Sheet transport device, automatic document transport device, and image forming apparatus

Abstract

To provide a sheet conveyance device capable of reducing cost required for collecting data for machine learning.SOLUTION: An ADF which is a sheet conveyance device, collects an operation sound during a sheet conveyance with a sound collection microphone which is a sound collection part (S4), and calculates a Mel frequency cepstral coefficient, which is a feature quantity that quantitatively expresses a characteristic of the operation sound based on a collected operation sound (S5). Then, at least a calculated Mel frequency cepstral coefficient and a conveyance result are linked and stored as a data table in a non-volatile memory serving as a storage part (S14).SELECTED DRAWING: Figure 8

Description

Technical Field

[0001] The present invention relates to a sheet conveying device, an automatic document conveying device and and an image forming apparatus to and is related thereto.

Background Art

[0002] Conventionally, there is known a sheet conveying device including a conveying member that conveys a sheet, a sound collecting unit that collects an operation sound during sheet conveyance, a feature amount extraction unit that extracts a feature amount that quantitatively represents the characteristics of the operation sound collected by the sound collecting unit, and a conveyance abnormality determination unit that determines whether a conveyance abnormality has occurred based on the feature amount.

[0003] In Patent Document 1, as the above sheet conveying device, for the purpose of suppressing the occurrence of conveyance abnormalities, a feature amount that quantitatively represents the characteristics of the operation sound collected by the sound collecting unit is given to a support vector machine, which is one of the supervised machine learning methods, and the feature amount is classified into any one of three classes: normal conveyance, document deformation, and paper feed slip. And when the feature amount is classified as sheet slip and document deformation by the support vector machine, it is regarded as a conveyance abnormality. When it is classified as sheet slip, heating of the conveying member is performed, and when it is classified as document deformation, the paper feed cover is opened.

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, there is a problem that it is costly to collect a huge amount of learning data used for machine learning.

Means for Solving the Problems

[0005] To solve the above-mentioned problems, the present invention comprises a conveying member for conveying a sheet, a sound collection unit for collecting operating sounds during sheet conveyance, a feature quantity extraction unit for extracting feature quantities that quantitatively represent the characteristics of the operating sounds collected by the sound collection unit, a conveyance abnormality determination unit for determining whether or not a conveyance abnormality occurs based on the feature quantities, a sheet detection unit for detecting sheets on the conveyance path, a conveyance abnormality detection unit for detecting a conveyance abnormality based on the detection results of the sheet detection unit, and a storage unit for storing a data table in which at least the feature quantities and the detection results by the conveyance abnormality detection unit are linked. When the transport abnormality determination unit determines that a transport abnormality has occurred, it temporarily stops the transport of the sheet, prompts the user to remove the cause of the transport abnormality, and determines whether to store the detection result by the transport abnormality detection unit as a transport abnormality in the storage unit, linked to the characteristic quantity at the time the transport abnormality determination unit determined that a transport abnormality had occurred, based on whether the time from when the user removes the sheet from the sheet tray until when the sheet is reset in the sheet tray is within a predetermined time. It is characterized by the following: [Effects of the Invention]

[0006] According to the present invention, the cost of collecting data for machine learning can be reduced. [Brief explanation of the drawing]

[0007] [Figure 1] A schematic diagram showing a copier according to this embodiment. [Figure 2] A partially enlarged diagram showing an enlarged view of a part of the internal structure of the image forming unit in the same copier. [Figure 3] A partially enlarged view showing a portion of the tandem section, which consists of four process units, in the image forming unit. [Figure 4] A perspective view showing the scanner and ADF of the copier. [Figure 5] An enlarged diagram showing the main components of the ADF along with the top of the scanner. [Figure 6] A block diagram showing part of the electrical circuits of the ADF and scanner. [Figure 7] A diagram illustrating the placement of the sound-collecting microphones in the ADF. [Figure 8] A flowchart of the operational status determination process performed by the ADF controller. [Figure 9] This figure shows an example of a data table stored in the non-volatile memory of the ADF. [Figure 10]A control flow diagram for query control, which prompts the user for an inquiry when the classifier determines that a transport error has occurred. [Figure 11] This diagram illustrates an example of the transitions between inquiry screens displayed on the operation display unit of a copier's control panel. [Figure 12] A flowchart for determining the operating state when transport anomaly prediction using a classifier is not performed. [Modes for carrying out the invention]

[0008] The following describes embodiments of the present invention applied to an electrophotographic photocopier (hereinafter simply referred to as a photocopier). First, the basic configuration of the copier according to this embodiment will be described. Figure 1 is a schematic diagram showing the copier according to this embodiment. This copier comprises an image forming unit 1, a blank paper supply device 40, and a document reading device 50. The document reading device 50 has a scanner 150 fixed on the image forming unit 1 and an ADF 51, which is a document transport device supported by the scanner 150.

[0009] The blank paper supply device 40 includes two paper feed cassettes 42 arranged in multiple stages within the paper bank 41, a feed roller 43 that feeds the transfer paper from the paper feed cassettes, and a separation roller 45 that separates the fed transfer paper and supplies it to the paper feed path 44. It also includes multiple transport rollers 47 that transport the transfer paper to the paper feed path 37 of the image forming unit 1. The transfer paper in the paper feed cassettes is then fed into the paper feed path 37 in the image forming unit 1.

[0010] Figure 2 is a partially enlarged diagram showing a part of the internal configuration of the image forming unit. The image forming unit 1, which is an image forming means, includes an optical writing device 2, four process units 3K,Y,M,C that form toner images of K,Y,M,C colors, and a transfer unit 24. It also includes a paper transport unit 28, a pair of registration rollers 33, a fixing device 34, a switchback device 36, a paper feed path 37, etc. A light source such as a laser diode or LED, which is arranged in the optical writing device 2, is driven to irradiate the four drum-shaped photoreceptors 4K,Y,M,C with laser light L. This irradiation forms an electrostatic latent image on the surface of the photoreceptors 4K,Y,M,C, and this latent image is developed into a toner image via a predetermined development process. The subscripts K,Y,M,C after the symbols indicate the specifications for black, yellow, magenta, and cyan.

[0011] Each process unit 3K, Y, M, and C supports a photoreceptor and various devices arranged around it as a single unit on a common support, and is detachable from the main body of the image forming unit 1. Taking the black process unit 3K as an example, it has a photoreceptor 4K and a developing device 6K for developing the electrostatic latent image formed on its surface into a black toner image. It also has a drum cleaning device 15 for cleaning the transfer residue toner adhering to the surface of the photoreceptor 4K after it has passed through the primary transfer nip for K, which will be described later. In this copier, the four process units 3K, Y, M, and C are arranged opposite each other along the endless movement direction of the intermediate transfer belt 25, which will be described later, in a so-called tandem configuration.

[0012] Figure 3 is a partially enlarged view showing a portion of the tandem unit consisting of four process units 3K, Y, M, and C. Note that the four process units 3K, Y, M, and C have almost identical configurations except for the different toner colors they use; therefore, the subscripts K, Y, M, and C attached to each unit are omitted in the figure. As shown in the figure, process unit 3 has a charging device 5, a developing device 6, a drum cleaning device 15, an anti-static lamp 22, etc., surrounding the photoreceptor 4.

[0013] As the photoreceptor 4, a drum-shaped one is used in which a photosensitive layer is formed by applying an organic photosensitive material having photosensitivity to a base tube such as aluminum. However, an endless belt-shaped one may also be used.

