. Therefore, since the operation sound S1 is not acquired by the sound acquisition unit 103 of the malfunction detection device 100 via another device or the like, the operation sound S1 is unlikely to be damaged in the process of being transmitted.

[0068] <Other embodiments> Next, another embodiment will be described. As described above, the operation sound S1 has a first motor sound S21 (see FIG. 7) which is the motor sound of the scan motor 33 of the head moving mechanism 30 when the scan motor 33 is driven, and a second motor sound S22 (see FIG. 7) which is the motor sound of the feed motor 23 of the medium moving mechanism 20 when the feed motor 23 is driven. In this embodiment, the learning model MD1 may be used depending on the type of operation sound S1 (here, the first motor sound S21 and the second motor sound S22).

[0069] FIG. 11 is a diagram showing a first learning model MD11. FIG. 12 is a diagram showing a second learning model MD12. In this embodiment, as shown in FIG. 11 and FIG. 12, the learning model MD1 has a first learning model MD11 and a second learning model MD12. As shown in FIG. 11, the first learning model MD11 is a model generated by machine learning using the first motor sound S21 as teacher data. In the first learning model MD11, the input is the first motor sound S21, and the output is the presence or absence of a malfunction of the printer 10 based on the scan motor 33 (hereinafter, also referred to as a malfunction of the scan motor 33). The malfunction of the scan motor 33 includes a malfunction of the scan motor 33 caused by a failure of another member. Here, the model generation unit 105 of the malfunction detection device 100 of FIG. 5 generates the first learning model MD11 by machine learning using the first motor sound S21 as teacher data. When the scan motor 33 is driven, the carriage 17 moves in the main scanning direction Y while sliding on the guide rail 15, which may cause a sliding sound S23 (see FIG. 7). That is, the sliding sound S23 may be generated simultaneously with the first motor sound S21. Therefore, the sliding sound S23 may be mixed into the first motor sound S21.

[0070] As shown in Fig. 12, the second learning model MD12 is a model generated by machine learning using the second motor sound S22 as training data. In the second learning model MD12, the input is the second motor sound S22, and the output is the presence or absence of a malfunction of the printer 10 due to the feed motor 23 (hereinafter also referred to as a malfunction of the feed motor 23). Malfunctions of the feed motor 23 include malfunctions of the feed motor 23 caused by failures of other components, etc. Here, the model generation unit 105 in Fig. 5 generates the second learning model MD12 by machine learning using the second motor sound S22 as training data.

[0071] In this embodiment, in the printer 10, the control device 60 in FIG. 5 controls the driving of the scan motor 33 and the feed motor 23. Therefore, the control device 60 can determine whether the operation sound S1 recorded by the recorder 55 of the printer 10 is either the first motor sound S21 or the second motor sound S22 based on the motor being controlled. Therefore, when transmitting the operation sound S1 from the communication unit 65 in FIG. 5 to the malfunction detection device 100, sound identification information indicating whether the operation sound S1 is the first motor sound S21 or the second motor sound S22 may be transmitted. The sound acquisition unit 103 in FIG. 5 can determine whether the operation sound S1 is the first motor sound S21 or the second motor sound S22 based on the sound identification information by receiving the sound identification information together with the operation sound S1.

[0072] In this embodiment, when the operation sound S1 is the first motor sound S21, the framing processor 107 in Fig. 5 performs framing processing on the first motor sound S21 at predetermined frame shift intervals to create multiple first frames. When the operation sound S1 is the second motor sound S22, the framing processor 107 performs framing processing on the second motor sound S22 at predetermined frame shift intervals to create multiple second frames.

[0073] 11, when the operating sound S1 is the first motor sound S21, the learning model processing unit 109 in Fig. 5 inputs multiple first frames of the first motor sound S21 to the first learning model MD11 and outputs a result of whether or not a malfunction has occurred in the printer 10 based on the scan motor 33. When it is output that a malfunction has occurred in the scan motor 33, the notification unit 111 in Fig. 5 notifies the user that a malfunction has occurred in the printer 10 due to the malfunction of the scan motor 33.

[0074] 12, when the operating sound S1 is the second motor sound S22, the learning model processing unit 109 in Fig. 5 inputs a plurality of second frames of the second motor sound S22 to the second learning model MD12 and outputs a result of whether or not a malfunction has occurred in the printer 10 based on the feed motor 23. When it is output that a malfunction has occurred in the feed motor 23, the notification unit 111 in Fig. 5 notifies the user that a malfunction has occurred in the printer 10 due to the malfunction of the feed motor 23.

