WO2021001996A1

WO2021001996A1 - Component mounting system

Info

Publication number: WO2021001996A1
Application number: PCT/JP2019/026655
Authority: WO
Inventors: 光孝稲垣; 茂人大山; 春菜成田
Original assignee: 株式会社Ｆｕｊｉ
Priority date: 2019-07-04
Filing date: 2019-07-04
Publication date: 2021-01-07
Also published as: JPWO2021001996A1; JP7257514B2

Abstract

This component mounting system is provided with a plurality of component mounting machines which are arranged in the conveying direction of a substrate and respectively have image capturing devices which capture images of the substrate. Furthermore, the component mounting system is provided with a learning device, and an inspection device. The learning device captures, by means of an image capturing device, images of a component-to-be mounted substrate and a component-mounted substrate, generates learning data of the component-to-be mounted substrate through machine learning on the basis of the captured image of the component-to-be mounted substrate, acquires teacher data based on the captured image of the component-mounted substrate, and generates learning data of the component-mounted substrate through machine learning on the basis of the teacher data and the captured image of the component-mounted substrate. The inspection device captures, by means of the image capturing device, an image of the component-mounted substrate, and performs inspection for the component-mounted substrate by performing image processing for recognizing the component in the captured image of the component-mounted substrate by using the learning data of the component-to-be mounted substrate and the component-mounted substrate.

Description

Component mounting system

This specification discloses the component mounting system.

Conventionally, a component mounting line including one or a plurality of component mounting machines, an appearance inspection device installed on the board carry-out side of the component mounting line and determining the quality of the mounted state of each component mounted on the circuit board, and a component. A component mounting system including a learning computer connected to a mounting line network and collecting and learning teacher data used for learning processing has been proposed (see, for example, Patent Document 1). The control device of each component mounting machine is a reconstructed super-resolution processing unit that estimates a high-resolution image from a low-resolution image captured during production, and the learning result of the learning process of the learning computer during production. It has a learning-type super-resolution processing unit that estimates a high-resolution image from a low-resolution image obtained by imaging a component based on the image. The control device of each component mounting machine estimates a high-resolution image by reconstructive super-resolution processing and processes the estimated high-resolution image to recognize the component until the learning process of the learning computer is completed. On the other hand, after the learning process of the learning computer is completed, the control device of each component mounting machine switches to the learning type super-resolution processing to estimate the high-resolution image and processes the estimated high-resolution image to process the component. recognize. In addition, the control device of each component mounting machine switches to the reconstructive super-resolution processing when it is determined that the learning result of the teacher data needs to be updated during the execution period of the learning-type super-resolution processing. A high-resolution image is estimated, and a learning result update request is sent to the learning computer. Then, the control device of each component mounting machine is reconstructed when the learning computer that receives the update request recollects the teacher data and relearns to update the learning result of the teacher data and completes the learning process. Switch from super-resolution processing to learning-type super-resolution processing.

Japanese Unexamined Patent Publication No. 2018-97731

By the way, in the component mounting system, when the board after the mounting operation is imaged (the board after the component is mounted) and the image processing for recognizing the component is performed in the captured image to inspect the board after the component is mounted. The inspection must be performed appropriately to ensure the quality of the board after component mounting.

In the present disclosure, when the image processing of the captured image of the board after mounting the component is performed using the learning data generated by machine learning based on the teacher data to inspect the board after mounting the component, the inspection is appropriately performed. The main purpose is to ensure the quality of the board after mounting the components.

The present disclosure has taken the following measures to achieve the above-mentioned main purpose.

The component mounting system of the present disclosure is
It is a component mounting system equipped with a plurality of component mounting machines that are lined up in the substrate transport direction and each has an imaging device that images a substrate.
The pre-component mounting board before the component mounting operation and the post-component mounting board after the mounting operation are imaged by the imaging device, and machine learning is performed based on the captured image of the pre-component mounting board. This is performed to acquire teacher data based on the captured image of the board after mounting the component, and machine learning is performed based on the acquired teacher data and the captured image of the board after mounting the component to generate learning data for component inspection. Learning device and
The component-mounted board is inspected by imaging the component-mounted board with the imaging device and performing image processing for recognizing the component in the captured image of the component-mounted board using the learning data for component inspection. Inspection equipment to be performed and
The gist is to prepare.