[0014] The developing device 6 is configured to develop a latent image using a two-component developer containing a magnetic carrier and a non-magnetic toner. It has a stirring unit 7 that conveys the two-component developer accommodated inside while stirring and supplies it to the developing sleeve 12, and a developing unit 11 for transferring the toner in the two-component developer carried on the developing sleeve 12 to the photoreceptor 4.

[0015] The stirring unit 7 is provided at a position lower than the developing unit 11, and has two conveying screws 8 arranged in parallel with each other, a partition plate provided between these screws, a toner concentration sensor 10 provided on the bottom surface of the developing case 9, and the like.

[0016] The developing unit 11 has a developing sleeve 12 facing the photoreceptor 4 through an opening of the developing case 9, a magnet roller 13 fixedly provided inside it, a doctor blade 14 whose tip approaches the developing sleeve 12, and the like. The developing sleeve 12 is a non-magnetic rotatable cylindrical shape. The magnet roller 13 has a plurality of magnetic poles arranged sequentially in the rotational direction of the sleeve from the position facing the doctor blade 14. These magnetic poles apply magnetic force to the two-component developer on the sleeve surface at a predetermined position in the rotational direction. Thereby, the two-component developer sent from the stirring unit 7 is attracted and carried on the surface of the developing sleeve 12, and a magnetic brush along the magnetic field lines is formed on the sleeve surface.

[0017] The magnetic brush is regulated to an appropriate layer thickness when passing through the position facing the doctor blade 14 as the developing sleeve 12 rotates, and then is conveyed to the developing area facing the photoreceptor 4. Then, the toner is transferred onto the electrostatic latent image by the potential difference between the developing bias applied to the developing sleeve 12 and the electrostatic latent image of the photoreceptor 4, contributing to development. Further, as the developing sleeve 12 rotates, it returns to the developing unit 11 again. After detaching from the sleeve surface under the influence of the repulsive magnetic field formed between the magnetic poles of the magnet roller 13, it is returned to the stirring unit 7. In the stirring unit 7, an appropriate amount of toner is replenished to the two-component developer based on the detection result by the toner concentration sensor 10. Note that, as the developing device 6, instead of using a two-component developer, a one-component developer that does not contain a magnetic carrier may be adopted.

[0018] As the drum cleaning device 15, a type that presses a cleaning blade 16 made of polyurethane rubber against the photoreceptor 4 is used, but other types may also be used. For the purpose of enhancing the cleaning performance, in this example, a type having a contact conductive fur brush 17 whose outer peripheral surface is rotatable in the direction of the arrow in the figure and contacts the photoreceptor 4 is adopted. This fur brush 17 also serves to scrape the lubricant from the solid lubricant into fine powder and apply it to the surface of the photoreceptor 4. A metal electric field roller 18 for applying a bias to the fur brush 17 is provided so as to be rotatable in the direction of the arrow shown in the figure, and the tip of the scraper 19 is pressed against it. The toner adhering to the fur brush 17 is displaced to the electric field roller 18 to which a bias is applied while rotating in contact with the fur brush 17 in the counter direction. Then, after being scraped off from the electric field roller 18 by the scraper 19, it falls onto the recovery screw 20. The recovery screw 20 conveys the recovered toner toward the end in the direction orthogonal to the drawing plane in the drum cleaning device 15 and delivers it to an external recycling conveyance device 21. The recycling conveyance device 21 sends the delivered toner to the developing device 6 for recycling.

[0019] The static elimination lamp 22 eliminates static electricity from the photoreceptor 4 by light irradiation. After the static elimination, the surface of the photoreceptor 4 is uniformly charged by the charging device 23, and then light writing is performed by the light writing device 2. The charging device 23 uses a charging roller to which a charging bias is applied, which is rotated while in contact with the photoreceptor 4. A non-contact charging device such as a scorotron charger may also be used.

[0020] In Figure 2 shown above, K, Y, M, C toner images are formed on the photoreceptors 4K, Y, M, C of the four process units 3K, Y, M, C by the process described above.

[0021] Below the four process units 3K,Y,M,C, a transfer unit 24 is positioned. This transfer unit 24 moves an intermediate transfer belt 25, which is stretched by multiple rollers, in an endless clockwise direction in the figure while keeping it in contact with the photoreceptors 4K,Y,M,C. This forms a primary transfer nip for K,Y,M,C where the photoreceptors 4K,Y,M,C and the intermediate transfer belt 25 come into contact. Near the primary transfer nip for K,Y,M,C, primary transfer rollers 26K,Y,M,C, positioned inside the belt loop, press the intermediate transfer belt 25 toward the photoreceptors 4K,Y,M,C. A primary transfer bias is applied to each of these primary transfer rollers 26K,Y,M,C by a power supply. As a result, a primary transfer electric field is formed in the primary transfer nip for K,Y,M,C, which electrostatically moves the toner image on the photoreceptors 4K,Y,M,C toward the intermediate transfer belt 25. As the intermediate transfer belt 25 moves endlessly in a clockwise direction in the diagram, it sequentially passes through the primary transfer nips for K, Y, M, and C. On the front surface of the intermediate transfer belt 25, the toner images are sequentially superimposed and transferred at each primary transfer nip. This superimposed primary transfer forms a four-color superimposed toner image (hereinafter referred to as the four-color toner image) on the front surface of the intermediate transfer belt 25.

[0022] In the diagram, below the transfer unit 24, a paper transport unit 28 is provided, which moves an endless paper transport belt 29 by stretching it between the drive roller 30 and the secondary transfer roller 31. The intermediate transfer belt 25 and the paper transport belt 29 are sandwiched between the secondary transfer roller 31 and the lower tension roller 27 of the transfer unit 24. As a result, a secondary transfer nip is formed where the front surface of the intermediate transfer belt 25 and the front surface of the paper transport belt 29 come into contact. A secondary transfer bias is applied to the secondary transfer roller 31 by a power supply. Meanwhile, the lower tension roller 27 of the transfer unit 24 is grounded. As a result, a secondary transfer electric field is formed in the secondary transfer nip.

[0023] A pair of register rollers 33 is positioned on the right side of the secondary transfer nip in the diagram. The transfer paper, sandwiched between the rollers, is fed to the secondary transfer nip at a timing that synchronizes with the four-color toner image on the intermediate transfer belt 25. Inside the secondary transfer nip, the four-color toner image on the intermediate transfer belt 25 is transferred to the transfer paper in one go due to the influence of the secondary transfer electric field and nip pressure, and combined with the white of the transfer paper, it becomes a full-color image. After passing through the secondary transfer nip, the transfer paper separates from the intermediate transfer belt 25 and is held on the front surface of the paper transport belt 29, and is transported to the fixing device 34 as it moves endlessly.

[0024] On the surface of the intermediate transfer belt 25 that has passed through the secondary transfer nip, residual toner that was not transferred to the transfer paper by the secondary transfer nip is attached. This residual toner is scraped off by a belt cleaning device that comes into contact with the intermediate transfer belt 25.

[0025] The transfer paper, transported to the fixing device 34, has the full-color image fixed by pressurization and heating within the fixing device 34. After that, it is sent from the fixing device 34 to the paper discharge roller pair 35 and then discharged outside the machine.