[0075] In this manner, in this embodiment, as shown in FIG. 11, when the operation sound S1 is the first motor sound S21, the first learning model MD11 is used. Therefore, since the first learning model MD11 is generated according to the characteristics of the scan motor 33, it is possible to easily detect a malfunction of the scan motor 33 from the first motor sound S21. Also, in this embodiment, as shown in FIG. 12, when the operation sound S1 is the second motor sound S22, the second learning model MD12 is used. Therefore, since the second learning model MD12 is generated according to the characteristics of the feed motor 23, it is possible to easily detect a malfunction of the feed motor 23 from the second motor sound S22. In this embodiment, by using the learning models MD11 and MD12 depending on the type of the operation sound S1, it is possible to easily identify which part of the printer 10 is malfunctioning. [Explanation of symbols]

[0076] 1 Printing System 5 Medium 10 Printers 13 Support stand 20 Media movement mechanism 23 Feed motor 30 Head movement mechanism 33 Scan Motor 40 Ink Head 55 Recorder 60 Control device 63 Acquisition Department 65 Communications Department 70 Terminals 100 Fault detection device 103 Sound acquisition section 105 Model Generation Unit 107 Frame Processing Unit 109 Learning model processing section 111 Notification Department MD1 Learning Model MD11 1st Learning Model MD12 Second Learning Model S1 Operation sound S11 Malfunctioning sound S12 Normal operation sound S13 Environmental Sound S21 1st motor sound S22 2nd motor sound

Claims

1. a sound acquisition unit that acquires an operating sound of the printer during printing; a model generation unit that generates a learning model by machine learning, the learning model having the operation sound acquired by the sound acquisition unit as training data, the operation sound as an input, and the presence or absence of a malfunction of the printer as an output; a framing processing unit that performs framing processing on the operation sound acquired by the sound acquisition unit at predetermined frame shift intervals to create a plurality of frames; a learning model processing unit that inputs the plurality of frames created by the framing processing unit into the learning model and outputs whether or not a malfunction has occurred in the printer; A printer failure detection device comprising:

2. The operation sound is a malfunctioning operation sound when a malfunction occurs in the printer; A normal operation sound of the printer without any malfunction; having 2. The printer malfunction detection device according to claim 1, wherein the model generation unit generates the learning model using the malfunction sound and the normal operation sound as the teacher data.

3. The printer malfunction detection device of claim 1, wherein the learning model processing unit determines that a malfunction has occurred in the printer when the proportion of frames output by the learning model as a malfunction in the printer is equal to or greater than a predetermined judgment proportion, for a plurality of frames.

4. The printer fault detection device according to claim 1 , wherein the model generation unit generates the learning model using the operation sound mixed with an environmental sound as the teacher data.

5. The printer fault detection device as described in claim 1, further comprising a notification unit that notifies a user that a fault has occurred in the printer when the learning model processing unit determines that a fault has occurred in the printer.

6. A printer failure detection device according to any one of claims 1 to 5, A printer; A printing system comprising:

7. The printer includes: a recorder for recording the operation sound of the printer; A control device; Equipped with The control device includes: an acquisition unit that acquires the operation sound recorded by the recorder; a communication unit that transmits the operation sound acquired by the acquisition unit to the malfunction detection device, The printing system according to claim 6 , wherein the sound acquisition unit of the malfunction detection device acquires the operation sound transmitted by the communication unit of the printer.

8. The printer includes: A support base for supporting the medium; An ink head that ejects ink; a head moving mechanism having a scan motor and configured to move the ink head in a main scanning direction relative to the support table; a medium moving mechanism having a feed motor and configured to move the medium supported by the support table in a sub-scanning direction relative to the ink head; Equipped with The operation sound is a first motor sound generated when the scan motor is driven; a second motor sound generated when the feed motor is driven; and having The model generation unit of the fault detection device A first learning model is generated by machine learning, the first motor sound is used as training data, an input is the first motor sound, and an output is a presence or absence of a malfunction of the printer based on the scan motor; and The printing system described in claim 6, wherein a second learning model is generated by machine learning, in which the second motor sound is used as training data, the second motor sound is used as an input, and the presence or absence of a malfunction of the printer based on the feed motor is used as an output.

9. The framing processing unit of the failure detection device A framing process is performed on the first motor sound at intervals of the frame shift to generate a plurality of first frames; and performing a framing process on the second motor sound at intervals of the frame shift to create a plurality of second frames; The learning model processing unit is A plurality of the first frames created by the framing processing unit are input to the first learning model, and a determination is made as to whether or not a malfunction of the printer has occurred based on the scan motor. And A printing system as described in claim 8, wherein a plurality of the second frames created by the framing processing unit are input into the second learning model, and an indication of whether or not a malfunction has occurred in the printer is output based on the feed motor.

10. A sound acquisition step of acquiring an operating sound of the printer during printing; a model generation process for generating a learning model by machine learning, the learning model having the operation sound acquired in the sound acquisition process as training data, the operation sound as an input, and the presence or absence of a malfunction of the printer as an output; a framing process step of performing a framing process on the operation sound acquired in the sound acquisition step at a predetermined frame shift interval to create a plurality of frames; a learning model processing step of inputting the plurality of frames created in the framing processing step into the learning model and outputting whether or not a malfunction has occurred in the printer; A method for detecting a malfunction in a printer.

Images