The component mounting system of this disclosure includes a learning device and an inspection device. The learning device images the pre-component mounting board and the post-component mounting board, performs machine learning based on the captured image of the pre-component mounting board, and acquires and acquires teacher data based on the captured image of the post-component mounting board. Machine learning is performed based on the teacher data and the captured image of the board after mounting the components to generate learning data for component inspection. The inspection device inspects the board after mounting the component by taking an image of the board after mounting the component and performing image processing for recognizing the component in the captured image of the board after mounting the component using the learning data for component inspection. In this way, by generating learning data for component inspection based on the captured images of the board before component mounting and the board after component mounting, the recognition accuracy when recognizing the component from the captured image of the board after component mounting can be improved. Can be enhanced. As a result, the quality of the board can be ensured by appropriately inspecting the board after mounting the components.

It is a block diagram which shows the outline of the structure of the component mounting system 1 of this embodiment. It is an external perspective view of the component mounting machine 10. It is explanatory drawing which shows the electrical connection relationship between the control device 60 and the management device 80 of a component mounting machine 10. It is a flowchart which shows an example of a component mounting process. It is a flowchart which shows an example of a learning process. It is a flowchart which shows an example of the relearning execution condition success / failure judgment processing. It is a flowchart which shows an example of an inspection process. It is a flowchart which shows the learning process of a modification. It is a flowchart which shows an example of the additional learning execution condition success / failure judgment processing.

Next, a mode for carrying out the invention of the present disclosure will be described with reference to the drawings.

FIG. 1 is a configuration diagram showing an outline of the configuration of the component mounting system 1 of the present embodiment. FIG. 2 is an external perspective view of the component mounting machine 10. FIG. 3 is an explanatory diagram showing an electrical connection relationship between the control device 60 and the management device 80 of the component mounting machine 10. In FIGS. 1 and 2, the left-right direction is the X-axis direction, the front-back direction is the Y-axis direction, and the up-down direction is the Z-axis direction.

As shown in FIG. 1, the component mounting system 1 includes a printing machine 2, a printing inspection machine 3, a plurality of (for example, 5) component mounting machines 10, and a management device 80 for managing the entire system. .. The printing machine 2 prints solder on the substrate S to form a circuit pattern. The printing inspection machine 3 inspects the state of the solder printed by the printing machine 2. The plurality of component mounting machines 10 perform a mounting operation of mounting the components on the board and perform a mounting inspection of whether or not the components are mounted on the board. The printing machine 2, the printing inspection machine 3, and the plurality of component mounting machines 10 are arranged side by side in the transport direction of the substrate S to form a production line.

As shown in FIG. 2, the component mounting machine 10 has a component supply device 21 for supplying components, a substrate transfer device 22 for transporting a substrate, a head 40 having a suction nozzle for sucking the components, and an X-axis of the head 40. It includes a head moving device 30 that moves in the direction and the Y-axis direction, and a control device 60 (see FIG. 3) that controls the entire mounting machine. In addition to these, the component mounting machine 10 images the parts camera 23 for capturing the suction posture of the component sucked by the suction nozzle, the nozzle station 24 accommodating the replacement suction nozzle, and the substrate S. It also has a mark camera 43 for this purpose.

The parts supply device 21 includes, for example, a tape reel around which carrier tape containing parts is wound at predetermined intervals, and a tape feeding mechanism that pulls out the carrier tape from the tape reel and sends it to a parts supply position by driving a drive motor. It is configured as a tape feeder. The component supply device 21 (tape feeder) is detachably attached to a feeder stand (not shown) included in the component mounting machine 10.

The board transfer device 22 includes a pair of conveyor rails arranged at intervals in the Y-axis direction, and by driving the pair of conveyor rails, the board is moved from left to right (board transfer direction) in FIG. Transport.