[0026] In Figure 1 shown above, a switchback device 36 is positioned below the paper transport unit 28 and the fixing device 34. This allows the transfer paper, after image fixing on one side has been completed, to have its path switched by a switching claw to the transfer paper reversing device, where it is reversed and enters the secondary transfer nip again. After the other side is subjected to secondary image transfer and fixing, the paper is discharged onto the output tray.

[0027] The scanner 150, fixed on the image forming unit 1, includes a first-surface fixed reading unit 151 as a first-surface reading means and a movable reading unit 152 as a first-surface reading means.

[0028] The movable reading unit 152, which serves as the first-side reading means, is positioned directly beneath the second contact glass, which is fixed to the upper wall of the scanner 150 casing so as to contact the original document MS. The optical system, consisting of a light source and reflective mirrors, can be moved in the left-right direction in the diagram. As the optical system moves from left to right in the diagram, the light emitted from the light source is reflected by the original document placed on the second contact glass, then passes through multiple reflective mirrors, and is received by the image reading sensor 153 fixed to the scanner body.

[0029] The first-side fixed reading unit 151, which serves as the first-side reading means, is positioned directly below the first contact glass, which is fixed to the upper wall of the scanner 150 casing so as to be in contact with the original document MS. As the original document MS, transported by the ADF 51 (described later), passes over the first contact glass, light emitted from the light source is sequentially reflected from the document surface and received by the image reading sensor after passing through multiple reflection mirrors. This allows scanning of the first side of the original document MS without moving the optical system consisting of the light source, reflection mirrors, etc.

[0030] Furthermore, scanner 150 also has a contact-type image sensor that reads the second side of the original MS document. This contact-type image sensor will be described later.

[0031] The ADF 51, mounted on top of the scanner 150, has a main body cover 52 that holds a document placement table 53 for placing the original document MS before scanning, a transport unit for transporting the original document MS, and a document stacking table 55 for stacking the scanned original document MS. As shown in Figure 4, it is supported by a hinge 159 fixed to the scanner 150 so that it can swing up and down. This swinging motion creates a door-like movement, exposing the first contact glass 154 and second contact glass 155 on the top surface of the scanner 150 when open. In the case of single-sided bound documents, such as books bound at one corner, the documents cannot be separated one by one, and therefore cannot be transported by the ADF 51. In the case of single-sided bound documents, after opening the ADF 51 as shown in Figure 4, the single-sided bound document with the pages to be scanned facing downwards is placed on the second contact glass 155, and then the ADF 51 is closed. Then, the image of that page is read by the movable reading unit 152 of the scanner 150, as shown in Figure 1.

[0032] On the other hand, in the case of a stack of multiple independent original documents MS simply stacked on top of each other, the original documents MS can be automatically transported one by one by the ADF 51 and sequentially read by the first-side fixed reading unit 151 in the scanner 150 and the contact-type image sensor in the ADF 51. In this case, after setting the stack of documents on the document tray 53, the copy start button is pressed. The ADF 51 then feeds the original documents MS from the stack placed on the document tray 53 from top to bottom, and transports them toward the document stack tray 55 while inverting them. During this transport process, immediately after inverting the original documents MS, they are passed directly above the first-side fixed reading unit 151 of the scanner 150. At this time, the image of the first side of the original documents MS is read by the first-side fixed reading unit 151 of the scanner 150.

[0033] Figure 5 is an enlarged diagram showing the main components of the ADF 51 together with the upper part of the scanner 150. Figure 6 is a block diagram showing a part of the electrical circuit of the ADF 51 and scanner 150. The ADF 51 includes a document setting section A, a separation transport section B, a resist section C, a turning section D, a first reading transport section E, a second reading transport section F, a paper output section G, a stacking section H, etc.

[0034] As shown in Figure 6, the ADF 51 has a controller 904 consisting of an ASIC (Application Specific Integrated Circuit), which can control various devices and sensors. The controller 904 is connected to a registration sensor 65, a document set sensor 63, a paper ejection sensor 61, abutment sensor 72, a document width sensor 73, a reading input sensor 67, and length sensors 57, 58. It is also connected to a paper feed motor 191, a transport motor 192, a pickup motor 193, and a paper ejection clutch 194. In addition, it is connected to a sound collection unit (a microphone 201), a side guide position detection sensor 806, and a storage unit (a non-volatile memory 807).

[0035] The sound-collecting microphone 201 captures sound during document transport. The side guide position detection sensor 806 is mounted to be movable in the width direction and detects the width direction position of the side guide 202 (see Figure 7), which abuts against the width direction edge of the document on the document placement table 53 to regulate the width direction position of the document. As the side guide position detection sensor 806, a known sensor such as a distance measuring sensor can be used. The non-volatile memory 807 can be an HDD, flash memory, etc., and stores data tables for use in machine learning, as will be described later. The data tables stored in this non-volatile memory 807 are periodically transmitted from the controller 904 to the main unit control unit 901, and then transmitted by the communication unit 907 of the image forming unit 1 to the server 905 of the copier maintenance center via the internet line 906.

[0036] The scanner 150 has a read control unit 903 consisting of a CPU (Central Processing Unit) and RAM (Random Access Memory). This allows it to control various devices and sensors inside the scanner 150. Furthermore, the read control unit 903 is connected to the ADF 51's controller 904 via an interface, and the read control unit 903 can also indirectly control various devices and sensors within the ADF 51 via the controller 904.

[0037] In Figure 5, the document setting unit A has a document placement table 53 on which a stack of document MSs is set. The separation and transport unit B separates and feeds the document MSs one by one from the set stack of document MSs. The resist unit C temporarily abuts the fed document MSs to align them before feeding them out. The turn unit D has a curved transport unit that curves in a C shape, and within this curved transport unit, it folds the document MSs and inverts their top and bottom. The first reading and transport unit E transports the document MSs on the first contact glass 154 and causes the first side of the document MSs to be read by the first side fixed reading unit 151, which is located inside the scanner below the first contact glass 154. The second reading and transport unit F transports the document MSs under the contact image sensor 95 and causes the second side of the document MSs to be read by the contact image sensor 95. The paper discharge unit G discharges the document MSs, whose images have been read on both sides, toward the stack unit H. Furthermore, the stacking unit H stacks the original documents MS on the original document stacking platform 55.

[0038] The original document MS is set with its leading edge resting on a movable document table 54, which is a sheet tray that can pivot in the directions of arrows a and b in the figure according to the thickness of the stack of original documents MS, and its trailing edge resting on a document support table 53. At this time, the position in the width direction is adjusted by abutting side guides against both ends of the document support table 53 in the width direction (the direction perpendicular to the paper plane). The original document MS set in this way pushes up a lever member 62 that is pivotably positioned above the movable document table 54. As a result, the document set sensor 63 detects the setting of the original document MS and transmits a detection signal to the controller 904. This detection signal is then sent from the controller 904 to the reading control unit 903 via an interface.

[0039] The document placement table 53 holds a first length sensor 57 and a second length sensor 58, which consist of a reflective photosensor or an actuator-type sensor that detects the length of the document MS in the transport direction. These length sensors detect the length of the document MS in the transport direction.

[0040] Above the stack of documents MS placed on the movable document table 54, a pickup roller 80 is provided, which is supported by a cam mechanism so as to be movable in the vertical direction (arrows c and d in the figure). This cam mechanism can be driven by a pickup motor 193 to move the pickup roller 80 up and down. When the pickup roller 80 moves upward, the movable document table 54 swings in the direction of arrow a in the figure, and the pickup roller 80 comes into contact with the uppermost document MS in the stack of documents MS. As the movable document table 54 rises further, the table rise sensor 59 eventually detects that the movable document table 54 has risen to its upper limit. As a result, the pickup motor 193 stops, and the upward movement of the movable document table 54 stops.