As shown in FIG. 2, the head moving device 30 includes a pair of X-axis guide rails 31, an X-axis slider 32, an X-axis actuator 33 (see FIG. 3), a pair of Y-axis guide rails 35, and a Y-axis. It includes a slider 36 and a Y-axis actuator 37 (see FIG. 3). The pair of Y-axis guide rails 35 are installed on the upper stage of the housing 11 so as to extend parallel to each other in the Y-axis direction. The Y-axis slider 36 is bridged over a pair of Y-axis guide rails 35 and moves in the Y-axis direction along the Y-axis guide rail 35 by being driven by the Y-axis actuator 37. The pair of X-axis guide rails 31 are installed on the lower surface of the Y-axis slider 36 so as to extend parallel to each other in the X-axis direction. The X-axis slider 32 is bridged over a pair of X-axis guide rails 31 and moves in the X-axis direction along the X-axis guide rail 31 by driving the X-axis actuator 33. A head 40 is attached to the X-axis slider 32, and the head moving device 30 moves the head 40 in the X-axis direction and the Y-axis direction by moving the X-axis slider 32 and the Y-axis slider 36. ..

The head 40 includes a Z-axis actuator 41 (see FIG. 3) that moves the suction nozzle in the Z-axis (up and down) direction, and a θ-axis actuator 42 (see FIG. 3) that rotates the suction nozzle around the Z-axis. By communicating a negative pressure source with the suction port of the suction nozzle, the head 40 can apply a negative pressure to the suction port to suck the parts. Further, the head 40 can release the suction of the parts by applying the positive pressure to the suction port by communicating the positive pressure source with the suction port of the suction nozzle.

As shown in FIG. 3, the control device 60 is configured as a microprocessor centered on the CPU 61, and includes a ROM 62, an HDD 63, a RAM 64, and an input / output interface 65 in addition to the CPU 61. These are electrically connected via a bus 66. The control device 60 includes a position signal from the X-axis position sensor 34 that detects the position of the X-axis slider 32, a position signal from the Y-axis position sensor 38 that detects the position of the Y-axis slider 36, and a mark camera 43. An image signal, an image signal from the parts camera 23, and the like are input via the input / output interface 65. On the other hand, from the control device 60, a control signal to the component supply device 21, a control signal to the substrate transfer device 22, a drive signal to the X-axis actuator 33, a drive signal to the Y-axis actuator 37, and a Z-axis actuator 41. The drive signal, the drive signal to the θ-axis actuator 42, the control signal to the parts camera 23, the control signal to the mark camera 43, and the like are output via the input / output interface 65. Further, the control device 60 is connected to the management device 80 so as to be capable of bidirectional communication, and exchanges data and control signals with each other.

The management device 80 is, for example, a general-purpose computer, and includes a CPU 81, a ROM 82, an HDD 83, a RAM 84, an input / output interface 85, and the like, as shown in FIG. These are electrically connected via the bus 86. An input signal is input to the management device 80 from an input device 87 such as a mouse or a keyboard via the input / output interface 85. Further, the image signal to the display 88 is output from the management device 80 via the input / output interface 85. The HDD 83 stores the production job of the substrate S. Here, in the production job of the board S, the production of which parts are mounted on the board S in what order in each component mounting machine 10, and how many boards S on which the components are mounted are manufactured. Includes schedule. The management device 80 generates a production job based on the data input by the operator via the input device 87, and transmits the generated production job to each component mounting machine 10 to produce the production job for each component mounting machine 10. Instruct the start of.

Next, the operation of the component mounting machine 10 in the component mounting system 1 of the present embodiment configured in this way will be described. In particular, the operation when the component is mounted on the substrate S and the mounted state of the mounted component is inspected will be described. FIG. 4 is a flowchart showing an example of the component mounting process executed by the CPU 61 of the control device 60. This process is executed when the management device 80 instructs the start of production with the production job. When the control device 60 is instructed to start production, the control device 60 performs the component mounting process according to the production job received from the management device 80.

When the component mounting process is executed, the CPU 61 of the control device 60 first controls the board transfer device 22 so that the board S is carried in (step S100). Subsequently, the CPU 61 determines whether or not the learning process described later is being executed (step S110), and when it is determined that the learning process is being executed, the mark camera 43 is placed above the carried-in substrate S. The mark camera 43 is controlled so as to control the head moving device 30 and image the substrate S (step S120), and the process proceeds to step S130. As a result, an image of the board before the component is mounted (board image before component mounting) is acquired. On the other hand, when the CPU 61 determines that the learning process is not being executed (learned), the CPU 61 skips step S120 and proceeds to step S130.

Next, the CPU 61 controls the head moving device 30 so that the suction nozzle comes above the component supply position of the component supply device 21, and performs a suction operation that controls the head 40 so as to suck the component to the suction nozzle (step). S130). Then, the CPU 61 controls the head moving device 30 so that the attracted component comes above the target mounting position of the substrate S (step S140), and controls the head 40 so that the component is mounted at the target mounting position of the substrate S. Perform the mounting operation (step S150). The process of step S140 is performed so that the attracted parts pass above the parts camera 23 and reach above the target mounting position. The CPU 61 takes an image of the sucked component when it passes above the parts camera 23, calculates the suction deviation amount of the component based on the obtained image, and mounts the target based on the calculated suction deviation amount. Correct the position. When the mounting operation is performed in this way, the CPU 61 controls the mark camera 43 so as to image the substrate S on which the components are mounted (step S160). As a result, an image of the board after the component is mounted (board image after component mounting) is acquired. Then, the CPU 61 controls the substrate transfer device 22 so that the substrate S is carried out (step S170), and ends this process.

Next, the learning process will be described. The learning process is a process of creating learning data used for mounting inspection of parts, and in the present embodiment, the learning process is repeatedly executed by the management device 80 at predetermined time intervals. FIG. 5 is a flowchart showing an example of the learning process executed by the CPU 81 of the management device 80. The learning process is repeatedly executed at predetermined time intervals while the above-mentioned component mounting process is being executed.

When the learning process is executed, the CPU 81 of the management device 80 first determines whether or not the learning is incomplete (step S200). When the CPU 81 determines that the learning is incomplete, it determines whether or not the board image before component mounting has been acquired (step S210). As described above, the image of the board before mounting the component is acquired by taking an image of the board S with the mark camera 43 before mounting the component on the board S in step S120 of the component mounting process. When the CPU 81 determines that the board image before component mounting has been acquired, the CPU 81 generates learning data for component inspection by machine learning (for example, a support vector machine or the like) based on the board image before component mounting with no component as teacher data (step S220). ). Here, the learning data for component inspection is generated by pairing the input pre-mounting substrate image of the component and the output inspection result (without components) and obtaining the relationship between the substrate image and the inspection result.

Next, the CPU 81 determines whether or not the board image has been acquired after mounting the components (step S230). As described above, the image of the board after mounting the components is acquired by taking an image of the board S with the mark camera 43 after mounting the components on the board S in step S160 of the component mounting process. When the CPU 81 determines that the board image after mounting the component has been acquired, the CPU 81 displays the board image after mounting the component on the display 88, accepts the input of the teacher data with or without the component (step S240), and waits for the teacher data to be input. (Step S250). The operator determines whether or not the component to be inspected is shown in the board image after mounting the component displayed on the display 88, and if it is determined that the component is visible, the operator inputs "with component". If it is determined that the parts are not shown, enter "No parts". When the CPU 81 inputs the teacher data of the presence or absence of parts, the CPU 81 generates (updates) the learning data for parts inspection by the above-mentioned machine learning based on the input teacher data and the board image after mounting the parts (step S260), and the teacher data. The acquisition number N of is incremented by a value of 1 (step S270). To generate learning data for component inspection, the relationship between the board image and the inspection result is obtained by pairing the input board image after mounting the component and the output teacher data of the inspection result (presence or absence of the component input by the operator). Is done by.

Then, the CPU 81 determines whether or not the acquisition number N of the teacher data is equal to or greater than the predetermined number Nref (step S280). Here, the predetermined number Nref is determined to be a necessary and sufficient number for ensuring the accuracy of the mounting inspection using the learning data. When the CPU 81 determines that the acquisition number N of the teacher data is less than the predetermined number Nref, the CPU 81 returns to step S210 and generates learning data for component inspection by machine learning based on the board image before component mounting and the board image after component mounting. The process of (updating) is repeated. When the CPU 81 determines in step S280 that the number of acquired teacher data N is equal to or greater than a predetermined number Nref in the iterative process, it determines that the learning has been completed (step S290), and ends this process.