[0041] The main unit operation unit 902, which consists of a numeric keypad and display located on the main body of the copier, allows the operator to perform key operations to set the reading mode, such as indicating whether it is in double-sided or single-sided reading mode, and to press the copy start key. In other words, the main unit operation unit 902 functions as a means of acquiring mode information to obtain information on whether it is in double-sided or single-sided reading mode. In addition, there is a thin paper mode for reading thin paper, and in thin paper mode, the document transport speed is slowed down overall compared to the normal reading mode.

[0042] When the copy start key is pressed, a document feed signal is sent from the main unit control unit 901 to the ADF 51 controller 904 via the I / F. Then, the pickup roller 80 is rotated by the forward rotation of the paper feed motor 191 and feeds the document MS from the movable document table 54.

[0043] When setting whether to use double-sided or single-sided scanning mode, it is possible to set all documents MS placed on the movable document table 54 to double-sided or single-sided scanning mode at once. Alternatively, it is possible to set the scanning mode individually for each document MS, for example, setting the first and tenth documents MS to double-sided scanning mode while setting the other documents MS to single-sided scanning mode.

[0044] The original document MS, fed by the pickup roller 80, enters the separation transport section B and is fed to a contact position with the paper feed belt 84. This paper feed belt 84 is stretched by a drive roller 82 and a driven roller 83, and is moved endlessly in a clockwise direction in the figure by the rotation of the drive roller 82 accompanying the forward rotation of the paper feed motor 191. A separation roller 85, which is rotated clockwise in the figure by the forward rotation of the paper feed motor 191, is in contact with the lower tension surface of this paper feed belt 84. At the contact point, the surface of the paper feed belt 84 moves in the paper feeding direction. In contrast, the separation roller 85 is in contact with the paper feed belt 84 with a predetermined pressure, and when it is in direct contact with the paper feed belt 84, or when only one original document MS is caught in the contact point, it moves along with the belt or the original document MS. However, when multiple original documents (MS) are caught in the contact area, the rotational force becomes lower than the torque limiter, so the machine rotates clockwise in the opposite direction to the rotational force. As a result, the original documents below the top are subjected to a moving force in the opposite direction to the paper feed by the separation roller 85, separating only the top original document from several other documents.

[0045] The original document MS, separated into individual sheets by the action of the paper feed belt 84 and the separation roller 85, enters the registration section C. Its leading edge is detected as it passes directly beneath the stop sensor 72. At this time, the pickup roller 80, which is driven by the paper feed motor 191, is still rotating, but as the movable document table 54 descends, it moves away from the original document MS, so the original document MS is transported solely by the endless movement force of the paper feed belt 84. The endless movement of the paper feed belt 84 continues for a predetermined time from the moment the leading edge of the original document MS is detected by the stop sensor 72. After that, the leading edge of the original document MS abuts against the contact point between the pull-out driven roller 86 and the pull-out driven roller 87, which rotates while in contact with it. With the leading edge of the original document MS abutting against the contact points of both rollers, the rear end of the original document MS is fed in the paper feeding direction, causing the original document MS to bend by a predetermined amount while its leading edge is positioned at the contact point. This corrects the skew (tilt) of the original document MS, ensuring that the document MS is positioned in a orientation aligned with the paper feed direction.

[0046] The pull-out drive roller 87 not only corrects the skew of the original document MS, but also transports the skew-corrected original document MS to the intermediate roller pair 66 downstream in the document transport direction. The drive roller 82 that tensions the pickup roller 80 and the paper feed belt 84, the pull-out drive roller 87, and the drive rollers of the intermediate roller pair are connected to the paper feed motor 191 via one-way clutches. The one-way clutch connected to the pull-out drive roller 87 and the drive rollers of the intermediate roller pair transmits driving force when the paper feed motor 191 reverses direction, and the one-way clutch connected to the drive roller 82 transmits driving force when the paper feed motor 191 rotates forward. Therefore, when the paper feed motor 191 reverses direction, the pull-out drive roller 87 and the drive rollers of the intermediate roller pair 66 begin to rotate, and the endless movement of the paper feed belt 84 stops. At this time, the rotation of the pickup roller 80 also stops.

[0047] The original document MS, fed out from the pull-out drive roller 87, passes directly beneath the document width sensor 73. The document width sensor 73 has multiple paper detection units, such as reflective photosensors, which are arranged in the document width direction (the direction perpendicular to the drawing plane). The width of the original document MS is detected based on which paper detection unit detects it. The length of the original document MS in the transport direction is detected based on the timing from when the leading edge of the original document MS is detected by the stop sensor 72 until the trailing edge of the original document MS is no longer detected by the stop sensor 72.

[0048] The leading edge of the document MS, whose width has been detected by the document width sensor 73, enters the turn section D and is caught between the contact points of the intermediate roller pair 66. The transport speed of the document MS by this intermediate roller pair 66 is set to be faster than the transport speed of the document MS in the first reading transport section E, which will be described later. This shortens the time it takes to feed the document MS to the first reading transport section E.

[0049] As the leading edge of the document MS is transported within the turn section D, it passes a position opposite the reading entrance sensor 67. When the leading edge of the document MS is detected by the reading entrance sensor 67, the document transport speed by the intermediate roller pair 66 is reduced until the leading edge is transported to the position of the reading entrance roller pair (the pair of 89 and 90) on the downstream side in the transport direction. In addition, as the transport motor 192 starts rotating, one roller in the reading entrance roller pair (89, 90), one roller in the reading exit roller pair 92, and one roller in the second reading exit roller pair 93 each begin rotating.

[0050] Within the turn section D, the original document MS is transported along a curved transport path between the intermediate roller pair 66 and the reading entrance roller pair (89, 90), during which time its upper and lower surfaces are reversed and its transport direction is reversed. The leading edge of the original document MS, having passed through the nip between the rollers of the reading entrance roller pair (89, 90), then passes directly beneath the resist sensor 65. When the leading edge of the original document MS is detected by the resist sensor 65, the document transport speed is reduced over a predetermined transport distance, and the transport of the original document MS is temporarily suspended just before reaching the first reading transport section E. A temporary suspension signal is also transmitted to the reading control unit 903 via the I / F.

[0051] When the reading control unit 903 receives a pause signal and transmits a start signal, the controller 904 controls the rotation of the transport motor 192 to resume until the leading edge of the document MS reaches the first reading transport unit E, and the transport speed of the document MS is increased to a predetermined transport speed. Then, at the moment when the leading edge of the document MS reaches the reading position of the first fixed-face reading unit 151, the controller 904 transmits a gate signal to the reading control unit 903 indicating the effective image area in the sub-scanning direction of the first surface of the document MS. This transmission continues until the trailing edge of the document MS leaves the reading position of the first fixed-face reading unit 151, and the first surface of the document MS is read by the first fixed-face reading unit 151. The timing at which the leading edge of the document MS reaches the reading position of the first fixed-face reading unit 151 is calculated based on the pulse count of the transport motor 192.