When the CPU 81 determines that the learning is completed in step S200, it determines whether or not the re-learning execution condition is satisfied (step S300). When the CPU 81 determines that the re-learning execution condition is not satisfied, the main process is terminated, and when it is determined that the re-learning execution condition is satisfied, the learning history of steps S220 and S260 is cleared and the learning history is cleared (step S310). , The number N of acquired teacher data is initialized to a value of 0 (step S320), and the process proceeds to step S210. As a result, until the number of acquired teacher data N becomes a predetermined number Nref or more, learning (re-learning) based on the pre-parts mounting board image and the teacher data without parts, and the post-part mounting board image and the component mounting are performed again. Learning (re-learning) based on the teacher data of the presence or absence of parts input by the operator who viewed the rear board image is executed, and the learning data for parts inspection is regenerated.

Here, the determination as to whether or not the re-learning condition is satisfied is performed by executing the re-learning execution condition success / failure determination process of FIG. In the re-learning execution condition success / failure determination process, the CPU 81 of the management device 80 mounts on the board S whether or not there is a re-learning instruction from the operator (step S400), whether or not the lot of the board S has been switched (step S410). It is determined whether or not the lot of parts has been switched (step S420), whether or not a predetermined time has elapsed since the last learning (step S430), and whether or not the production job has been edited (step S440). If any of steps S400 to S440 is negative, the CPU 81 determines that the relearning execution condition is not satisfied (step S450), ends this process, and one of them is positive. If it is determined, it is determined that the re-learning execution condition is satisfied (step S460), and this process is terminated. The reason for re-learning when the lot of the substrate S or the component is switched is that the image of the substrate S or the component may change before and after the switching. Further, the reason why the re-learning is performed when a predetermined time has elapsed since the previous learning is that the image of the substrate S and the parts may change with the aging of the production environment such as lighting. Further, when the production job is edited, it is relearned because the image of the part may change before and after the dimension, shape, mounting position, etc. of the part are edited.

The CPU 81 stores the lot information of the board S in advance in the HDD 83 in association with the board ID that identifies the board S, and when the board S is carried into the component mounting machine 10, it corresponds to the board ID of the board S. The lot information is read out to determine whether or not the lot of the substrate S has been switched. Then, when the CPU 81 determines that the lot switching of the board S has occurred, the CPU 81 executes re-learning, and identifies the learning data for component inspection generated by the re-learning to identify the learning data (learning ID) and the board ID. Is stored in the HDD 83 in association with. Further, the CPU 81 stores the lot information of the component in advance in the HDD 83 in association with the component ID that identifies the component, and when the component is mounted on the substrate S carried into the component mounting machine 10, the component of the component is mounted. The lot information corresponding to the ID is read out to determine whether or not the lot of parts has been switched. Then, when the CPU 81 determines that the lot switching of the parts has occurred, the CPU 81 executes re-learning and stores the parts inspection learning data generated by the re-learning in the HDD 83 in association with the learning ID and the board ID. The learning data for component inspection stored in this way is used for the next inspection process.

Next, the mounting inspection of parts using the learning data for parts inspection will be described. FIG. 7 is a flowchart showing an example of the inspection process executed by the CPU 81 of the management device 80. This process is repeatedly executed at predetermined time intervals while the above-mentioned component mounting process is being executed. Here, the CPU 81 may execute the inspection process after waiting for the learning by the learning process described above to be completed, or if the learning data for component inspection is generated, the learning of the component is completed. The inspection process may be executed without waiting (before the teacher data is available).