[0052] After the document MS has passed through the first reading and transport section E, it passes through the reading exit roller pair 92 (described later), and its leading edge is detected by the paper discharge sensor 61. When the single-sided reading mode is set, reading the second side of the document MS by the contact-type image sensor 95 (described later) is unnecessary. When the leading edge of the document MS is detected by the paper discharge sensor 61, the driving force of the transport motor 192 is connected to the paper discharge roller pair 94 by the paper discharge clutch 194, and the lower paper discharge roller in the paper discharge roller pair 94 is rotated clockwise in the figure. Furthermore, based on the pulse count of the transport motor 192 since the leading edge of the document MS was detected by the paper discharge sensor 61, the timing for the trailing edge of the document MS to exit the nip of the paper discharge roller pair 94 is calculated. Based on this calculation result, the paper discharge clutch 194 cuts the driving force of the transport motor 192 and stops the paper discharge roller pair 94.

[0053] On the other hand, when the double-sided reading mode is set, the timing from when the leading edge of the document MS is detected by the paper ejection sensor 61 until it reaches the contact-type image sensor 95 is calculated based on the pulse count of the transport motor 192. At that timing, the controller 904 transmits a gate signal to the reading control unit 903 indicating the effective image area in the sub-scanning direction on the second side of the document MS. This transmission continues until the trailing edge of the document MS leaves the reading position of the contact-type image sensor 95, and the second side of the document MS is read by the contact-type image sensor 95.

[0054] The contact-type image sensor 95 (CIS), used as the second-side reading means, has a coating applied to its reading surface to prevent vertical streaks caused by adhesive foreign matter adhering to the original document MS. A second reading roller 96 is positioned opposite the contact-type image sensor 95 as a document support means that supports the original document MS from the non-reading side (first side). This second reading roller 96 prevents the original document MS from lifting at the reading position by the contact-type image sensor 95 and also functions as a reference white area for acquiring shading data from the contact-type image sensor 95. In this copier, a second reading roller 96 is used as the document support means that supports the original document at the position opposite the contact-type image sensor 95, but a guide plate-like device may also be used.

[0055] Figure 7 is an explanatory diagram regarding the arrangement of the sound-collecting microphone 201. (a) shows the arrangement of the sound-collecting microphone 201 when the document is set on the back side of the device, and (b) shows the arrangement of the sound-collecting microphone 201 when the document is set on the center side of the device. (c) is a view of the area near the pickup roller 80 from the axial direction. As shown in Figure 7, in this embodiment, the sound-collecting microphone 201 is attached to the paper feed cover 98 and is positioned upstream of the pickup roller 80 in the document transport. In this embodiment, the purpose is to collect the operating sound during document feeding, so it is positioned around the pickup roller 80 which is driven during paper feeding. Note that the position of the sound-collecting microphone 201 is not limited to this, and it may be placed in a suitable location where the operating sound of the paper feeding / separation operation can be collected well. In the figure, 80a is a holder 80a that holds the pickup roller 80 so that it can move toward and away from the document placed on the movable document table 54.

[0056] When feeding a document, the pickup roller 80 transports the document against the frictional force between the document directly below it or the movable document table 54, making it prone to the document slipping relative to the pickup roller 80. This slippage may prevent the document from being transported to the designated position within the specified time, potentially resulting in a paper jam.

[0057] If a paper jam occurs due to a feeding error after the document has been transported to a certain extent into the ADF, it may become difficult to remove the document, and there is a risk of damage to the document, such as wrinkles or tears, occurring during removal.

[0058] Furthermore, due to user negligence, a stack of multiple MS documents bound together with staples or clips may be placed on the document tray 53. When such a stack of bound documents is fed, there is a risk of a paper jam at the separation section, which is the contact point between the separation roller 85 and the paper feed belt 84. Also, if the leading edge of a stack of multiple MS documents bound together with metal pieces such as staples or clips enters the separation section, the top document in the stack will continue to be transported by the paper feed belt 84. However, the second and subsequent documents will be subjected to a transport force that pushes them back by the separation roller 85. As a result, a large stress is applied to the areas bound by metal pieces, which may cause tears, folds, wrinkles, or other damage to the documents.

[0059] Therefore, in this embodiment, it is determined from the operating sound during document feeding whether or not there is a risk of a paper jam occurring. If there is a risk of a paper jam occurring, the document transport is stopped to prevent a paper jam from occurring and to prevent damage to the document.

[0060] Figure 8 is a flowchart showing the operational state determination process performed by the controller 904. When the main unit control unit 901 transmits a document feed signal and the reading mode set by the user using the main unit operation unit 902 to the controller 904, the controller 904 determines whether the reading mode is normal mode or thin paper mode (S1).

[0061] Next, the controller 904 provisionally determines the width of the document set on the document tray 53 based on the detection result of the side guide position detection sensor 806 (S2). Next, the controller 904 starts the paper feeding / separation operation (S3), and at the same time, the sound collection microphone 201 collects the sound of the operation during document feeding (S4). The controller 904 sequentially cuts the collected sound data into frames of a predetermined length and calculates the Mel-frequency cepstrum coefficient, which is a characteristic quantity of the sound (S5). In other words, in this embodiment, the controller 904 functions as a feature extraction unit.

[0062] The Mel-frequency cepstrum coefficient is a feature that weights the spectral envelope of speech (frequency characteristics derived from vocal tract components) with Mel frequencies that take into account human frequency perception characteristics, and is commonly used as a feature for speech recognition.

[0063] The "Mel frequency" mentioned above is a frequency transformed to take into account the human characteristic of pitch perception, where sounds near the lower limit of the audible range are perceived as higher pitched, and sounds near the upper limit are perceived as lower pitched. The "cepstrum" is obtained by taking the logarithm of an audio signal (waveform data), transforming it into a frequency spectrum using the Fourier transform, and then performing an inverse Fourier transform on the resulting logarithm to return it to spacetime. The "cepstrum" is also called the "inverse Fourier transform of the logarithm of the spectrum" (spectrum of the spectrum), and it separates the fine structure of the spectrum (subtle changes) from the gradual changes of the spectrum (spectral envelope).

[0064] As mentioned above, the Mel-frequency cepstrum coefficient is the spectral envelope of sound, which is the frequency characteristic derived from the vocal tract components. However, the spectral envelope of the operating sound during normal paper feeding differs from that of the operating sound during abnormal feeding, such as when slipping occurs or when transporting a stack of bound documents. Therefore, it is possible to classify normal and abnormal paper transport using the Mel-frequency cepstrum coefficient.

[0065] In this embodiment, Mel-frequency cepstrum coefficients are used as feature quantities that quantitatively represent the characteristics of the operating sound used for determining the operating state, but this is not the only option. For example, known sound feature quantities such as linear prediction coefficients may be used as feature quantities for determining the operating state of the operating sound.

[0066] The calculated Mel-frequency cepstrum coefficients are subjected to a classifier, which is a transport anomaly detection unit using a support vector machine, a type of supervised machine learning method, to determine the operating state (S6). Support vector machines linearly separate and classify multi-dimensional data into multiple classes based on a trained model. In this embodiment, the Mel-frequency cepstrum coefficients, which are feature quantities of the paper feeding operation sound collected by the sound-collecting microphone 201, are classified into one of two classes: normal transport or abnormal transport. Support vector machines can achieve high generalization performance with a small amount of training data and can distinguish between normal and abnormal transport with high accuracy.

[0067] In this embodiment, there are multiple classifiers corresponding to transport parameters that affect the operating noise, such as document thickness, document width, and document transport speed. Each classifier classifies the above-mentioned features into normal transport and abnormal transport using different trained models obtained based on multiple training data separated by document thickness, document width, and document transport speed.

[0068] The audio signal (waveform data) of the operating sound tends to have different signal component strengths for each frequency depending on the document thickness, document width, and transport speed. For example, the operating sound of feeding thin paper tends to have stronger high-frequency signal components compared to feeding plain paper. Also, the overall signal component strength tends to increase with increasing document width. Furthermore, the higher the transport speed, the stronger the high-frequency signal components tend to be.

[0069] Thus, the audio signal (waveform data) of the operating sound differs depending on the document thickness, document width, and transport speed. Therefore, by preparing multiple classifiers, each with different pre-trained models corresponding to the document thickness, document width, and transport speed, the accuracy of distinguishing between normal and abnormal transport can be improved.

[0070] In this embodiment, the document thickness (plain paper / thin paper) and document transport speed are determined from the reading mode determined in step S1, and the document width is determined in step S2. The controller 904 determines the classifier to be used from the information obtained in steps S1 and S2.

[0071] As will be described later, the above-mentioned trained model is a model that was trained on server 905 of the copier maintenance center based on a data table collected from multiple devices used in the market.

[0072] If the classifier classifies the Mel frequency cepstrum coefficient as abnormal transport (paper jam), the paper feeding / separation operation is terminated (S31) because there is a risk of a paper jam, and the user is asked whether or not to continue document transport (S32). The operation related to the inquiry to the user will be described later.

[0073] In this way, by predicting the occurrence of abnormal feeding (paper jam) based on the operating sound during document feeding and stopping document feeding, it is possible to prevent paper jams from occurring. Also, by stopping document feeding before the document reaches the separation section, it is possible to easily remove the document. Therefore, it is possible to prevent paper jams from occurring in the recessed parts of the ADF, which would make it difficult to remove the document, and to prevent wrinkles and tears from occurring on the document when it is removed. Furthermore, it is possible to prevent multiple documents bound with staples from entering the separation section and causing a jam, thereby preventing tears and folds on the document. Therefore, it is possible to prevent damage to the document such as tears, folds, and wrinkles, and protect the document.

[0074] On the other hand, if the classifier determines that the document is being transported normally, the paper feeding / separation operation continues (S7), followed by a pull-out operation (S9) and a registration stop process (S10). During this pull-out operation (S9), the width of the document MS is determined by the document width sensor 73. After that, the system waits for the reception of a read start signal from the read control unit 903 (S11), and after reception (Y in S11), the document transport is resumed and the read transport process is performed (S12).

[0075] In parallel with the series of transport control steps S7 to S11, the controller 904 monitors whether a document has arrived by a predetermined time using the sheet detection unit, which consists of the reading input sensor 67, the registration sensor 65, and the paper discharge sensor 61. If the reading input sensor 67, registration sensor 65, and paper discharge sensor 61 do not detect a document by the predetermined time, the controller interrupts the document transport operation, assuming that a paper jam has occurred (S21). In other words, in this embodiment, the controller 904 functions as a transport abnormality detection unit.

[0076] Then, when the paper output sensor 61 detects the document by a predetermined timing and the document is ejected to the document stacking tray 55, the controller 904 determines the transport result to be "normal" (S13). Next, the Mel frequency cepstrum coefficient calculated in step S5, the document thickness and transport speed based on the reading mode determined in step S1, the document width determined in step S9, and the transport result ("normal") determined in step 13 are stored in the non-volatile memory 807 as a linked data table (S14).

[0077] On the other hand, if a paper jam occurs and the document transport operation is interrupted (S21), the controller 904 determines the transport result as "jam" (S22). The Mel frequency cepstrum coefficient calculated in step S5, the document thickness and transport speed determined in step S1, the document width determined in step S9, and the transport result ("jam") determined in step 22 are then stored in the non-volatile memory 807 as a linked data table (S14). If a non-delivered / stuck jam is detected during the paper feeding / separation operation (S7), or if a non-delivered / stuck jam is detected before the width of the document MS is determined by the document width sensor 73, the document width provisionally determined in step S2 is stored in the non-volatile memory 807.

[0078] Figure 9 shows an example of a data table stored in the non-volatile memory 807. As shown in Figure 9, for each transported document, the Mel frequency cepstrum coefficient, document thickness, transport speed, document width, and transport result are stored as a single row in the non-volatile memory 807.

[0079] The data table shown in Figure 9, stored in the non-volatile memory 807, is periodically transmitted from the controller 904 to the main unit control unit 901. Then, the communication unit 907 of the image forming unit 1 transmits it via the internet line 906 to the server 905 at the copier maintenance center (see Figure 6). The server 905 receives the data tables from multiple devices used in the market, and a vast amount of these data tables accumulates on the server 905.

[0080] As shown in Figure 9, the data table contains Mel-frequency cepstrum coefficients, which are characteristic features of the operating sound, and the corresponding transport results as ground truth data, and includes the training data necessary for machine learning. In addition, this data table includes transport parameters such as document thickness, document width, and document transport speed for sorting the training data.

[0081] The device developers categorize training data, consisting of Mel-frequency cepstrum coefficients and their corresponding transport results, based on transport parameters that affect operating noise, such as document thickness, document width, and document transport speed, which are stored in the data table accumulated in the server 905. Then, machine learning is performed using the categorized training data to generate a new trained model. This trained model is obtained by machine learning using a large amount of training data accumulated in the server 905, and its performance in distinguishing between normal and abnormal transport can be improved.

[0082] The newly generated trained model is sent from the server 905 to each copier via the internet line 906. The newly generated trained model received by the communication unit 907 of the image forming unit 1 is transmitted from the main unit control unit to the controller 904, and the trained model of the classifier contained in the controller 904 is updated with the newly generated trained model. This makes it possible to accurately determine the document transport status from the operating sound.

[0083] In this embodiment, training data, in which Mel-frequency cepstrum coefficients and transport results are linked, is stored in the non-volatile memory 807. The training data stored in the non-volatile memory 807 can then be collected from multiple devices used in the market via an internet connection. This allows for the collection of vast amounts of training data without the developer having to prepare a large number of devices and keep them running for extended periods, thereby reducing the cost of training data collection.

[0084] Furthermore, by linking the above-mentioned training data to the transport parameters (document thickness, transport speed, document width, etc.) that affect the operating noise and storing them in the non-volatile memory 807, a trained model corresponding to the above transport parameters can be generated. This allows for the determination of whether transport is normal or abnormal based on the trained model corresponding to the above transport parameters, and enables accurate prediction of abnormal transport from the operating noise. In this embodiment, document thickness, transport speed, and document width are used as transport parameters, but the transport parameters used can be appropriately selected depending on the configuration of the device, etc. For example, elements that affect the operating noise, such as the material of the document or the condition of the pickup roller 80 (deterioration, etc.), may be used as transport parameters.

[0085] Figure 10 is a control flow diagram of the inquiry control that prompts the user for information when the classifier determines that there is a transport error, and Figure 11 is a diagram illustrating an example of the transition of the inquiry screen displayed on the operation display unit such as the touch panel of the main unit operation unit 902. As explained using Figure 8, when the classifier classifies the Mel frequency cepstrum coefficient as a transport abnormality, the paper feeding / separation operation is temporarily suspended and user inquiry control is performed. When query control is performed, the controller 904 first displays a query screen as shown in Figure 11(a) on the operation display unit of the main unit operation unit 902 via the main unit control unit 901 (S41).