When the inspection process is executed, the CPU 81 of the management device 80 first determines whether or not the board image has been acquired after mounting the components (step S500) and whether or not the learning data for component inspection exists (step S510). Judge each. If it is determined that the board image has not been acquired after the component is mounted or the learning data for component inspection does not exist, the CPU 81 determines that the component inspection cannot be performed, and ends this process. On the other hand, when the CPU 81 determines that the component-mounted board image is acquired and the component inspection learning data exists, the CPU 81 recognizes the component in the component-mounted board image using the component inspection learning data (image). By performing the processing), a mounting inspection is performed to determine whether or not the component to be inspected is mounted on the substrate S (step S520). Here, in the mounting inspection, the corresponding learning data for component inspection is read from the HDD 83 based on the identification information (board ID and component ID) of the board S and the component to be inspected, and the read learning data (recognition model) is used. It is performed depending on whether or not the component can be recognized from the board image after mounting the component. The CPU 81 determines whether or not the recognition of the component is successful by the recognition process (step S530), and if it determines that the component recognition is successful, determines that the component is mounted normally (step S540). When the present process is terminated and it is determined that the recognition of the component has failed, it is determined that a mounting error has occurred (step S550), and the present process is terminated.

Here, the correspondence between the main elements of the embodiment and the main elements of the invention described in the column of disclosure of the invention will be described. That is, the mark camera 43 corresponds to the image pickup device, the component mounting machine 10 corresponds to the component mounting machine, the component mounting system 1 corresponds to the component mounting system, and the CPU 81 of the management device 80 that executes the learning process serves as the learning device. The CPU 81 of the management device 80 that executes the inspection process corresponds to the inspection device. Further, the HDD 83 of the management device 80 corresponds to the storage device. Further, the head 40 corresponds to the head, and the head moving device 30 corresponds to the head moving device.

It is needless to say that the present invention is not limited to the above-described embodiment, and can be implemented in various embodiments as long as it belongs to the technical scope of the invention of the present disclosure.

For example, in the above-described embodiment, the CPU 81 of the management device 80 clears the learning history and initializes the acquisition number N of the teacher data to a value 0 when the re-learning condition is satisfied in the learning process. Re-learning is performed by machine learning based on teacher data until the acquired number N becomes a predetermined number Nref or more. However, the CPU 81 may perform additional learning to additionally acquire teacher data and update the learning data without clearing the learning history. FIG. 8 is a flowchart showing a learning process of a modification executed by the CPU 81 of the management device 80. Of the steps of the learning process of the modified example, the same step numbers as those of the learning process of the embodiment are assigned the same step numbers, and the description thereof will be omitted because they are duplicated. In the learning process of the modified example, when the CPU 81 determines that the learning is completed in step S200, it determines whether or not the additional learning condition is satisfied (step S300B). When the CPU 81 determines that the additional learning condition is not satisfied, the present process is terminated, and when it is determined that the additional learning condition is satisfied, the number N of teacher data acquisitions is set to a predetermined value α less than the predetermined number Nref. Decrement (step S320B), the process proceeds to step S210, and additional learning is performed. In the additional learning, the teacher data is additionally acquired and the parts are inspected until the acquired number N of the teacher data becomes the predetermined number Nref or more, that is, until the teacher data is newly acquired the predetermined value α, without clearing the learning history. This is done by updating the training data.

Here, the determination as to whether or not the additional learning condition is satisfied is performed by executing the additional learning execution condition success / failure determination process shown in FIG. In the additional learning execution condition success / failure determination process, the CPU 81 of the management device 80 determines whether or not there is an additional learning instruction from the operator (step S400B) and whether or not the lot of the substrate S is switched, as in the relearning execution condition success / failure determination process. (Step S410), whether the lot of parts to be mounted on the board S has been switched (step S420), whether a predetermined time has passed since the last learning (step S430), and whether the production job has been edited. (Step S440), respectively. If any of steps S400B and S410 to S440 is negative, the CPU 81 determines that the additional learning execution condition is not satisfied (step S450B), terminates this process, and affirms any of them. If it is a specific determination, it is determined that the additional learning execution condition is satisfied (step S460B), and this process is terminated.

In the above-described embodiment, in the inspection process, the CPU 81 inspects whether or not the component is actually mounted on the substrate S after the mounting operation is performed as a mounting inspection using the learning data for component inspection. did. However, the CPU 81 may inspect the substrate S for mounting deviation (deviation in the XY direction or deviation in the rotation direction) as a mounting inspection using the learning data.

When the component mounting machine 10 includes a plurality of blocks (child boards) and mounts a board in which information (skip information) indicating whether or not to mount the components is set for each block, "no skip" is displayed. Parts are mounted on the set blocks, and parts are not mounted on the blocks for which "with skip" is set. Learning before mounting components on a board containing a plurality of blocks is performed for all blocks regardless of whether or not skipping is performed. On the other hand, learning after component mounting is performed only on the block for which no skip is set, and not for the block for which skip is set. As a result, only the mounted parts can be learned more reliably.

As described above, the component mounting system of the present disclosure is a component mounting system including a plurality of component mounting machines that are arranged in the substrate transport direction and each has an image pickup device that images a substrate, and performs a mounting operation for mounting components. The image pickup device captures an image of the component mounting pre-board before the component mounting and the component mounting post-board after the component mounting operation, and machine learning is performed based on the captured image of the component mounting pre-board to perform machine learning. A learning device that acquires teacher data based on the captured image of the above and generates learning data for component inspection by performing machine learning based on the acquired teacher data and the captured image of the board after mounting the component, and the component mounting. An inspection device that inspects the post-component mounting substrate by imaging the rear substrate with the imaging device and performing image processing for recognizing the component in the captured image of the component mounting post-board using the learning data for component inspection. And, the gist is to prepare.

The component mounting system of this disclosure includes a learning device and an inspection device. The learning device images the pre-component mounting board and the post-component mounting board, performs machine learning based on the captured image of the pre-component mounting board, and acquires and acquires teacher data based on the captured image of the post-component mounting board. Machine learning is performed based on the teacher data and the captured image of the board after mounting the components to generate learning data for component inspection. The inspection device inspects the board after mounting the component by taking an image of the board after mounting the component and performing image processing for recognizing the component in the captured image of the board after mounting the component using the learning data for component inspection. In this way, by generating learning data based on the captured images of the board before component mounting and the board after component mounting, it is possible to improve the recognition accuracy when recognizing the component from the captured image of the board after component mounting. .. As a result, the quality of the board can be ensured by appropriately inspecting the board after mounting the components.

In such a component mounting system of the present disclosure, when a predetermined re-learning condition is satisfied, the learning device images the pre-parts mounting board and the post-parts mounting board with the image pickup device, and before the component mounting. Machine learning is performed based on the captured image of the board, and the teacher data based on the captured image of the board after mounting the component is reacquired, and the machine learning is performed based on the re-acquired teacher data and the captured image of the board after mounting the component. Is performed to regenerate the learning data for component inspection, and the regenerated learning data for component inspection is generated in association with the identification information of the board after mounting the component or the identification information of the component mounted on the board after mounting the component. Stored in a storage device, the inspection device images the board after mounting the component with the imaging device, and acquires identification information of the board after mounting the component or identification information of the component mounted on the board after mounting the component. The learning data for component inspection corresponding to the acquired identification information is read from the storage device, and the read learning data for component inspection is used to perform image processing for recognizing the component in the captured image of the board after mounting the component. Therefore, the substrate may be inspected after mounting the components. By doing so, it is possible to maintain good recognition accuracy of the parts. Here, the re-learning conditions are as follows: when instructed by the operator, when the lot of the board is switched, when the lot of the parts to be mounted is switched, when a predetermined time has elapsed since the last learning was executed, or when production is performed. It may be a condition that is satisfied when the job is changed.

Further, in the component mounting system of the present disclosure, when a predetermined additional learning condition is satisfied, the learning device images the pre-component mounting board and the post-component mounting board with the imaging device, and mounts the component. Additional learning is performed based on the captured image of the front board, the teacher data based on the captured image of the board after mounting the component is reacquired, and the teacher data is added based on the reacquired teacher data and the captured image of the board after mounting the component. Learning may be performed to update the learning data for the component inspection. By doing so, it is possible to maintain good recognition accuracy of the parts. Here, the additional learning conditions are as follows: when instructed by the operator, when the lot of the board is switched, when the lot of the parts to be mounted is switched, when a predetermined time has elapsed since the last learning was executed, or when production is performed. It may be a condition that is satisfied when the job is changed.

Further, in the component mounting system of the present disclosure, the inspection device may determine whether or not the component to be inspected is mounted on the component mounting board as the inspection of the component mounting board.

Further, in the component mounting system of the present disclosure, the component mounting machine includes a head capable of collecting the component and a head moving device for moving the head, and the imaging device includes the head by the head moving device. It may be provided so as to be movable together with.