[0086] In Figure 11(a), the "Continue Reading" button is disabled, and the user can only press the "Cancel" button. Note that although the "Continue Reading" button is displayed in Figure 11(a), it is also possible to display only the "Cancel" button.

[0087] When the user removes the document from the document tray 53 according to the instructions displayed on the inquiry screen shown in Figure 11(a), the document set sensor 63 detects that the document is not set. Once the document set sensor 63 detects that the document is not set, the controller 904 starts timing using the timer (S42). Also, the inquiry screen displayed on the touch panel of the main unit operation unit 902 is updated to the inquiry screen shown in Figure 11(b) via the main unit control unit 901 (S43).

[0088] In the inquiry screen shown in Figure 11(b), the "Continue Reading" button is disabled, and the user can only press the "Cancel" button. Alternatively, Figure 11(b) could also be designed so that only the "Cancel" button is displayed, without the "Continue Reading" button.

[0089] The user checks whether the document is stapled according to the instructions displayed on the inquiry screen shown in Figure 11(b). If it is stapled, the user removes the staples and places the document on the document tray 53. When the document is placed on the document tray 53, the document set sensor 63 detects that the document has been placed. Once the document set sensor 63 detects that the document has been placed, the controller 904 stops the timer (S44). Then, via the main unit control unit 901, the inquiry screen displayed on the touch panel of the main unit operation unit 902 is updated to the inquiry screen shown in Figure 11(c) (S45). In the inquiry screen shown in Figure 11(c), the "Continue Reading" button is enabled, allowing the user to press either the "Continue Reading" or "Cancel" button.

[0090] When the user presses the "Continue Reading" button on the inquiry screen shown in Figure 11(c), the controller 904 checks whether the timer's timing is equal to or greater than the predetermined time (S46). This timer's timing is the time from when the user removes the document from the document tray 53 until the document is reset on the document tray 53. If the document does not contain staples or the like, the document is reset immediately after the check, so the timing is shorter and less than the predetermined time. On the other hand, if the document is bound with staples or the like, the staples or the like must be removed from the document before it is reset on the document tray 53. Therefore, the time from when the document is removed from the document tray 53 until it is reset on the document tray 53 becomes longer and is equal to or greater than the predetermined time.

[0091] Therefore, if the timer's timing is longer than the predetermined time (Yes in S46), the controller 904 determines that the staples have been removed from the stapled document. If a stapled document is transported, it will jam in the separation unit, and the transport result is recorded as "jam" (S47). The document thickness and transport speed determined in step S1 of the control flow in Figure 8, the document width tentatively determined in step S2, and the transport result ("jam") in step S47 are then linked and stored in the non-volatile memory 807 (S48). On the other hand, if the timer's timing is within the predetermined time (No in S46), the inquiry control is terminated.

[0092] Furthermore, when the user presses the "Continue Reading" button on the inquiry screen shown in Figure 11(c), the controller 904 resumes transporting the document. At this time, the control flow shown in Figure 12 is executed, and the document is transported without using the classifier to predict transport anomalies. However, as shown in S4 of Figure 12, the operation sound is collected by the sound-collecting microphone 201, and the Mel-frequency cepstrum coefficient is calculated in the background (S5). Then, similar to the flow shown in Figure 8, the Mel-frequency cepstrum coefficient, document thickness, transport speed, document width, and transport result are linked and stored in the non-volatile memory 807 as a data table (S14). This makes it possible to acquire training data even when document transport is resumed.

[0093] Furthermore, in this embodiment, the user can set whether or not to perform abnormal transport prediction using the classifier by operating the main unit operation unit 902. Even when abnormal transport prediction is not performed, the flow shown in Figure 12 described above is carried out to collect operating sounds and calculate the Mel-frequency cepstrum coefficients. Then, training data consisting of the Mel-frequency cepstrum coefficients and transport results is acquired.

[0094] In the above description, an example of applying the present invention to the ADF51 as a sheet feeding device was explained, but the present invention can also be applied to a transport device that transports the transfer paper of the image forming unit 1.

[0095] The above is just one example; each of the following embodiments produces its own unique effects. (Aspect 1) The system includes transport members such as a pickup roller 80 for transporting sheets such as original MS; a sound collection unit such as a sound collection microphone 201 for collecting operating sounds during sheet transport; a feature extraction unit such as a controller 904 for extracting feature quantities such as Mel frequency cepstrum coefficients that quantitatively represent the characteristics of the operating sounds collected by the sound collection unit; a transport anomaly determination unit such as a classifier that determines whether or not a transport anomaly occurs based on the feature quantities; a sheet detection unit (in this embodiment, a reading entrance sensor 67, a resist sensor 65, and a paper discharge sensor 61) for detecting sheets on the transport path; a transport anomaly detection unit such as a controller 904 that detects transport anomalies based on the detection results of the sheet detection unit; and a storage unit such as a non-volatile memory 807 that stores a data table in which at least the feature quantities and the detection results by the transport anomaly detection unit are linked. To train a machine learning model to classify features into normal or abnormal transport, training data is needed in which features are linked to the correct transport results. Traditionally, this training data was obtained, for example, by developers operating equipment in a laboratory. Improving the classification performance of machine learning requires a massive amount of training data. As mentioned above, obtaining a massive amount of training data by developers operating equipment requires operating the equipment for long periods of time or preparing a large number of devices, which presents the challenge of high costs for collecting such a large amount of training data. In contrast, in Embodiment 1, a data table is stored in the memory unit, which links feature quantities that quantitatively represent the characteristics of the operating sound during sheet transport extracted by the feature quantity extraction unit with the detection results of the transport anomaly detection unit. By storing a data table in the memory unit that links the feature quantities usable as training data with the transport results, it becomes possible to send the data table stored in the memory unit to the developer from many devices used in the market, for example, by using an internet connection. This makes it possible for the developer to collect a vast amount of the above-mentioned training data from devices used in the market without having to run a large number of devices in a test laboratory for a long period of time. Therefore, the cost of collecting a vast amount of training data can be reduced compared to when the developer runs a large number of devices in a test laboratory for a long period of time to collect training data.

[0096] (Aspect 2) In Embodiment 1, the data table also includes transport parameters that affect the operating noise (in this embodiment, sheet thickness, transport speed, and sheet width). According to this, as described in the embodiment, it becomes possible to classify training data, in which features and detection results from the transport anomaly detection unit are linked based on transport parameters, and perform machine learning. This makes it possible to generate a trained model that corresponds to the transport parameters.

[0097] (Aspect 3) In embodiment 1 or 2, when a transport anomaly detection unit such as a classifier determines that a transport anomaly has occurred, the transport of the sheet is temporarily suspended, and the transport anomaly detection unit prompts the user to remove the cause of the transport anomaly (in this embodiment, the removal of staples). According to this, as described in the embodiment, users such as a user can, based on the content of the inquiry, remove the staples or other factors causing the transport abnormality and reset the sheet such as the document.

[0098] (Aspect 4) In embodiment 3, when it is determined that the user has removed the cause of abnormal transport of the sheet and instructed the sheet to be transported again (in this embodiment, when the time from removing the original sheet to resetting the sheet is longer than a predetermined time, it is determined that the cause of abnormal transport has been removed), the detection result by the transport abnormality detection unit is recorded as a transport abnormality and stored in the storage unit, linked to the feature quantities extracted at that time. According to this, useful training data can be obtained even if no transport abnormalities such as paper jams occur, as described in the embodiment.