This disclosure can be used in the manufacturing industry of component mounting machines and component mounting systems.

1 Parts mounting system, 2 Printing machine, 3 Printing inspection machine, 10 Parts mounting machine, 11 Housing, 21 Parts supply device, 22 Board transfer device, 23 Parts camera, 24 Nozzle station, 30 Head moving device, 31 X-axis guide Rail, 32 X-axis slider, 33 X-axis actuator, 34 X-axis position sensor, 35 Y-axis guide rail, 36 Y-axis slider, 37 Y-axis actuator, 38 Y-axis position sensor, 40 head, 41 Z-axis actuator, 42 θ Axis actuator, 43 mark camera, 60 control device, 61 CPU, 62 ROM, 63 HDD, 64 RAM, 65 input / output interface, 66 bus, 80 management device, 81 CPU, 82 ROM, 83 HDD, 84 RAM, 85 input / output Interface, 86 bus, 87 input device, 88 display, S board.

Claims

It is a component mounting system equipped with a plurality of component mounting machines that are lined up in the substrate transport direction and each has an imaging device that images a substrate.
The pre-component mounting board before the component mounting operation and the post-component mounting board after the mounting operation are imaged by the imaging device, and machine learning is performed based on the captured image of the pre-component mounting board. This is performed to acquire teacher data based on the captured image of the board after mounting the component, and machine learning is performed based on the acquired teacher data and the captured image of the board after mounting the component to generate learning data for component inspection. Learning device and
The component-mounted board is inspected by imaging the component-mounted board with the imaging device and performing image processing for recognizing the component in the captured image of the component-mounted board using the learning data for component inspection. Inspection equipment to be performed and
A component mounting system equipped with.
The component mounting system according to claim 1.
When a predetermined relearning condition is satisfied, the learning device takes an image of the pre-component mounting board and the post-component mounting board with the imaging device, and machine learning is performed based on the captured image of the pre-component mounting board. Is performed, and the teacher data based on the captured image of the board after mounting the component is reacquired, and machine learning is performed based on the re-acquired teacher data and the captured image of the board after mounting the component to perform learning data for component inspection. Is regenerated, and the regenerated learning data for component inspection is stored in the storage device in association with the identification information of the board after mounting the component or the identification information of the component mounted on the board after mounting the component.
The inspection device images the board after mounting the component with the imaging device, acquires the identification information of the board after mounting the component or the identification information of the component mounted on the board after mounting the component, and obtains the acquired identification information. After the component is mounted, the corresponding learning data for component inspection is read from the storage device, and the read learning data for component inspection is used to perform image processing for recognizing the component in the captured image of the board after the component is mounted. Inspect the board,
Component mounting system.
The component mounting system according to claim 2.
The re-learning conditions are specified by the operator, the lot of the board is switched, the lot of the parts to be mounted is switched, the predetermined time has passed since the last learning was executed, or the production job is changed. It is a condition that is satisfied when it is done,
Component mounting system.
The component mounting system according to any one of claims 1 to 3.
When a predetermined additional learning condition is satisfied, the learning device takes an image of the pre-component mounting board and the post-component mounting board with the imaging device, and additional learning is performed based on the captured image of the pre-component mounting board. , And re-acquire the teacher data based on the captured image of the board after mounting the component, and perform additional learning based on the re-acquired teacher data and the captured image of the board after mounting the component to learn for the component inspection. Update the data,
Component mounting system.
The component mounting system according to claim 4.
The additional learning conditions are specified by the operator, the lot of the board is switched, the lot of the parts to be mounted is switched, the predetermined time has passed since the last learning was executed, or the production job is changed. It is a condition that is satisfied when it is done,
Component mounting system.
The component mounting system according to any one of claims 1 to 5.
The inspection device determines whether or not a component to be inspected is mounted on the component-mounted board as an inspection of the component-mounted board.
Component mounting system.
The component mounting system according to any one of claims 1 to 6.
The component mounting machine includes a head capable of collecting the component and a head moving device for moving the head.
The image pickup device is provided so as to be movable together with the head by the head moving device.
Component mounting system.