[0099] (Aspect 5) In any of embodiments 1 to 4, even when the transport abnormality determination unit such as the controller 904 does not perform an abnormality determination, the sound collection unit such as the sound collection microphone 201 collects the operating sound, the feature quantity extraction unit such as the controller 904 extracts feature quantities based on the operating sound collected by the sound collection unit, and stores the extracted feature quantities and the detection result from the transport abnormality detection unit in a storage unit such as the non-volatile memory 807. According to this, as described in the embodiment, useful training data can be obtained even when the transport abnormality determination unit does not perform abnormality determination.

[0100] (Aspect 6) In any of embodiments 1 to 5, the operating sound is the operating sound of the feeding rollers, such as the pickup roller 80, which feeds the sheets placed on the sheet tray. According to this, as described in the embodiment, it becomes possible to determine whether or not abnormal transport will occur in the abnormality detection unit before the sheets are transported to the separation unit. This prevents damage to the sheets, such as tearing or wrinkling, that may occur when the stapled sheet bundles are transported to the separation unit, thus protecting the sheets.

[0101] (Aspect 7) An automatic document transport device is provided which includes a document sheet transport unit for transporting document sheets such as original documents, and which transports the document sheets to an image reading unit using the document sheet transport unit. In this device, one of the sheet transport devices from embodiment 1 to 6 is used as the document sheet transport unit. This method helps to suppress jam formation and prevent damage such as folds, wrinkles, and tears from occurring in the manuscript.

[0102] (Pattern 8) An image forming apparatus that forms an image on a sheet transported by a sheet transport unit, comprising a sheet transport unit as described in embodiments 1 to 6, or an automatic document transport unit such as an ADF as described in embodiment 7.

[0103] (Aspect 9) In embodiment 8, the copier has a transmission / reception unit such as a communication unit 907 that transmits and receives data with an external device such as a server 905 at the copier maintenance center. The transmission / reception unit transmits a data table to the external device and receives determination parameters from the external device for the transport abnormality determination unit to determine whether or not there is a transport abnormality. According to this, external devices such as Server 905 can receive databases from multiple devices used in the market, enabling the efficient collection of vast amounts of training data. Furthermore, by receiving judgment parameters such as trained models generated by machine learning from the vast amount of training data sent from multiple devices used in the market from the external device, it is possible to use the received judgment parameters to determine whether or not a transport anomaly will occur.

[0104] (Aspect 10) In embodiment 9, the determination parameters are a trained model obtained by machine learning on an external device such as the server 905 of the copier maintenance center based on a data table, and the transport abnormality determination unit uses the trained model to determine whether or not an abnormal transport occurs. According to this, it is possible to determine transport abnormalities with high accuracy. [Explanation of Symbols]

[0105] 1: Image forming unit 50: Document scanning device 51: ADF 53: Manuscript Stand 57: First length sensor 58: Second sensor 61: Paper output sensor 63: Document Placement Sensor 65: Resist Sensor 67: Read Inlet Sensor 72: Buttock sensor 73: Document width sensor 80: Pickup Roller 84: Paper feed belt 85: Separation roller 86: Pull-out driven roller 87: Pull-out drive roller 92: Read-out exit roller pair 93: Second Read Exit Roller 94: Paper output roller pair 95: Contact-type image sensor 96: Second reading roller 150: Scanner 201: Sound-collecting microphone 202: Side Guide 806: Side guide position detection sensor 807: Non-volatile memory 901: Main Unit Control 902: Main Unit Control Panel 903: Read Control Unit 904: Controller 905: Server 906: Internet connection 907: Communications Department MS: Manuscript [Prior art documents] [Patent Documents]

[0106] [Patent Document 1] Japanese Patent Publication No. 2021-181376

Claims

1. A conveying member for transporting sheets, A sound collection unit that collects operating noises during sheet transport, A feature extraction unit extracts feature quantities that quantitatively represent the characteristics of the operating sound collected by the sound collection unit, A transport abnormality determination unit determines whether or not a transport abnormality occurs based on the aforementioned feature quantities, A sheet detection unit that detects sheets along the transport path, A transport abnormality detection unit detects transport abnormalities based on the detection results of the sheet detection unit, The system includes a storage unit that stores a data table in which at least the feature quantities and the detection results from the transport anomaly detection unit are linked, When the transport abnormality detection unit determines that a transport abnormality has occurred, it temporarily stops the transport of the sheet and prompts the user to remove the cause of the transport abnormality. A sheet transport device characterized in that, based on whether the time from when the user removes a sheet from the sheet tray until the sheet is set back into the sheet tray is within a predetermined time, the transport abnormality detection unit determines whether to store the detection result as a transport abnormality in the storage unit in association with the characteristic quantity at which the transport abnormality was determined to have occurred.

2. In the sheet conveying device according to Claim 1, A sheet transport device characterized in that, if the time from when the user removes a sheet from the sheet tray until when the sheet is re-set in the sheet tray is longer than a predetermined time, the transport abnormality detection unit detects the result as a transport abnormality, and the transport abnormality determination unit stores it in the storage unit in association with the characteristic quantity at which it determined that a transport abnormality occurred.

3. In the sheet conveying device according to claim 2, A sheet transport device characterized in that, when the time from when the user removes a sheet from the sheet tray until when the sheet is set back into the sheet tray is longer than a predetermined time, the device determines that the user has removed the cause of abnormal transport and instructed the sheet to be transported again, the detection result by the transport abnormality detection unit is treated as a transport abnormality, and the transport abnormality determination unit stores it in the storage unit in association with the characteristic quantity at which it determined that a transport abnormality occurred.

4. In the sheet conveying device according to Claim 1, The sheet conveying device is characterized in that the data table also includes conveying parameters that affect the operating noise.

5. In the sheet conveying device according to Claim 1, A sheet conveying device characterized in that, even when the conveying abnormality determination unit does not perform an abnormality determination, the sound collection unit collects the operating sound, the feature quantity extraction unit extracts the feature quantity based on the operating sound collected by the sound collection unit, and the extracted feature quantity is linked to the detection result by the conveying abnormality detection unit and stored in the storage unit.

6. In the sheet conveying device according to claim 1, A sheet conveying device characterized in that the aforementioned operating sound is the operating sound of a feed roller that feeds sheets placed on the sheet tray.

7. It is equipped with a document sheet transport unit that transports document sheets, In an automatic document transport device that transports a document sheet to an image reading unit using the document sheet transport unit, An automatic document transport device characterized in that the aforementioned document sheet transport unit uses the sheet transport device described in claim 1.

8. In an image forming apparatus that forms an image on a sheet conveyed by a sheet conveying unit, An image forming apparatus characterized in that the sheet transport unit comprises the sheet transport device described in claim 1, or the automatic document transport device described in claim 7.

9. In the image forming apparatus according to claim 8, It has a transmitting / receiving unit that transmits and receives data with external devices, The image forming apparatus is characterized in that the transmitting / receiving unit transmits the data table to the external device and the transport abnormality determination unit receives determination parameters from the external device for determining whether or not there is a transport abnormality.

10. In the image forming apparatus according to claim 9, The aforementioned determination parameters are trained models obtained by machine learning using the external device based on the aforementioned data table. The image forming apparatus is characterized in that the transport anomaly determination unit determines whether or not a transport anomaly occurs from the feature quantities using the trained model.

Images